Characterization of Three Novel IMP Metallo-β-Lactamases, IMP-89, IMP-91, and IMP-96, and Diverse blaIMP-Carrying Accessory Genetic Elements from Chinese Clinical Isolates

ABSTRACT Three novel imipenemase (IMP)-type metallo-β-lactamases (MBLs), referred to as IMP-89, IMP-91, and IMP-96, were detected in three clinical isolates from China. Antimicrobial susceptibility tests indicated these novel enzymes were resistant to most β-lactams, and IMP-96 with a Ser262Gly mutation had higher activity against meropenem than its point mutant. We then collected sequence data on all 91 available IMP variants for phylogenetic analysis. To further analyze the genetic environment of blaIMP, an extensive comparison was applied to nine accessory genetic elements (AGEs), including six sequenced blaIMP-carrying AGEs in this study and three others from GenBank. These nine AGEs were divided into three groups: three IncpJBCL41 plasmids, Tn6417 and its two derivatives, and three Tn6879-related integrative and conjugative elements (ICEs). All blaIMP genes in this study were captured by class 1 integrons. In the integrons, blaIMP genes usually coexisted with other resistance genes, which further impeded clinical antibacterial treatment. The emergence of new IMP variants and the diversity and complexity of their genetic environment make the prevention and control of drug-resistant strains critical, requiring serious attention from clinical and public health management departments. IMPORTANCE The spread of IMP-type MBLs has increased dramatically in recent years. We discovered three novel IMP variants from three clinical isolates in China. We summarized the classification and evolutionary relationship of all available IMP variants. Moreover, we detailed the genetic characteristics of blaIMP-carrying accessory genetic elements in five clinical isolates. Given the risk of rapid and extensive spread of blaIMP genes, we suggest that continuous surveillance is crucial to combat the acquisition and transmission of blaIMP genes by bacteria, which can impede clinical therapy effectiveness.

In this study, five IMP-producing clinical isolates from China were fully sequenced. Three novel IMP variants, namely IMP-89, IMP-91, and IMP-96, were detected. The three enzymes were viable and conferred resistance to nearly all antibiotics tested. IMP-96 showed an increased MIC value specifically toward meropenem in Escherichia coli because of its critical Ser262Gly substitution. A genomic phylogenetic analysis of 91 available IMP variants suggested the three novel IMP variants belong to three different phylogenetic groups. We further performed a genetic analysis of nine AGEs, including six bla IMP -carrying AGEs sequenced from the five clinical isolates in this study and three prototype AGEs from GenBank. Thus, this study identifies three novel IMP variants and provides a deeper understanding of the genetic diversification of IMP-encoding AGEs, highlighting the emergence and spread of the resistance genes bla IMP .

