Emergence of Hybrid Resistance and Virulence Plasmids Harboring New Delhi Metallo-β-Lactamase in Klebsiella pneumoniae in Russia

The emergence of carbapenem-resistant hypervirulent Klebsiella pneumoniae (CR-hvKp) is a new threat to healthcare. In this study, we analyzed nine CR-hvKp isolates of different sequence-types (ST) recovered from patients with nosocomial infections in two hospitals in Saint Petersburg. Whole-genome sequencing showed that eight of them harbored large mosaic plasmids carrying resistance to carbapenems and hypervirulence simultaneously, and four different types of hybrid plasmids were identified. BLAST analysis showed a high identity with two hybrid plasmids originating in the UK and Czech Republic. We demonstrated that hybrid plasmids emerged due to the acquisition of resistance genes by virulent plasmids. Moreover, one of the hybrid plasmids carried a novel New Delhi metallo-beta-lactamase (NDM) variant, differing from NDM-1 by one amino acid substitution (D130N), which did not provide significant evolutionary advantages compared to NDM-1. The discovery of structurally similar plasmids in geographically distant regions suggests that the actual distribution of hybrid plasmids carrying virulence and resistance genes is much wider than expected.


Introduction
Klebsiella pneumoniae (Kpn) is one of the leading causative agents of hospital-and community-acquired infections worldwide. Multidrug-resistant (MDR) and hypervirulent (hvKP) genetic lineages, as well as a recently emerging trend toward their convergence, are the greatest dangers associated with Kpn [1]. Among MDR genetic lineages, the greatest threat is associated with resistance to carbapenems due to the production of carbapenemases. Metallo-β-lactamases (MBLs) are of particular concern since they are insensitive to clinically available inhibitors, including diazobicyclooctanes, thereby limiting the therapeutic options [2].
There are several evolutionary pathways for the formation of Kpn genetic lineages that demonstrate high virulence and resistance to carbapenems, all of which are associ-ated with horizontal gene transfer. Two of these pathways are related to the acquisition of additional plasmids through MDR or hvKP isolates. According to [3], MDR Kpn is more likely to acquire virulence genes than hvKP is to acquire drug-resistance genes due to the ability of MDRs to easily acquire additional genes. However, the majority of carbapenem-resistant hvKP (CR-hvKP) reported to date emerged because of the acquisition of carbapenemase-encoding plasmids (mainly bla KPC-2 and bla NDM-1 ) by hvKP [4]. The reason for the prevalence of this evolutionary pathway is probably the fact that virulence plasmids like pLVPK are nonconjugative which is confirmed by their predominant distribution within a limited number of genetic lines.
Since New Delhi metallo-beta-lactamase (NDM) was first reported in Sweden [5], 31 variants of NDM-type carbapenemases have been identified. The evolution of NDMs results from amino acid substitutions (AAS) or insertions that possess different hydrolytic activity toward antibiotics, which affects the treatment strategy [6]; hybrid plasmid formation and evolution of resistance probably occur simultaneously. Thus, nowadays NDM variants are evolving towards enhanced stability, Zn(II) binding affinity [7], or the ability to function as mono-zinc enzymes with high catalytic efficiency [8].
In a previous study, we described the emergence of Kpn sequence type (ST) 395 and ST147 isolates in an oncology hospital carrying virulence and bla NDM-1 genes simultaneously; however, we were unable to pinpoint the location of these genes [9]. In another study from Russia, the authors were also unable to accurately determine the location of the resistance and virulence genes [10]. Shaidulina et al. used Illumina MiSeq and ONT MinION systems and succeeded in localizing the virulence genes and bla OXA-48 on different plasmids. To the best of our knowledge, no reports of CR-hvKp isolates with confirmed hybrid plasmids coharboring genes of carbapenemases and virulence genes have been published in Russia so far [11]. In the present study, we aimed to describe the emergence of Kpn isolates of different STs (ST15, ST147, ST395, and ST874) with hybrid plasmids co-harboring virulence genes and bla NDM in the hospitals in St. Petersburg. Moreover, in one of the isolates, bla NDM was identified as a new variant (bla NDM-29 ).

