Convergence of carbapenem resistance and hypervirulence in a highly-transmissible ST11 clone of K. pneumoniae: An epidemiological, genomic and functional study

Abstract Co-occurrence of hypervirulence and KPC-2 carbapenem resistant phenotypes in a highly-transmissible ST11 clone ofKlebsiella pneumoniae has elicited deep concerns from public health stand point. To address this puzzle, we conducted a large-scale epidemiological, clinical and genomic study of K. pneumonia ST11 clones with both hypervirulence and carbapenem resistance in two tertiary hospitals in Zhejiang province. Most of the patients (15/23) were diagnosed with exclusively carbapenem-resistant K. pneumoniae (CRKP) infections. Ten death cases were reported, some of which are due to the failure of antibiotic therapies. As a result, we identified one new rare sequence types (ST449) to KPC-2-producing CRKP, in addition to the dominant ST11. These clinical isolates of K. pneumoniae are multi-drug resistant and possess a number of virulence factors. Experimental infections of wax moth larvae revealed the presence of hypervirulence at varied level, suggesting the complexity in bacterial virulence factors. However, plasmid curing assays further suggested that the rmpA2-virulence plasmid is associated with, but not sufficient for neither phenotypic hypermucoviscosity nor virulence of K. pneumoniae. Intriguingly, all the rmpA2 genes were found to be inactive due to genetic deletion. In total, we reported 21 complete plasmid sequences comprising 13 rmpA2-positive virulence plasmids and 8 blaKPC-2-harboring resistance plasmids. In addition to the prevalent pLVKP-like virulence plasmid variants (~178kb), we found an unexpected diversity among KPC-2-producing plasmids whose dominant form is IncFII-IncR type (~120kb), rather than the previously anticipated version of ~170kb. These findings provide an updated snapshot of convergence of hypervirulence and carbapenem resistance in ST11 K. pneumoniae.

Very recently, Chen and coworkers described five cases of lethal infections caused by KPC-2-bearing ST11 K. pneumoniae with hypervirulence in the Second Affiliated Hospital, Zhejiang University [26]. This suggested the emergence of a carbapenemresistant hypervirulent clone of K. pneumoniae [27][28][29]. This finding was unexpected, but not without any precedent. Retrospectively, similar cases were reported in sporadic cases and referenced herein [22,30,31]. The genetic mechanism by which classic ST11 CRKP strains gain hypervirulence relies on the horizontal acquisition of a pLVPK-like virulence plasmid [1,16,26]. In this study, we attempted to bridge this knowledge gap. We have conducted a retrospective study of patients (23 in total) admitted to two different tertiary hospitals (namely the Second Affiliated Hospital and Lishui People's Hospital) in the Zhejiang province of China, from 2016 to 2017. Combined string tests and experimental infections, our data of virulence plasmid curing suggests that rmpA2 plasmid is associated with, but not sufficient for hypervirulence of K. pneumoniae, underscoring the complexity in virulence determinants rather than previously-recognized rmpA2 plasmid alone. Along with the observation of Gu et al [26]., this study provides additional insights into the diversified convergence of hypervirulence and carbapenem resistance in the highly-transmissible ST11 clone of K. pneumoniae.

