Emergence of the third-generation cephalosporin-resistant hypervirulent Klebsiella pneumoniae due to the acquisition of a self-transferable blaDHA-1-carrying plasmid by an ST23 strain

1 State Key Laboratory of Microbial Metabolism, Joint International Research Laboratory of Metabolic & Developmental Sciences, School of Life Sciences & Biotechnology, Shanghai Jiao Tong University, Shanghai 200030, China 2 Department of Critical Care Medicine, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai 200025, China 3 Department of Laboratory Medicine, Jinshan Hospital Fudan University, Shanghai 201508, China 4 Department of Clinical Microbiology, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai 200025, China 5Department of Laboratory Medicine, the Second People’s Hospital of Lianyungang, Jiangsu 222023, China 6Shanghai Key Laboratory of Veterinary Biotechnology, Shanghai 200240, China


Ethics and consent Strains and primers
String test for the hypermucoviscosity phenotype The virulence of transconjugants with pRJA166a is similar to that of recipient PCR detection of pRJA166a in collected of K. pneumoniae clinical isolates Table S1. Bacterial strains used in this study. Table S2. Oligonucleotides used in this study. Table S3. General genome features of the K. pneumoniae RJA166, NTUH-K2044, RJF293 and HS11286. Table S4. Putative virulence factor genes detected on the RJA166 chromosome and the plasmid pRJA166b.

Ethics and consent
The present study is observational, and uses human sputum samples for the ex vivo experiments.
Sampling from patients is routinely performed in the clinical treatment. For this study, verbal informed consent was obtained from the volunteers. The bacterial samples and data sheets were anonymized.
The study protocol, including the procedure for obtaining verbal informed consent, was approved by the ethics committee of the Ruijin Hospital affiliated with Shanghai Jiao Tong University, Shanghai, China.
Protocols of mouse experiments were approved by the same ethic committee.

Strains and primers
K. pneumoniae RJA166 was isolated from sputum specimen from a patient in Ruijin Hospital, Shanghai, China. A 47-year-old male patient, with a sign of a cough for over two months and diagnosed with mitral regurgitation, was sent to the cardiac surgery ward on April 18th, 2015. The patient underwent mitral valve repair and coronary artery bypass graft surgery by cardiopulmonary bypass on April 27th, and then, the patient was transferred to the intensive care unit (ICU). However, exploratory thoracotomy was performed because of an unusual volume of bloody drainage on April 28th. Then, the patient coughed up with excessive sputum, and the chest X-ray showed thickness or turbulence of texture in bilateral lungs, with a white blood cell count as high as 19.77×10 9 /L on May 3rd. A routine microbiological culture of the sputum sample was performed and K. pneumoniae strain RJA166 was isolated on May 4th in the ICU ward. The patient received cefuroxime, vancomycin, and cefotaxime antibiotics during hospitalization, and was discharged from hospital on May 8th.
K. pneumoniae clinical strain RJF293 was collected from a blood sample 1 , and its hygromycin-resistant derivative K. pneumoniae RJF293H was constructed using a previously described method 2 . The detailed information of strains and primers used in this study are listed in Table S1 and   Table S2, respectively.

String test for the hypermucoviscosity phenotype
The string test to determine the hypermucoviscosity phenotype was performed on colonies grown overnight on a blood agar plate at 37°C. The colonies were touched with a loop and pulled upwards, and a string of 5 mm or longer was considered to be a positive result 3 .

PFGE, MLST, and K-typing
The pulsed-field gel electrophoresis (PFGE) protocol used was based on the PulseNet 1-day standardized PFGE protocol for Escherichia coli O157:H7 and Salmonella 4 . The MLST was performed according to the method described on the BIGSdb website of the Institute Pasteur (http://bigsdb.pasteur.fr/klebsiella/klebsiella.html). Additionally, detection of the capsular serotype of K.
pneumoniae was conducted using the primer sets listed in a previous publication 5 .

