A total of 104 non-duplicate Acinetobacter spp. isolates were obtained between January 2020 and April 2021 in the Affiliated Hospital of Southwest Medical University. Bacterial identification was confirmed by matrix-assist laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) (Bruker, Bremen, Germany) and 16s rRNA sequencing (Supplementary Datasheet 1). The antimicrobial susceptibility profiles were tested by MicroScan Walk-Away 96 Plus system (Siemens, Germany). The minimum inhibitory concentrations (MICs) of meropenem, imipenem, tigecycline, and colistin were determined using the broth microdilution method, and results were interpreted following the Clinical and Laboratory Standards Institute 2020 standards (CLSI, 2020). Escherichia coli ATCC25922 and Pseudomonas aeruginosa ATCC27853 were used as a quality control. ATCC19606 (lab-WT) was used as the reference strain in this study.

Genomic DNA of Acinetobacter pittii AP2044 strain was extracted using a DNA extraction kit (Qiagen, Hilden, Germany) and then sequenced using the Illumina NovaSeq 6000 PE150 (Illumina Inc, San Diego, CA, United States) and nanopore platforms (Sangon Biotech, Shanghai, China). Read sequences were de novo assembled using Canu workflow (v1.7) (Koren et al., 2017). Prokka (v1.10) was utilized to predict coding genes, tRNA, and rRNA in the assembled genome Seemann, 2014. All genomic data were uploaded in the National Center for Biotechnology Information (NCBI) database under the accession number CP087716-CP087718. A novel OXA variant was then identified and termed OXA-1045 (accession no. OL790815). The ResFinder¹ and NCBI BLAST² were used to determine the acquired resistance genes. The GC Content Calculator web tool³ was used to measure the GC ratio. Additionally, the IS finder⁴ and VFDB⁵ databases were used to determine the insertion sequence (IS) elements and virulence genes. The genomic island (GI) sequences were predicted based on GI prediction software packages (IslandPATH-DIMOB) (Hsiao et al., 2003). Chromosome comparison was performed using BLAST Ring Image Generator (BRIG) in the default settings (Alikhan et al., 2011).

As previously mentioned, mating experiments were conducted in broth and on filters using Escherichia coli J53 AizR (an azide resistant strain of J53) as the recipient at 37°C (Fu et al., 2012). Potential transconjugants were selected on Luria-Bertani (LB) broth agar plates containing 0.5 mg/L of imipenem and 180 mg/L of sodium azide (Xiang et al., 2020). Transconjugants were verified by MALDI-TOF MS and polymerase chain reaction (PCR), respectively.

Sequence alignment of OXA-1045 with OXA-213 was investigated using Clustal Omega (Sievers et al., 2011) and ESPript 3.0⁶. The secondary structure of OXA-1045 β-lactamase was predicted using the JPred4 online tool⁷, which is based on neural networks. To investigate the phylogenetic relationship of OXA-213-like proteins, we collected a total of 78 OXA amino acid sequences from Beta-Lactamase DataBase (BLDB)⁸ (accessed 11 March 2022). Protein sequence alignment of these aa sequences and of OXA-1045 was computed using Clustal Omega. This alignment was used to reconstruct the maximum-likelihood phylogeny by MEGA 7.0 (Kumar et al., 2016).

To evaluate the impact of resistance-determinant genes on MICs, bla_OXA–1045 and bla_OXA–213 were cloned into the vector pET28b (MiaoLingBio, Wuhan, China). To make pET28b-OXA1045 and pET28b-OXA213, PCR amplification and vector pET-28b were digested with BamHI and XhoI and then ligated to the pET-28b vector (Invitrogen, Carlsbad, California, United States). As previously described, the generated plasmid was chemically converted into E. coli strain BL21 (Sigma-Aldrich, St. Louis, MO, United States) (Liu et al., 2021). Potential transformants containing pET28b-OXA1045 were identified on LB agar plates (Sigma-Aldrich, St. Louis, MO, United States) containing 20 mg/L of kanamycin (TransGen, Beijing, China). PCR primers PET28AVF2/PET-VF were used to screen colonies on plates, followed by Sanger sequencing (Supplementary Datasheet 1). The empty vector pET-28b was turned into BL21 for use as a control.