Clinical description of CRKP-positive inpatients
Two tertiary hospitals in Zhejiang Province were involved in this clinical study, namely the Second Affiliated Hospital of Zhejiang University (Hangzhou City) and Lishui People's Hospital (Lishui City, approximate 270 km south of Hangzhou City). In total, 23 patients (19 males & 4 females) that were admitted to the hospitals between late Dec. 2016 and Dec. 2017, were selected for this study (Figure 1 and S1-S3).
Among them, 17 patients were aged 16-80 years (15 males & 2 females) and admitted to the general ICU with single ward isolation (40 beds in-service of ICU with single ward isolation which treats approximately 1,500 patients each year) of the Second Affiliated Hospital of Zhejiang University, in Hangzhou City, from August to December of 2017 (Figs S1-S2). The remaining 6 patients (4 males & 2 females) were admitted in Lishui People's Hospital, Lishui City, between late Dec. 2016 and Oct. 2017 (Fig. S3). These patients had suffered various forms of trauma caused by traffic accidents, falling objects, or acute spontaneous cerebral hemorrhages. Following varied surgeries, all the patients were subjected to antimicrobial treatment [either carbapenem alone or in the combination with other alternative antibiotics or mechanical ventilation (Figure 1, S1-S3) when necessary.
In the Second Affiliated Hospital of Zhejiang University, the length of hospital stays (LOS) of patients in ICU with single ward isolation varied from 0 days to 100 days during which six of 18 inpatients died. Among them, 7 individuals carried CRKP cultures in two or more sites, and the remaining 10 individuals were diagnosed with CRKP infections only in one site, five of which died (Supplementary Text and Figs S1-S2). As for the inpatients in Lishui People's Hospital, the length of hospital stays varied from 5 days to one year. As expected, most of the patients developed pneumonia and CRKP strains were consistently isolated from each patient specimen (Table S1). Four out of six hospitalized patients suffered the failure of antibiotic treatment and died (Fig. S3).
In terms of treatment, the most common antibiotic regimens were combination therapies based on carbapenems. If the MIC value of carbapenems exceeded 16 in patients, carbapenems would not be considered. Tigecycline (in 9 cases) and polymyxin B (in 6 patients) were selected according to the patient's situation and economic condition due to their high price while other options include amikacin, sulperazone, and fosfomycin. Patient 7's condition rapidly deteriorated. After a bloodstream infection, the bacteria rapidly invaded the liver and gall bladder to form a liver and pericholecystic abscess. The patient developed DIC (Disseminated Intravascular Coagulation) and died rapidly. Finally, we successfully collected 19 CRKP isolates from 23 patients mentioned above (Table S1), including 5 positive sputum specimens, 5 positive urine samples, 3 positive blood samples, 2 positive feces samples and 4 other samples.

Genetic characterization of CRKP
Using MacConkey agar medium supplemented with imipenem, 19 non-redundant carbapenem-resistant K. pneumoniae strains of interest were recovered from the above clinical specimens (Figure 1 and Figs S1-S3). Among them, 22 isolates of CRKP were sampled from 17 patients admitted into general ICU with single ward isolation, the Second Affiliated Hospital of Zhejiang University (Tables S1-S2 & Figs S1-S2). Except for two patients who had multiple bacterial isolates each, i.e., 4 strains for 'Patient 7ʹ and 3 strains for 'Patient 8ʹ (Figure 1), all other inpatients were labeled with a single CRKP isolate (Table S1-S2 & Figs S1-S2). The six isolates (one per patient) of CRKP from Lishui People Hospital separately corresponded to K202, K204, K211, K212, K213, and K214 (Table S1 and Fig. S3). The identities of all these strains as K. pneumoniae were confirmed by both MALDI-TOF mass spectrometry and 16S rDNA sequencing. Antibiotic susceptibility assays revealed that they are consistently resistant to 6 classes of (10 different kinds of) antibiotics like Imipenem and Amikacin (Tables S3-S4), indicating that they are multi-drug resistant K. pneuminiae. The fact that they are all positive for bla KPC-2 in a PCR screen, allowed us to believe the phenotypic resistance to carbapenem is due to the presence of bla KPC-2 (Figure 2, S6A and S7A).
Given the results of PCR screening and sequence typing, 19 representative strains (13 strains from the Second Affilated Hospital of Zhejaing University, and 6 strains from Lishui People's Hospital, in Tables S1-S3) were selected for genetic mapping and further genomic sequencing. Except that one strain is of non-ST11 genetic background (i.e., ST449 for K196), multilocus sequence typing (MLST) revealed that all the other 18 strains are exclusively grouped into ST11 (Figure 2). In general agreement with a recent observation by Gu et al. [26], ST11 was the dominant clone containing bla KPC-2 in hospitalized patients in these two hospitals during 2016-2017.

