Genetic Diversity of Carbapenem-Resistant Enterobacteriaceae (CRE) Clinical Isolates From a Tertiary Hospital in Eastern China

The prevalence of carbapenem-resistant Enterobacteriaceae (CRE) is increasing globally, with different molecular mechanisms described. Here we studied the molecular mechanisms of carbapenem resistance, including clonal and plasmid dissemination, of 67 CRE isolates collected between 2012 and 2016 from a tertiary hospital in Eastern China, an CRE endemic region. Species identification and susceptibility testing were performed using the BD Phoenix Automated Microbiology System. Isolates were characterized by PCR (for carbapenemases, ESBLs, AmpC and porin genes), multilocus sequence typing (MLST), pulsed-field gel electrophoresis (PFGE), and conjugation transfer experiments. Selected blaKPC-2 -harboring plasmids were subjected to next-generation sequencing using the Illumina Miseq platform. Among the 67 CRE isolates, 42 Klebsiella pneumoniae, 10 Serratia marcescens, 6 Enterobacter cloacae, 2 Raoultella ornithinolytica, 2 K. oxytoca, 1 K. aerogenes, and 4 Escherichia coli isolates were identified. Six different carbapenemases were detected, including blaKPC-2 (n = 45), blaKPC-3 (n = 1), blaNDM-1 (n = 6), blaNDM-5 (n = 1), blaIMP-4 (n = 2), and blaVIM-1 (n = 2); blaOXA-48-like genes were not detected. One E. cloacae strain possessed both blaNDM-1 and blaKPC-3, while two E. cloacae isolates harbored blaNDM-1 and blaVIM-1. ESBLs (CTX-M, SHV, and TEM) and/or AmpC (CMY, DHA, and ACT/MIR) genes were also identified in 59 isolates, including 13 strains that lacked carbapenemases. Several insertions or stop codon mutations were found within porin genes of K. pneumoniae, E. coli and S. marcescens isolates, both with and without carbapenemases. The 42 K. pneumoniae isolates belonged to 12 different sequence types (ST), with ST11 being the most common, while the 6 E. cloacae isolates comprised 4 different STs. The 10 S. marcescens all shared the same PFGE pulsotype, suggestive of clonal spread. Complete plasmid sequencing and PCR screening revealed both intra-strain and inter-species spread of a common blaKPC-2-harboring plasmid in our hospital. Taken together, our study revealed extensive genetic diversity among CRE isolates form a single Chinese hospital. CRE isolates circulating in the hospital differ significantly in their species, STs, porin genes, carbapenemase genes, and their plasmid content, highlighting the complex dissemination of CRE in this endemic region.

spread. Complete plasmid sequencing and PCR screening revealed both intra-strain

INTRODUCTION
Enterobacteriaceae are among the most common pathogenic Gram-negative bacteria (GNB), causing various community-and healthcare-acquired infections. Nowadays, multidrug resistant GNB (MDR-GNB) are increasingly described in clinical settings, and carbapenems are regarded as the most effective antibiotic therapy for infections caused by MDR-GNB. However, as a result of clinical use of carbapenems since the late 1980s, the occurrence of carbapenem-resistant Enterobacteriaceae (CRE) has been increasingly reported worldwide, including in China (Gupta et al., 2011;Nordmann and Poirel, 2014). According to reports by the China Antimicrobial Resistance Surveillance System (CARSS), the detection rate of CRE in China increased from 2005 to 2014, demonstrating a continuous upward trend and suggesting a worsening situation (Hu et al., 2016).
Carbapenem resistance in Enterobacteriaceae can arise through distinct molecular mechanisms, mainly via the production of carbapenemases, but also as a consequence of outer membrane porin dysfunction coupled with hyperproduction of AmpC cephalosporinases or extended-spectrum β-lactamases (ESBLs) (Bush and Jacoby, 2010;Bush and Fisher, 2011). Carbapenemases are a group of β-lactamases that are capable of hydrolyzing carbapenem antibiotics, in addition to cephalosporins and other β-lactam antimicrobials. Three major class of carbapenemases are widespread globally in clinical CRE isolates, including class A (mainly KPC), class B (VIM, NDM, and IMP) and class D (OXA-48 and its variants, etc.) (Ambler, 1980;Hall et al., 2003;Nordmann and Poirel, 2014). Notably, carbapenemase genes are primarily carried by large conjugative plasmids, thereby facilitating horizontal transfer of carbapenem resistance among different bacterial strains and species. As mentioned above, another common mechanism of carbapenem resistance involves the combination of porin dysfunction with hyper-production of AmpC (e.g., CMY, DHA, and ACT) or ESBLs (e.g., TEM, SHV, and CTX-M) (Logan and Weinstein, 2017). The lack of the production of porins can preclude diffusion of antibiotics through bacterial membranes, along with the action of ESBL and AmpC enzymes, thereby producing the phenotype of carbapenem resistance in Enterobacteriaceae (Paterson and Bonomo, 2005;Jacoby, 2009;Bush and Jacoby, 2010). Unlike carbapenemases, porin dysfunction-associated resistance is not able to spread through horizontal transfer, but may disseminate via clonal expansion.
Dissemination of CRE usually demonstrates geographical and temporal variation in specific global regions. Here we characterized CRE clinical strains collected from a tertiary hospital in eastern China between August 2012 and August 2016. The genetic relatedness, antimicrobial susceptibility, and carbapenem-resistance mechanisms of these CRE isolates were examined in detail.

