In August 2023, we isolated a meropenem-unsusceptible strain M2 from a sludge sample at a commercial cattle farm in Hebei, China.

PCR amplification and Sanger sequencing verified that this strain co-carried bla _NDM and tet(X) genes. This strain was classified as Acinetobacter genus using MALDI-TOF-MS and 16S rDNA. Subsequently, this strain was sequenced using whole-genome sequencing (WGS) on the Oxford Nanopore platform (long-read sequencing technology). The analysis confirmed 99.03% completeness and 0.89% contamination in this genome assembly (Acinetobacter sp. A2 genome assembly ASM4853755v1 - NCBI - NLM). To confirm the species of this genome, the Average Nucleotide Identity (ANI) match was performed using the NCBI annotation service. The result found that no genome showed >95% ANI with this genome (Acinetobacter sp. A2 genome assembly ASM4853755v1 - NCBI - NLM). The search on the DSMZ (Deutsche Sammlung von Mikroorganismen und Zellkulture) platform using the type strain genome server (TYGS) identified this genome as a potential new species, with the best match A. seohaensis DSM 16313 (34.1% dDDH), A. towneri DSM 14962 (33%), and A. indicus CIP 110367 (24%). This genome best matched A. towneri DSM 14962 = CIP 107472 (0.98559 Z-Score, 87.32% ANI), A. towneri DSM 14962 = CIP 107472 DSM 14962(0.98498, 87.27%), and A. tibetensis Y-23 (0.94622, 77.69%) on the JSpeciesWS platform using the tetra correlation search (TCS). Subsequently, we employed Mash (a k-mer-based rapid sequence alignment tool) on the Pathogenwatch platform. The analysis revealed that this genome best matched NZ_JAAZQX010000010.1, with a mash distance of 0.0147665, a p-value of 0, and matching hashes of 579/1000. We found that the NZ_JAAZQX010000010.1 was one of the contigs in the genome, GCA_012371325.1 (this is the only genome identified as Acinetobacter sp. A2 in the NCBI genome database). Thus, the ANI calculation on the EZBioCloud platform was executed for GCA_012371325.1 and genome M2. We found that these genomes exhibited 98.35% ANI. Furthermore, the digital DNA-DNA hybridization (dDDH) calculation of these genomes was performed on the DSMZ platform using the genome-to-genome distance calculator (GGDC). The analysis revealed that these genomes exhibited 85.5% dDDH. The thresholds of ANI and dDDH are usually applied to define genomic species (95–96 and 70 % for ANI and dDDH, respectively) (Meier-Kolthoff et al., 2013). However, 98.35% ANI and 85.5% dDDH are beyond defined thresholds, suggesting genome M2 and GCA_012371325.1 are very close but may belong to different species (Riesco and Trujillo, 2024). Furthermore, Acinetobacter sp. A2 belongs to an unclassified species of the Acinetobacter genus in the family Moraxellaceae (NCBI taxonomy database). Thus, it is classified only at the genus level, not at the species level. As a result, although strain M2 is closest related to Acinetobacter sp. A2 (Taxonomy ID: 362457, Acinetobacter sp. A2 - NCBI - NLM), it should be defined as a novel species under the Acinetobacter genus.

The antimicrobial susceptibility testing showed that the strain M2 was resistant to meropenem (16 mg/L), ampicillin (128 mg/L), ceftazidime (>1024 mg/L), cefepime (128 mg/L), gentamicin (128 mg/L), kanamycin (512 mg/L), fosfomycin (512 mg/L), imipenem (16 mg/L), ertapenem (32 mg/L) and tetracycline (16 mg/L) ( Table 1 ). However, it was susceptible to aztreonam, chloramphenicol, colistin, ciprofloxacin, sulfamethoxazole, azithromycin, doxycycline, tigecycline, amikacin and omadacycline ( Table 1 ).

The previous report regarding successful conjugation in the Acinetobacter plasmid remains rare. However, successful conjugation in the A. indicus plasmid was observed in our laboratory. In this study, we used the same parameters and tried new ones. However, the horizontal transfer of the plasmid carrying bla _NDM was not observed.