Unexpected complexity in CRKP with hypervirulence
Even though the string test is a phenotype indicative of hypermucoviscosity, it is helpful, but not sufficient to conclude the correlation between phenotypic hypervirulence and hvKP strains [32]. Prior to string tests (Figs S4A-F), 6 representative hvKP isolates with rmpA2-positive plasmids were subjected to plasmid curing, which are confirmed with the rmpA2-, and KPC-2-specific PCR detection (Fig. S4G). These included four isolates (K185, K195, K199, and K235)  (Table S5) were applied to detect genetic determinants of virulence and antibiotic resistance in series of clinical strains ( Table S1). The positive results of PCR are colored orange. * indicates the truncated version of rmpA2, which is due to the deletion -causing frameshift mutations and premature stop codons.
Using a G. mellonella infection model (Figure 3), we observed that i) among the aforementioned six hvKP isolates (Fig. S4), only two strains (K185 and K199) can be significantly attenuated via the elimination of  Figure 3c). Meanwhile, the hvKP strain of K202 is unusual in that its low virulence is relatively lower than most of the CRKP isolates lacking the virulence plasmids ( Figure 3c). It is clear that no frank correlation occurs between rmpA2 positivity and LD50. Surprisingly, although having lost the string test phenotype, strain K204-PC retains virulence in Galleria model. Therefore, this points out the complexity in phenotypic hypervirulence of hvKP, and suggests that rmpA2 is not a single virulence determinant. Indeed, plasmid-borne aerobactin is also an important virulence factor.

Genomics of rmp2-harboring virulence plasmids
The criteria applied in the selection of virulence plasmids candidate mainly relies on the distinct pattern of virulence determinants displayed in multiplex PCR assay (Figure 2). Therefore, 10 unique clinical rmpA2-positive isolates (Fig. S5A) were selected for further S1-PFGE analyses (Fig. S5B), together with Southern blot with the rmpA2 probe (Fig. S5C). Among them, 6 isolates (K185, K187, K192, K195, K198, and K199) were collected from the Second Affiliated Hospital of Zhejiang University, and the other 4 strains (K202, K204, K211, and K214) are derived from the Lishui People's Hospital (Figure 2). Similar to those reported by Gu et al. [26], our Southern blot analysis suggested an estimated size of 170-210 kb for the rmpA2harboring plasmids (Fig. S5C). As expected, Illumina next-generation sequencing revealed that all the plasmid backbones (e.g., pK185-rmpA2) were around 178kb in length ( Table 1 and S6). Using Plasmid Namely, they refer to Y4 strain with high virulence, WNX-1 with an intermediate virulence, and WNX-5 with the relativelylow virulene. The two baselines is indicated (i.e., red line is cutoff for low virulence and gray line is curoff for high virulence). PC denotes plasmid curing. CRKP (purple triangle), hvKP (black circle) and plasmid curing mutants (red square) have been evaluated. The data here suggests the strains/speciesdependent role of virulence plasmids in various clinical K. pneumoniae isolates. F i n d e r 1 . 3 ( h t t p s : / / c g e . c b s . d t u . d k / s e r v i c e s / PlasmidFinder-1.3/), we further confirmed that these rmpA2-containing plasmids belong to the "IncHI1B-IncFIB" replicon incompatibility type, analogous to those of the paradigm virulence plasmids pLVPK and pVir-CR-HvKP4 (Table 1).
A total of 19 clinical K. pneumoniae isolates were applied in plasmid isolations and the resultant plasmids of interest were subjected to Illumina next-generation sequencing. As a result, we assembled 21 complete plasmid sequences and 7 plasmid contigs (Table 1 and S6). The completely assembled 13 rmpA2-positive plasmids and 8 KPC-2-producing plasmids ( Table 1). Seven incomplete sequences corresponded to 2 large rmpA2-harboring fragments and 5 bla KPC-2 -carrying contigs ( Table 1). Not surprisingly, plasmid sequencing revealed that 15 clinical strains (such as K185 and K204) carry a bla KPC-2 -positive plasmid along with a rmpA2-expressing plasmid ( Table 1), which corroborated our PCR assays (Figure 2) and Southern blot results (Figs S5-S7). In addition, we demonstrated that 4 strains (namely K196, K212, K213, and K237) exclusively retain a single KPC-2-producing resistance plasmid, without the rmpA2-bearing virulence plasmid ( Table 1), which is generally consistent with PCR-based virulence factor profiles (Figure 2).
Comparative genomics of these rmpA2-positive plasmids showed that they consistently possess over 99% similarity (with 99% query coverage; Figure 4a) to pVir-CR-HvKP4 (Acc. no.: MF437313) and seem to be heterogeneous variants of this plasmid which was recently reported to be circulating in the Second Affiliated Hospital of Zhejiang University [26]. In addition to rmpA2, both iucABCD (an aerobactin synthesis operon) and iutA gene (that codes the outer membrane ferric aerobactin receptor) are also present in a number of sequenced plasmids (Figure 4a and b). Similar scenarios were also seen in other virulence plasmids detected in ST11 strains circulating in China [33,34]. This further validated the profile of virulence factors seen in our multiplex PCR assay (Figure 2). Even though they align well to the paradigm ~230kb long virulence plasmid pLVPK (Acc. no.: AY378100) [13], these virulence plasmids consistently lacked an approximately 50kb of region that carried the other mucoid regulator-encoding gene rmpA and iroBCDN, a gene cluster of salmochelin production ( Figure 4b).
These results provide genomic evidence for the discovery of epidemic virulence plasmids circulated in Hospitals of Zhejiang Province (Figure 4). Moreover, it seems likely that they are pLVPK-like derivatives "NA": not applied;/, not detected generated by a genetic deletion event in response to an unknown selective pressure.