Antimicrobial susceptibility test and molecular typing
Bacterial antimicrobial susceptibility was determined using the VITEK 2 Compact system (bioMérieux, Marcy l'Etoile, France). The result was interpreted according to the guidelines of the Clinical and Laboratory Standards Institute guidelines (Twenty-Fourth informational Supplement M10-S26, 2016).

Killing assay on Galleria mellonella larvae
The virulence of RJA166 was assessed by a killing assay on G. mellonella larvae as per the protocol designed by Insua et al. with minor modification 6 . Larvae weighing 200-400 mg were selected as subjects for infection. Bacteria was cultured overnight in a shaking incubator at 37℃. The bacteria were subcultured by inoculating 30μl overnight culture into 3ml of fresh LB broth (1:100). In a pilot experiment, a culture with an OD600 of 1.0 was determined to contain 10 9 CFU/ml. Thus, when the OD600 of the subculture reached 1.0, 100 μl of the cells were pelleted and resuspended with 900 μl of phosphate-buffered saline (PBS). Ten microlitres of the resuspension, which was equal to 10 6 CFU, was injected into the last right proleg of the larvae using Hamilton precision syringes (Hamilton company). A group was composed of 10 larvae, and each experiment was performed in triplicate. After injection, each group of worms was placed in a 90 mm Petri dish and incubated at 37℃. The mortality was monitored and recorded every 24 hours for a total of 72 hours. The larvae were considered dead when they did not respond to the mechanical stimuli.

Mouse lethality assay
Five-week-old female BALB/c mice were used for virulence assessment, as described previously 7 . Briefly, the overnight cultures of RJA166 to be tested were inoculated into fresh LB broth at 1:100, and grown for 3 hours in a 37°C shaking incubator. Then, the cells were harvested, washed once and re-suspended in saline (0.9%) to an OD600 of 1.0 (approximately 10 9 CFU/ml). Serial dilution was performed to obtain dilutions containing 10 2 CFU/ml, 10 3 CFU/ml, 10 4 CFU/ml and 10 5 CFU/ml for the determination of the median lethal dose (LD50). 100μl of the bacterial suspension was injected intraperitoneally into the mice. All mice were monitored daily for survival. Five mice were involved in each group for LD50 determination.

Antibiotic treatment of the plague in mice
Thirty-six female BALB/c mice aged 5-6 months were intraperitoneally infected with 1.5×10 3 CFU of K. pneumoniae RJA166. The mice then were randomly assigned into 3 groups for different treatment, there were 12 mice in each group. Twelve hours post infection, the mice in the different groups were injected intraperitoneally with ceftazidine (40 mg/kg, q12 h), imipenem (20 mg/kg, q12 h) and 0.9% saline (q12 h). The treatment was stopped when the first death was observed. Two mice in each group were randomly selected for bacterial load determination after the first administration of antibiotics, and the remaining 10 mice were used for survival analysis. The mortality was recorded every 24 hours, for a total of 14 days.

Bacterial load determination
Two mice in each group were sacrificed 12 hours after the first administration of antibiotics. The kidney, spleen and liver of each mouse were sampled and homogenized in 0.9% saline. DNA was extracted from 0.1 g of each homogenate was using the DNeasy Blood & Tissue Kit (Qiagen, catalog No. 69504) according to the user manual. The bacterial load of the tissue was determined by quantitative real-time PCR (qPCR), calculating the relative amount of bacterial gyrase gene gyrA to the murine β-actin gene. The gyrase gene and β-actin gene were amplified using the primer sets 166gyrA-qF/R and Bactin-qF/R (Table S2). qPCR was performed using the SYBR Premix Ex Taq II Kit (Takara, catalog No. RR820L) as per the user manual.