RESULTS
Cloning and heterologous expression of bla IMP-89 , bla IMP-91 , and bla IMP-96 . The complete genome of each of the five IMP-producing clinical isolates from China was sequenced (Table S1 in the supplemental material). Three novel IMP variants, namely IMP-89, IMP-91, and IMP-96 (Fig. S1), were detected from three different clinical isolates: Pseudomonas putida NY5709, P. aeruginosa NY3045, and Stenotrophomonas spp. NY11291, respectively. The open reading frame of the three novel variants IMP-89/91/96, their point mutants IMP-26/14/8, and the representative IMP variant IMP-1 (used as a positive control) were cloned into a kanamycin-resistant pUC57 vector and then transformed into E. coli TOP10 to obtain the seven corresponding E. coli electroporants, pUC57K-bla IMP-89/26/91/14/96/8/1 -TOP10. Strains encoding IMP-89, IMP-91, and IMP-96 enzymes were active and had resistance to cephem (cefazolin, ceftazidime, and cefoxitin) and carbapenem (meropenem) relative to the negative-control strains TOP10, TOP10/pUC57K, and ATCC 25922 but remained susceptible to aztreonam. IMP-89 and IMP-96 were also resistant to carbapenem (imipenem) but remained susceptible to penam (ampicillin). In contrast, IMP-91 could hydrolyze ampicillin but was susceptible to imipenem. The hydrolytic activity of IMP-89/91 was relatively poor against the antibiotics tested compared with IMP-26/IMP-14. However, the E. coli TOP10 expressing IMP-96 had increased resistance toward meropenem with a higher MIC value than the strain expressing bla IMP-8 (Table 1). Molecular representation of the structure of  Imipenem  Ampicillin  Aztreonam  TOP10/pUC57K-IMP-1  256  16  128  32  4  16  #1  TOP10/pUC57K-IMP-89  512  128  128  32  8  16  #1  TOP10/pUC57K-IMP- (Table  S2), including the three novel IMP variants identified in this study. All IMP variants could be assigned to seven groups (G1 to G7), with significant phylogenetic distance between each group. Classification results were also verified by pairwise comparison of the amino acid sequences of IMPs ( Fig. S2 and Table S3), which showed that the genetic identity of IMPs in each group was generally $90%. IMP-89, IMP-91, and IMP-96 belonged to groups G1, G5, and G4, respectively. We further considered the countries where these IMP variants first appeared (Fig. S3). The endemic spread of IMP-type enzymes had been reported in Japan and China (18). Overall, these data suggested complex genetic mutations and spread of IMP-type MBLs.
A collection of nine AGEs for detailed sequence comparison. We identified two bla IMP -carrying plasmids and four bla IMP -carrying chromosomal AGEs from the five isolates we collected. A detailed sequence comparison analysis applied to these six AGEs together with three prototype AGEs from GenBank (Table 2) revealed that these AGEs are composed of three distinct groups: (i) three Inc pJBCL41 plasmids pJBCL41, pNY5709-IMP, and pNY11382-IMP; (ii) three related ICEs Tn6879, Tn6880, and Tn6881; and (iii) Two Tn6417 derivatives Tn7450 and T6417RE NY3045 -2. Elements T6417RE, T1403RE, and T6483RE (see below) could not be recognized as intact transposons due to the truncation of relevant core transposition modules.
Genetic environment of bla IMP-89 in NY5709. bla IMP-89 was identified in pNY5709-IMP from the P. putida isolate NY5709. A bla IMP-34 -carrying plasmid pNY11382-IMP was found from the Pseudomonas isolate NY11382. These two plasmids (Fig. S4) were identified as belonging to the Inc pJBCL41 group because they had identical repA genes sharing $95% nucleotide identity to repA IncpJBCL41 . Found in P. shirazica FFUP_PS_41,

Novel IMP Variants and Accessory Genetic Elements
Microbiology Spectrum pJBCL41was reported to be a 498,516-bp megaplasmid with limited similarity to publicly identified plasmids (19), therefore, was proposed to be a novel incompatibility group, Inc pJBCL41 . pJBCL41 represented the Inc pJBCL41 reference plasmid, because it had the most complete Inc pJBCL41 backbone with the least exogenous insertions. All Inc pJBCL41 plasmids from public databases were megaplasmids, with lengths .490 kb.
The complete nucleotide sequence of pNY5709-IMP was 636,818 bp in length. To the best of our knowledge, pNY5709-IMP is the largest multidrug resistance plasmid reported to date in Pseudomonas.
Genetic environment of bla IMP-91 in NY3045 and bla IMP-96 in NY11291. We did a detailed sequence comparison of Tn6417 and three of its derivatives: bla IMP-96 -carrying Tn7450 detected in the Stenotrophomonas strain NY11291 and bla IMP-91 -carrying T6417RE NY3045 -2 and T6417RE NY3045 -1 (found upstream of T6417RE NY3045 -2) (Fig. 5) from the P. aeruginosa strain NY3045. The prototype ICE Tn6417 was initially described in P. aeruginosa DHS01 (22) and contained a core backbone structure of attL/R, int, cpl, rlx (relaxase), and an F-type T4SS gene set. The three Tn6417 derivatives were inserted into three different chromosomal locations and exhibited modular variations across their backbones: (i) xerC-to-orf1068 region from Tn6417 was absent in all three derivatives; (ii) Tn7450 had its unique orf1965-to-orf630 and orf114-to-orf642 regions; and (iii) only T6417RE NY3045 -1 contained orf1371. Each of these derivatives carried different main accessory modules: a bla IMP-96 -carrying concise class 1 integron In2092 (see below) in Tn7450, a 7.0-kb Tn6532 No therapeutic options exist to treat Stenotrophomonas infections with b-lactams. Based on in vitro activity, the "drug of choice" for Stenotrophomonas has long been trimethoprim-sulfamethoxazole (SXT) (23); however, our MIC experiments (Table S1) showed that NY11291 had high-level resistance to SXT (MIC $ 320). Tn7450 from NY11291 harbored the sul1 gene at the 39 end of the concise class 1 integron In212