Diversity of bla KPC-2 -bearing plasmids
The bla KPC-2 -specific molecular assays (PCR detection and Southern blot, Figs S6-S7) of 18 K. pneumoniae isolates showed clearly that at least 6 kinds of bla KPC-2positive plasmids were disseminated in these clinical CRKP isolates, whose physical size ranges from ~10kb to ~130kb (Table S6 and Figs S6-S7). Given that KPC-2 is transmitted by diversified plasmids in patients hospitalized in Zhejiang province, it is of much interest to gain genomic insights into these carbapenem resistance plasmids. As a result, plasmid sequencing returned 9 complete plasmid sequences (57.798--128.447kb) and 4 unclosed large contigs (120-180kb) of bla KPC-2 -carrying plasmids ( Table 1). Among them, the IncFII/IncR-type bla KPC-2 -bearing plasmids (118.504--128.447kb) seemed to be prevalent and they were found to exist in strains that also carried the IncHI1B-IncFIB type, rmpA2-harboring hypervirulent plasmids (176.-874-178.537kb) in 90% (8/9) of sequenced K. pneumoniae isolates (like K187, in Table 1 and Figure 5a). Sequence analysis and annotation indicated that the bla KPC-2 -encoding plasmids carry additional antibiotic resistance determinants including extended-spectrum β-lactamase genes (bla TEM-1 , bla KPC-2 , bla CTX-M-65 , bla CTX-M-90 ), aminoglycoside resistance gene (rmtB), phenicol resistance gene (catA2), etc. (Figure 5 and Table S7). This finding can explain the phenotypic multidrug resistance of these clinical K. pneumoniae strains (Tables S2-S3). It is common that various mobile elements, mostly IS26, are located upstream and downstream of the aforementioned antimicrobial resistance genes ( Figure 5) and might play a role in the AMR horizontal transfer. Genomic alignment of the eight IncFII/R type bla KPC-2 -harboring plasmids suggested that they share nearly 99% identity with each other with query coverage ranging from 87% to 95% (Figure 5a). Although that they align well with a recently-recognized bla KPC-2 -positive plasmid, pKPC2_0200030, from West China University (~105kb, Acc. no.: CP031720), our KPC-2-containing plasmids, like pK187_KPC, exhibit varied level of query coverage (85%-89%). This might hint that they are all longer versions with discontinuous inserts like the short plasmid pKPC2_0200030 (Figure 5a). Also, it is clear that a cassette surrounding bla KPC-2 (ISKpn26bla KPC-2 -ISKpn27-ΔTn3-IS26) is extremely conserved amongst these plasmids ( Figure 5). However, linear genomic alignment pointed out that the insert direction in pK186_KPC and pKPC2_0200030 are opposite (Figure 5b). Presumably, it is due to active transposition between plasmids and highlights a risk that bla KPC-2 can spread across various isolates of ST11 K. pneumoniae.