Whole genome sequencing and annotation
Since none of the ST23 K. pneumoniae strain with multidrug resistance have been completely sequenced to date, we decided to explore the RJA166 antibiotic resistance determinants via whole genome sequencing (WGS). The RJA166 strain was grown overnight at 37°C to stationary phase in LB medium, and the total DNA was isolated from harvested cells. The genome sequence was determined by using Illumina Miseq short-read sequencing (2×251 paired-end reads for a 400-bp library) and PacBio RSII sequencing (4 Kb template). Then, the short and long reads were assembled using SPAdes 3.9 (ref. 8), resulting in four circular replicons (a chromosome and three plasmids). Finally, the genome sequence was annotated using NCBI Prokaryote Genome Annotation Pipeline version 2.0. Additionally, the putative virulence factor, antibiotic resistance determinants, and mobile genetic elements were predicted using VRprofile 9 .
The RJA166 chromosome is 5,348,368 bp in size with a G+C content of 57.6%. It contains 5,354 annotated protein-coding sequences (CDSs). A total of 39 putative virulence factor genes (clusters) were detected on the chromosome (Table S4), including the gene clusters coding for allantoin metabolism (all), enterobactin synthesis (fep, ent), pilus (yag), type 3 fimbriae (mrk), lipopolysaccharide synthesis (lpx) and iron ABC transporter (kfu). In addition, twenty-five putative acquired antibiotic resistance determinants are encoded by the RJA166 chromosome (Table S5). The RJA166 chromosome also putatively carries two prophages, two integrative and conjugative elements, and two type VI secretion system gene clusters. Notably, one likely intact 60.3-kb prophage (coordinate 1,411,855 to 1,472,192 bp) is unique to the RJA166 strain and absent in NTUH-K2044, RJF293, HS11286, and other completely sequenced bacterial genomes currently available in GenBank.
The largest plasmid pRJA166a is 230,606 bp in size and its G+C content is 45.8%. It carries four acquired antibiotic resistance genes ( Figure 3), namely, blaDHA-1, qnrB4, sul1, and blaSHV-12. An intact tra gene cluster coding for a type IV secretion system belonging to type F was also predicted on pRJA166a.
The plasmid also encodes a relaxase (RJA_28205) belonging to the MobH family. The putative oriT (coordinate: 141,312-142,312 bp) contains a conserved nic motif: 5′-CATCCTG^C-3′ (^: putative nic-cleavage site) that has been seen in MobH-associated oriT. In addition, a type II toxin-antitoxin locus (TA) was also detected using TAfinder 10 , which might aid the stable inheritance of the plasmid by host.
This TA locus codes for a RelE-like toxin protein (RJA_28095) and a related antitoxin protein containing a PHD domain (RJA_28090). These results theoretically confirmed the conjugative ability of pRJA166a 11 . Notably, pRJA166a has a mosaic-like accessory resistance region. For convenience of description and discussion, this accessory resistance region was divided into three putative modules defined on the basis of sequence homology, G+C content, and/or the flanking mobile elements ( Figure 6): (i) Module I (3.3 kb, 61.1% G+C) harbours blaSHV-12 and shares 100% full-length nucleotide identity with a region in pKOX_R1, and is flanked by ISKpn19 and IS26; (ii) Module II (11.5 kb, 51.3% G+C) seemed to cover a conserved blaDHA-1 context and is next to the 3'-CS of the class I integron and ISCR1 (type I insertion sequences with a common region); and (iii) Module III (1.5 kb, 55.9% G+C) is adjacent to another IS26 and encodes a glutathione dehydrogenase and an esterase and shares 96% nucleotide identity with pT5282-mphA of Enterobacter cloacae T5282 (GenBank accession no: KY270852) 12 . Such a mosaic-like compilation of various functional modules could involve an initial invasion by a mobile element, containing, for example, a component of module II, which could be followed by recombinational promiscuity in disparate IS elements. This suggests that the accessory resistance region of pRJA166a could be an example of the mechanism by which a bacterial host successfully acquires the peripatetic genetic information of antibiotic resistances from multiple sources.
Like other ST23 hvKP strains, RJA166 also harbours a virulence plasmid, pRJA166b, which is 229 kb in size and has a G+C content of 50.1% ( Figure S2a). The mucoviscosity regulator genes rmpA and rmpA2 genes were identified, with no frameshift mutation, on this plasmid. Additionally, the gene clusters coding for the biosynthetic pathway of salmochelin (iroBCDN) and aerobactin (iucABCD, shiF), and for iron uptake regulation (fecIRA), were located on this virulence plasmid. No type IV secretion system (T4SS) or gene cluster encoding for other DNA-transfer-associated machinery were predicted to be present on pRJA166b, indicating that the virulence plasmid is non-transferable.
RJA166 carries the third plasmid, pRJA166c, with a size of 111,083 bp and a G+C content of 49.0% ( Figure S2b). Homologous plasmids were found in other K. pneumoniae strains, including HS11286 and AATZP. By using VRprofile 9 , the plasmid pRJA166c was also found to carry a predicted prophage gene  Dilutions may be used, if necessary. In the present study, K. pneumoniae RJA166 was used as the donor while the three different recipients were K. pneumoniae RJF293H, NTUH-K2044IT, and HS11286YZ6 (Table S1), all of which are hygromycin-resistant. Thus, 200 µg/ml hygromycin and 50 µg/ml ceftazidime were added to the LB agar for selection of transconjugants. Experiments were performed in three separate rounds. In addition, each round was repeated in triplicate. The results are presented as (mean±SD).
When transconjugants were observed, three colonies were selected randomly and cultured for verification. PCR with primers (Table S2), unique to the blaDHA-1 gene, the backbone of pRJA166a and the chromosomes of RJA166, RJF293H, NTUH-K2044IT and HS11286YZ6, were employed to confirm the transfer of pRJA166a from donor to recipients. XbaI-PFGE and S1-nuclease PFGE were also performed for further confirmation of the RJF293H transconjugants.