MICs of ampicillin, ampicillin-sulbactam, piperacillin, piperacillin-tazobactam, oxacillin, cefazolin, cefoxitin, cefuroxime, ceftazidime, cefotaxime, imipenem, and meropenem for the transformants containing pET28b-OXA1045 (BL21:pET28b-OXA1045) and pET28b-OXA213 (BL21:pET28b-OXA213) were determined by the broth microdilution method. MICs for ampicillin in the presence of 4 mg/L of sulbactam were also determined based on the methods used to establish MICs for piperacillin-tazobactam.

The biofilm formation ability of AP2044 strain and lab-WT was determined by the crystal violet staining method and field-emission scanning electron microscope (FE-SEM) as described previously with minor adjustment. Briefly, the strains were cultured overnight and adjusted to a final OD₆₀₀ of 0.1. A total of 20 μL of each bacterial suspension and 180 μL of LB broth (Haibo, Qingdao, China) were inoculated into a 96-well polystyrene microtiter plate (Costar#3524, Corning, United States) in triplicate. After incubation at 37°C overnight, the plate was washed with phosphate-buffered saline (PBS, Solarbio, Beijing, China) to remove planktonic cells, and the plate was stained with crystal violet (Solarbio, Beijing, China) and solubilized with 95% ethanol (v/v), after which its absorbance was measured at 570 nm.

In addition, the biofilm formation capacity of each strain was used for SEM analysis. Biofilms were cultured for 24 h on coverslips as described above and planktonic cells were removed using PBS. First, the cells were fixed with 2.5% glutaraldehyde in PBS for 4 h. Second, the cells were gently washed with PBS twice. The samples were then gradually treated with ethanol (30, 50, 70, 80, 90, and 100%) for 20 min at each concentration. The dried samples were coated with platinum and visualized using FE-SEM (Thermo Fisher Scientific, Waltham, Massachusetts, United States).

Prior to all injections, Galleria mellonella larvae were stored at 20°C for 4 h to increase susceptibility to infection. A total of 10 μL of bacterial suspension of AP2044 strain and lab-WT were injected into the last left proleg of G. mellonella larvae and incubated at 37°C. Larvae were scored for survival every 24 h for 7 days of incubation. Then, 15 larvae were injected at a concentration of 10⁶CFU/mL of each strain, which was repeated three times. A negative control group (10 μL of PBS buffer were injected) was set for each batch of experiments. Larvae were identified as dead when failure to move in response to external touch was noted (Kim et al., 2021).

Total RNA extraction and reverse transcription were conducted according to a previously described protocol (Tang et al., 2020). The quantitative reverse transcription-PCR (qRT-PCR) was used to detect the expression level of efflux pump-related genes (Supplementary Datasheet 1). The 2 ^–ΔΔCt method was used to calculate the fold change of mRNA expression. The relative gene expression level was compared to the control sample (tigecycline-susceptible A. pittii, TSAP) (Salehi et al., 2021), which was assigned a value of 1 arbitrary unit. All assays were performed in triplicate in three independent cultures.

A total of 69 carbapenem-resistant Acinetobacter spp. samples were obtained from sputum (40, 58%), urine (13, 18.8%), secreta (6, 8.7%), pleural fluid (6, 8.7%), and blood (4, 5.8%). The majority of them (35, 50.7%) was collected from the ICU. Among these clinical isolates, 36 (52.2%) were considered as extensively drug-resistant (XDR, the isolates are resistant to all antimicrobial classes, except colistin and/or tigecycline). Surprisingly, only the AP2044 strain was resistant to tigecycline. The MICs of the AP2044 strain increased drastically compared to the MICs of ATCC19606 among all tested antibiotics (Table 1). The increases were eightfold for gentamicin, 16-fold for piperacillin, more than 32-fold for ciprofloxacin, 64-fold for tigecycline more than 256-fold for carbapenems, more than 64-fold for tetracycline, and in the range more than 32–256-fold for cephalosporins. However, MICs of colistin were almost the same between the two strains, which meant that they are all susceptible to colistin. The AP2044 strain was recovered from a patient with a lung tumor who underwent multiple surgical and invasive procedures at the ICU. With previous treatments of ceftazidime, sulperazon, and moxifloxacin, the patient was polymicrobial-positive, which included Pseudomonas aeruginosa, A. pittii, and Stenotrophomonas maltophilia.