Discussion
The data shown here is in general agreement with the statement by Gu et al. [26] that i) occurrence of carbapenem resistance is conferred by KPC-2 in clinical isolates of K. pneumoniae in Zhejiang Province; and ii) the dominant KPC-2-producing K. pneumoniae belongs to ST11 in the Second Affiliated Hospital of a. Genomic analyses of the thirteen pK211_rmpA-like virulence plasmids (roughly 178kb) recovered from ST11 carbapenemresistant hypervirulent K. pneumoniaeGenomic sequences of the above 13 plasmids were aligned against two known virulence plasmids (pLVPK of the K. pneumoniae strain CG43 [219kb, Acc. no.: AY378100] and pVir-CR-HvKP4 [178kb, Acc. no.: MF437313]). The circular map was generated with the BLAST Ring Image Generator. b. Colinear genome alignment amongst pK211_rmpA2, pLVPK (Acc. no.: AY378100) and pVir-CR-HvKP4 (Acc. no.: MF437313)The program Easyfig was applied in comparative genomics. Colored arrows indicate ORFs and the shaded region reflects sequence similarity. The virulence associated genes are indicated in red.
Zhejiang University, 2016-2017. Despite our recent discovery of co-occurrence of the other carbapenemase-encoding gene bla NDM-5 the transferable colistin resistance determinant mcr-1 on two distinct plasmids from the isolates of Lishui People's Hospital [35], our multiplex-PCR screen cannot detect the co-occurrence of carbapenem-resistance gene bla KPC-2 with mcr-1 in these clinical isolates. Of being noteworthy, this sampling in the same hospital, in 2017, allowed us to identify one rare sequence types (ST449) of K. pneumoniae carrying KPC-2 (Figure 2). More intriguingly, this unique isolates of K. pneumoniae (ST449 for strain K196, Figure 2) only carry a KPC-2-positive plasmid (~90kb for pK196-KPC in K196, Fig. S6C), but no virulence plasmids. It is plausible that the lack of virulence plasmid in the two strains (Figure 2), explains in part why the value of their LD50 is relatively-higher than certain hypervirulence plasmidcarrying strains like K187 and K192 (Figure 3c).
In general, an intact rmpA2 gene is 639bp long, which encodes RmpA2 regulator of 212aa in full length. Unlike the scenario seen with pVir-CR-HvKP4 having the intact rmpA2 version [26], Sanger sequencing elucidates that an identical-truncated version of rmpA2 consistently occurs in all the 15 isolates collected in this study (Fig. S8). Along with an "A" deletion in the poly-tract at the position 353, the two "G" deletion in the poly-G tract at position 285-286 (Fig. S8A&B), results in the frameshift and premature stop codon, giving a predicted polypeptide product of 99 amino acids (Fig. S8C). Because of lacking DNA-binding domain at C-terminus, the truncated version of RmpA2 is presumably nonfunctional. Not surprisingly, the discovery of such rmpA2 variant appears in our bacterial species. This is because that it has already been recorded in NCBI database, involving 40 other K. pneumoniae isolates. Among them, 37 isolates arise from China (e.g.: 13 isolates from Hunan Province,   [37]). The remaining three isolates are separately derived from USA (Acc. no.: CP037743), Japan (Acc. no.: AP019549), and South Korea (Acc. no.: CP030924). The functional inactivation of rmpA2 might indicate an ongoing microevolution of K. pneumoniae colonizing in Zhejiang Province, China.
Our experimental infections with wax moth larvae confirmed that bacterial virulence differ across these isolates (Figure 3). Several rmpA2-negative CRKP isolates (like K212, K213, and K237) gave comparable intermediate virulence in relative to certain HvKP strains (e.g., K211, K232, and K239) (Figure 3c). Of note, Galleria possibly is not the best model to test the relationship between hypermucoviscosity and virulence. We only found 10% (2/20) of the representative HvKP isolates are positive in string test, which is quite lower than 100% (5/5) in a recent report of Gu et al. [26]. This argues that rmpA2-virulence plasmid is a necessary, but not sufficient for the formation of hypermucoviscosity, a virulence-associated phenotype. Not surprisingly, the rmpA2-virulence plasmid curing diminished the phenotypic hyper-mucoviscosity in the string tests for the two strains K119 and K204 (Figs S4C-D). However, the elimination of rmpA2-virulence plasmid only attenuated virulence of K199 (Figure 3b-C), rather than K204 (Figure  3a and C). This underscored that rmpA2 virulence plasmid can't be a sole determinant of hypervirulence in K. pneumoniae. Moreover, the fact that bacterial hypervirulence varies greatly amongst different virulence plasmidsbearing isolates allowed us to believe that the role of virulence plasmid is strain-dependent and associated with the context of K. pneumoniae isolates. Because that i) hypervirulence is a complex phenotype associated with but not limited to hyper-mucoviscosity, (i.e., hyper-mucoviscosity is not synonmous with hypervirulence in certain rmpA2-bearing K. pneumoniae [38]); and ii) hypermucoviscosity is dependent on production of biofilm, siderophores (especially aerobactin), fimbriae adhesion, and polysaccharides, we further speculated that our isolation might represent a heterogeneous/diversified population of HvKP. In fact, this hypothesis was supported at the level of plasmids since most (if not all) were harbored in these clinical isolates (Figure 2-5). In general consistence with the scenario assigned to HvKP4 described by Gu and coworkers [26], we found that 75% (15/20) of our K. pneumoniae isolates coharbors both a rmpA2-virulence plasmid and a carbapenem-resistance plasmid (Table 1 and Figure 2). Indeed, all the hypervirulence plasmids we determined here, are ~178 kb in length, and possess 99% identity to pVir-CR-HvKP4 (Acc. no.: MF437313) reported by Gu et al. [26]. This indicated that they are a series of shortened variants from the original virulence plasmid, pLVPK (~219 kb, Figure 4) [13]. Therefore, we concluded that the plasmid determinants of hypervirulence are relatively-conserved, but retain the potential for microevolution in current isolates of CRKP in Zhejiang's hospitals.
In contrast, the KPC-2-harboring plasmids with carbapenem resistance displayed an unexpected diversity ( Table 1 and Figure 5 & S6-S7). We also noted that the dominant form of bla KPC-2 -carriers in Zhejiang's hospitals is IncFII/IncR-type plasmid with the size of ~120kb ( Table 1), rather than ~178kb IncFII/IncR-type reported by Gu et al. [26]. Given that a polyclonal ST11-KPC plasmid scenario is observed, we anticipated that several parallel events of rmpA2 plasmid acquisition might have occurred, giving rise to at least 2 clones ST11-KPC-rmpA2 clones, rather than the only one described by Gu et al [26]. In addition, 8 resistance plasmids with complete plasmid sequences are localized at the size ranging from 118.504kb to 128.447kb (Table 1). Obviously, the complexity of KPC-2-positive plasmids in ST11 CRKP supplemented the conclusion of Gu et al. [26] that the emergence of ST11 carbapenem-resistant hypervirulent K. pneumoniae is due to a single genetic event, an acquisition of a pLVPK-like virulence plasmid (~170kb) by a classic ST11 CRKP presumably carrying a dominant one pKPC-CR-HvKP4 (~170kb). Probably, it has already arisen purely through chance and continues to circulate due to improved fitness costs. It is unusual that bla KPC-2 coexists with rmpA2 in a hybrid plasmid pKP70-2 (~240kb) from HvKP [39]. Evidently, it might be more complicated than we earlier anticipated.

