The A. veronii strain, HD6454, was isolated from the raw sewage in sewage conduit under washbasin in digestive department of the Second Affiliated Hospital of Soochow University (Suzhou, China) in July 2019. The detailed isolation methodology was described previously (Wen et al., 2022). Bacterial species identification was carried out using Matrix-Assisted Laser Desorption/Ionization Time of Flight Mass Spectrometry (MALDI-TOF-MS). The presence of carbapenemase genes (bla_KPC, bla_NDM, bla_VIM, bla_{OXA–48–like}, and bla_IMP), mcr genes and tmexCD-toprJ was determined by PCR amplification as described previously (Lv et al., 2020; Tang et al., 2020).

Bacterial genomic DNA was isolated using the Omega Bio-Tek Bacterial DNA Kit (Doraville, GA, USA), and sequenced from a sheared DNA library with average size of 10 kb on a Nanopore PromethION platform (Oxford Nanopore Technologies, OX, UK), as well as a paired-end library with an average insert size of 350 bp on a NovaSeq sequencer (Illumina, CA, USA). A hybrid assembly was then conducted by Unicycler 0.4.9¹ using both paired-end short Illumina reads and the long Nanopore reads.

The sequence type (ST) of the A. veronii strain HD6454 was identified according to the online multi-locus sequence typing (MLST) scheme.² Amino acid sequences of MCR variants and MCR-3-like variants were aligned using ClustalW in MEGAX 10.1.8 (Kumar et al., 2018). Unrooted maximum-likelihood phylogenetic trees were further generated using MEGAX 10.1.8 with a bootstrap iteration of 1,000.

Open reading frames (ORFs) and pseudogenes were predicted using RAST 2.0 (Brettin et al., 2015) combined with BLASTP/BLASTN searches (Boratyn et al., 2013) against the UniProtKB/Swiss-Prot database (Boutet et al., 2016) and the RefSeq database (O’Leary et al., 2016). Annotation of resistance genes, mobile elements, and other features were carried out using the online databases including CARD (Jia et al., 2017), ResFinder 4.1 (Zankari et al., 2012), Danmel (Wang et al., 2022), ISfinder (Siguier et al., 2006), and Tn Number Registry (Tansirichaiya et al., 2019). Multiple and pairwise sequence comparisons were performed using BLASTN. Gene organization diagrams were drawn through scripts from Danmel (Wang et al., 2022), and displayed using Inkscape 1.0.³ Prediction of protein secondary structure was performed using ESPript 3.0 (Robert and Gouet, 2014) and displayed using Phyre 2.0 (Kelley et al., 2015).

The construct of recombinant plasmid was performed through seamless cloning by ClonExpress Ultra One Step Cloning Kit (Vazyme, China). The details were described in Supplementary text, and the primers used herein were shown in Supplementary Table 1. The resulting recombinant plasmids were transformed through heat shock into Escherichia coli DH5α, and 100 μg/ml ampicillin was used for the transformant selection. The transformants DH5α/pUC18, DH5α/pBAD24 and DH5α/pBAD24-mcr-1.1 were also constructed as controls. Successful transformants were confirmed by PCR, and Sanger sequencing on a 3730XL sequencer (ABI, Boston, MA, USA). Induction of the pBAD24 vector was performed using MH II broth (Cation-Adjusted) supplemented with 0.4% L-arabinose as previously described (Kieffer et al., 2019).

The susceptibility of colistin and tigecycline was carried by the minimum inhibitory concentrations (MICs) method. Results were interpreted according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST).⁴ The susceptibility of other antimicrobial agents (Supplementary Table 2) was carried by using the Phoenix System-M50 automatic microbiology analyzer (BD, USA). Results were interpreted according to the 2020 Clinical and Laboratory Standards Institute (CLSI) guidelines. The susceptibility test through MIC method was repeated three times to ensure the accuracy of the result. The E. coli ATCC 25922 was used as the quality control.

