On 10 October 2021, a patient (52-year-old male) was admitted to the neurosurgery department, due to sudden unconsciousness and vomiting for three and a half hours. The diagnosis was spontaneous subarachnoid hemorrhage, hemorrhage in the third, fourth and lateral ventricles. After admission, right ventricle drainage was performed. Fluid replacement, brain protection, prevention of bleeding infection, and neurological rehabilitation were given. On the eighth day of admission, the patient was considered intracranial infection with the body temperature 37.9°C and persistent unconsciousness, and a carbapenem-resistant A. caviae strain 211703 was isolated from the patient’s cerebrospinal fluid drainage sample initially identified by Vitek 2. Later, bacterial species identification using average nucleotide identity (ANI) analysis (http://www.ezbiocloud.net/tools/ani) based on genome sequences finally proved that 211703 belongs to A. caviae (Richter and Rossello-Mora, 2009).

The drug minimum inhibitory concentrations (MICs) of 211703 were determined by BioMérieux VITEK2 ( Table 1 ). The antibiotic susceptibility test results were determined by the Clinical and Laboratory Standards Institute (CLSI) guidelines (2021). The production of NDM carbapenemase in 211703 was detected on an NG-Test CARBA 5 (NG Biotech, Guipry, France), a rapid diagnostic test based on the immunocolloidal gold technique ( Figure S1 ).

Conjugal transfer experiments were carried out with rifampicin-resistant Escherichia coli EC600 being used as a recipient, and the strain 211703 as a donor. The donor and recipient strains were grown in three milliliters (mL) brain heart infusion (BHI) broth overnight at 37°C. And then, 50 μL of donor strain culture was mixed with 500 μL of recipient strain culture (v:v = 1:10) and 4.5 mL of fresh BHI broth. In addition, 100 μL of the mixture was applied onto a cellulose filter membrane (pore size, 0.22 μm) already placed on a BHI agar plate. After incubation at 37°C for 16h-18h, the filter membrane was taken out and vortexed in 1 mL of BHI broth. The vortex mixtures were plated on BHI agar plates containing 2 mg/L imipenem and 1,500 mg/L rifampicin for the selection of the transconjugants. However, repeated conjugation experiments failed to acquire transconjugants.

Bacterial genomic DNA of A. caviae 211703 was isolated using the Gentra Puregene Yeast/Bact. Kit (Qiagen, Valencia, CA). The TruePrepTM DNA Library Prep Kit V2 and the SQU-LSK109 Ligation Sequencing kit were used for libraries preparation separately. After the preparation of the library was completed, it was separately sequenced on an Illumina HiSeq X Ten platform (Illumina Inc., San Diego, CA, USA) and GridION X5 platform (Oxford Nanopore, UK). Raw data from the HiSeq X Ten platform and the GridION X5 platform were trimmed to obtain the high-quality clean reads (clean data) by Canu v1.8 (https://canu.readthedocs.io/en/latest/index.html). The paired-end short Illumina reads and the long Nanopore reads were assembled de novo utilizing Unicycler v0.4.8.0 (https://github.com/rrwick/Unicycler).

A total of 62 sequences of A. caviae sequenced at three levels (scaffold, chromosome, and complete) were downloaded from NCBI (last accessed on 15 Aug 2022), which were isolated from various sources from 2004 to 2021 ( Table S1 ). Assembly method of 211703 genome that were used in the phylogenetic analysis was Unicycler v0.4.8.0. MUMmer v3.1 was used for alignments against the reference genome to create a core genome alignment (Delcher et al., 2003). The core-genome length for phylogenetic analysis was 4.5-Mb. A total of 394,778 single nucleotide polymorphisms (SNPs) in the backbone regions were identified and extracted, and a maximum-likelihood phylogenetic tree was constructed based on the SNPs’ dataset ( Figure 1 ). Matrix of pairwise SNP distances among studied genomes was shown in Table S2 . The phylogenetic tree and related background information (collection date, location, isolation source, and host) were shown using the Interactive Tree of Life (iTOL) programs (Letunic and Bork, 2021).

