Frequent convergence of mcr-9 and carbapenemase genes in Enterobacter cloacae complex driven by epidemic plasmids and host incompatibility

ABSTRACT Convergence of mcr and carbapenemase genes has been sporadically detected in Enterobacter cloacae complex (ECC) with an upward trend. However, the state of the epidemic and underlying mechanism of such convergence has been poorly understood. In this study, the co-occurrence of MCR and carbapenemases was systematically analyzed in 230 clinical ECC isolates collected between 2000 and 2018 together with a global dataset consisting of 3,559 ECC genomes compiled from GenBank. We identified 48 mcr-9/mcr-10-positive isolates (MCR-ECC) (20.9%) in our collection, and a comparable ratio of MCR-ECC (720/3559, 20.2%) was detected in the global dataset. A high prevalence of carbapenemase-producing MCR-ECC (MCR-CREC) was further identified in the MCR-ECC of both datasets (16/48, 33.3%; 388/720, 53.9%), demonstrating a frequent convergence of mcr-9/10 and carbapenemase genes in ECC worldwide. An epidemic IncHI2/2A plasmid with a highly conserved backbone was identified and largely contributed to the dissemination of mcr-9 in ECC worldwide. A highly conserved IncX3-type NDM-1-carrying plasmid and IncN-type IMP-4-carrying plasmid were additionally detected in MCR-CREC isolated in China. Our surveillance data showed that MCR-CREC emerged (in 2013) much later than MCR-ECC (in 2000), indicating that MCR-CREC could be derived from MCR-ECC by additional captures of carbapenemase-encoding plasmids. Tests of plasmid stability and incompatibility showed that the mcr-9/mcr-10-encoding plasmids with the NDM-1-encoding plasmids stably remained in ECC but incompatible in Escherichia coli, suggesting that the convergence was host-dependent. The findings extend our concern on the convergence of resistance to the last resort antibiotics and highlight the necessity of continued surveillance in the future.


Introduction
Clinical infections caused by multidrug-resistant (MDR) bacteria with an increasing trend have become a critical threat to the public health network [1]. In particular, the dramatic increase of carbapenem resistance worldwide has greatly compromised the efficacy of carbapenems and prompted renewed attention to the importance of the last-line antibiotics colistin (polymyxin E). However, with the heavy consumption of colistin in veterinary medicine and clinical settings, colistin-resistant isolates have emerged globally [2][3][4], causing a therapeutic challenge in the clinical setting.
E. coli and K. pneumoniae are the major reservoirs for most of the mcr genes, while mcr-9 and mcr-10 have been frequently reported in Enterobacter cloacae complex (ECC) [14,16,22,23]. ECC is an important nosocomial pathogen capable of causing a wide variety of infections, such as pneumonia, urinary tract infections, and septicemia [24]. Molecular typing methods have identified more than 20 species/phylogenetic clusters in the complex [25]. Epidemiological studies show that Enterobacter hormaechei is the predominant species among ECC in the clinical setting, contributing to increased morbidity and mortality in hospitalized patients, particularly when infected by carbapenem-resistant isolates [24,26].
Recently, the co-occurrence of carbapenemases and MCR in a single strain has been reported sporadically [27,28], especially the combination of genes encoding metallo-beta-lactamases and MCR-9. Up to date, such genotype (i.e. coexistence of carbapenemase genes and mcr-9) has been found in Enterobacter spp., Klebsiella spp., Escherichia spp., and Citrobacter spp. [17,22,[29][30][31]]. An MDR E. hormaechei isolate co-harboring bla NDM-1 and mcr-9 genes was identified in a retrospective screen of carbapenemase-and MCR-producing Enterobacterales recovered from a tertiary hospital in Hangzhou, China [30]. Two carbapenemase genes (bla IMP-4 and bla NDM-1 ) were simultaneously detected in an mcr-9-positive E. hormaechei strain [32]. In particular, a high incidence of carbapenemases, especially metallo-beta-lactamases, in mcr-9-positive genomes was detected in Enterobacter spp. collected from a tertiary hospital recently [33]. These data consistently suggest that Enterobacter spp. could be an important reservoir for carbapenemase genes and mcr-9. However, the underlying mechanism involved has been poorly understood. Herein, we aimed to systematically study the co-existence of mcr and carbapenemase genes in ECC isolates collected between 2000 and 2018 in a tertiary hospital of China together with a global dataset retrieved from GenBank. The plasmidom encoding mcr and carbapenemase genes were dissected, and their incompatibility and stability were tested to explore the co-occurrence mechanism.