Clinical Manifestations and Antimicrobial Susceptibility of Recovered Isolates
Eight hypermucoviscous bla NDM -positive Kpn isolates, and one hypermucoviscous but bla NDM -lacking isolate were included in the study (Table 1). One isolate was recovered from a patient in a general hospital and six from patients in an oncology hospital. All oncology patients had solid tumors. Three patients experienced liver failure. The clinical details of the patients are shown in Supplementary Material Table S1. Most isolates were obtained from sites that were not associated with invasive diseases. Two isolates attributed to different clonal lines were recovered from one patient.
The Kpn isolates were resistant to all the tested cephalosporins, carbapenems, fosfomycin, ceftazidime/avibactam, and a trimethoprim-sulfamethoxazole combination, but were susceptible to polymyxin, tigecycline, and aztreonam/avibactam. In addition, they showed miscellaneous resistance to aminoglycosides. The antimicrobial agents and phenotypes of all the tested isolates are listed in Supplementary Material Table S3.

General Sequencing Results
On average, 69,569 reads were generated by Oxford Nanopore with a base call accuracy of Q11 per isolate (average read length: 3.5 kbp). The output from the Illumina platform was, on average, 1,354,011 paired-end reads (average read length: 300 bp) with a sequencing quality of Q36. The hybrid assemblies provided an average of 145.96-fold coverage. The contig number (>1000 bp) of the latter ranged from three to six and yielded an average total length of 5,875,483 bp, with the longest corresponding to a chromosome (approx. 5.3 Mbp). The average N50 size for all isolates was 5,019,528 bp, and the mean GC content was 56.6%.

Plasmids
All the studied isolates carried one to four plasmids, including a large (~311-422 kb) virulence plasmid. Sequence analysis revealed that these multireplicon plasmids were a fusion of IncFIB and IncHI1B backbones and carried both antimicrobial resistance (AMR) and virulence genes. The phvKpST395_2024 and phvKpST395_NDM-1_1657 hybrid plasmids had an additional IncR replicon.
The closest matching plasmids from a global database were pKpvST383L (GenBank accession no. CP034201.2) and pKpvST147B_virulence (GenBank accession no. CP040726.1) of the Kpn strain from two different London hospitals [12], as well as p51015_NDM_1 (GenBank accession no. CP050380.1) from the Czech Republic. The plasmids content, as revealed by whole-genome sequencing (WGS), is shown in Supplementary Material Table S2. Virulence plasmids of 1970_kpn, 2024_kpn, and 2501_kpn were found to be shorter (~315 kb), mostly because they lacked the tellurium resistance cassette in comparison to plasmids in the other hvKp. A schematic representation of the circular genome maps of mosaic plasmids of CR-hvKp Kpn and linear alignment are shown in Figure 1 and Supplementary Materials, Figure S1a, respectively.

