Stool Samples of Acute Diarrhea Inpatients as a Reservoir of ST11 Hypervirulent KPC-2-Producing Klebsiella pneumoniae

China has been experiencing a rapid increase in the number of nosocomial infections caused by carbapenem-resistant Klebsiella pneumoniae ST11 (ST11-CRKP) for decades. The emergence of hypervirulent ST11-CRKP (ST11-CR-HvKP) strains is expected to become a serious public health issue in China, considering that carbapenem resistance and virulence have converged in an epidemic clone. K. pneumoniae strains that colonize the human intestinal tract may become a reservoir of virulence and carbapenemase-encoding genes. Here, we first characterized the genotypes and antimicrobial phenotypes of ST11-CR-HvKP strains isolated from diarrheal stool samples of inpatients in Zhejiang Province, China. Active surveillance approaches based on the findings of the present study should be implemented, particularly in intensive care units, to combat the spread of ST11-CR-HvKP and to improve treatment.

C arbapenem-resistant Klebsiella pneumoniae (CRKP), which is associated with high mortality rates of up to 50%, has been identified by the World Health Organization as a critical-priority organism (1,2). Hypervirulent K. pneumoniae (HvKP) causes lifethreatening infections, and the rmpA and rmpA2 genes are associated with its pathogenicity (3). Recently, a fatal outbreak of sequence type 11 (ST11) carbapenemresistant, hypervirulent K. pneumoniae (ST11-CR-HvKP) in China generated great concern from the public health community (4). The identification of this ST11-CR-HvKP strain demonstrated that carbapenem resistance and virulence had converged in an epidemic clone that could become a serious public health issue (5). Subsequently, cases of ST11-CR-HvKP infection have been documented in diverse regions in China (6)(7)(8). These reports suggest that this clone is widespread in China; however, the underlying mechanism that enables its wide dissemination remains unclear.
Intestinal tract colonization by carbapenemase-producing Enterobacteriaceae (CPE) may lead to a subsequent nosocomial infection in at-risk patients (9)(10)(11). We hypothesized that a relatively silent CPE gastrointestinal reservoir could exist in inpatients with diarrhea and could potentially be involved in the transmission of ST11-CR-HvKP. Accordingly, the aims of the present study were to investigate the prevalence of CPE carriage among inpatients with diarrhea in a teaching hospital in China over 1 year, to identify ST11-CR-HvKP reservoirs, and to explore the genomic complexity of the highly transmissible ST11-CR-HvKP clone in Zhejiang Province.
Patients in medical intensive care units (ICUs) typically experience a prolonged hospital stay, which results in exposure to potential risk factors. Not surprisingly, therefore, most of the colonization cases were observed among patients from medical ICUs (Table S1). Colonization was detected in 13 wards: 49% (43/87) of cases were from the ICU, 9% were from the hepatobiliary and pancreatic surgery ward, 8% were from the emergency ICU, and the remaining 33% were from 10 other units. Forty-two (76%) of the inpatients were male, and the median inpatient age was 64 years (range, 16 to 97 years). Interestingly, correlations were detected between antibiotic exposure (including linezolid, a ␤-lactam-␤-lactamase inhibitor combination, and carbapenems), surgical history, and being CPE positive (Table 1).
To better define the population structure of the 65 CRKP isolates, we further investigated their comprehensive molecular characteristics. Remarkably, multilocus sequence typing (MLST) analysis revealed that the K. pneumoniae isolates belonged to four STs, indicating low diversity ( Fig. 1 and Table S1). ST11 was the predominant ST (58/65; 89%), followed by ST37 (n ϭ 5), ST15 (n ϭ 1), and ST107 (n ϭ 1). A notable feature was that patient P1 (single individual) carried two STs, ST37, which was isolated from three different samples, and ST11, which was isolated from eight samples, over the entire study period. Furthermore, 10 patients in the study carried multiple isolates. Several major pulsotypes were found via pulsed-field gel electrophoresis (PFGE) analysis, which is consistent with the MLST results; however, genetic diversity was observed in the profiles (Fig. S3). Roary matrix-based gene sequence analysis generated a large pangenome of 10,355 gene clusters across 65 full genomes (Fig. S4). A maximum likelihood phylogenetic tree demonstrated that the 65 strains were partitioned into three major clades (Fig. 1). Seven non-ST11 isolates were clustered into two separate clades, and the 58 ST11 isolates were grouped into three clusters, despite their high similarity. An analysis of the wzi locus revealed three different K types (KL105, KL47, and K64) among the ST11 isolates, indicating genetic diversity among these isolates. Of note, ST11 isolates were detected in patients from 12 wards. A review of the cases revealed that these individuals had no contact with one another, lowering the probability of nosocomial transmission.