Materials and methods
Bacterial isolates, mcr-positive-ECC (MCR-ECC) bacteria identification, and clinical data collection ECC isolates were collected at a tertiary hospital of China between 2000 and 2018, and have been species-typed previously [34]. The presence of mcr genes including mcr-1 ∼ mcr-10 were assessed by using primers reported previously [16]. The mcr positive isolates were determined as MCR-ECC. Metadata including patients' gender and age, dates of specimen collection, and specimen types were recorded.

Antimicrobial susceptibility testing
The minimum inhibitory concentrations (MICs) of 14 antibiotics were evaluated according to the guidelines of the Clinical and Laboratory Standards Institute (CLSI) (M100-S30, 2020). MICs of colistin and tigecycline were determined by the broth dilution method, and the results were interpreted according to European Committee on Antimicrobial Susceptibility Testing (EUCAST) (version 10.0) criteria (https://eucast. org/clinical_breakpoints/). The other 12 antibiotics were determined using the agar dilution method. E. coli ATCC 25922 and Pseudomonas aeruginosa ATCC 27853 were used as quality control standards. MCR-ECC isolates resistant to imipenem and/or meropenem (MIC ≥ 4 mg/L) were determined as carbapenemase-producing MCR-ECC (MCR-CREC).

Conjugation assay
In order to determine whether colistin and carbapenem resistance was transferable, conjugation assays were carried out in LB broth with using E. coli EC600 as the recipient according to a method previously described [16]. Briefly, logarithmic-phase cultures of donor and recipient cells were added to fresh LB broth and incubated overnight without shaking. Transconjugants were selected on LB agar plates containing 600 mg/L rifampicin plus 2 mg/L meropenem/imipenem and/or 2 mg/L colistin. The presence of carbapenemase and/or mcr genes in transconjugants was confirmed using PCR and Sanger sequencing, and the species of transconjugants was confirmed using MALDI-TOF Mass Spectrometry (Hexin Instrument Co., Ltd, Guangzhou, China).

Plasmid stability assay
Plasmid stability assay was performed as previously described with slight modifications [47]. The donor strains were grown in antibiotic-free LB broth. Cultures grew at 37°C in a shaking bath, followed by 1:1000 dilution in fresh antibiotic-free LB broth, and passaged for 10 successive days. Cultures of day 10 were serially diluted and plated on antibiotic-free LB agar plates, and incubated overnight. Fifty colonies were randomly picked from plates for each isolate. The presence of carbapenemase and mcr genes was validated by colony-PCR.

Plasmid incompatibility assay
Plasmid incompatibility assay was performed as previously described [47]. Briefly, the transconjugants bearing both carbapenemase-encoding and MCRencoding plasmids were cultured in antibiotic-free LB broth and were grown at 37°C in a shaking bath.
Overnight cultures were serially diluted and plated on antibiotic-free LB agar. Ninety-four colonies of each culture were selected for the detection of carbapenemase and mcr genes by PCR. Plasmids were considered incompatible when more than 80% of colonies lost either or both of carbapenemase-encoding and MCR-encoding plasmids.

Accession numbers
The genome sequences have been deposited into GenBank. The detail accession numbers are listed in Table S1.
Among the 48 MCR-ECC isolates, 26 were resistant to colistin (54.2%), 15 to tigecycline (31.2%), and 16 to carbapenems (imipenem and/or meropenem). At least four isolates were resistant to two of the three classes (colistin + tigecycline: 4; colistin + carbapenems: 5; tigecycline + carbapenems: 8), and two were resistant to all. The MICs of the 48 MCR-ECC isolates are summarized in Table 2 and Table S1. Of note, all MCR-CREC isolates were detected in 2013∼2018, which is much later than the emergence of MCR-ECC (in 2000). We, therefore, suppose that MCR-CREC might be derived from MCR-ECC by capturing the carbapenemase genes in our collection.
The mcr-9 gene detected on the eight long-read sequenced genomes (four ST133, three ST418, and one novel ST) located on IncHI2/2A plasmids with sizes ranging from 274,120 bp to 334,517 bp. Pairwise comparison of the eight circularized plasmids showed that the nucleotide similarity is accordant with the ST that plasmids carried by the same clone exhibited > 90% coverage and > 90% identity compared with each other, and pECC116-MCR-9 carried by the novel ST isolate shared < 70% coverage and < 90% coverage with the other plasmids (except for pECC45-MCR-9) (Figure 2A). Synteny analysis showed that the diverse regions among the eight plasmids were mainly caused by mobile genetic elements ( Figure 2B), suggesting that they might have originated from a common ancestor with a conserved backbone, and had further co-evolved with the chromosome.
The results indicate that mcr-9 dissemination among ECC was highly mediated by an IncHI2/ 2A epidemic plasmid.

