Molecular characterization of carbapenem-resistant and virulent plasmids in Klebsiella pneumoniae from patients with bloodstream infections in China

ABSTRACT Bloodstream infections (BSIs) caused by carbapenem-resistant Klebsiella pneumoniae (CRKP) are potentially life-threatening and an urgent threat to public health. The present study aims to clarify the characteristics of carbapenemase-encoding and virulent plasmids, and their interactions with the host bacterium. A total of 425 Kp isolates were collected from the blood of BSI patients from nine Chinese hospitals, between 2005 and 2019. Integrated epidemiological and genomic data showed that ST11 and ST307 Kp isolates were associated with nosocomial outbreak and transmission. Comparative analysis of 147 Kp genomes and 39 completely assembled chromosomes revealed extensive interruption of acrR by ISKpn26 in all Kp carbapenemase-2 (KPC-2)-producing ST11 Kp isolates, leading to activation of the AcrAB-Tolc multidrug efflux pump and a subsequent reduction in susceptibility to the last-resort antibiotic tigecycline and six other antibiotics. We described 29 KPC-2 plasmids showing diverse structures, two virulence plasmids in two KPC-2-producing Kp, and two novel multidrug-resistant (MDR)-virulent plasmids. This study revealed a multifactorial impact of KPC-2 plasmid on Kp, which may be associated with nosocomial dissemination of MDR isolates.


Introduction
Bloodstream infections (BSIs) caused by Enterobacterales have become increasingly life-threatening, leading to a mortality rate as high as 48% [1,2]. Carbapenems remain one of the first-line of therapeutic agents for BSIs. Therefore, the emergence of carbapenemasemediated resistance represents a serious public health threat [3]. Carbapenem resistance has been associated with increased length of hospital stay and mortality of BSI patients [4]. Carbapenemase-encoding plasmids can be transferred among various Enterobacterales via horizontal gene transfer (HGT) and disseminated in hospitals [5]. As a result, carbapenem-resistant Enterobacterales (CRE) have been reported worldwide [6]. Klebsiella pneumoniae is a clinically important species and causes serious nosocomial infections such as septicemia, pneumonia, urinary tract infection, surgical site infection, and soft tissue infection [7]. In China, carbapenem-resistant Klebsiella pneumoniae (CRKP) accounts for about 64% of CRE infections [8]. Nonetheless, the characteristics of carbapenemase-encoding plasmids and their interactions with the host bacterium are not fully understood.
The hypervirulent variant of Kp (HvKp) has been increasingly reported in association with plasmidmediated virulence loci rmpA/rmpA2, iuc, and iro [9,10]. The rmpA/rmpA2 genes encode proteins regulating capsule production in Kp, while iucABCDiutA and iroBCDN are responsible for the biosynthesis of siderophores aerobactin and salmochelin, respectively [11]. Carbapenem-resistant and virulence plasmidcarrying Kp is associated with excess morbidity and mortality in China [12]. The emergence and dissemination of carbapenem-resistant HvKp (CR-HvKp) are of great concern due to the combination of virulence and lack of treatment options.
Herein, we conducted integrated epidemiological and genomic analysis to infer the resistomes, virulence determinants, and the phylogenetic relationship between CRKP and CSKP isolates. We demonstrated that ISKpn26 insertion contributed to the MDR phenotypes in all the ST11-bla KPC-2 Kp by blocking the expression of AcrAB-TolC repressor acrR. Furthermore, we identified novel MDR-virulent plasmids due to ongoing recombination in Kp, representing a significant health threat in terms of both disease and treatment.

Bacterial isolates
We collected 425 Kp isolates from the blood of BSI patients from nine tertiary hospitals in Guangdong province, China, between 2005 and 2019 (Table S1). In cases with multiple positive blood cultures, we only included the first positive blood culture. Preliminary species identification was achieved by MALDI-TOF MS (BrukerDaltonik GmbH, Bremen, Germany) and 16S rRNA sequencing. Ethical approval for this study was given by Zhongshan School of Medicine of Sun Yat-sen University under approval number 068.
Antimicrobial susceptibility testing, s1-PFGE, and Southern blotting The minimum inhibitory concentrations (MICs) were determined for 15 antibiotics for all isolates using the agar dilution method with the exception of colistin, which used the broth dilution method following the Clinical and Laboratory Standards Institute (CLSI) guidelines [13]. MIC determinations were also carried out with fixed concentration (100 μg/mL) of the efflux pump inhibitor 1-(1-naphthylmethyl)-piperazine (NMP) against 20 antibiotics among all the ST11bla KPC-2 strains. The plasmid location of the carbapenem encoding gene was determined by S1-nuclease digestion and pulsed-field gel electrophoresis (S1-PFGE), followed by Southern blotting hybridizations with a bla KPC-2 probe [14].