Conclusions
Very recently, the discovery that rmpA2-harboring plasmid hypervirulence converges with KPC-2 carbapenem resistance in a highly-transmissible ST11 clone of Klebsiella pneumoniae raised concerns among public health professionals and social communities [26]. However, the description of CR-HvKP isolations is incomplete. This is primarily due to the limited availability of clinical cases and genomic sequences of the causative agents. To address this issue, we conducted a large-scale epidemiological, clinical, and genomic study in two tertiary hospitals in Zhejiang Province. Overall, our finding provides additional insight into the diversity of CR-HvKP isolates in Zhejiang province. 7 new major discoveries are described as follows: i) 23 patients admitted to hospitals were clinically described, most of which had undergone surgery for various traumas and then subjected to antimicrobial treatment of carbapenem alone (or combined with other alternative antibiotics); ii) we discovered one new rare sequence types (namely ST449), in addition to the prevalent ST11 sequence type; iii) as for these newly-isolated KPC-2-producing CRKP, they all displayed the complexity of multidrug resistance as well as genetic heterogeneity in virulence factor profiles; iv) to gain detailed genomic insights into the convergence of hypervirulence and carbapenem resistance, we completed plasmid sequencing of 21 plasmids consisting of 13 rmpA2-positive virulence plasmids and 8 bla KPC-2 -harboring resistance plasmids; v) the rmpA2 regulator were found to be inactive variants here; vi) we observed that all the virulence plasmids are pLVKP-like variants of ~178kb long, and proposed the complexity of plasmid-encoded virulence factors in the context of K. pneumoniae hypervirulence; and vii) the unexpected diversity among these KPC-2-producing resistance plasmids were identified, whose dominant form belongs to IncFII-IncR type plasmid of ~120kb, not the previouslyanticipated version of ~170kb. In summary, our findings benefit our understanding on the convergence of hypervirulence and carbapenem resistance in ST11 K. pneumoniae, providing additional genomic insights into the diversity of dominant clones in Zhejiang province.