Identification of Carbapenem-Resistant Enterobacteriaceae Isolates
Sixty-seven unique (one isolate per patient) CRE clinical isolates were retrospectively collected from the Second Affiliated Hospital of Soochow University between August 2012 and August 2016. In this study, carbapenem resistance was defined as resistance to meropenem or imipenem based on 2016 Clinical and Laboratory Standards Institute (CLSI) guidelines (CLSI, 2016). The isolates were collected from various sources, including sputum (n = 48), urine (n = 14), blood (n = 2), catheter (n = 1), ascites (n = 1), and drainage fluid (n = 1). Species identification was performed using the Phoenix 100 Automated Microbiology System (Becton-Dickinson, United States), and confirmed by 16S rRNA sequencing (Weisburg et al., 1991). This study was approved by the institutional review board (IRB) of The Second Affiliated Hospital of Soochow University. The clinical isolates were retrospectively collected, and patient data were not included in this study.

Pulsed-Field Gel Electrophoresis (PFGE)
Raoultella ornithinolytica and Serratia marcescens strains (for which no MLST schemes Are available) Were further investigated by PFGE using a CHEF Mapper Power Module instrument (Bio-Rad, United States). In brief, genomic DNA Was digested With Xba I, and then electrophoresed Under the following conditions: voltage 6 V/cm, running time 18-19 h, temperature 14 • C, and pulse times of 5-40 s (R. ornithinolytica) and 5-20 s (S. marcescens). Salmonella strain H9812 Was used as a control strain and size marker. Clonal relatedness Between strains Was evaluated based on the criteria proposed by Tenover et al. (1995).

Plasmid Sequencing and Screening
Conjugation transfer experiments were performed with selected bla KPC−2 -harboring strains and rifampicin-resistant E. coli EC600. Experiments were carried out using mixed broth culture method as described previously (Chen et al., 2014a). Transconjugants were identified by detecting resistance genes using PCR. Plasmid DNA from E. coli EC600 transconjugants harboring single plasmids was extracted using a Qiagen Plasmid Midi Kit (Qiagen, Valencia, CA, United States), and sequenced using the Illumina Miseq system (Illumina, United States) (Du et al., 2016). Sequencing reads were assembled de novo into contigs using SPAdes (Bankevich et al., 2012), then manually inspected using Geneious 9.1 2 ; and gaps were closed by PCR and Sanger sequencing.
A PCR mapping strategy was developed to detect a common bla KPC−2 -harboring pSZF_KPC/p628-KPC-like plasmid sequenced in this study. The scheme includes six individual PCR reactions. PCR-I was designed to target the region spanning IncFII replicon gene repA and its downstream DNA methylase gene, while PCR-V was designed to target the junction of second replicon gene repB and its upstream parA. PCR-II, III, and PCR-IV were designed to target the bla KPC−2 neighboring regions of rep-klcA, klcA-bla KPC−2 , and bla KPC−2 -IS26, respectively (Figure 1). PCR-VI was used to detect the traX-finO junction in pSZF_KPC/p628-KPC-like plasmids. The oligonucleotide primer target regions for identification of pSZF_KPC/p628-KPC-like plasmids are shown in Figure 1, and primer sequences are listed in Table 1.