Galleria mellonella infection model
The virulence of target strains was determined using the wax moth (G. mellonella) larvae model [12,15]. Three doses of 1×10⁴, 1×10 5 , 1×10 6 CFU each with ten worms per group were tested. 1×10⁴ CFU was used for the injection. Controls included a PBS injection group, one group receiving no dose, a non-virulent control using E. coli MG1655, and a highlyvirulent control using HvKP4 as previously reported [12]. The larvae were incubated at 37°C in a darkroom and the survival rate was recorded every 12 h for seven days. The experiments were conducted in duplicate.

Quantitative real-time PCR (qPCR)
The experimental procedures for qPCR were modified from a previous report [16]. The total RNA of Kp strains was extracted using the bacteria RNA Extraction Kit (Vazyme Biotech, China). Reverse transcription was performed using Goldenstar TM RT6 cDNA Synthesis Kit Ver.2 (Beijing TsingKe Biotech, China). The qPCR assay was conducted using the Bio-Rad IQ thermocycler and Master qPCR Mix-SYBR (Beijing TsingKe Biotech, China) for three biological replicates and three technical replicates. Calculation of 2 -ΔΔ CT using 16S rRNA as the reference was used to determine the relative transcript levels for each target gene of acrA, acrB, and acrR. The primers used to amplify each gene are listed in Table S2.

Phylogenetic analysis
For each de novo assembly, coding sequences were predicted using Prodigal v2.6 [28] and annotated using Prokka v1.13.3 [29]. Core genes were identified and used to build the core genome using Roary v3.12 [30] with the -e -mafft setting to create a concatenated alignment of core genomic CDS. SNP-sites (https:// github.com/sanger-pathogens/snp-sites) was used to extract the core-genome SNPs (cgSNPs) [31]. The clonal strains differed by fewer than four cgSNPs [12]. Recombinogenic regions were removed with Gubbins v2.3.4 [32]. To construct a maximum likelihood phylogeny of the sequenced isolates, RAxML v8.2.10 was used with the generalized time-reversible model and a GTRGAMMA distribution to model site-specific rate variation [33]. We used iTOL [34] to visualize and edit the phylogenetic tree.

Data availability
All the 154 whole-genome sequenced data have been deposited in the NCBI database BioProject: PRJNA550041. A total of 34 bla KPC-2 -carrying or virulence plasmid sequences were deposited in the Gen-Bank database and assigned the accession numbers MT269819-MT269852.

Kp is the dominant CRE species from BSI patients
A total of nine tertiary hospitals were involved in this study, with a median of 1,851 beds (IQR 1,375-2,851). Of 425 Kp isolates from patients with BSIs collected between 2005 and 2019, 72 were CRKP. The CRKP isolates were most frequently susceptible to colistin (71/72, 99%), followed by tigecycline (60/72, 83%), and amikacin (38/72, 53%) ( Figure S1A). The carriage of rmtB and armA contributed to the resistance of amikacin and other aminoglycoside antibiotics, while the resistance of quinolones was mostly attributed to plasmid-mediated qnrB and qnrS1 genes. Furthermore, the MICs were significantly higher among CRKP for multiple antibiotics ( Figure S1B; Table S4).
By maximum likelihood (ML) phylogenetic analysis derived from the core genome SNPs among the 154 Kp isolates, four phylogroups of KpI (Kp, n = 147), KpII (Klebsiella quasipneumoniae, n = 5), KpIII (Klebsiella quasivariicola, n = 1), and KpIII (Klebsiella variicola, n = 1) were observed. The population structure of the 147 KpI isolates, including 76,408 SNPs extracted from 3,617,098 bp sequences concatenated from 3,780 core genes was explored. Phylogenetic analyses revealed a deep branching and scattered population structure that was broadly classified into distinct phylogenetic lineages ( Figure 1B). In contrast, the 76 CSKP isolates were unclustered and intermingled with the 71 CRKP isolates. Notably, all the ST11-CRKP isolates were clustered as the dominant phylogroup with limited nucleotide divergences among isolates belonging to the same capsular types. However, the other three ST11-CSKP isolates were grouped into two sub-lineages that were phylogenetically distal from the ST11-KL47/KL46 group. Two of these three isolates belonged to KL111, and the other isolate belonged to KL15. Upon the enrollment of reference genomes, the ST11-bla KPC-2 strains from multiple provinces in China were clustered together ( Figure S2). Furthermore, the two ST11-KL111 isolates in the collection clustered together with an ST11-KL64 isolate from Singapore, two isolates from Germany, and one isolate from Spain, while the ST11-KL15 strain was clustered together with KL15 strains from Europe [18,19].