Highly conserved NDM-1-and IMP-4-carrying plasmids carried by MCR-CREC in China
All of the 11 hybrid assembled genomes carried a bla NDM-1 gene, and an IncX3-type NDM-1-carrying circularized plasmid with the size of about 53 kb was identified in each genome, with limited sequence variations (> 98% coverage and > 90% nucleotide identity) (Figure 4), suggesting the occurrence of cross-species horizontal transfer for the plasmid. The IncX3 plasmid additionally carried a bleomycin resistance gene ble MBL and a beta-lactam resistance gene bla SHV-12 . Blasting the representative pECC27-NDM against NCBI nt database identified 34 genomes sharing > 90% coverage and identity, and all of them were carried by Enterobacterales, of which 32 were isolated from China (Table S6), indicating that the IncX3-type NDM-1-carrying plasmid was an epidemic plasmid in Enterobacterales in China.
An IncN-type bla IMP-4 -encoding circularized plasmid with the size of about 43 kb was detected in the four hybrid assembled genomes carrying a bla IMP-4 gene (ECC45, ECC46, ECC47, and ECC48) with highly limited sequence variations (100% coverage and > 99% nucleotide-acid identity). The four isolates belong to ST133, suggestive of a clonal dissemination. pECC45-IMP-4 was almost identical to pIMP-SZ1502 (GenBank accession no. KU051707) carried by an E. coli strain (99.9% identity and 100% coverage) ( Figure S1). A quinolone resistance gene qnrS1 colocated on the bla IMP-4 -encoding plasmid. A pECC45-IMP-4-like plasmid was also detected in the remaining four IMP-4-positive genomes by mapping the reads to pECC45-IMP-4 (100% coverage and > 99% nucleotide-acid identity). These data indicate that the emergence of MCR-CREC was derived from MCR-ECC by capturing an IncX3-type NDM-1-carrying and/or an IncN-type IMP-4-encoding plasmid in China.
The mcr-10 of ECC275 was detected on a 35,290-bp contig encoding an IncFIB-type replicon. Mapping the sequencing reads of ECC275 to the three circularized plasmids (pECC59-MCR-10, pECC60-MCR-10, and pECC72-MCR-10) resulted in a low coverage ranging between 24.44% and 53.08%. When mapping the 70 mcr-10-positive genomes retrieved from GenBank to the three circularized plasmids, only four showed ≥ 90% coverage to pECC72-MCR-10 (Table S7). These results indicate that the structure of mcr-10 plasmids might be highly diverse.
Conjugation assays were performed to evaluate the transferability of plasmids encoding MCR and carbapenemases. The mcr plasmids carried by most of MCR-ECC isolates were self-transmissible to E. coli EC600, including 32 mcr-9 and four mcr-10 isolates, and all plasmids encoding carbapenemase-encoding genes can be transferred successfully. The carbapenemase and mcr genes were able to be transferred simultaneously to EC600 from all MCR-CREC isolates.
The stability of plasmids encoding carbapenemases and MCR was evaluated in the 16 MCR-CREC by a 10-day passage. Fifty colonies of each isolate were tested (see details in method). All of the mcr-, bla NDM-1 -, and bla IMP-26 -harboring plasmids displayed high stability, as shown by the retention rates of over 80% at the end of the passages. While bla IMP-4 -harboring plasmids showed low retention rates (10%∼74%) in six strains (ECC45, ECC47, ECC48, ECC52, ECC54, and ECC188), and ECC45 lost its IMP-4 activity after 10 days with 10% retention rate, suggestive of the instability for bla IMP-4 -harboring plasmids in ECC (Table S8).
Plasmid incompatibility was further assessed in the transconjugants of the 16 MCR-CREC isolates. Ninety-four colonies of each isolate were tested (see details in method). The plasmids encoding bla NDM-1 or bla IMP-26 remained in all colonies overnight culture, while those with mcr or bla IMP-4 were completely lost (Table S9). These results suggest that the mcr and carbapenemase-producing plasmids were incompatible in E. coli EC600. The observed plasmid incompatibility is consistent with the current epidemiological data that in E. coli mcr-9 and mcr-10 was rarely detected and bla NDM-1 was prevalent.