Novel IMP Variants and Accessory Genetic Elements
Microbiology Spectrum (see below). The sul1 genes have been reported to associate with the class 1 integrons and contribute to the resistance to SXT (24). Three related ICEs Tn6879, Tn6880, and Tn6881. Whole-genome sequencing of strain NY5709 confirmed the presence of another bla IMP gene in a chromosome-borne AGE Tn6881. We also detected bla IMP-1 -carrying Tn6880 in the Pseudomonas strain NY5710. Then, a detailed sequence comparison was applied to Tn6880, Tn6881, and the prototype ICE Tn6879. Tn6879 was initially found in Pseudomonas spp. LTGT-11-2Z (25). All three ICEs were integrated 92-bp upstream of the chromosomal gene queC (queuosine biosynthesis protein) and shared the core backbone markers attL/R, int, rlx, and cpl (Fig. 6). However, the backbone of Tn6879 had a unique fdhA-to-pobA region; correspondingly, Tn6880 and Tn6881 carried their unique orf573 and orf351, respectively. Each of the three ICEs carried a single accessory module integrated at the same site. Tn6879 had no resistance genes, while Tn6880 and Tn6881 harbored bla IMP-1 -carrying concise class 1 integrons In1771 and In1768, respectively (see below).
Newly identified or designated AGEs. There were eight newly identified chromosomal AGEs: (i) three were directly integrated into the chromosomes and included the three ICEs Tn6880, Tn6881, and Tn7450; and (ii) the remaining five were the inner components of the above elements directly inserted, including the four integrons In1771, In1768, In1792, and In2092, and the one IS element ISStes1. There were seven newly identified AGEs from plasmids, including the one unit transposon Tn7454, the four integrons In1769, In1782, In1783, In2089, and the two IS elements ISPpu36 and ISPsp18. Additionally, three AGEs were newly designated (first designated in this study, but with previously determined sequences): one ICE Tn6879, one unit transposon Tn7453, and one IS element ISPsp17.

DISCUSSION
IMP-type enzymes are the first acquired MBLs in Gram-negative bacilli (26). The number of reports on the characterization of novel IMP in various bacterial species has increased worldwide (27). Here, we report on three novel variants of IMP named IMP-89, IMP-91, and IMP-96. These novel IMP variants can hydrolyze nearly all tested antibiotics. IMP-89 and IMP-91 are less resistant than their point mutations IMP-26 and IMP-14, respectively. Notably, the amino acid sequence of IMP-96 differs from its point mutation IMP-8 by the amino acid substitution S262G, resulting in increased resistance against meropenem. The substitution S262G may play an important role in increasing resistance against certain carbapenems in E. coli, as the substitution disrupts a hydrogen-bonding interaction with the base of the L3 loop, altering its conformation (28). This substitution may have arisen due to the selective pressure caused by the use of carbapenem.
To date, 91 IMP variants have been reported worldwide. We first constructed a phylogenetic tree containing the three new variants. Based on the phylogenetic tree and the pairwise comparison of amino acid sequences, IMPs fall into seven groups, and IMP-89, IMP-91, and IMP-96 phylogenetically belong to the groups of G1, G5, and G6, respectively. IMPs are clearly not monophyletic with 1 to 22% sequence dissimilarity. Novel IMP variants continue to be identified in Japan (n = 23) and China (n = 13). One report indicates increased use of certain carbapenems in Japan in recent years (29), and China has a high rate of antibiotic use for hospital inpatients (30). The rapid evolution and observed spread of IMP highlight a growing threat that needs to be taken seriously.
Few existing reports show diverse bla IMP -carrying AGEs or provide a systematic summary for them. This study presents the full sequences of six AGEs carrying bla IMP genes, including two plasmids and four chromosomal AGEs that can be divided into three distinct groups. The Inc pJBCL41 is a novel incompatibility group. The four chromosomal AGEs belong to two different categories: two Tn6879-related ICEs Tn6880 and Tn6881; two Tn6417-related derivatives Tn7450 and T6417RE NY3045 -2. Our results indicate a wide dissemination of bla IMP by AGEs, which may become common vehicles mediating the spread of antimicrobial resistance.
Many resistance genes in addition to bla IMP coexist in these AGEs. The results of the antimicrobial susceptibility test (Table S1) show that all five strains remain highly drug- Novel IMP Variants and Accessory Genetic Elements Microbiology Spectrum resistant. bla IMP genes usually coexist in the integrons with other resistance genes, such as the aminoglycoside resistance genes aacA4 and aad1, class D b-lactamase gene bla OXA , and phenicol resistance gene catB3, resulting in multidrug resistance that increases the difficulty of finding a clinical antibacterial treatment. In the conserved region of class 1 integrons, the sul1 gene also contributes to the problem of drug resistance. Stenotrophomonas is typically treated with SXT (31). The harboring of the sul1 gene in the bla IMP-96 -carrying integron In2092 is the key to the increase of SXT resistance in the Stenotrophomonas strain NY11291. Clinical class 1 integrons play a crucial role in spreading antibiotic resistance among bacterial pathogens, and further investigations are needed. The combination of two MBLs encoded by the bla IMP gene and another gene in one strain is becoming more common. Notably, we report the largest multidrug resistance plasmid ever sequenced in the Pseudomonas genus, pNY5709-IMP, which encodes two MBLs, IMP-89, and VIM-2. pNY5709-IMP belongs to Inc pJBCL41 plasmids and carries substantial numbers of diverse accessory modules.
In conclusion, five IMP-producing clinical strains were collected from Chinese hospitals, and these strains showed high resistance to various broad-spectrum b-lactams. Three novel IMP variants, IMP-89, IMP-91, and IMP-96 were detected. The presence of S262G in IMP-96 leads to increased resistance against the carbapenem meropenem. Rational and effective clinical use of antibiotics requires investigating the diversity of hydrolysis capacities in IMPs. A phylogenetic reconstruction using all available bla IMP protein sequences shows the evolutionary relationship of the three variants with other IMPs. Then, a detailed sequence comparison of three groups of nine AGEs (including six bla IMP -carrying AGEs sequenced in this study) was performed. IMP genes are often situated within class 1 integrons harbored on broad-host-range plasmids or chromosomal AGEs. This work provides a deeper insight into the bioinformatics and epidemiology of IMP and bla IMP -carrying accessory genetic elements. The emergence of new IMP variants, and the diversity and complexity of their genetic environment make the prevention and control of drug-resistant strains more challenging, and clinical and public health management departments will need to pay attention to drug-resistant genes.