Using in silico analysis, we identified 14 rmpA-and 19 rmpA2-positive ST11-CRKP isolates. Interestingly, S1 nuclease PFGE (S1-PFGE) and Southern blot analyses clearly showed that the ST11-CR-HvKP isolates carried at least four types of rmpA-and rmpA2-positive plasmids, ranging in size from approximately 146 kb to approximately 218 kb (Fig. 4). The rmpA and rmpA2 genes coexisted on the same plasmid in 14 isolates. We used Pacific Biosciences (PacBio) sequencing to generate five complete rmpA-and rmpA2-carrying IncHI1B plasmid sequences, which aligned well with the virulence plasmid pLVPK (GenBank accession no. AY378100), a 219-kb plasmid that harbors a set of virulence genes, including iroBCDN, iucABCD, rmpA, and rmpA2 (Fig. 5). a Univariate analysis; categorical variables were compared using a 2 test. b Multivariate analysis of variables (P Յ 0.20) in the univariate analysis was performed using stepwise backward logistic regression. Odds ratios (ORs) and their 95% confidence intervals (CIs) were calculated. c P values were calculated based on a comparison between the ages of Ն60 and Ͻ60 years using a 2 test.

FIG 1
Core-genome phylogeny of the CRKP isolates, associated MLST data, capsule locus genotype, ward of isolation, virulence genes present, and isolation date. The maximum likelihood phylogeny tree is based on single-nucleotide polymorphisms in the core genomes of 65 CRKP isolates. The scale bar indicates nucleotide divergence. The origins of the isolates are shown in different colors. Genome clusters are shaded using different colors. The presence of the rmpA and rmpA2 genes is indicated. The curved lines indicate isolates from the same individual. The red asterisks indicate the isolates that were analyzed by PacBio sequencing. EICU, emergency intensive care unit.

DISCUSSION
It is well recognized that intestinal tract colonization by Gram-negative bacteria may lead to a subsequent nosocomial infection in high-risk patients (12). In a recent large cohort study, Frencken et al. (13) demonstrated that rectal carriage of a Gram-negative pathogen is strongly associated with an increased risk of ICU-acquired bacteremia, suggesting that routine surveillance for gut colonization by Gram-negative bacteria may have clinical implications. Therefore, we designed the present study as a prospective survey to evaluate gut colonization by CPE and the associated risk factors among inpatients with diarrhea. We also performed an in-depth characterization of the gut flora CPE isolates. ST11 is the most prevalent ST among CRKP isolates in China and accounts for up to 59.8% of CRKP isolates (14). We previously reported that 89.9% (124/138) of CRKP strains isolated from a university-affiliated hospital in 2014 carry the bla KPC-2 gene (15). Furthermore, K. pneumoniae ST11 undergoes rapid clonal expansion, resulting in regional outbreaks (16), which poses a challenge to patients, clinicians, and public health. The first ST11-CRKP strain in China was isolated at the First Affiliated Hospital of Zhejiang University (FAHZU) in 2004 (17). In the past few decades, China has witnessed the emergence and subsequent rapid increase in the number of ST11-CRKP nosocomial infections (14). HvKP is rarely resistant to antibiotics. The occurrence of ST11-CR-HvKP threatens the effective management of infections caused by this clinically untreatable pathogen, which could become a serious public health problem and even cause a global CR-HvKP epidemic (5). However, the role of the gut microbiota in the rapid spread of ST11-CR-HvKP has not yet been elucidated. Experimental evidence from the present study, as well as from retrieved clinical data, indicated a disturbing rate of carriage of ST11-CRKP (8.6%) among hospitalized patients with acute diarrhea, particularly in ICU wards. Furthermore, the findings of the present study suggest that stool samples could be a major reservoir of ST11-CR-HvKP. Active surveillance of nondiarrheal patients is required to compare their rates of CPE colonization with those of diarrheal patients and to investigate whether the implementation of contact precautions or other interventions would improve outcomes.