Genetic context of MCR and carbapenemase genes
At least five types of mcr-9 genetic contexts were identified in the 44 genomes (Figure 5A), and a module IS903B (intact or fragmented)-mcr-9 was detected in all of them, implying the role of IS903B in the transmission of mcr-9. Type I genetic context was identical Figure 3. Detection of pECC116-MCR-9, pECC47-MCR-9, and pECC65-MCR-9 in 44 mcr-9-carrying MCR-ECC. The percentage length of virulence plasmid sequences are obtained by mapping short reads of the 44 isolates to the three mcr-9 plasmids (pECC116-MCR-9, pECC47-MCR-9, and pECC65-MCR-9) used as references. The existence of plasmid is defined by that isolates having short reads mapped to ≥90% of the reference plasmid length. The isolates are clustered according to coverages using the Pearson method. STs of isolates are indicated, and NA represents the novel ST.
to that of pW17-1 (CP031102), and compared with the other types it exclusively encodes a two-component regulator qseB/C probably involved in mcr-9 expression, an ATPase gene, and an orf with unknown function at downstream of wbuC (encoding a cupin fold metalloprotein). Type II might be derived from type I by the deletion of orf-ATPase-qseB/C, which is almost identical to that detected on p17277A_477 (CP043927). Compared with type II, IS26-wbuC was replaced by IS3000Δ-IS1R in type III as detected in pMRVIM0813 (KP97507). Different from the three types, pcoS-pcoE-rcnA-rcnR locating at the mcr-9 upstream flanking was absent in type IV and V, and a gene encoding a hypothetical protein was detected in type IV instead. Note that, we were unable to determine the intact mcr-9 genetic environments in 34 draft genomes due to the short mcr-9-bearing contigs available for the comparison.
The genetic context of mcr-10 detected on pECC59-MCR-10 and pECC60-MCR-10 was identical (ISkpn26-xerC-mcr-10-orfs-IS4321R), while on pECC72-MCR-10 genes locating at upstream and downstream of xerC-mcr-10 were replaced by those encoding hypothetical proteins without any insert sequences. An ISEc36 interrupted by the insertion of an ISKpx1 and a truncated ISKpx1 were located downstream of mcr-10 in ECC275. The genetic contexts of mcr-10 identified in this study are different from those described previously ( Figure 5B) [16]. A highly conserved genetic environment of bla NDM −1 was found in all isolates that an ISAba125 interrupted by the insertion of an IS5 lied upstream of bla NDM-1 , and the fragment ble MBL -trpF-dsbD-cutA-groS-groL-orf lied downstream of bla NDM-1 ( Figure  S3). The bla IMP-4 was harboured by a class I integron in the four genomes with an identical structure (an IS26-truncated intI, bla IMP-4 , a group II intron reverse transcriptase/maturase, a mobilization protein MobC, and an IS6100) ( Figure 5C).