AMR Determinants
CR-hvKp isolates, together with those described in the previous section, were found to possess an array of AMR genes. In addition to the highly prevalent chromosomal quinoxaline resistance (oqxAB) and fosfomycin resistance genes (fosA), the chromosome of the KpvST383_NDM_OXA-48 strain also contained catA1, tet(A), mph(A), and the chromosome of 2501_kpn contained aac(3)-IId. Apart from the bla SHV [13], bla CTXM-15 was found on the ISEcp1-bla CTX-M-15 element. Class A extended-spectrum beta-lactamases were located on different mobile genetic elements (MGEs) harbored through fusion (IncFIB/IncHI1B, IncFIB/IncFII) or IncR backbone plasmids. bla OXA-9 and bla TEM-1A were prevalent in either Tn6238-or IS26-dependent translocatable units integrated into IncR or IncHI1B/IncFIB plasmid types, or class 1 integrons. bla OXA-48 was associated with inverted Tn1999.2 on the IncL plasmid, bla OXA-244 , in turn, with IS1R-mediated MGE. bla OXA-1 together with catB3 and aac(6')-Ib-cr were linked to IS26-mediated insertion sequences positioned on fusion or IncR backbone plasmids. A variety of MGEs containing bla TEM-1b were integrated into the latter types of plasmids in the same manner.
All the isolates were found to carry bla NDM-1 except for 2024_kpn, which did not contain any MBL genes. In most of the isolates, bla NDM was integrated into the IncFIB/IncHI1B plasmid, while in the KpvST147B_SE1_1_NDM isolate, KpvST383_NDM_OXA-48 carried bla NDM-5 . In addition, WGS revealed bla NDM-29 in 1970_kpn, which was carried on a 311-kb mosaic plasmid referred to as phvKpST147_NDM-29.
Antibiotics 2021, 10, 691 5 of 13 resistance cassette in comparison to plasmids in the other hvKp. A schematic representation of the circular genome maps of mosaic plasmids of CR-hvKp Kpn and linear alignment are shown in Figure 1 and Supplementary Materials, Figure S1a, respectively.

Enzymes
Amino Acids bla NDM-29 was located on a Tn125-like hypothetical transposon (Supplementary Material Figure S1b) [14]. In phvKpST147_NDM-29, bla NDM-29 was flanked upstream by ISAba125-ISspu2-ISAba125 [15][16][17][18] and downstream by IS26, turning into an IS26-dependent composite transposon structure. Additionally, aphA6 was part of the remnant of a composite transposon originally with two terminal direct copies of ISAba14 [19]. Downstream of bla NDM-29, coding for ble-iso-tat-dct, was truncated at dct via the insertion of IS26. IS26s were found at different positions in the NDM genetic environment, which indicated increased activity and multiple independent acquisitions of IS26 [20]. However, as part of the IS26dependent composite transposon, IS26 played a key role in the mobilization of composite transposons between different plasmids or chromosomes [14]. This genetic context was identical to that of NDM-containing plasmids in the Czech Republic and the UK.