Not surprisingly, in the present study, the majority of colonization cases were observed among patients in the medical ICU. Recently, a systematic review of the clinical epidemiology of CRE identified 13 risk factors associated with CRE acquisition (18). These are medical devices, carbapenem use, invasive procedures, ICU admission, cephalosporin use, glycopeptide use, CRE exposure, underlying diseases, quinolone use, ␤-lactam use, mechanical ventilation, demographic patient characteristics, and exposure to hospital care. The findings of the present study are consistent with those observations, although we focused on gut colonization by CPE and not on CPEassociated infections. To the best of our knowledge, an association between linezolid exposure and CPE colonization in patients has not been reported to date. The lack of such an association might be explained by the wide use of linezolid, which eliminates Gram-positive bacteria and disrupts the gut microflora, thus promoting gut colonization by Enterobacteriaceae, especially by CPE strains.
A previous investigation speculated that the emergence of ST11-CR-HvKP occurred due to a single genetic event in which a pLVPK-like virulence plasmid was acquired by an ST11-CRKP strain that presumably carried a common plasmid such as pKPC-CR-HvKP4 (4). The observed complexity in the existence of virulence plasmids and resistance plasmids in the ST11-CRKP isolates indicates that this process was more complicated than was previously anticipated, as it may have involved multiple ST11 K. pneumoniae lineages and a variety of virulence plasmids.
This study has several limitations. First, our study is limited by its single-institution design and 1-year study period; therefore, the findings of this study should be interpreted with caution and might not be generalizable to other hospitals. Second, although we identified 19 fecal ST11-CR-HvKP colonization isolates, none of the colonized patients had ST11-CR-HvKP infections. Third, the findings probably reflect the status of diarrheal patients, i.e., individuals with a disrupted gut microbiota. A previous investigation described the high prevalence and mortality rate of ST11-CR-HvKP meningitis in FAHZU (8), and K64 was also the most common serotype detected in our study. Future investigations addressing the transition from ST11-CR-HvKP carriage to infection in high-risk patients are crucial for the management of such infections.
In conclusion, our genomic epidemiological investigation demonstrated that stool samples of hospitalized patients served as a reservoir of CPE isolates, particularly ST11-CR-HvKP. Because of the serious clinical outcomes of CRKP infections, more information is needed to understand the potential risk of ST11-CR-HvKP spreading by inpatients and to develop measures for the surveillance or control of these risks.

Study design.
A prospective, observational cohort study involving inpatients with diarrhea was performed between February 2016 and February 2017 (see Fig. S1 in the supplemental material) at the First Affiliated Hospital of Zhejiang University (FAHZU), China. FAHZU is the largest tertiary teaching hospital in Zhejiang Province, with 2,500 beds. Enterobacteriaceae isolates cultured from the first collected clinical samples of the respective inpatients were collected. Duplicate isolates (the same species isolated from material from the same patient within 2 weeks of the first positive culture) were excluded. All CPE isolates underwent antimicrobial susceptibility testing, molecular typing, and whole-genome sequencing (WGS). Clinical data for all participants were extracted from the clinical and medical record system using centralized queries. Information on demographic data, characteristics at admission, surgical intervention, hospital-acquired infections, and antibiotic exposure during the hospital stay were collected for CPE carriers and noncarriers.