Discussion
Carbapenems and colistin are the last resort antibiotics used for severe infections caused by multidrug-resistant bacteria. The worrisome fact is the emergence and spread of bacteria co-producing carbapenemases and mcr genes, largely limiting clinical treatment strategies and worsening outcomes. Understanding the occurrence and mechanism of such convergence of resistance to the last-resort antibiotics is imperative for making tailored strategies to control its further spread. In this study, we aimed to answer this question by systematically studying the co-existence of mcr and carbapenemase genes in ECC isolates collected between 2000 and 2018 in a tertiary hospital of China together with a global dataset retrieved from GenBank.
mcr-9 has been detected in multiple species of Enterobacterales mainly isolated from clinical samples worldwide, e.g. E. coli [49], S. enterica [13], ECC [17], and K. pneumoniae [50]. In contrast, mcr-10 was almost hosted in ECC and sporadically isolated from environment [16], animal [51] and clinical samples [52]. Previous studies have shown that mcr-9 and mcr-10 were frequently carried in ECC by using single centre or endemic data [14,16,22,23]. To our knowledge, this is the first report demonstrating that mcr-9/ mcr-10 were prevalent in ECC globally with the ratio of ca. 20%, and mcr-9 was much more prevalent than mcr-10 (18.5% vs 2.0%). The mcr-9 gene was firstly identified in a clinical S. enterica serotype Typhimurium strain isolated in 2010 [13], and mcr-10 was in an Enterobacter roggenkampii clinical strain recovered in 2016 [14]. In this study, the earliest mcr-9 and mcr-10 isolates were obtained in 2000 and 2018, respectively, implying that mcr-9 might have emerged much earlier than mcr-10. We further pinpoint that E. hormaechei was the major host of mcr-9. Accumulating epidemiological data suggest that E. hormaechei is the predominant species among ECC frequently obtaining multi-drug resistance and causing infections in the clinical setting [26,53,54]. Therefore, the high prevalence of mcr-9 in E. hormaechei would further challenge the clinical treatment.
Different from the other mcr variants, the activities of mcr-9 and mcr-10 to colistin are much weaker in numerous isolates [13,29,55]. The first reported isolate carrying mcr-9 or mcr-10 was susceptible to colistin with an MIC of 0.25-0.5 mg/L, despite both genes did confer colistin MICs of > 2.5 mg/L when overexpressed in the laboratory strains [13,14]. A recent study found that mcr-9 expression was inducible in the presence of colistin when located upstream of the two-component system qseB/qseC, and the mcr-9-harboring isolates were susceptible to colistin when lacked the qseB/qseC regulatory operon [56]. In this study, qseB/qseC was detected at downstream of mcr-9 in seven isolates resistant to colistin, further supporting the role of qseB/qseC in the induction of mcr-9-mdiated colistin resistance. While others lacking this two-component system at downstream of mcr-9 also showed colistin resistance with even higher MIC values (≥ 32 mg/L), and no functional mutations (i.e. PhoP D56 and PhoQ H277 ) were detected in their chromosomes, suggestive of unknown mechanism involved.
Our analysis showed that the dissemination of mcr-9 worldwide was mainly mediated by an IncHI2/2A epidemic plasmid. This is consistent with a previous study that IncHI2 was identified to be the dominant replicon type (90.1%; 64/71) of the mcr-9-carrying plasmids disseminating worldwide [57]. Notably, IncHI2/2A type plasmids are highly associated with multidrug resistance genes, which raising a concern that the mcr-9-carrying epidemic plasmid could become multidrug resistant in the future. In contrast, a highly diversity was found for the mcr-10 plasmids that three different replicons (IncFIA, IncFIA-II, and IncFIB) were detected in the four mcr-10-carrying plasmids, and only one genome of the global dataset showed ≥ 90% coverage with one of the four plasmids. This is accordant with the finding of a recent study [33], and may be one of the causes for the less prevalence of mcr-10 observed here.
Of greater concern, a high carbapenem resistance rate was detected in our MCR-ECC collection and also the global dataset, accounting for 33.3% (16/48) and 53.89% (388/720), respectively. These data demonstrate that the convergence of carbapenemase and mcr-9/mcr10 genes was highly frequent in ECC. The discrepancy of plasmid stability observed between bla IMP-4 -harboring and bla IMP-26 -harboring plasmids might be caused by their different hosts that the host of bla IMP-4 -harboring plasmids belongs to ST133, while that of bla IMP-26 -harboring plasmid belongs to ST171. Our surveillance data showed that MCR-CREC emerged later than MCR-ECC in China, and the two populations shared prevalent clones (ST133 and ST418), we thus suspect that MCR-CREC might have been derived from MCR-ECC recently through obtaining plasmids encoding carbapenemases in clinical settings. This can be further supported by that the two prevalent carbapenemases detected in our collection, i.e. NDM-1 and IMP-4, were carried by an IncX3-type and an IncN-type epidemic plasmid with very limited sequence variations, respectively. Additionally, our results underpin that such convergence could be dependent on the host preference of plasmids, since mcr-9/mcr-10-encoding plasmids can be remained in ECC stably but not in E. coli though the plasmids encoding carbapenemase were stable in both.
Of clinical and epidemiological concern, various combinations of resistance to the last resort antibiotics were detected in MCR-ECC, and two of them were even resistant to colistin, carbapenems, and tigecycline. Such convergence would largely worsen clinical outcomes in the future, although the tigecycline resistance determinants could not be defined in this study.
Additionally, a majority of MCR-CREC belonged to epidemic clones, e.g. ST171, ST93, ST133, and ST418 [48,58,59], and the plasmids encoding MCR or carbapenemases were able to remain stable in ECC and selftransmissible as shown here. Despite that it remains unclear whether the resistance convergence in these epidemic clones has established evolutionary advantages, continuous surveillance is imperative to prevent them from being high-risk clones.
In summary, our study revealed a high prevalence of mcr-9 and carbapenemase genes co-existing in ECC, and the convergence was driven by epidemic plasmids. The data suggest that MCR-CRE could be derived from MCR-ECC, and multiple epidemic clones have mediated the dissemination of MCR-CRE worldwide, highlighting that effective measures should be taken to control its further spread.