MATERIALS AND METHODS
Bacterial strains, identification, and detection of carbapenemase-encoding genes. Five bla IMPcarrying clinical isolates (Table S1) were screened from nonduplicate carbapenem-resistant strains collected from different public hospitals in China. The presence of known MBL genes was investigated using PCR assays. Bacterial species identification was done using genome sequence-based average nucleotide identity analysis (http://www.ezbiocloud.net/tools/ani) (32). Bacterial antimicrobial susceptibility was tested by BioMérieux Vitek 2 and interpreted as per the Clinical and Laboratory Standards Institute (CLSI) guidelines (33).
Sequencing and sequence assembly and annotation. Bacterial genomic DNA was isolated using the UltraClean Microbial kit (Qiagen, Hilden, Germany) and sequenced from a paired-end library with an average insert size of 350 bp (range from 150 to 600 bp) on a HiSeq sequencer (Illumina, CA), as well as a shared DNA library with an average size of 15 kb (range from 10 to 20 kb) on a PacBio RSII sequencer (Pacific Biosciences, CA). The paired-end short Illumina reads were used to correct the long PacBio reads with the software proovread (34), and then corrected PacBio reads were assembled de novo using SMARTdenovo (https://github.com/ruanjue/smartdenovo). The sequencing data were checked using NanoPack29 (35) and FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc). Further sequence data mining was performed as described previously (36)(37)(38).
Phylogenetic analysis. Amino acid sequences were aligned using Clustal Omega 1.2.2 (39), and a neighbor-joining phylogenetic tree was constructed from aligned sequences using MEGA X 10.2 (40) with a bootstrap iteration of 1,000. The phylogenetic tree was subsequently visualized and modified using iTOL version 6 (https://itol.embl.de/).
Cloning experiments and MIC measurements. The bla IMP-89/91/96 coding regions together with their 450-bp upstream (promoter) regions and 300-bp downstream (terminator) regions from strains NY5709/ NY3045/NY11291, respectively, were cloned into the cloning vector pUC57-Kan (pUC57K). Similarly, the coding region of each of the other bla IMP variants together with the above promoter-proximal regions and terminator-proximal regions were synthesized and cloned into pUC57K. Each resulting recombinant plasmid was transformed through electroporation into E. coli TOP10, generating the relevant electroporant. Electroporant selection was done with 4 mg/mL meropenem (for bla IMP ). Bacterial antimicrobial susceptibility was tested using the classic broth microdilution method, and interpreted as per the 2020 CLSI guidelines (33).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.