Bacterial isolation and identification. Fecal samples (1.0 g) were diluted in 5 ml of sterile Luria-Bertani broth and cultured overnight at 37°C. The cultures grown overnight were plated on MacConkey agar supplemented with 1 mg/liter meropenem for 18 to 24 h at 37°C to isolate potential carbapenemresistant Enterobacteriaceae (CRE) strains. Colonies with different morphologies were repeatedly streaked on MacConkey agar with meropenem to obtain pure isolates. Bacterial species were identified by matrix-assisted laser desorption ionization-time of flight mass spectrometry and 16S rRNA gene sequencing using universal prokaryotic primers. CRE isolates were confirmed by susceptibility testing in the presence of imipenem, ertapenem, and meropenem via a broth microdilution method (19). Genomic DNA was extracted from cultures grown overnight using the Qiagen (Hilden, Germany) blood/tissue kit. CPE isolates were identified by detecting carbapenemase-encoding genes (bla NDM , bla KPC , bla IMP , and bla VIM ). The genes were amplified by PCR, with 1 l of genomic DNA from each isolate as a template, and the nature of the amplicons was then confirmed by sequencing (20).
Characterization of HvKP. The presence of the virulence-associated genes rmpA (regulator of mucoid phenotype A), rmpA2 (activator for capsule biosynthesis), iucA, and iutA (21,22) was examined in HvKP strains by PCR. The PCR products were first visualized by 1% agarose gel electrophoresis, and the nature of the amplicons was then confirmed by direct Sanger sequencing. Furthermore, the hypermucoviscous phenotype was analyzed by using the string test, as described elsewhere (23). The cutoff criterion for positivity in the string test was a viscous string length of Ͼ5 mm.
Risk factor analysis. To investigate the risk factors associated with fecal carriage of CPE, carriers were compared with noncarriers in terms of exposure to different variables. Categorical variables were expressed as percentages and compared using the chi-squared test or two-tailed Fisher exact test, as appropriate. Independent predictors for CPE were examined by logistic regression analysis. Variables with a P value of Յ0.2 (univariate analysis) were included in a logistic regression model to identify the variables with either a negative or a positive impact on the CPE colonization of inpatients with acute diarrhea. The strength of associations was determined by calculating the odds ratios (ORs) and 95% confidence intervals (CIs). Variables with a two-tailed P value of Ͻ0.05 were considered statistically significant. All statistical analyses were performed using SPSS version 24.0.
Antimicrobial susceptibility testing. MIC values of 14 antimicrobial agents (amikacin, aztreonam, cefotaxime, cefpirome, ceftazidime, ciprofloxacin, colistin, fosfomycin, gentamicin, imipenem, meropenem, piperacillin-tazobactam, tobramycin, and tigecycline) were determined by an agar dilution method, except for colistin and tigecycline, for which the broth microdilution method was used. The results were interpreted using Clinical and Laboratory Standards Institute guideline breakpoints (19). Standard reference strains of Escherichia coli, ATCC 25922, and K. pneumoniae, ATCC 700603, were used for quality control.
Molecular typing of KPC-2-producing K. pneumoniae isolates. The genetic relatedness of K. pneumoniae isolates was assessed by pulsed-field gel electrophoresis (PFGE). Briefly, DNA fragments were separated by 1% agarose gel electrophoresis (SeaKem gold agarose; Lonza, USA) at 14°C and 6 V/cm and with an alternating pulse at 120°, with a 2-to 40-s pulse-time gradient, for 22 h in 0.5ϫ Tris-boric acid-EDTA buffer using a Chef apparatus (Bio-Rad, USA). Salmonella enterica serotype Braenderup H9812 was used as a size marker (24). The PFGE pattern dendrogram was constructed using BioNumerics version 7.6 (Applied Maths, Sint-Martens-Latem, Belgium) with UPGMA (unweighted pair group method using average linkages) clustering. Isolates with a similarity cutoff of Ն80% were considered to be a pulsotype.