Other Mechanisms Associated With Carbapenem Resistance
As described above, PCR failed to identify any carbapenemases among 13 out of 67 CRE strains, including 5 K. pneumoniae, 3 E. cloacae and 5 S. marcescens, suggesting that other mechanisms may have contributed to the phenotypic carbapenem resistance identities. ORFs are portrayed by arrows and colored according to predicted gene function: orange arrows indicate plasmid scaffold regions; green arrows denote genes associated with the tra locus; dark blue arrows indicate replication-associated genes; Red arrows denote antimicrobial and mercury resistance genes; and yellow arrows indicate accessory genes. Small black arrowheads above the plasmids indicate the locations of primers used for PCR screening (primer sequences are shown in Table 1). among these isolates. We therefore examined ESBL-encoding genes (bla TEM , bla SHV , and bla CTX−M ), AmpC-encoding genes (bla CMY , bla DHA , and bla ACT ) and outer membrane porin genes. ESBL and AmpC genes were tested in all isolates, while outer membrane porin genes were examined in K. pneumoniae (ompK35 and ompK36), as well as in E. coli, E. cloacae, and S. marcescens (ompF and ompC). Fifty-nine isolates were found to carry at least one ESBL and/or AmpC gene, including 13 non-carbapenemase-producing strains. Specifically, 51 (76.1%) strains were found to carry bla CTX−M genes, including bla CTX−M−65 (n = 19), bla CTX−M−14 (n = 19), bla CTX−M−9 (n = 7), bla CTX−M−15 (n = 3), and bla CTX−M−3 (n = 3). These were found in several species, including K. pneumoniae (n = 32), S. marcescens (n = 10), E. cloacae (n = 6), E. coli (n = 2), and R. ornighinolytica (n = 1). In addition, 30 strains were positive for bla SHV−12 , most of which were K. pneumoniae (26/30, 86.7%), while 21 K. pneumoniae harbored bla DHA−1 , 6 E. cloacae possessed bla ACT , and two E. coli were positive for bla CMY−2 .
Outer membrane porin gene sequence analysis showed that 27 K. pneumoniae isolates harbored ompK35 and/or ompK36 mutations. ompK35 mutations (n = 26) were exclusively due to premature stop codons, while ompK36 mutations included glycine-aspartic acid (GD) insertions at amino acid positions 134-135 (n = 23), IS10 insertions (n = 2), and stop codons (n = 1). Five non-carbapenemase-producing K. pneumoniae CRE contained at least one outer membrane porin gene mutant (ompK35 or ompK36) while also harboring ESBL genes bla CTX−M or bla SHV−12 , which likely explains the carbapenem resistance among these isolates. Sequence analysis of ompF and ompC genes in S. marcescens showed that they all possess mutated ompF, with premature stop codons at amino acid position 72. One noncarbapenemase-producing E. cloacae isolate was also found to carry an ompF mutation (stop codon), in addition to bla ACT and bla CTX−M−9 . However, two non-carbapenemase-producing E. cloacae CRE isolates did not display mutations in either ompF or ompC, although they were found to harbor ESBL genes (either bla CTX−M−14 or bla CTX−M−3 ). We suspect additional mechanisms, such as efflux pumps or penicillin-binding protein modifications, may contribute to carbapenem resistance in the latter two isolates. Therefore, carbapenem resistance among 11 out of 13 non-carbapenemase-producing CRE isolates may be

Distribution of MLST Sequence Types and PFGE Patterns
Multilocus sequence typing results showed that 42 K. pneumoniae belonged to 12 different STs, with ST11 being the most common (25/42, 59.5%). All ST11 isolates possessed the same OmpK35 stop codon, while 23 of them harbored the 134-135 GD OmpK36 mutant. The two K. oxytoca isolates belonged to ST135 and ST180, while the 4 E. coli strains were assigned to ST1488, ST3234, ST167, and ST354. The six E. cloacae also comprised 4 STs, including ST231 (n = 3), ST120 (n = 1), ST421 (n = 1), and ST97 (n = 1). Two R. ornithinolytica and 10 S. marcescens strains were further analyzed by Xba I-PFGE. The results showed that all 10 S. marcescens strains shared the same PFGE pattern, suggesting clonal spread (Supplementary Figure 1). In addition, the ten S. marcescens isolates were collected within 7 months in 2013 from three closed wards (respiratory, neurology, and renal wards) and mostly from respiratory samples (n = 9), suggestive of the likelihood of a small S. marcescens outbreak.
As aforementioned, all S. marcescens harbored the same OmpF mutant, and among them, 5 were also found to carry bla KPC−2 . By contrast, the two R. ornithinolytica isolates displayed unrelated PFGE patterns (data not shown).