ORFs and pseudogenes of 211703 genome were predicted using RAST 2.0 (Brettin et al., 2015). Further detailed dissection and manual annotation were performed using BLASTP/BLASTN (Boratyn et al., 2013). Annotation of resistance genes, MGEs, and other features was carried out using online databases such as CARD (Alcock et al., 2020), ResFinder (Zankari et al., 2012), ISfinder (Siguier et al., 2006), and INTEGRALL (Moura et al., 2009), and the Tn Number Registry (Roberts et al., 2008). Alignments with homologous sequences of ICE harboring on 211703 were performed by using the BRIG tool (Alikhan et al., 2011). Multiple and pairwise sequence comparisons were performed using MUSCLE 3.8.31 and BLASTN (Edgar, 2004). Comparison diagrams of ICEs/IMEs regions were drawn using Inkscape 1.1 (https://inkscape.org/en).

The complete sequence of A. caviae 211703 (chromosome c211703) was submitted to the GenBank database, under accession number CP092181. The Illumina reads of A. caviae 211703 was deposited as a SRA in the GenBank database (accession number PRJNA934671).

Antimicrobial susceptibility tests were performed on A. caviae 211703. The results showed that 211703 exhibited resistance to a broad range of tested antimicrobials except aztreonam, amikacin, tigecycline, and minocycline ( Table 1 ). The production of NDM carbapenemase was confirmed by immunocolloidal gold technique, which was consistent with the result of antibiotic susceptibility tests ( Figure S1 ). Corresponding to the results of multidrug resistance profile, 15 different types of resistant genes were detected from A. caviae 211703 ( Table S3 ), which confer resistance to aminoglycosides (arr2 and aphA6), β-lactams (bla _MOX-6, bla _NDM-1, and bla _RSA-1), macrolides (mph(A), mph(E), and msr(E)), quinolones (qnrVC1), sulphonamides (sul1), and other antibiotics.

Based on the genome of 211703 and all A. caviae strains available from GenBank of NCBI whose assembly levels are scaffold, chromosome, or complete, phylogenetic analysis was performed to explore the evolutionary relationships and potential associations (Sayers et al., 2021). A total of 62 A. caviae strains from GenBank were included, and the related information of these strains were collected and verified manually ( Table S1 ). From the background information, the main collection sites of these strains were China (18/62), Japan (10/62), Brazil (9/62), Thailand (7/62), and the USA (6/62). Among these 62 strains, the chromosome sequence of A. caviae WP8-S18-ESBL-04 (accession number AP022254, GCF_014169735.1, the standard strain of A. caviae) was used as the reference. Strain OnP3.1 (accession number CP050851, GCF_017310215.1, the standard strain of A. hydrophila) was used as the outgroup. After multiple sequence alignment, a total of 394,778 SNPs located in the core genome regions were identified and extracted. Using these SNPs’ dataset, a maximum likelihood (ML) phylogenetic tree was constructed. Figure 1 showed the phylogenetic tree and marked basic background information of each node/strain such as the collection date, location, isolation source, and host. The most closely related strain to 211703 was GCF_019711275.1, which was isolated from a rectal swab in Thailand in 2019. They both belonged to the same branch on the evolutionary tree.

The whole genome sequence of A. caviae 211703 was obtained by high-throughput sequencing. Genome sequencing result revealed that A. caviae 211703 contained only chromosome (c211703, accession number CP092181) with a length of 4,783,384 bp, and no plasmid was found in its genome after verification ( Table 2 ). The mean G+C content of c211703 was 61.8% and 4,400 ORFs were predicted on c211703. Based on the six key housekeeping genes (gltA, groL, gyrB, metG, ppsA, and recA) identified on c211703, MLST typing method proved that 211703 belongs to ST928.

Among the 16 resistance locus existing on c211703, 15 of them were concentrated on one ICE (6/16, including bla _NDM-1) and one IME (9/16, including bla _RSA-1) ( Table S3 ; Figure S2 ). Further, detailed genome annotation identified the location and genetic context of ICE and IME (designated as Tn7413a). In order to explore the similarities and differences in the acquisition of resistance genes and the genetic environment, precise comparisons were performed with related ICEs and IMEs, respectively.

The IME Tn7413a carrying bla _RSA-1 was 35,097 bp in length, located on A. caviae 211703 from 2,715,927 bp to 2,751,023 bp ( Figure S2 ). Accurate annotation and genomic dissection revealed that Tn7413a was bordered by a pair of 18-bp attL/attR (attachment site at the left/right end) ( Figure 2 ). According to genes function, Tn7413a can be divided into backbone regions and accessory regions. Backbone regions with a length of 16.2 kb were consisted by genes such as int, repC, hipA, and hipB. Integron In2148 inserted into the backbone regions, as the accessory region of Tn7413a.