Nosocomial outbreak and transmission caused by ST11 and ST307 CRKP
Considering the strains within each phylogroup of ST11, ST20, and ST307 differed by a few cgSNPs ( Figure S3), their link to a nosocomial outbreak of infection was investigated. Out of the eight neonatal infections caused by ST307 CRKP, seven infants were admitted to the pediatric intensive care unit (PICU) in the same hospital ( Figure 2). The duration of the infections lasted from 28 days to 63 days, and the seven ST307 CRKP were isolated within one week of admission to the hospital. (Table S1). Pairwise SNP analysis of the seven ST307 CRKP showed that  . Epidemiological data and clonal identification of ST11, ST307, and ST20 Kp isolates. All the ST11 (n = 34), ST20 (n = 13), and ST307 (n = 8) isolates were enrolled and each block represented a strain which was ordered by the sampling date from the same hospital within each group of STs. For the clonal panel, clonal strains were indicated when the difference of core-genome SNPs fewer than four. Each number in the legend represented one type of collections of clonal strains, and the strain marked the same number/color represented that they are clonal-related strains. six of them differed by fewer than 4 cgSNPs, indicating that these strains originated from a single clone. The remaining ST307 strain from this hospital exhibited a difference of 10-12 cgSNPs compared to the other six isolates. The integrated genomic and epidemiological analysis suggested that ST307 CRKP strains were linked to a nosocomial outbreak of infection. The 13 ST20 Kp strains were collected between 2014 and 2018 across three different hospitals. The isolates from the same hospital had fewer cgSNPs differences and multiple clones were identified across the isolates (Figure 2). Furthermore, 21 out of the 34 ST11 Kp isolates were collected from the same hospital from 2013 to 2018 and the majority of the patients were admitted to the MICU (67%, 14/21) (Figure 2; Table S1). 14 out of the 21 strains were related to multiple clones with cgSNPs ≤ 4. However, since ST11 Kp is prevalent in China, and given the limited strain numbers and time lag between patient presentations, both nosocomial transmission events and independent introductions in each hospital were possible.
We further determined the presence of IS elements in the chromosome for those genomes. Notably, we found that 30 out of 39 genomes had an insertion of ISKpn26 (1,196 bp, IS5 family) within the acrR gene, all of which were ST11-bla KPC-2 strains. Among the 30 ISKpn26 sequences, ten SNPs were detected and consisted of G+173A, T+176C, A+180G, C+181 T, C +182G, C+184A, T+185A, G+647A, G+698 T, and G +950A. In each strain, the ISKpn26 sequence within acrR was identical to the ISKpn26 located in the bla KPC-2 -carrying plasmid. ISKpn26 insertion was located in the target site of a 4-bp (CTAG) direct repeat (DR) at +276 bp of acrR, with a target site duplication producing another copy of DR at the boundaries of ISKpn26 after its transposition ( Figure  3B). Besides the 29 strains with the consistent insertion of ISKpn26 forming two ΔacrR fragments, one isolate harboured an additional insertion causing the loss of the first 276-bp sequence of acrR and the duplicated DR. Although the remaining ST11-bla KPC-2 isolate lacked ISKpn26 within acrR, it harboured two copies of DR and a 4-bp (GTTC) sequence belonging to ISKpn26, indicating that ISKpn26 had been inserted in acrR. By searching all the additional newly assembled Kp genomes (n = 117), ISKpn26 insertion in acrR was not detected. By searching all the 669 fully assembled Kp genomes in the public database, the intact ISKpn26 insertion in acrR was found in 85 isolates. The vast majority of these isolates were collected in China and 80 were ST11 strains (Table S6). qPCR showed that ISKpn26 interruption blocked the expression of acrR (P = 0.032, t-test), while the relative expression of acrB was enhanced by a 12.6-fold change (P = 0.002, t-test) ( Figure 3C). The susceptibility testing among all the ST11-bla KPC-2 isolates revealed significantly reduced MICs for multiple antimicrobial agents; namely tigecycline, ciprofloxacin, colistin, piperacillin-tazobactam, nitrofurantoin, ofloxacin, and chloramphenicol in the presence of the efflux pump inhibitor NMP (P < 0.05) ( Figure 4). Next integration of ISKpn26 into other regions of the chromosome was assessed. It was found that the strains which harboured plasmids with both ISKpn26 and bla KPC-2 , had a significantly higher mean number of ISKpn26 in the chromosome (15 ± 0.4 vs. 6 ± 2.4, P < 0.001, ttest) ( Figure 3D).
Among the two KPC-2-producing strains ST23 BSI030 and ST11 BSI074, pBSI030_vf harbours the virulence genes of iuc, iro, iutA, rmpA and rmpA2. This plasmid backbone showed similarity to the classic virulence plasmid pLVPK (>99% nucleotide identity, 93% coverage) but with multiple inverted regions ( Figure 5A). Furthermore, pBSI030_vf carried an 11,716-bp fragment containing HigB/HigA toxin/antitoxin system that was not present in pLVPK. The pBSI074_vf backbone was similar to pBSI030_vf (>99% identity, 97% coverage) and pLVPK (>99% identity, 90% coverage) but lacked iroBCD and rmpA2 genes, and rmpA and iroN genes were truncated by ISKpn26. Rearrangement of multiple IS elements in pBSI074_vf also resulted in the difference in the recently identified pVir-CR-HvKP4 in China ( Figure 5A). Using the G. mellonella infection model, we demonstrated that all four strains harbouring virulence plasmids are highly virulent compared with the hypervirulent strain HvKP4 ( Figure 5B).