Conjugal transfer experiment was carried out with sodium azide-resistant E. coli J53 as the recipient, and the HD6454 isolate as the donor. In brief, overnight cultures of the HD6454 and J53 were mixed at a ratio of 1:1 in LB broth, and the mixture was then spotted on a hydrophilic nylon membrane filter with a 0.45 μm pore size (Millipore) that was placed on LB agar plate and then incubated for mating at 37 C for 18 h. Bacteria were washed from filter membrane and spotted on LB agar plate, and 200 μg/ml sodium azide (for J53) together with 2 μg/ml imipenem (for bla_KPC) was used for selecting an E. coli transconjugant carrying bla_KPC.

The complete chromosome and plasmid sequences of HD6454 were submitted to GenBank under accession numbers CP079823-CP079826, respectively.

The A. veronii strain HD6454 was isolated from the raw sewage in sewage conduit under washbasin of digestive department, and belonged to a novel ST variant ST1016. The strain harbored a 5,029,035-bp chromosome genome and three plasmids: pHD6454-KPC, pHD6454-2, and pHD6454-3 (Table 1). Meanwhile, HD6454 carried 18 important ARGs, including bla_KPC–2, mcr-3.17, and tmexC3.2-tmexD3.3-toprJ1b gene cluster. Additionally, a mcr-3-like gene was found downstream of the mcr-3.17 gene. This mcr-3-like gene was further designated as mcr-3-like3 as previously described (Shen et al., 2018). Susceptibility testing results showed that HD6454 was multidrug-resistant, including carbapenems (Supplementary Table 2). However, despite carrying mcr-3.17 and tmexC3.2-tmexD3.3-toprJ1b gene cluster, HD6454 was susceptible to colistin (2 mg/L) and tigecycline (0.125 mg/L) based on the EUCAST clinical breakpoints (see footnote 4).

The MCR-3.17 shared 89.24% amino acid sequence similarity to MCR-3.1 (accession number KY924928), while MCR-3-like3 shared 84.29% amino acid sequence similarity to MCR-3.1. The amino acid sequence of MCR-3.17 shared the highest similarity (93.15%) with MCR-3.6 (accession number MF598076) among all MCR-3 variants. The phylogenetic analysis also showed that, compared with other MCR-3 variants, the evolution relationship between MCR-3.6 and MCR-3.17 was the closest (Supplementary Figure 1). Nevertheless, MCR-3.17 was still different from MCR-3.6 in predicted protein secondary structure (Supplementary Figure 2). Previous studies showed that MCR proteins were predicted to have two conservative domains, the membrane-anchored domain (approximate residues range for MCR-3 was 1–172) and the soluble catalytic domain (approximate residues range for MCR-3 was 173 to the last) (Yin et al., 2017; Carroll et al., 2019). Compared with MCR-3.6, 33 mutations in amino acid sequences of the membrane-anchored domain of MCR-3.17 were identified, and further led the loss of two β-sheet areas in protein secondary structure, while MCR-3.17 also had four mutations in amino acid sequences of the soluble catalytic domain, and further led the increase of a 3₁₀-helix area in protein secondary structure (Supplementary Figure 2).

In order to determine the function of mcr-3.17 and mcr-3-like3 in mediating colistin resistance, the transformants DH5α/pUC18-mcr-3.17, DH5α/pUC18-mcr-3-like3 and DH5α/pUC18-mcr-3.17&mcr-3-like3 were obtained. However, MIC of these three transformants displayed no difference, compared with MIC of DH5α/pUC18 (Table 2). Considering probable variations in the promoter activities of mcr-3.17 and mcr-3-like3 in HD6545, the transformants DH5α/pBAD24-mcr-3.17 and DH5α/pBAD24-mcr-3-like3 were further constructed. In addition, transformants DH5α/pBAD24 and DH5α/pBAD24-mcr-1.1 were also constructed as controls. The antimicrobial susceptibility testing showed that DH5α/pBAD24-mcr-3.17 presented an MIC of 4 mg/L after induction of arabinose, while DH5α/pBAD24-mcr-1.1 had an MIC of 16 mg/L. On the contrary, DH5α/pBAD24-mcr-3-like3 presented a MIC of 1 mg/L after induction of arabinose (Table 2). This result suggested that mcr-3.17, but not mcr-3-like3, could confer low-level resistance to colistin.