Illumina sequencing and sequence assembly. To characterize the genetic features and resistome of KPC-2-producing K. pneumoniae isolates, WGS was performed. Genomic DNA was extracted from cultures grown overnight using the Gentra Puregene Yeast/Bact kit (Qiagen). The extracted DNA was evaluated by 1% (wt/vol) agarose gel electrophoresis, and its concentration and purity were determined using a NanoDrop 2000 UV-visible (UV-Vis) spectrophotometer (Thermo Scientific, Waltham, MA, USA) and a Qubit version 2.0 fluorometer (Thermo Scientific). DNA was stored at Ϫ20°C until further processing. The sequencing library was prepared by using a Nextera XT kit (Illumina, San Diego, CA, USA) and sequenced using the Illumina HiSeq X 10-PE150 platform (Illumina). A-tailed fragments were ligated with paired-end adaptors and PCR amplified with a 500-bp insert. A mate pair library with an insert size of 5 kb was used for library construction by Beijing Novogene Bioinformatics Technology Co., Ltd. PCR adaptor reads and low-quality reads from the paired-end and mate pair library were filtered during a quality control step using the Novogene pipeline. Paired reads were then assembled into scaffolds using Velvet version 1.2.10 (25).
Long-read Pacific Biosciences sequencing and assembly. To elucidate the genetic environment of K. pneumoniae isolates carrying the rmpA and rmpA2 genes, five rmpA-and rmpA2-positive isolates were selected based on the size of the rmpA-and rpmA2-carrying plasmids and analyzed by PacBio sequencing. DNA extracted from isolates L39, L201, L388, L482, and L491 was sequenced by using long-read single-molecule real-time (SMRT) sequencing technology and the PacBio (Menlo Park, CA, USA) RS II platform. Briefly, the extracted high-quality double-stranded DNA (10 g) was sequenced using P6-C4 chemistry, PacBio RS II instrumentation, and a complexed 20-kb SMRTbell library. Unicycler was used for a hybrid assembly of three bacterial genomes from a combination of Illumina short reads and PacBio long reads (26). To generate the best assembly, Unicycler was operated in three different modes (conservative, normal, and bold) that alter the cutoff for a minimum acceptable bridge quality.
Bioinformatics analysis and phylogenomic computations. To determine the clonal lineages, the sequence types (STs) of KPC-2-producing K. pneumoniae isolates were determined by multilocus sequence typing (MLST) of WGS data, in accordance with protocols described on the Institut Pasteur website (http://bigsdb.pasteur.fr/klebsiella/). ResFinder version 2.1 (http://cge.cbs.dtu.dk/services/ resfinder) was used to identify antimicrobial resistance genes. Plasmid Finder version 1.3 was used to identify the plasmid incompatibility type (27). The virulence loci in the assembled genome sequences were identified using the BIGSdb Klebsiella genome database. A heat map of virulence loci was generated using Genesis software version 1.7.7. The core genes in the genomes of KPC-2-producing K. pneumoniae strains were identified using Prokka (28) and Roary (29). Maximum likelihood-based phylogenetic reconstruction was performed with RAxML version 8.2.10 using the generalized time-reversible (GTR) ϩ ⌫ nucleotide substitution model (30). One hundred bootstrap replicates were evaluated to determine branch support. A maximum likelihood phylogenetic tree based on the core single-nucleotide polymorphism alignments was generated using FastTree (31).
Characterization of the rmpA-and rmpA2-carrying plasmids. S1 nuclease pulsed-field gel electrophoresis (S1-PFGE) and Southern blot analysis were performed to estimate the size of the rmpA-and rmpA2-carrying plasmids (32). Plasmid sequences were assembled from WGS data using plasmidSPAdes (33) and annotated using the RAST tool (34). The sequences of representative plasmids were compared against plasmid sequences deposited in the National Center for Biotechnology Information database using BLAST and plotted by using the BLAST Ring Image Generator (35).
Data availability. The complete whole-genome sequences of 65 CRKP isolates have been deposited in the GenBank database under BioProject accession no. PRJNA390758.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.

ACKNOWLEDGMENTS
We thank the personnel of the clinical microbiology laboratory at FAHZU for assistance with collecting samples.
This study was partially funded by grants from the National Key Research and Devel-