The genome of AP2044 was assembled into three contigs of 4,248,736 bp, consisting of one chromosomal backbone and two plasmids. The chromosome, pAP2044-1, and pAP2044-2 had 3,921,810, 283,349, and 43,577 bp and 39, 39.4, and 39.14% G + C content, respectively (Table 2). The genome had 18 rRNA operons, 74 tRNAs, and 4,128 predicted protein coding sequences. Indeed, conjugation experiments of the AP2044 strain failed to transfer the plasmid into an E. coli recipient. Antimicrobial resistance can be acquired through horizontal gene transfer (HGT) of antibiotic resistance genes (ARGs) and plasmids are critical for HGT and serve as a support for other mobile genetic elements (MGEs) (Vrancianu et al., 2020). The plasmid carrying bla_NDM–1, pAP2044-1, is similar to plasmid pXBB1-9 (accession no. CP010351.1) from Acinetobacter johnsonii and pALWED1.1 (accession no. KX426227.1) from Acinetobacter lwoffii. pAP2044-2 plasmid shows similarity among the sequenced plasmids to p1_010059 from Acinetobacter junii (accession no. CP028798.2) and pACI-235c from Acinetobacter sp. (accession no. CP026414.1). We speculated that pAP2044-1 and pAP2044-2 plasmids have narrow bacterial host spectrums, which may have a dominant transmission in Acinetobacter spp. The genome analyses demonstrated that the AP2044 strain possessed a total of 12 ARGs linked to five different classes of antibiotics, including carbapenem, β-lactam, aminoglycoside, macrolide, and sulfonamide (Figure 1A) (Table 3). The GC content of a gene can be compared to the whole genome of an organism to determine if that gene comes from that organism. If this is the case, the two GC content profiles are likely to be the same (Evans and Amyes, 2014). Furthermore, the GC ratios (50–60%) of several ARGs, including bla_NDM–1, aac(3″)-IIb, aph(3″)-Ib, aph(6)-Id, and sul2, were much greater than those of the chromosomal backbone (39.46%), implying that the ARGs were acquired from other bacterial species. Therefore, we used the Center for Genomic Epidemiology database⁹ to examine the homology of the ARGs to see if they originated from other bacterial species. Eight of the 12 ARGs appeared to be originating from other pathogens, with gene homologies exceeding 99% identities (Table 3).

Bacterial survival tactics against aminoglycoside antibiotics include altering enzymes to inactivate aminoglycosides, increasing efflux pump production, lowering membrane permeability, and interfering with aminoglycoside binding via modification of 16S ribosomal RNA (Su et al., 2018). Interestingly, the AP2044 genome has four distinct aminoglycoside resistance genes, conferring remarkable cell resistance to a wide variety of aminoglycoside antibiotics (eightfold greater than ATCC19606, Table 1, and Supplementary Datasheet 2). Two genes [aac(3″)-IIb and aph(6)-Id] were most likely transmitted from E. coli, whereas one aph(3″)-Ib gene was likely transferred from Shigella flexneri. This might add phosphate, acetyl, adenyl, or methyl groups to aminoglycosides to alter them. Macrolide antibiotics inhibit protein synthesis by targeting the bacterial ribosome (Vázquez-Laslop and Mankin, 2018). Furthermore, the msr(E) gene in the AP2044 strain’s genome, which comes from Pasteurella multocida via HGT, may protect ribosomes from macrolides. Under antibiotic selection pressure, random mutations, ARG acquisition via HGT, and activation of mobile DNA elements are viable mechanisms (Salehi et al., 2021). Surprisingly, the XDR A. pittii AP2044 strain most likely acquired many ARGs via HGT, resulting in high levels of resistance to the majority of the antibiotics tested (Table 1). Additionally, a total of four different IS elements next to the ARGs are found in AP2044 plasmids. Several different IS elements were identical (100%) to ISAlw125, IS17, and ISAba2 of Acinetobacter spp., which are adjacent to the aph(3′)-VIa,aph(3″)-Ib, and sul2 genes. The aac(3″)-IIb gene was located between two IS30 family elements. As a result, the AP2044 strain’s remarkable XDR capacity was related to the enormous amount of ARGs and IS elements in its genome.