In2148 was a novel concise class 1 integron containing primary components 5’-conserved region (5’-CS), 3’-conserved region (3’-CS), and a gene cassette array (GCA) qnrVC1–arr2–bla _RSA-1–dfrA1–gcuC. The bla _RSA-1 gene was found to be located in the GCA of In2148, most likely due to the integration of an integron, which facilitated its spread. In the downstream of 3’-CS, there was a reverse putative resistance unit IS26–mph(A)–IS6100 unit carrying mph(A), encoding proteins conferring resistance to macrolides (Partridge et al., 2018). In the upstream of IS26–mph(A)–IS6100 unit, IS26 was shared by IS26–mph(E)–IS26 unit followed by a truncated tniA encoding DD(35)E transposase TniA. A pair of 25-bp IRi (inverted repeat at the integrase end) and IRt (inverted repeat at the tni end) flanked In2148. On the outside of IRi and IRt, there were 5-bp DRs (direct repeats, target site duplication signals for transposition), respectively.

To investigate the structural variations among these IMEs, a BLAST analysis was performed using the backbone region sequences of Tn7413a as a reference. BLAST analysis revealed that a total of seven IMEs in the GenBank database were similar to Tn7413a, sharing 100% coverage and >99% identity for gene int from Tn7413a. These seven IMEs were included in this analysis, and the gene structures of these IMEs were detailed annotated manually. Due to their structural similarity to Tn7413a, these seven IMEs were designated as Tn7413b-h as listed in Table S4 .

Tn7413b-h ranged in length from 29,639 bp to 59,408 bp, and were all distributed in Aeromonas. Similar to Tn7413a, the boundaries of these seven IMEs were a pair of 18-bp attL/attR. The backbone regions of Tn7413b-h were almost identical to that of Tn7413a, but differed in the accessory regions ( Figure 2 ). In all the seven IMEs except Tn7413d, there was only one accessory region with integron as the primary content. Two insertion sequences insertions in Tn7413d resulted in its backbone gene hsdS being interrupted into three parts. In Tn7413b-d and Tn7413f-h, In2148 in Tn7413a was replaced by In27. In27 was a concise class 1 integron, carrying GCA dfrA12–gcuF–aadA2 (Partridge et al., 2009; Partridge, 2011). In Tn7413e, In2148 in Tn7413a was replaced by In384, a concise class 1 integron carrying GCA dfrA12–gcuF. In Tn7413c-e, the insertion sequence ISAve3 inserted upstream of the integron. For all these seven IMEs Tn7413b-h, there was the putative resistance unit chrA–orf98 unit downstream of the 3’-CS of the integron they carried. Further, the chrA–orf98 unit was all followed by an IS26–mph(A)–IS6100 unit, which was same as in Tn7413a. For Tn7413c-e and Tn7413h, all contained single or multiple Tn4352, a composite transposon involved in kanamycin resistance (Wrighton and Strike, 1987). For Tn7413b, the most identical IME related to Tn7413a, its integron In27 contained the IS26–mph(E)–IS26 unit, which was identical to Tn7413a. Tn7413g contained a putative resistance unit ISCR2–floR unit. ISCR elements were bounded by an origin (oriIS) downstream and a terminus (terIS) upstream, responsible for capturing resistance genes (Partridge, 2011; Partridge et al., 2018). In Tn7413h, there was a novel composite transposon newly designated Tn7419, flanked by 8-bp DRs both sides. Both ends of Tn7419 were insertion sequence ISAhy4, and the middle part were cmeA, cmeB, and cmeC, encoding membrane fusion proteins related to RND efflux system. Meanwhile, downstream of Tn7419, there were IS26–catA2–IS26 unit and IS26–tetA(D)–IS26 unit sharing IS26, carrying resistance genes catA2 and tetA(D), respectively.