The WGS revealed that strain M2 contained one circular chromosome genome (2873922 bp, 41.5% GC content) and two circular plasmid genomes (pM2-1: 70176 bp, 40.2%; pM2-2-NDM-1: 64147 bp, 38.7%) ( Table 2 ; Figure 1 ). The GC content of pM2-2-NDM-1 was lower than chromosome, hinting that they originated from different hosts. The GC content can serve as a marker to measure the origin of plasmids, indicating the potential horizontal transfer of plasmids (Tang et al., 2021; Long et al., 2025). The chromosome of the strain M2 carried multiple ARGs, including tet(X3), aph(6)-Id, aph(3’’)-Ib, and sul2 ( Figure 1 ). In addition to ARGs, this chromosome carried MRGs against arsenic, mercury, and copper, as well as genes encoding VFs related to bacterial adhesion, invasion, enzyme, immune evasion, and serum resistance ( Table 2 : Figure 1 ). The pM2–1 did not carry ARGs but harbored the genes related to copper resistance and the CusS-CusR two-component regulatory system (TCS) ( Table 2 : Figure 1 ). The pM2-2-NDM-1 carried various ARGs, such as bla _NDM-1, aac(3)-IId, aph(6)-Id, aph(3’)-Ia, aph(3’’)-Ib, aph(3’)-VI, msr(E), mph(E), sul2, and tet(39) ( Table 2 : Figure 1 ).

The analysis of the pM2–1 and pM2-2-NDM-1 on the Pathogenwatch platform (homology-based alignment) revealed that they could not be identified as any known replicon type. Subsequently, the analysis of these plasmids on the Galaxy platform using MOB-typer revealed that pM2-2-NDM-1 was typed to rep_cluster_481 and predicted as conjugative, and pM2–1 was typed to rep_cluster_1656 and predicted as mobilizable. Furthermore, these plasmids were searched on the PLSDB database. The result revealed that pM2-2-NDM-1 exhibited the highest identity (>99.3%) with five plasmids. Among these, the sizes of three plasmids <5000 bp could not be identified as any known replicon type. The remaining two plasmids (NZ_CP051870.1 and NZ_CP051876.1) with the same size of 48239 bp were sourced from A. baumannii, typed to rep_cluster_481 and predicted as conjugative. However, these plasmids did not carry bla _NDM or tet(X). In addition, the search of pM2–1 found that no plasmid exhibited identity.

The analysis of the pM2-2-NDM-1 on the ICEfinder platform identified a putative ICE region of 52688 bp (4758–57445 bp) containing the origin site of DNA transfer (oriT, 29162–29538 bp), T4SS (virB11, virB10, virB9, virB8, virB5, virB6, virB4, virB3, virB2, and virB1), T4CP (Type IV Coupling Proteins), relaxase (mobC and MOBP), integrase, and ARGs (bla _NDM-1, aac(3)-IId, aph(3’)-Ia and tet(39)) ( Figure 1 ). Furthermore, six pdif sites were identified in the pM2-2-NDM-1, and they formed five pdif modules including three pdif-ARGs modules (pdif-aac(3)-IId-aph(3’)-Ia, pdif-tet(39) and pdif-msr(E)-mph(E)), pdif-ser (a serine recombinase) module and pdif-hp (hypothetical protein) module ( Figures 1 , 2 ). These modules collectively build a pdif module island. This inevitably leads to sharing the internal C/D or D/C sites to form two types of pdif modules (Blackwell and Hall, 2017), one flanked by a C/D and a D/C site, and the other type flanked by a D/C and a C/D site. We searched the pdif-aac(3)-IId-aph(3’)-Ia module in the NCBI database and found that only one plasmid (pRp428) was highly similar (coverage 99% and identity 100%) to this module. The pdif-tet(39), pdif-msr(E)-mph(E) and pdif-ser modules were reported in pS30-1 (Blackwell and Hall, 2017). Thus, we used these plasmids and pM2-2-NDM-1 to construct the linear comparison ( Figure 2 ) and identified pdif sites ( Table 3 ). The pdif-aac(3)-IId-aph(3’)-Ia, pdif-tet(39), pdif-msr(E)-mph(E), pdif-ser and pdif-hp modules were identified in pRp428. Among these, the pdif-tet(39), pdif-ser, and pdif-hp modules were identical (100% coverage and identity) to those in pM2-2-NDM-1. The pdif-aac(3)-IId-aph(3’)-Ia module exhibited 98.64% identity to that in pM2-2-NDM-1. A gap of 178 bp existed in the non-coding region between left-hand IS15 and aph(3’)-Ia. The pdif-tet(39) and pdif-ser modules in pS30–1 were identical (100% coverage and identity) to those in pM2-2-NDM-1. Interestingly, the pdif-msr(E)-mph(E) module in pRp428 and pS30–1 only carried msr(E)-mph(E) gene pair. This structure was commonly seen in previous reports (Blackwell and Hall, 2017; Shao et al., 2023). However, the pdif-msr(E)-mph(E) module in pM2-2-NDM-1 additionally carried ISAjo2 and higA-higB. The pdif-higA-higB and pdif-ISAjo2 modules were independently present in pS30-1. However, the C/D and D/C sites flanking higA-higB and ISAjo2 were not found in pM2-2-NDM-1. In addition, the D/C site is located upstream of msr(E) in the pdif-msr(E)-mph(E) modules of pRp428 and pS30-1. However, the C/D site is located upstream of ISAjo2 in the pdif-msr(E)-mph(E) module of pM2-2-NDM-1. The analysis of pdif sites ( Table 3 ) found that the base variant appeared in the D/C sites of the pdif-ser module of pS30-1. The C/D and D/C sites of the remaining pdif modules of pRp428 and pS30–1 were identical to those in pM2-2-NDM-1.