Functional Analysis of bla ndm-29
The susceptibility test results showed that the clinical Kpn isolate, 1970_kpn, exhibited high-level resistance to all examined carbapenems, cephalosporins, aminoglycosides, fluoroquinolones, and a piperacillin/tazobactam combination, but was susceptible to aztreonam. The antimicrobial agents and phenotypes of all the tested isolates are listed in Supplementary Materials, Table S4. Antimicrobial susceptibility test results initially suggested the presence of an MBL, which was confirmed by PCR targeting NDM-like sequences.
The expression of bla NDM-29 and bla NDM-1 in E. coli XL10-Gold, compared to the control strain, resulted in reduced susceptibility to third-and fourth-generation cephalosporins, carbapenems, and piperacillin/tazobactam. All the transformants were susceptible to aztreonam. Both pEGFP-N3 transformants exhibited significantly higher resistance to gentamicin compared to the pJET1.2 transformants, which proved that the first vector contained an aminoglycoside resistance gene.
Under conditions designed to mimic zinc(II) scarcity by including the metal chelator ethylenediaminetetraacetic acid (EDTA), all tested isolates showed a significant decrease in minimal inhibitory concentration (MIC) values for broad-spectrum cephalosporins and carbapenems. Transformants containing NDM-29 showed the same susceptibility to third and fourth generation cephalosporins and carbapenems as NDM-1. Both pJET1.2 transformants demonstrated significantly higher resistance to ampicillin compared to the pEGFP-N3 transformants, even under zinc-limiting conditions, which is a selective marker for ampicillin resistance.
A different substitution at the 130th position presented in NDM-14 (D130G) and NDM-8 (D130G, M154L). Based on our findings, the evaluation of the impact of the D130N substitution on the hydrolytic activity toward β-lactam antibiotics revealed that transformants of bla NDM-29 exhibited the same β-lactam resistance as the transformants of bla NDM-1 , unlike D130G which has been found to confer higher resistance to meropenem and imipenem than bla NDM-1 and indirectly influence on the active site [21]. The stability changes at the 130th position (∆∆Gu = −0.3330) is neutral in comparison with NDM-1 according to EASE-MM. The comparison of NDM-1 and NDM-29 structures is represented in Supplementary Materials, Figure S2.
In contrast, for NDM-7 and NDM-4, the leucine residue at position 154 (M154L) extended the hydrolytic activity toward carbapenems [22,23] and for NDM-4 and NDM-15, the metal-binding [7]. Which is in agreement with [8] showing resistance >3-4 dilution increases in MIC values for two or more antibiotics in zinc-depleted condition of NDM-4, 8, 7, 15. A233V, in turn, was not only a stabilizing substitution, not affecting zinc-binding abilities [7], but also appeared to further increase the carbapenemase activity under zinclimiting growth conditions [24]. The D130N amino acid change also presented in NDM-29, 7, 19. We suggest that the asparagine residue at the 130th position does not have an effect on the formation of the active site.
Silver resistance gene (silA), usually located on the virulence plasmids, was found on the chromosome of 1971_kpn, 2512_kpn, KpvST147B_SE1_1_NDM, and KpvST383_NDM_OXA-48 strains. In KpvST147B_SE1_1_NDM, the ecp locus was mobilized by putative IME. In Kpn 51015 and 2512_kpn strains, an additional siderophore system (ybt), its receptor locus (fyu), and iron regulatory protein (irp) genes, acquired by putative ICE with T4SS, were integrated into the chromosome.

Restriction Modification and TA Systems
The type II methyltransferase M.Kpn34618Dcm was present on the chromosomes of the analyzed isolates. The chromosomes of 1659_kpn and 2566_kpn isolates carried the type I restriction enzymes StySKI and 1970_kpn M.Sen1921I methyltransferase (type I). As for the mosaic plasmids, a DNA cytosine methyltransferase was found in all isolates except for 2024_kpn.
Using a collection of Pfam toxin profiles embedded in SLING, we identified 21 toxin groups (18 of type II and three of type IV (CbtA, Cpta, AbiEii)). TA systems were highly prevalent in all isolates of Kpn, with a median of 17 loci per genome. GNAT_acetyltran, HD, HipA, Cpta, and Fic toxin domains were highly prevalent on the chromosomes of the analyzed Kpn isolates. Different combinations of HigB, CcdB, CbtA, ParE, PIN, Gp49, RelE, Polyketide_cyc2, YdaT, PIN, and AbiEii toxin domains were also found.
Equally important is the vapBC of the type II TA system that has been identified on all mosaic plasmids. In addition, higAB (PIN domain) was integrated into the hybrid plasmids of 1657_kpn and 2512_kpn. PIN, ParE, RelE, Gp49 were also observed on different plasmids besides mosaic ones together with HicA, NTP_transf_2, PemK, YoeB, CbtA, and ANT toxin domains.