Interestingly, the AP2044 strain had a high level of carbapenem resistance and carried the bla_NDM–1 gene. The presence of NDM, a widespread metallo-β-lactamase (MBL) in Acinetobacter was noteworthy (Perez et al., 2007). The increased number and diversity of MBLs in Acinetobacter spp. indicated a concerning trend in the global emergence of resistance in this pathogen (Dortet et al., 2014). In Acinetobacter, the bla_NDM-type genes were located on either the plasmid or chromosome (Wong et al., 2017). Nevertheless, the AP2044 strain carried two copies of bla_NDM–1 genes, which are located on both the plasmid and chromosome. The bla_NDM–1 gene is typically located between two copies of the ISAba125 element in NDM-producing A. baumannii, forming a composite transposon termed Tn125 (Bonnin et al., 2012, 2014; Krahn et al., 2016). Except for the insertion of an IS91 family transposon and deletion of insE, the genetic context of bla_NDM–1 on the chromosome was similar to that of Tn125. Indeed, by the late 1970s, a worldwide lineage of A. baumannii had gained resistance to conventionally known antibiotic families. Further resistance was acquired in transposon lineages in the 1980s as new antibiotics became available (Blackwell et al., 2016). Transposon Tn125 proved to be the primary vehicle for the spread of bla_NDM–1 in Acinetobacter spp. (Poirel et al., 2012). Current observations suggested that the bla_NDM–1 gene originated from an unknown environmental bacterial progenitor species and was integrated into the chromosome of Acinetobacter spp. The bla_NDM–1-bearing Tn125 transposon was most likely derived from such Acinetobacter spp. and then transferred to broad-host-range plasmids before being horizontally transferred to Enterobacteriaceae and P. aeruginosa (Nordmann et al., 2012; Bonnin et al., 2014). Of greater interest was that this NDM-1 producer carried the structure (IS30-bla_NDM–1-ble_MBL-trpF-ORF-Y-family DNA polymerase-umuD-ORF-ORF-ORF-ISAha3) surrounding the bla_NDM–1 gene (Figure 1B), which was similar to that found in pAcsw19-2 (Sichuan, Luzhou) (accession no. CP043309.1). These two plasmids originate in the same area, and the personnel mobility is substantial. Moreover, sewage is the origin of pAcsw19-2. Several investigations have suggested that it could be a major source of resistance genes as well as a hotspot for transmitting resistance genes and MGEs to clinical microorganisms. Moreover, the possibility of gene cluster transfer should be considered due to the diversity of the bla_NDM–1 gene environment. In addition, a massive resistance island in the Acinetobacter’s genome (Adams et al., 2008) could acquire additional genetic entities for resistance from other bacterial species. Genomic island GI_AP2044-4 (42,734 bp) carried the chromosome-borne bla_NDM–1. Sequence analysis showed that GI_AP2044-4 had 90% query cover and 98.8% sequence similarities with the DNA sequence of A. pittii strain ST220 chromosome (accession no. CP029610.1) genome (Supplementary Datasheet 3).