BLASTn analysis in the GenBank database found that pM2-2-NDM-1 was similar to pRp428 (100% identity and 90% coverage), pAb-D10a-a (99.91% identity and 68% coverage), pDT01139C (99.07% identity and 67% coverage), pFk2-7 (98.5% identity and 68% coverage), and pAR3 (99.21% identity and 66% coverage) ( Figure 3A ). Among these, three plasmids (pRp428, pAb-D10a-a, and pDT01139C) were sourced from A. baumannii, pAR3 from A. radioresistens, and pFk2–7 from an unidentified strain. These plasmids and pM2-2-NDM-1 were used to build an alignment analysis ( Figure 3A ). The analysis found that the conserved region contained oriT and the genes encoding T4SS, T4CP, and relaxase. The non-conserved regions included ARGs, pdif-ARG modules, and IS elements. Only pDT01139C carried bla _NDM in the genomes from the public database. Thus, pM2-2-NDM-1 and pDT01139C genomes were used to build a comparison analysis of the bla _NDM skeleton ( Figure 3B ). The analysis found that the trpF in the bla _NDM-1 framework of pDT01139C consisted of 636 bp. The trpF in the bla _NDM-1 framework of pM2-2-NDM-1 consisted of 114 bp. Furthermore, no IS element could be identified downstream of the bla _NDM-1 framework in pDT01139C.

Three genomes were classified as Acinetobacter sp. A1 (GCA_012371315.1), A2 (GCA_012371325.1), and A3 (GCA_012371415.1) in the NCBI genome database. A total of 51 genomes closely related to genome M2 were retrieved from the JSpesiesWS platform (TCS) and DSMZ platform (TYGS). Subsequently, the above genomes and genome M2 were used to build a phylogenetic tree ( Figure 4 ). Phylogenetic analysis revealed that this tree was diverged into three branches, highlighting three different ancestors. In the branch one, Acinetobacter sp. A1-A3 (A1-A3), A. towneri (DSM 14962 and DSM 14962 CIP 107472), and A. seohaensis (DSM 16313) were diverged from a clade, indicating their high homology. Among this, genomes M2 and A2 were diverged from a leaf node, suggesting they originated from a most recent common ancestor. Branch length quantifies accumulated genetic divergence from the most recent common ancestor. Genome M2 has higher genetic variability than A2 (branch length, 0.025 (M2) > 0.00707 (A2)), hinting that it undergoes prolonged independent evolution and accelerated evolution derived by selective pressure. Like the Acinetobacter sp. A2, the Acinetobacter sp. A1 (Taxonomy ID: 401467) and Acinetobacter sp. A3 (Taxonomy ID: 2725492) belong to unclassified species of the Acinetobacter genus in the family Moraxellaceae.