Study design
In 2017, a small-scale outbreak of carbapenem-resistant hypervirulent Klebsiella pneumoniae (CR-hvKP) was recorded in an integrated ICU of the Second Affiliated Hospital of Zhejiang University (Hangzhou, China). We therefore intended to investigate whether CR-hvKP infections were becoming increasingly prevalent among ICU patients. Overall, clinical strains included in this study referred to carbapenem-resistant Klebsiella pneumoniae (CRKP) strains sampled by clinical laboratories from patients admitted to two independent ICUs (the Second Affiliated Hospital of Zhejiang University, and Lishui People's Hospital), from Dec. 2016 to Dec. 2017.

Ethics statement
Clinical data were recorded and approved by the local ethics committee from both the Second Affiliated Hospital, Zhejiang University School of Medicine (2018-01-F) and the Lishui People's Hospital (2017-005-01).

Clinical Isolates, Identification and Growth Condition
All the clinical isolates of Klebsiella pneumoniae (K. pneumoniae) were collected from the patients admitted to two public Grade-A Tertiary Hospitals in Zhejiang Provinces, China. This includes the Second Affiliated Hospital, Zhejiang University, and Lishui People's Hospital, Lishui City, respectively. Prior to direct sequencing of 16S rDNA PCR products, bacterial identification was conducted routinely using matrixassisted laser desorption/ionization-time of flight mass spectrometry (MALDI-TOF MS) [40]. The isolated K. pneumoniae (Table S1) were cultivated overnight on either MacConkey agar media [or Luria-Bertani Broth (LB) agar plates] or LB liquid medium at 37°C for further sequence typing and plasmid isolation.

Assays for antimicrobial susceptibility
The K. pneumoniae strains of interest were subjected to antimicrobial susceptibility tests comprising 13 different antibiotics such as Imipenem and Ertapenem (Tables S3-S4). The minimum inhibitory concentration (MIC) was routinely determined in accordance with the general guidance of Clinical and Laboratory Standards Institute. The MIC values were interpreted in terms of the breakpoints recommended by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) [41,42].

String test
To address the variation of phenotypic muco-viscosity, an indication of bacterial virulence, all the K. pneumoniae isolates (Table S1) were inoculated on Columbia blood (5%) agar plates (Oxoid, Thermo Scientific) and maintained at 37° C overnight. In general, a toothpick was used to touch and pull a single colony upwards. The cutoff criteria of being positive in the string tests denotes the >5 mm length of a vicious string.

Multi-locus sequence typing
Genetic diversity of K. pneumoniae was examined with multi-locus sequence typing (MLST). In general, 7 sets of DNA fragments derived from house-keeping genes were PCR amplified with specific primers (Table S5), as routinely described in the protocol-specific in the online MLST database for K. pneumoniae. The gel-purified PCR products were subjected to further DNA Sanger sequencing. Lastly, the sequence data with various allelic profile was submitted to the MLST database (https://www.pasteur.fr/fr/mlst), giving final sequence type.