Plasmid stability assay
The stability of the blaDHA-1-carrying plasmid pRJA166a was tested in both its original host and in the RJF293H transconjugants, as per the previously described protocol 16

Fitness cost of the acquisition of the blaDHA-1-carrying plasmid pRJA166a
It has been reported that the acquisition of resistance plasmids and expression of the AmpC gene could come at an extra fitness cost to bacteria 17,18 . In the head-to-head competition, the ratio of the pRJA166a/RJF293H transconjugant (Tc1) to RJF293H decreased to only 0.02 after 24 hours, with an input ratio at 0.6 ( Figure 5a). These results indicated that the acquisition of the large resistance plasmid pRJA166a (231 kb in size) lowered the fitness of the MDR hvKP.

The virulence of transconjugants with pRJA166a is similar to that of recipient
We also compared the growth curve and virulence of pRJA166a-free RJF293H and the pRJA166a/RJF293H transconjugants (Tc1-3) from the filter mating described above. RJF293H and the pRJA166a/RJF293H transconjugants had similar growth rates, which showed that the acquisition of pRJA166a did not affect the primary metabolism too much (Figure 5b). In the killing assay on the greater wax moth larvae, RJF293H and the transconjugants also showed similar virulence (Figure 5c, p = 0.2695 by two-way ANOVA).

PCR detection of pRJA166a in collected of K. pneumoniae clinical isolates
We had collected twenty-three hypermucoviscous K. pneumoniae isolates, and forty-six cKP isolates in the Ruijin Hospital during the same time period as RJA166 (September 2014 to March 2016). We thus performed PCR to detect the presence of pRJA166a among these isolates using three sets of the primers specific to pRJA166a (Table S2). No isolates, except RJA166, were seen to harbour the blaDHA-1-carrying plasmid pRJA166a.     where i was the level of BLASTp identities of the region with the highest Bit score expressed as a frequency of between 0 and 1, lm the length of the highest scoring matching sequence (including gaps) and lq the query length. If there were no matching sequences with a  c.