The bla _NDM-1 framework consisted of ISAba125-bla _NDM-1-ble _MBL-trpF with a size of 2426 bp, located in the pM2-2-NDM-1 ( Figure 1 ). We found that an aph-(3’)-VI was located upstream of bla _NDM-1 framework. Two intact ISAba14 (1282 bp, orienting the same direction) flanking the structure of aph-(3’)-VI-ISAba125-bla _NDM-1-ble _MBL-trpF formed a rolling-circle-like structure of 5966 bp, ISAba14-aph-(3’)-VI-ISAba125-bla _NDM-1-ble _MBL-trpF-ISAba14 ( Figure 1 ). ISFinder analysis revealed that this ISAba14 (the group IS150 of family IS3) had 99% identity with the reference sequence (1280/1282 bp) from A. baumannii. The BLASTn analysis of this rolling-circle-like structure in the NCBI database found that the plasmids of multiple bacterial species were high similar (>99% identity and coverage), including Acinetobacter genus (A. baumannii, A. nosocomialis, A. baylyi, A. johnsonii, A. junnii, A. lwoffii, A. pittii, A. schindleri, A. soli, and A. towneri), Citrobacter genus (Citrobacter freundii and Citrobacter werkmanii), Providencia genus (Providencia rettgeri and Providencia stuartii), E. coli, Enterobacter hormaechei, and Klebsiella pneumoniae. These plasmids (one plasmid genome selected from each bacterial species) and the pM2-2-NDM-1 were used to construct a comparison analysis ( Figure 5A-1 ). Meanwhile, ISFinder was used to search downstream of the bla _NDM skeleton. The analysis found that ISAba14 frequently emerged upstream of the bla _NDM skeleton in the plasmids from the public database but rarely downstream of the bla _NDM skeleton. Furthermore, the BLASTn analysis on the UniProt database found that an intact trpF consisted of 639 bp, which could be identified in the plasmids from the public database. However, the trpF in the bla _NDM framework of pM2-2-NDM-1 was truncated by ISAba14 with a size of 114 bp. Tn7382 is derived from Tn125 and encompasses seven open reading frames (aph-(3’)-VI, ISAba125, bla _NDM-1 , ble, iso, TAT, cutA) enclosed by two direct copies of ISAba14 (Hamed et al., 2022). It was found in multiple chromosomes (TP2, TP3, AbBAS-1, and CI300) from A. baumannii and exhibited structural similarity to the rolling-circle-like structure in pM2-2-NDM-1. Thus, a linear comparison was built ( Figure 5A-2 ). The analysis found three genes (iso, TAT, cutA) between ble and downstream ISAba14 in Tn7382. However, only one gene (fragmentary trpF) was located in the same site of pM2-2-NDM-1. Furthermore, an IS4 or ISAba33 was located upstream of the left-hand ISAba14. A similar location was not observed in pM2-2-NDM-1.

The tet(X3) framework, ISVsa3-XerD-tet(X3)-res-ISVsa3, with a size of 5819 bp, was located in the chromosome of strain M2. Two ISVsa3 flanked the structure of XerD-tet(X3)-res. The upstream ISVsa3 was a fragmentary element of 715 bp, and the downstream ISVsa3 was an intact element of 1494 bp. ISFinder analysis revealed that ISVsa3, belonging to the ISCR family, originated from Vibrio salmonicida. The BLASTn analysis of this tet(X3) framework in the NCBI database found that the similar frameworks were mainly identified in the Acinetobacter genus and E. coli. The plasmid genomes with >99% identity and coverage to this framework were selected to construct the alignment analysis ( Figure 5B ). The analysis found that in addition to tet(X3), the ISVsa3 also mediated the formation of tet(X4) and tet(X5) frameworks. Most ISVsa3-mediated tet(X3) frameworks were sourced from the Acinetobacter genus. An intact ISVsa3 was commonly located downstream of tet(X). However, either a fragmentary or intact ISVsa3 was located upstream of tet(X). In addition to ISVsa3, an IS26 frequently appeared upstream of tet(X). An IS element, IS1006, that is rarely reported to be associated with the tet(X) family emerged upstream of the left-hand ISVsa3. Furthermore, ISVsa3 mediated the rolling-circle-like structures carrying tet(X3) (p34AB) or tet(X4) (p47EC).

The sludge was collected from a commercial beef cattle farm in Hebei province, China. The strain was isolated from MacConkey (Huan Kai Microbial, China) agar plates supplemented with 2 μg/mL meropenem and cultured at 37°C for 16 hours. PCR amplification and Sanger sequencing verification were performed for bla _NDM and tet(X) genes (Wang et al., 2017; He et al., 2019).

The long-read sequencing was executed on the Oxford Nanopore platform. Briefly, the genome was sequenced using Oxford Nanopore and the DNBSEQ platform. The resulting corrected reads were carried out by hybrid assembly in combination with DNBSEQ short reads. The assembled genome was checked for completeness and contamination using the CheckM (v1.2.3) of the NCBI annotation service.