The mcr-3.17 and mcr-3-like3 genes were located in a 46.5-kb mcr-3.17 region in the chromosome of HD6454 (Figure 1). The mcr-3-like3 gene was located immediately downstream of mcr-3.17, the nucleoside bases between them was 66 bp. Compared with the chromosome of a reference A. veronii strain JC529 (accession number CP058912), a 6.9-kb backbone region was replaced by the mcr-3.17 region in HD6454. This mcr-3.17 region contained a truncated Tn6378, a 25.3-kb region with multiple IS elements, and the “core mcr-3.17 region,” including truncated sulP gene, mcr-3.17 gene and mcr-3-like3–to–orf1128 region.

To further probe how this 46.5-kb mcr-3.17 region originated, a detailed genetic dissection analysis was applied to compare the genetic structure of HD6454 with additional 10 mcr-3.6 (n = 7) or mcr-3.17 (n = 3)-harboring genomes download form the GenBank (Supplementary Tables 3, 4). Similar core mcr-3.17 regions could be found in three mcr-3.17-carried contig fragments from Aeromonas allosaccharophila Z9-6 (Shen et al., 2018), and A. veronii CN17A0120 and ADV102 (Rangel et al., 2019; Figure 2). Truncated or complete usp–sulP region and truncated or complete mcr-3-like3–to–orf1128 regions could be found upstream and downstream of the mcr-3.17 gene from all mcr-3.17 regions in this study, respectively. However, due to the limited length of these contig assemblies, whether these core mcr-3.17 regions were carried by complete genetic elements or not could not be determined. Fortunately, an integrative and mobilizable element (IME) from A. sanarellii NS1 was found to share at least 96.71% identity with these core mcr-3.17 regions (Figure 2). This IME harbored mcr-3.6 and mcr-3-like3 genes, and was newly designated Tn7360. Similar usp–sulP region and truncated mcr-3-like3–to–orf1128 region could also be found in Tn7360. Therefore, the so-called core mcr-3.17 region was further named as Tn7360-related region. Moreover, except Tn7360, all mcr-3.6 and mcr-3-like3 genes were carried by IME Tn7361a/b (Shi et al., 2020) or related genetic elements and regions (Figure 3). The backbone sequence of Tn7361a/b was 95.60% identical to the backbone sequence of Tn7360 with 76% coverage. Tn7361a/b also harbored the complete mcr-3-like3–to–orf1128 region. Meanwhile, Tn7360 and Tn7361a/b were all integrated within the thyA gene with 6-bp direct repeats (DRs). In addition, truncated Tn7360 fragments were identified in other genetic elements, including composite transposon Tn6518 (Wang et al., 2020b) from A. veronii w55 and IME Tn6868 from A. hydrophila WP7-S18-ESBL-06.

In summary, all mcr-3.6, mcr-3.17 and mcr-3-like3 characterized in this study were located within IME Tn7360/Tn7361 or related genetic elements and regions from Aeromonas species.

A tnfxB3-tmexCD3-toprJ1-like gene cluster was found in the chromosome of strain HD6454 initially. The sequence of this cluster was of 100% identical to the one carried by chromosome of A. caviae WCW1-2 (accession number CP039832) (Wang et al., 2021b). Furthermore, according to the naming scheme previously described (Wang et al., 2021a): (1) compared with the tnfxB3 gene from Proteus cibarius strain SDQ8C180-2T (accession number CP073356), one mutation was identified in the tnfxB3-like gene from HD6454 (predicted to encode Thr46Ala), and this tnfxB3-like gene was thus named tnfxB3.2; (2) compared with the tmexD3 gene from strain SDQ8C180-2T, one mutation was identified in the tmexD3-like gene from HD6454 (predicted to encode Val56Glu). This tmexD3-like gene was also different from the tmexD3.2 gene in strain P. mirabilis SDY9C89-2 (accession number MZ004963), so it was named tmexD3.3; (3) the tmexC3-like gene and toprJ1-like gene from HD6454 were identical to the tmexC3.2 gene from Pseudomonas aeruginosa strain AHM8C91AI (accession number JAGSOC000000000) and the toprJ1b gene from strain SDQ8C180-2T, respectively. In general, this tnfxB3-tmexCD3-toprJ1-like gene cluster from HD6454 was identified as tnfxB3.2-tmexC3.2-tmexD3.3-toprJ1b finally.