Sequencing and Screening of bla KPC−2 -Harboring Plasmids
Since KPC-2 is the most common carbapenemase found in our hospital, we sought to determine the complete sequences of bla KPC−2 -harboring plasmids spreading through our institution. We selected 4 bla KPC−2 -harboring strains, including two K. pneumoniae (ST8 and ST11), one E. coli (ST3234), and one S. marcescens for plasmid conjugation and complete plasmid sequencing.
The bla KPC−2 -harboring plasmid (subsequently named pSZN_KPC) isolated from K. pneumoniae ST8 belonged to incompatibility group N. Plasmid pSZN_KPC is 65,604 bp in length, with an average G+C content of 53.3 %, and contains 88 predicted open reading frames (ORFs) (Figure 1). A BLAST search of the pSZN_KPC plasmid sequence against the GenBank database 3 showed that pSZN_KPC exhibits a high degree of identity to the previously published IncN plasmid pECN580 (accession no. KF914891) from an E. coli strain found in China (Chen et al., 2014a), with 100% query coverage and overall 99% nucleotide identity. The major difference is that the region harboring antirestriction protein gene ardA in plasmid pSZN_KPC was located downstream of gene ccgCD (Figure 1), likely as a consequence of recombination.
The bla KPC−2 -harboring plasmids from K. pneumoniae ST11 strain Kp715, E. coli ST3234 strain Ec732, and S. marcescens 3 http://blast.ncbi.nlm.nih.gov/Blast.cgi strain Sm703 were highly similar (each differed by only 4 SNPs), and belonged to the same IncFII incompatibility group. This plasmid (subsequently named pSZF_KPC) is 106,201 bp in length with a G+C content of 53.2 %, and harbors 122 predicted ORFs. Comparative sequence analysis for plasmid pSZF_KPC showed that it is highly similar (100% query coverage and >99% nucleotide identity) to p628-KPC (accession no. KP987218) from a K. pneumoniae isolate collected in our hospital in 2010 (Wang et al., 2015). Compared to p628-KPC, the main difference in pSZF_KPC is that there is an additional ISKpn18 element, with IS4321 located downstream of the mcr operon (Figure 1). The sequences of pSZN_KPC and pSZF_KPC have been deposited in GenBank under the accession numbers MH917122 and MH917123, respectively. The plasmid sequence results described above suggest that pSZF_KPC/p628-KPC-like plasmids have been spreading in our hospital since at least 2010, undergoing horizontal transmission into different species. We therefore developed a set of PCR assays to screen for the presence of pSZF_KPC/p628-KPClike plasmids among our 67 CRE isolates. The PCR results showed that 27 (64.2%) of the 42 bla KPC−2 positive strains harbored pSZF_KPC/p628-KPC-like plasmids. The 27 strains encompassed different species and STs, including K. pneumoniae ST11 (n = 14), ST774 (n = 3), ST1107 (n = 2), ST211 (n = 1), ST655 (n = 1), and ST218 (n = 1); S. marcescens (n = 4); and E. coli ST3234 (n = 1). These pSZF_KPC/p628-KPC-like plasmid-harboring isolates were collected in 8 different wards, spanning from 2012 to 2016, indicating the frequent horizontal transfer of this common plasmid among Enterobacteriaceae in our hospital.