Previously, two β-lactamase genes located on the chromosome were identified by WGS. One gene encoding an ADC-25-like cephalosporinase and another gene encoding a novel OXA variant were determined to have an 89% aa identity (243/273 aa) and 100% coverage (273/273 aa) compared to OXA-213. The aa sequences of OXA enzymes are quite diverse, and a cutoff of 73.1% of aa identity has recently been proposed as a criterion for dividing OXA subfamilies (Yoon and Jeong, 2021). Therefore, the novel OXA variant belonged to the OXA-213-like subfamily. The Pathogen Detection group at GenBank’s National Center for Biotechnology Information has awarded it the number OXA-1045 (accession no. OL790815) (Evans and Amyes, 2014; Yoon and Jeong, 2021). Sequence alignment of OXA-1045 with OXA-213 revealed 30 aa changes and the secondary structure of OXA-1045 contained nine α helixes and six β sheets (Figure 2). Acinetobacter isolates have shown complex interactions with multiple mechanisms of resistance to carbapenems, and the production of naturally occurring OXAs has been the most frequently observed. The predominance of OXAs (OXA-23, OXA-24 or –40, OXA-51, OXA-58, and OXA-143) is the major reason for phenotypic resistance to carbapenems, which have been detected in many parts of the world (Adams-Haduch et al., 2011; Principe et al., 2014; Kamolvit et al., 2015; Labarca et al., 2016). Among these carbapenem-hydrolyzing OXA-type lactamases, bla_OXA–23 is regarded as an intrinsic gene of Acinetobacter radioresistens, OXA-51 is intrinsic to A. baumannii and the OXA-134 variant is intrinsic to Acinetobacter schindleri and A. lwoffii (Poirel et al., 2008; Turton et al., 2012; Périchon et al., 2014). OXA-213-like enzymes have been identified to be intrinsic to A. calcoaceticus and have been subsequently detected in A. pittii (Figueiredo et al., 2012; Tietgen et al., 2021). Phylogenetic analysis of OXA-213-like proteins identified two distinct subgroups within the OXA family. The first group was linked to A. pittii and the second group to A. calcoaceticus (Supplementary Figure 1).

The genetic environment of bla_OXA–1045 from the AP2044 strain is shown in Figure 1C. The fxsA gene, which is situated upstream of the bla_OXA–1045 gene, encoded the cytoplasmatic membrane protein. A potential redox protein-coding gene osmC-like, a transketolase protein-coding gene tkt, and an S adenosylmethionine synthase-coding gene metK were positioned upstream of fxsA. All bla_OXA–1045 downstream genes were yncA, ruvC, nnrD, and queG, which encoded the N-acetyltransferase family protein, crossover junction endodeoxyribonuclease, bifunctional NAD(P) H-hydrate repair enzyme, and epoxyqueuosine reductase, respectively. The genetic context of bla_OXA–1045 showed the closest similarity with that of bla_OXA–417 in a BLAST search, which was naturally found on the chromosome of A. pittii and belonged to the bla_{OXA–213–like} family. In addition, there was no mobile element found in the surrounding region of bla_OXA–1045. The above findings implied that the initial location of bla_OXA–1045 was in the chromosome. The presence of an IS upstream of the gene, which acted as a powerful promoter, can boost the production of OXAs (Turton et al., 2006). Hence, there was no evidence of OXA-1045 overproduction.

Among the antibiotics tested, only the MICs of ampicillin, piperacillin, cefazolin, cefoxitin, cefuroxime, ampicillin-sulbactam, and piperacillin-tazobactam for the transformant containing pET28b-OXA1045 (BL21:pET28b-OXA1045) were increased by ≥ 2-fold, as compared to those for the transformant containing pET-28b (BL21:pET28b). In comparison with OXA-213, OXA-1045 elevated the MICs of piperacillin and piperacillin-tazobactam slightly, suggesting that OXA-1045 had a greater impact on piperacillin. Noteworthy, the MICs of ceftazidime, cefotaxime, meropenem, and imipenem for both transformants remained the same as the acceptor strain, demonstrating that the OXA-213 resistance profile to cephalosporins and carbapenem was similar to that of OXA-1045 (Table 4). A previous study has illustrated that the production of all OXAs led to a significant increase in all carbapenem MICs in A. baumannii, while no elevation in MICs was observed in E. coli (Tietgen et al., 2021). Indeed, we could not rule out the possibility that the effect on MICs is a result of endogenous OXA cooperation with vector-expressed OXA-213-like variants in Acinetobacter spp. The results indicated that carbapenemase activity of tested OXAs may be related to the species-dependent effect.