In order to determine the function of tmexC3.2-tmexD3.3-toprJ1b in mediating tigecycline resistance, the transformants DH5α/pUC18-tmexC3.2-tmexD3.3-toprJ1b was obtained. The antimicrobial susceptibility testing showed that DH5α/pUC18-tmexC3.2-tmexD3.3-toprJ1b presented an MIC of 0.5 mg/L to tigecycline, while DH5α/pUC18 had an MIC of 0.125 mg/L (Table 2). This result suggested that tmexC3.2-tmexD3.3-toprJ1b could confer low-level resistance to tigecycline.

The tmexCD1-toprJ1 gene cluster was firstly identified in the structure “int1-int2-hp1-hp2-tnfxB1-tmexCD1-toprJ1,” and this structure was further carried by Tn5393 (Lv et al., 2020). Subsequently, this structure along with its attachment site at the left/right end (attL/R) was defined as relaxosome-missing IME Tn6855 (Yu et al., 2021b). Further genetic dissection analysis showed that the tnfxB3.2-tmexC3.2-tmexD3.3-toprJ1b gene cluster from HD6454 was located within an IME Tn6855 variant, which was integrated in the umuC gene of the novel IME Tn7379 with 6-bp DRs (Figure 4). This Tn6855 variant differed from Tn6855 by the substitution of tnfxB1-tmexCD1-toprJ1 gene cluster to tnfxB3.2-tmexC3.2-tmexD3.3-toprJ1b. Additionally, some other tnfxB3-tmexCD3-toprJ1b gene clusters were also located within similar Tn6855 variants (Figure 4), and integrated within the umuC genes as previously reported (Wang et al., 2021a,b).

The bla_KPC–2 was carried by a 51.66-kb IncP-6 plasmid, which was assigned the name pHD6454-KPC. The modular structure of pHD6454-KPC was divided into the backbone and three accessory modules (ISPa19, Tn5563b and the bla_KPC–2 region) which were resulted from exogenous DNA regions insertion at different sites of the backbone (Supplementary Figure 3). The plasmid pHD6454-KPC shared 99.99% nucleotide identity to the IncP-6 reference plasmid p10265-KPC from Pseudomonas aeruginosa 10,265 with 76% coverage (accession number KU578314) (Dai et al., 2016), while the backbone of them were 99.98% nucleotide identical, along with 99% coverage. This showed that these two plasmids were mainly different in accessory modules (Figure 5). The unit transposon Tn5563b in pHD6454-KPC differed from Tn5563a in p10265-KPC by the insertion of a merT-harboring region (inserted region_merT, 12.90 kb) (Figure 6A). The inserted region_merT was flanked by 9-bp DRs, and may originated from the transposition and homologous recombination of an unknown-type plasmid. Additionally, the bla_KPC–2 region in pHD6454-KPC was composed of a truncated bla_KPC–2-carrying unit transposon Tn6296 (Wang et al., 2017) and a truncated unit transposon Tn6376b, and was almost identical to the one in p10265-KPC (Figure 6B).

The genomic dissection analysis showed that pHD6454-KPC lost the tra module, which was consistent with other IncP-6 plasmids as described previously (Dai et al., 2016; Hu et al., 2019). Repeated conjugation attempts failed to transfer pHD6454-KPC from strain HD6454 into J53, which matched the sequence analysis result.