PCR-aided search for virulence and antibiotic resistance genes
Multiplex PCR systems were established to detect the putative genes associated with either virulence or carbapenem resistance in K. pneumoniae. The five loci of hypervirulent-K. pneumoniae (hvKp) we examined here referred to rmpA (regulator of mucoid phenotype A), iroN, rmpA2 (an activator for capsule biosynthesis), iucA, and iutA, respectively (Table S2). In addition to the mobile colistin resistance gene mcr-1 [43], five distinct genetic determinants encoding carbapenemases were also detected (Table S2), namely bla KPC-2 , bla NDM-5 , bla VIM , bla IMP , and bla OXA-48 [23]. After visualization of these PCR products separated with agarose (1%) gel electrophoresis, they were further confirmed by direct DNA sequencing.

S1-PFGE and southern blot
To reveal physical location of the bla KPC-2 (and/or rmpA2) determinant in clinical isolates of K. pneumoniae, S1-Pulsed Field Gel Electrophoresis (S1-PFGE) was conducted, along with Southern blot as recently described [44,45]. Overnight culture of different K. pneumoniae strains (Table S1) was imbedded in 1.0% agarose gel plugs (at the ratio 1:1), which was followed by the protease K treatment. Subsequently, S1 nuclease (Takara, Dalian, China) was applied to treat agarose gel plugs, prior to the separation of DNA fragments by PFGE [44,45]. Then, Southern blot was routinely performed in accordance with the manufacturer's protocol. In particular, two types of probes used here referred to digoxigenin-labeled PCR products separately targeting bla KPC-2 and rmpA2.

Plasmid curing
To determine potential effects of virulence plasmids on K. pneumoniae infection (Table S6), plasmid curing was conducted with the treatment of sodium dodecyl sulfate (SDS) as described by El-Mansi et al. [46]. Six K. pneumoniae isolates carrying rmpA2-positive virulence plasmids were examined here, which included four strains (K195, K204, K214, and K235) from Second Affiliated Hospital, Zhejiang University, and two isolates (K185 and K199) from Lishui People's Hospital (Table  S1). Prior to plasmid curing, various level of SDS (from 0.5%, 2.5%, to 5.0%) was supplemented into liquid LB culture and 2 μg/ml meropenem was added into LBA plates. In brief, 100 μl of log-phage culture was inoculated into 10 ml of liquid SDS-containing LB media, and kept overnight with shaking at 27°C. Then, the resultant culture was diluted, plated on the meropenem-selecting LB agar plates, and incubated overnight at 37°C. Finally, single colonies were subjected to genetic screening of the plasmid-curing derivatives with rmpA2 -and bla KPC-2 -specific PCR assays (Table S5 and Fig. S4).

Galleria mellonella infection
The virulence of K. pneumoniae was evaluated using the infection model of wax moth larvae (Galleria mellonella) as described earlier [47,48]. G. mellonella (~300 mg each) was purchased from Tianjin Huiyude Biotech Company (Tianjin, China), and divided into series of experimental groups (10/group). Larvae was challenged with log-phase cultures of K. pneumoniae at appropriate level of colonyforming units (CFU). As for survival curves, G. mellonella were infected with 10 4 CFU of K. pneumoniae and recorded 72 h post-infection. To further measure LD50 (lethal dose 50%), larvae was injected with different concentrations of K. pneumoniae ranging from 10 3 , 10 4 , 10 5 , to 10 6 CFU. All the experimental infections were conducted in triplicates.

Plasmid sequencing and assembly
Plasmids were extracted from clinical K. pneumoniae isolates using the Qiagen plasmid Midi-prep kit (Qiagen, Germany). After quality control on agarose gel, plasmids were used to construct a DNA library using KAPA Hyper Prep Kit used for Illumina platform (Roche, Basel, Switzerland). The resultant libraries were sequenced with the Hiseq X ten PE150 sequencer platform (Illumina, USA), giving a pool of 150 bp pairedend reads. The reads were trimmed and then assembled into a contig by the SPAdes Genome Assembler (version 3.11.0). A BLASTN against every contig containing resistance (and/or virulence) gene was carried out to find out best-hit plasmids. In light of the best-hit plasmids, possible gaps between contigs were closed with PCR combined with Sanger sequencing. All the plasmids with complete plasmid sequences have been deposited into GenBank ( Table 1).