Discussion
The Kp isolates have evolved separately in distinct clonal groups, including the virulent clonal ST23, and the MDR clonal ST258 and ST11 [36]. It is important to identify carbapenem-resistant hypervirulent Kp and to understand the risk of transmission. Among the Kp strains from BSI patients, potentially novel, differently sized virulence plasmids associated with high mortality of G. mellonella were identified. Two KPC-2-producing strains (ST11 BSI074 and ST23 BSI030) harbouring virulence plasmids were detected, and pBSI074_vf shared high similarity to pBSI030_vf. Considering that a virulence plasmid was rarely found in ST11 strains but was prevalent in ST23 isolates in our collection (100%) and other sources [10], pBSI074_vf was likely acquired from ST23 isolates. Notably, two MDR-virulent plasmids harbouring virulence markers iucABCDiutA plus multiple ARGs were identified. These plasmids also possess plasmid transfer-associated genes that probably facilitate their dissemination. MDR-virulent plasmids were rare but an increasing number of studies have reported such strains in the past few years [35,37,38,39]. The emergence and spread of MDR-virulent plasmids as a result of ongoing recombination is a significant potential health threat in terms of both disease and treatment.
ST11 Kp is the dominant sequence type in China [12], which was distinct from the prevalence in other countries in South and Southeast Asia [35]. In Europe, ST11 Kp was accounting for approximately 10% of the clinical Kp population, and bla NDM-1 and bla OXA-48 are the dominant genes in ST11 CRKP [18]. In this study, the extensively disseminated ST11-bla KPC-2 Kp was continuously identified in BSIs patients. The nosocomial transmission has been detected in some hospitals. The bla KPC-2 -carrying plasmids are highly diverse, and ongoing integration of additional resistance genes has been identified as causing the transmission of MDR plasmids. Another finding is the extensive integration of ISKpn26 into acrR, which was observed in all the ST11-bla KPC-2 genomes, leading to the deactivation of acrR and increased expression of acrB. ISKpn26like insertion in mgrB has been found in ST258 Kp which was responsible for colistin resistance [40]. ISKpn26 insertion in acrR has only been reported in an ST11-bla KPC-2 Kp in Taiwan [41]. We found that the intact ISKpn26 insertion in acrR mostly happened in strains from Asia. Furthermore, a replicative-like transposition [42] produced 14-18 copies of ISKpn26 in the chromosome. A strong link between KPC-2 plasmid-located ISKpn26 and ISKpn26 insertion into acrR indicates that bla KPC-2 plasmid is the reservoir for ISKpn26. The acrR gene is the local repressor of resistance-nodulation-division (RND) efflux pump AcrAB-TolC, which is critical to acquire antimicrobial resistance (AMR) [43,44], and is involved in virulence [45]. We found that the presence of NMP increased the susceptibility of multiple antibiotics, including the last-resort antibiotic tigecycline to treat CRE infections. These data demonstrated that ISKpn26 insertion contributed to the MDR phenotypes in ST11-bla KPC-2 Kp by blocking the expression of AcrAB-TolC repressor acrR. The interaction between the bla KPC-2 plasmid and the ST11 Kp host may, therefore, facilitate nosocomial dissemination and transfer of AMR.
In conclusion, our results revealed a widespread ST11-bla KPC-2 lineage of CRKP from BSI patients in China. In addition to contributing to AMR spread, KPC-2 plasmids can further interact with the host and alter some host-dependent social traits, such as the provision of ISKpn26 to insert into acrR, and therefore, up-regulate AcrAB-TolC multidrug efflux pump, leading to the increased MDR phenotypes of the host bacterium. Furthermore, we identified novel MDR-virulent plasmids due to ongoing recombination in Kp; therefore, representing a significant health threat in terms of both disease and treatment.