The biofilm formation capacity was measured in the Lab-WT and AP2044 strains. The OD₅₇₀ values for the Lab-WT and negative control were 0.84 ± 0.12 and 0.14 ± 0.008, respectively. The OD₅₇₀ value for the AP2044 strain was 0.177 ± 0.045, which was a weak biofilm producer. Moreover, the SEM result was consistent with the OD₅₇₀ values obtained by crystal violet staining (Figures 3A,B). The AP2044 strain presented mucoid phenotype, with moist colonies and an elevated surface (Supplementary Figure 2). Mucoid phenotype formation may influence the virulence of pathogenic microorganisms to varying degrees, which has allowed to make significant strides in characterizing the determinants of pathogenic mechanisms in P. aeruginosa and Klebsiella pneumoniae (Dennis et al., 2018; Ding et al., 2022). A previous study has demonstrated that mucoid A. baumannii strains were more virulent than non-mucoid isolates (Shan et al., 2021). Therefore, we analyzed the virulence of AP2044 strain by developing a Galleria mellonella infection model. As shown in Figure 3C, such virulence of AP2044 strain was comparable to that of the Lab-WT, which is well known for its lack of virulence (Khalil et al., 2021). The association between the virulence and mucoid phenotype in Acinetobacter spp. warrant further investigation. Indeed, the capacity of A. baumannii to form biofilm facilitated its survival and persistence in hospital environments (Donlan and Costerton, 2002; Gaddy et al., 2009). This, in turn, contributed to the extensive spread of this pathogen across the globe. Many virulence factors have been implicated in the initial adhesion process of biofilm (Zeighami et al., 2019). Likewise, biofilm development is one of the basic virulence traits of clinical isolates (Khalil et al., 2021). Mahmoud et al. reported biofilm formation as a potent virulence factor in A. baumanni, with the strong biofilm producers exhibiting a much greater ability to kill G. mellonella larvae than the moderate and weak biofilm producers (Khalil et al., 2021). Accordingly, AP2044 strain was a weak biofilm producer and a low virulence strain. Several previous studies have demonstrated that the majority of XDR and pandrug-resistant (PDR) clinical isolates were weak or non-biofilm producers, which is consistent with the present findings (Qi et al., 2016; Li et al., 2021). There may be a metabolic cost caused by high-level antibiotic resistance, which has been shown to cause a decrease in virulence (Roux et al., 2015). However, the emergence of carbapenem-resistant hypervirulent A. baumannii (CR-hvAB) strains presents significant challenges for public health and infection control (Li et al., 2020).

Compared to the reference strain, quantitative analysis demonstrated that AP2044 expressed 1. 54-, 0. 67-, and 2.05-fold more adeB, adeJ, and adeG genes, respectively (Figure 3D). In particular, the transcription level of adeB and adeJ in AP2044 was significantly overexpressed than that in TSAP (t-test, P < 0.01). The efflux pump plays a vital role in both biofilm formation and antibiotic resistance, particularly in tigecycline resistance (Lee et al., 2020). The tigecycline is one of the last resort options for XDR strain infection treatment (Wong et al., 2017). Acinetobacter has shown superior resistance to almost all available systemic antibiotics and demonstrates an XDR phenotype. Therefore, overcoming antibiotic resistance is the primary challenge of treating Acinetobacter infections (Wong et al., 2017). AdeABC in particular has been demonstrated to influence antibiotic sensitivity and to contribute to tigecycline resistance (Ruzin et al., 2007; Roca et al., 2011). The AdeRS two-component system, which consists of a sensor kinase and a response regulator, is in charge of expressing the transcription of the AdeABC efflux pumps. The aa changes or IS element insertion in the AdeRS two-component system can boost the transcription level of AdeABC efflux pumps (Yoon et al., 2013; Lucaßen et al., 2021). AdeIJK is regulated by the TetR-like repressor AdeN, whose overproduction results in antibiotic resistance and contributes to tigecycline resistance. The present findings indicated that AdeABC and AdeIJK overexpression was the cause of tigecycline resistance, which is consistent with previous studies. A previous study have demonstrated that the AdeABC and AdeIJK efflux systems contributed to tigecycline resistance in a synergistic manner (Damier-Piolle et al., 2008).