CRISPR-Cas System
Although the prevalence of IncF plasmids was recognized among strains in the absence of CRISPR-Cas systems [25], type I-E CRISPR-Cas locus with one CRISPR array was identified in the chromosomes of 1659_kpn, 1970_kpn, and 2566_kpn, and type I-E* CRISPR system with one CRISPR array in 2501_kpn. It consisted of eight genes (cas1,2,3,5, cse1,2,3,4) of the cas operon and one CRISPR array with 43 identical spacers, and the mosaic plasmids of all the analyzed isolates did not contain protospacers. We found the occurrence of a type I-E* CRISPR system with two identical CRISPR arrays (10 + 9 spacers) in KpvST147B_SE1_1_NDM and 5105 strain chromosomes. The isolate KpvST383_NDM_OXA-48 harbored the type I-E CRISPR-Cas locus with one CRISPR array consisting of 15 spacers, the last of which coincided with the last spacer of the CRISPR array with 43 spacers.
Similarly, a CRISPR array with 16 spacers was found on the aforementioned mosaic plasmids not adjacent to cas genes (12 spacers on phvKpST395_2024). One spacer of all the mosaic plasmid CRISPR arrays contained a Gifsy-2 prophage sequence located on the chromosomes of all the aforementioned strains (except for KpvST147B_SE1_1_NDM and 5105). Similarly, the hybrid plasmids carried another spacer that matched the traL of the Kpn conjugative IncF plasmid. This ensured the barrier for acquiring the IncF mosaic plasmid, which prevented the decrease in fitness due to the probable presence of multiple AMR plasmids.

Discussion
Here, we showed the emergence of Kpn isolates of different STs (ST15, ST147, ST395, and ST874) with hybrid plasmids coharboring virulence genes and bla NDM in the hospitals in St. Petersburg.
Alignment of four distinct plasmid types, phvKpST147_NDM-29, phvKpST395_2024, and pKpvST147B_virulence showed the diversification of plasmid branches (Supplementary Materials, Figure S1d). We suggest the existence of a progenitor of the phvKpST395_2024 plasmid of the IncFIB/IncHI1B replicon type with a full hypervirulence gene cluster (predecessor 1). Possible evolutionary pathways that generated four distinct types of mosaic plasmids were as follows: (1) the acquisition of Tn-125-like (such as study isolates 1970_kpn, 1971_kpn, 2501_kpn) or (2) tellurium resistance gene cluster (such as pKpvST147B_virulence in KpvST147B_SE1_1_NDM strain) by predecessor 2 following the acquisition of the region containing Tn1548 by predecessor 1; (3) an intermediate type of mosaic plasmid produced by the acquisition of Tn125-like genes from the IncFIB plasmid by the type 2 plasmid, as seen in Supplementary Materials, Figure S1c (such as study isolates 1659_kpn, 2471_kpn, 2512_kpn, 2566_kpn); (4) deletion of the region containing genes associated with hypervirulence (iutA, iucBCD, p rmpA2, cobW, pagO, ydjA) and fusion with the IncR plasmid (phvKpST395_2024 in such as the isolate 2024_kpn). A schematic representation of the possible evolutionary pathway is shown in Supplementary Materials, Figure S3.
We speculated that, given the simultaneous presence of pKpvST147B_virulence, carrying the tellurium resistance cluster, and pKpvST147B_NDM_1 carrying a Tn125-like sequence within KpvST147B_SE1_1_NDM in the UK, this strain was the predecessor of the convergence of Te resistance and Tn125 in one vector, found in phvKpST395_NDM-1_1657, phvKpST147_NDM-1_1659, phvKpST874_NDM-1_2471, phvKpST395_NDM-1_2512, and phvKpST147_NDM-1_2566. The linear alignment of phvKpST147_NDM-29, pKpvST147B_ virulence, and pKpvST147B_NDM_1 is shown in Supplementary Materials, Figure S1c. These perfect mosaic structure hybrid plasmids converging MDR and hv biphenotypes within a single vector are a shortcut to the evolutionary success of these superbugs.
Our data are consistent with the previously published hypothesis for the emergence of hybrid plasmids stating the acquisition of resistance genes by a virulent plasmid [4].
Because of recombination following transmission, convergent plasmids, which simultaneously enable virulence and resistance, and have enhanced genetic plasticity, were generated. All NDM-producing CR-hvKp isolates were from multiple STs and different geographical regions, suggesting the convergence of the resistance and virulence genes in conjugative plasmids. Mosaic structures were formed upon the fusion of carbapenemaseencoding genes and virulence plasmids via MGEs [4]. Despite the high identity of the analyzed mosaic plasmids, each particular lineage of Kpn was associated with a different mobilome, suggesting an IS-mediated combination that results in such hybrid plasmids.
TA systems have not been studied well in the Kpn species complex. Overall, TA systems are associated with the stabilization of MGE-containing virulence genes and/or AMR genes and maintenance of plasmids via postsegregational killing [26]. Our data are in agreement with previous findings of high prevalence of II TA systems among virulence Antibiotics 2021, 10, 691 9 of 13 plasmids [27]. VapC toxin containing a pilus retraction protein (PilT) N-terminal (PIN domain) acts as a ribonuclease that cleaves RNA molecules, thereby reducing the rate of translation, and vapB encodes a matching antitoxin. The role of vapBC is related to bacterial pathogenicity [28]. HigB is a ribosome-dependent mRNA endoribonuclease that inhibits cell proliferation, whereas the higA antitoxin prevents the cessation of cell growth.
In the present study, we also identified CRISPR array sequences. A possible role of the CRISPR-Cas system on the mosaic plasmid is to contribute to the homologous recombination of AMR and virulence determinants in a single vector and probable promotion of cointegration plasmid formation [29] because host plasmids are fused per se.
In light of the high sequence identity of the isolated array repeats on the mosaic plasmids to the repeats in the complete CRISPR-Cas systems of closely related genomes, and the absence of transposable elements in the vicinity of isolated arrays, the most probable route of the evolution of the isolated CRISPR array is the loss of cas genes in a CRISPR-Cas locus [30]. It remains unclear whether the array isolated in this study is functional; however, it is known to be active through trans utilization by Cas proteins. Similarly, it is not apparent why CRISPR positive isolates did not recognize IncF plasmid as a foreign DNA.
Moreover, in one of the isolates, bla NDM was identified as a new variant (bla . Contrary to what we found here, in another study, D130N added stability against degradation and tolerance to Zn scarcity in NDM-7, unlike NDM-4 [7]. D130N suggested the formation of a separate branch in the evolution of NDM enzymes. The evolution of β-lactamases is usually aimed at a gradual increase in the AMR. On the other hand, evolutionary paths may not lead to the formation of highly effective branches and can be the result of a random mutation. In this case, we cannot be sure whether the mutation at position 130 was a step in the evolutionary process or an independent event. In addition, amino acids presented in NDM-29 were not located in what has been identified as the active site of NDM enzyme or the amino acid residue that binds to the zinc ions (His120, His122, His189, His250, Cys208, and Asp124) [31], despite its close occurrence.