DISCUSSION
At the present time, KPC, NDM and OXA-48 are the most common carbapenemases worldwide (Nordmann and Poirel, 2014;Sugawara et al., 2016). KPCs are most frequently identified in K. pneumoniae from the United States, China, Colombia, Israel, Greece, and Italy, while NDMs are primarily found in K. pneumoniae, E. coli and Enterobacter spp. from the Indian subcontinent, and OXA-48-like carbapenemases in K. pneumoniae and E. coli from North Africa and Turkey (Nordmann and Poirel, 2014). In addition, the spread of CPEs has been associated with several high-risk clones. One notable example involves the global spread of KPCs, which has been largely associated with K. pneumoniae clonal group 258 (CG258) strains, of which ST258 is the most predominant KPC-producing K. pneumoniae clone in North America, while ST11 is most common in East Asia, especially China (Patel and Bonomo, 2013;Chen et al., 2014b;Zhang et al., 2017).
China, in particular eastern China, is regarded as one of the primary global endemic regions for CRE (Zhang et al., 2017). In this study, we phenotypically and genetically characterized the CRE isolates collected from an eastern Chinese hospital, and investigated the molecular mechanisms underlying carbapenem resistance. Our study revealed several interesting findings.
Firstly, the CRE isolates were recovered from seven different Enterobacteriaceae species. Although carbapenem resistance has been frequently identified in K. pneumoniae, Enterobacter spp., and E. coli, it is fairly uncommon in other species such as R. ornithinolytica and S. marcescens. R. ornithinolytica is mostly recovered from the environment and rarely causes severe infections in humans; nevertheless several reports have described the emergence of carbapenem-resistant R. ornithinolytica in China (Zhou et al., 2013;Qin et al., 2014;Yang et al., 2018). Notably, in this study we identified two carbapenem-resistant R. ornithinolytica, harboring bla KPC−2 and bla IMP−4 , respectively, suggesting that different carbapenemase plasmids have spread into R. ornithinolytica. In contrast, carbapenem resistance in S. marcescens has been historically associated with a specific group of carbapenemases, SMEs. However, in this study, none of the ten S. marcescens strains were found to carry bla SME , and the observed carbapenem resistance was likely due to the mutations of OmpK35 porin encoding genes, as well as production of CTX-M-14 and KPC-2. Interestingly, only five of the ten S. marcescens strains were found to harbor bla KPC−2 , although PFGE results suggested that the spread of carbapenem-resistant S. marcescens was largely clonal. We suspect it is likely that the S. marcescens OmpK35 mutant further acquired a bla KPC−2 plasmid (e.g., pSZF_KPC/p628-KPC-like).
In this study, K. pneumoniae was the most common CRE species, accounting for 62.7% (42/67) of all CRE isolates. Among these, ST11, a member of the epidemic CG258 clone, was the predominant ST (59.5%, 25/42), which is consistent with the molecular epidemiology described in other regions of China (Zhou et al., 2013;Zhang et al., 2017;Yang et al., 2018). However, in our study carbapenem-resistant K. pneumoniae were found in 12 different STs, including some STs rarely associated with carbapenem resistance (e.g., ST774 and ST1107). Our plasmid screening results showed that the diversity of STs was largely due to the frequent transfer of a common bla KPC−2 vector into different K. pneumoniae genetic backgrounds.
Secondly, our study revealed diverse molecular mechanisms of carbapenem resistance within our hospital, albeit with carbapenemase production the primary cause. Diverse types of carbapenemases, including KPC-2, KPC-3, NDM-1, NDM-5, IMP-4, and VIM-1, were identified, whereas in most Chinese hospitals, usually only KPC and/or NDM carbapenemases are prevalent (Hu et al., 2014). It is noteworthy that four different classes of carbapenemases were found in our hospital, suggesting that different carbapenemase producing plasmids/strains are spreading locally. Notably, however, ∼20% strains were noncarbapenemase producers. It is therefore likely that besides carbapenemases, other mechanisms such as porins defects and production of ESBLs (TEM, SHV, and CTX-M) or AmpC β-lactamases (CMY, DHA, and ACT) significantly contributed to carbapenem resistance in our hospital. Meanwhile, it is worth noting that 51.9% (28/54) of carbapenemase-producing isolates also harbored additional porin gene mutations, potentially rendering antimicrobial treatment more challenging in comparison to strains without porin mutations (Clancy et al., 2013).
Lastly, we identified a widespread bla KPC−2 -harboring plasmid vector within our institution. Several pSZF_KPC/p628-KPC-like plasmids were identified within distinct K. pneumoniae clones (STs), as well as in different species (K. pneumoniae, S. marcescens and E. coli). Our PCR-based screening of bla KPC−2 -positive K. pneumoniae isolates revealed that this plasmid is widely disseminated in our hospital and was found in nearly two third (22/37) of KPC-2 positive isolates, highlighting the importance of plasmid horizontal transfer in the dissemination of KPC. More importantly, this plasmid was identified in different K. pneumoniae STs, as well as other species, suggesting that intra-strain and inter-species transfer of this plasmid have significantly contributed to the spread of carbapenem resistance in our hospital.
In conclusion, our study revealed extensive genetic diversity among CRE isolates from a single Chinese hospital. Both clonal expansion (e.g., K. pneumoniae ST11 and S. marcescens) and plasmid horizontal transfer (i.e., pSZF_KPC/p628-KPC-like plasmids) were identified. Different carbapenemases classes and outer membrane porin defects were found in several species. CRE isolates circulating in our hospital differ significantly in their species, STs, porin genes, carbapenemase genes, and plasmid content, highlighting complex dissemination of CRE within our hospital. Further studies are required to understand the factors underlying the genetic diversity of CRE in our hospital, in order to effectively control further spread of carbapenem resistance.

AUTHOR CONTRIBUTIONS
MM, HW, and PX contributed to work, data analysis, and manuscript preparation. SN, JL, JM, Y-WT, and BK prepared the manuscript. XX contributed to work and data analysis. XZ and HZ analyzed the data. HD and LC contributed to study design, data analysis, and manuscript preparation.