The identification of bacterial genus was performed by MALDI-TOF-MS and 16s rRNA (Kim et al., 2010; Singhal et al., 2015). The identification of bacterial species was executed on the NCBI database using ANI match of the NCBI annotation service, on the DSMZ platform (Leibniz Institute DSMZ: Welcome to the Leibniz Institute DSMZ) using the type strain genome server (TYGS), and on the JSpeciesWS platform (JSpeciesWS - Taxonomic Thresholds) using tetra correlation search (TCS). Furthermore, we used the mash match tool to identify bacterial species on the Pathogenwatch platform (Pathogenwatch | A Global Platform for Genomic Surveillance) (Argimón et al., 2021). The ANI calculation was executed on the EZBioCloud platform (ANI Calculator | Ezbiocloud.net) using the OrthoANIu algorithm (Yoon et al., 2017). The GGDC platform (ggdc.dsmz.de/ggdc.php#) was used to calculate the dDDH value.

The minimum inhibitory concentration (MIC) of 20 antibiotics (meropenem, aztreonam, ampicillin, ceftazidime, cefepime, gentamicin, chloramphenicol, colistin, kanamycin, fosfomycin, ciprofloxacin, sulfamethoxazole, azithromycin, tetracycline, doxycycline, tigecycline, imipenem, ertapenem, amikacin and omadacycline) was performed using microdilution of Mueller-Hinton broth (Huan Kai Microbial, China). The testing concentration range is 2 mg/L to 1024 mg/L. Results were determined according to Clinical and Laboratory Standards Institute (CLSI) documents M100-S34 (2024). Since there is no established breakpoint for tigecycline and omadacycline resistance in Acinetobacter spp., their breakpoint was determined by FDA-defined interpretive criteria for Enterobacteriaceae (R≥ 8 μg/mL for tigecycline and R≥ 16 μg/mL for omadacycline).

Conjugation assays were conducted based on the filter mating method using E. coli J53 (sodium azide-resistant) and Salmonella LGJ2 (rifampicin-resistant) as recipients. A donor-recipient mixture of 10:1 (the ratio 10:1 of the donor (M2) to the recipient (J53 or LGJ2)) was incubated on a 0.22-μm filter membrane at 35°C for 16h, then moved to MacConkey plates supplemented with donor resistance (0.5 mg/L meropenem) and recipient resistance (100 mg/L sodium azide or 200 mg/L rifampicin). The ratio 5:1 and 15:1 of the donor to the recipient were supplementarily tested. At least three attempts were made for each parameter, and three parallel experiments were made for each attempt.

Core genomes were extracted using Roary (GitHub - sanger-pathogens/Roary: Rapid large-scale prokaryote pan genome analysis). Recombination was filtered using Gubbins (GitHub - nickjcroucher/gubbins: Rapid phylogenetic analysis of large samples of recombinant bacterial whole genome sequences using Gubbins). Filtered polymorphic sites were employed to build a tree on the PhyML platform (GitHub - stephaneguindon/phyml: PhyML – Phylogenetic estimation using (Maximum) Likelihood). The iTOL (iTOL: Interactive Tree Of Life) visualized this tree with the corresponding features of each genome.

ResFinder (v4.6.0) was used to screen all known acquired AMRs (Bortolaia et al., 2020). The threshold of ARG identification was set to 90%, and the minimum length was set to 80%. Plasmid typing was carried out on the Pathogenwatch platform using homology-based alignment and on the Galaxy (Galaxy | China) platform using MOB-typer. The PLSDB database (PLSDB) was used to analyze the plasmid characterization further. The VFDB was used to search for the genes encoding VFs (VFDB: Virulence Factors of Bacterial Pathogens) (Liu et al., 2022). Genome annotation was performed using the RAST genome annotation service (RAST Server - RAST Annotation Server) (Overbeek et al., 2014), with manual refinement using ORFfinder (ORFfinder Home-NCBI), UniPort (UniProt), and ISFinder (ISfinder) (Siguier et al., 2006). The pdif module was identified using PdifFinder (Home) (Shao et al., 2023). ICEfinder (ICEfinder) was employed to detect the oriT, integrase, relaxase, T4CP, T4SS, and the other associated components (28). The circular and linear comparisons were created using BRIG (ver. 0.95) and Easyfig (ver. 2.2.5).