Furthermore, in order to explore the genetic environment differences of bla _RSA-1, the genetic environment of the previously reported bla _RSA-1, which belonged to Aeromonas spp. strain 1402 was also included in this study (Marathe et al., 2018). The results showed that the gene fragment carrying bla _RSA-1 in former report was an IME with a length of 50,823 bp, which was also bounded by a pair of 18-bp length attL/attR ( Figure 2 ). Backbone regions of Tn7412 also contained genes such as int, hipA, and hipB, but other genes in the backbone region were quite different from Tn7413a and Tn7413b-h. The accessory region of Tn7412 was mainly composed of concise class 1 integron In1357, containing GCA qacG–bla _RSA-1–dfrA1–gcuC. Coincidentally, bla _RSA-1 was also located in the gene cassette of the integron, consistent with Tn7413a. After the 3’-CS of In1357, there were chrA–orf98 unit and IS26–mph(A)–IS6100 unit.

ICE of A. caviae 211703 was located on chromosome from 3,757,651 bp to 3,865,845 bp with a length of 108,195 bp, bounded by a pair of 19-bp attL/attR ( Figure S2 ). Genomic annotation revealed that ICE of A. caviae 211703 can also be divided into backbone regions and accessory regions according to their functions ( Figure 3 ). The backbone regions consisted of two parts, including backbone of maintenance harboring gene int and backbone of conjugal transfer. Accurate annotation demonstrated that there were two accessory regions on A. caviae 211703, ars locus and bla _NDM-1 region. ars locus were a series of resistance genes encoding proteins related to arsenic resistance. bla _NDM-1 region from ICE of 211703 was formed by multiple recombination mediated by ISCR14-like ( Figure 4 ). Unlike common structures of genetic contexts harboring bla _NDM-1 previously reported (ΔTn125 harboring bla _NDM-1 captured by ISCR1) (Wu et al., 2019; Luo et al., 2021), the bla _NDM-1 genetic context was formed by the insertion sequence ISCR14-like mediated capture of ΔTn125 harboring bla _NDM-1 and ble _MBL. Downstream of ΔISAba125 were another ISCR14-like with floR, tetR(G), and tetA(G), which together constituted the end of the bla _NDM-1 region from ICE of 211703. There were four ISCR14-like elements upstream of the ISCR14-like captured ΔTn125, which indicated that multiple insertions or recombination occurred.

BLAST analysis revealed that a total of 21 ICEs in the GenBank database were almost identical to ICE of A. caviae 211703 (100% coverage and >95% identity for gene int from ICE of A. caviae 211703) ( Table S5 ). The 21 ICEs ranged in length from 88,228 bp to 112,147 bp. Most of these ICEs (14/21) were distributed in Pseudomonas, where ICE of Pseudomonas juntendi PP-2463 was highly identical to ICE of A. caviae 211703. P. juntendi PP-2463 also isolated from the same city (Taizhou, China) with A. caviae 211703, which indicated that ICE of A. caviae 211703 may be transferred from Pseudomonas to Aeromonas. Only one of the 21 ICEs originated from Aeromonas (ICE of NUITM-VA2). As shown in Figure 3 , most of these ICEs (17/21) maintained high concordance with the ICE of 211703 in the backbone region of maintenance, ars locus, and the backbone region of conjugal transfer, and the differences mainly occurred in the accessory region where the bla _NDM-1 region was located.

Among the 21 ICEs, the ICE of P. juntendi PP_2463 (sharing the most similar accessory regions with the ICE of A. caviae 211703) and the ICE of A. caviae NUITM-VA2 (the only one ICE from A. caviae), were specially annotated and analyzed. Detailed annotation of their bla _NDM-1 region and comparison with bla _NDM-1 region from ICE of 211703 were performed ( Figure 4 ). Results showed that the genetic contexts of bla _NDM-1 region from ICE of PP-2463/NUITM-VA2 shared a common structure in upstream(Δorf5–sul1–ΔqacED1–ISCR14-like) with bla _NDM-1 region from ICE of 211703, following by a replacement of gene bla _PME-1. The genetic structures identified included an ISCR14-like ΔTn125 carrying the bla _NDM-1 gene, as well as a ΔISCR14-like–bla _PME-1–ISCR14-like structure, which was not present in the ICE carrying bla _NDM-1 in strain 211703. The only difference found between the bla _NDM-1 regions from ICE of PP-2463 and NUITM-VA2 was an insertion of aphA6–ISCR14-like, which was also present in the bla _NDM-1 region from ICE of 211703. Despite this variation, all three bla _NDM-1 regions shared a common structure downstream, consisting of floR–tetR(G)–tetA(G). However, there was an additional insertion of the ISStes1 sequence downstream of the bla _NDM-1 regions in both ICE of PP-2463 and NUITM-VA2.