Cloning
bla NDM-29 and bla NDM-1 cloning was conducted as described in Supplementary Material Text S1.

Genomic Sequencing
Nine isolates were characterized via Illumina MiSeq sequencing (Nextera XT libraries, paired-end 300-bp reads), and reads were assembled de novo into contigs using the SPAdes algorithm v.3.10.1. Only contigs longer than 5 kb were analyzed. Long reads were obtained using an Oxford Nanopore MinION Sequencer (SQK-LSK109 and flow cell R9.4.1). Min-ION reads were based on the Guppy software (v4.0.11) available from Oxford Nanopore technologies. The de novo hybrid assembly of both short Illumina reads, and long MinION reads were performed using Unicycler (v0.4.7).
The virulence and AMR score were calculated as shown previously [40], which reflects the accumulation of loci associated with the increased risk of clinically relevant AMR or hypervirulence.

MURINE Model of Sepsis
The virulence of eight isolates was assessed via a mouse lethality assay, and LD 50 values were determined as described previously [9].

Conclusions
The discovery of structurally similar plasmids in geographically distant regions suggests that the actual distribution of hybrid plasmids carrying virulence and resistance genes is much wider than expected. The convergence of virulence and AMR determinants in a single vector, together with the rapid evolution of NDMs, poses a global threat to healthcare systems.