Acinetobacter baumannii clinical strain 13A297 was isolated in 2013 from a urine sample collected from a patient hospitalized in France. Species identification was performed by MALDI-TOF mass spectrometry and confirmed by whole genome sequencing. Clinical A. baumannii strains 19A2597 and 13A297, producing the NDM-1 and NDM-9 β-lactamase, respectively, were selected for cloning of the respective genes into the pABEC vector and then transferred into E. coli DH5α and A. baumannii CIP70.10 (Vuillemenot et al., 2022). Mating out assays were carried out using the rifampicin-resistant BM4547 strain derived from CIP70.10 as the recipient strain (Potron et al., 2014). The transconjugants were selected on Mueller-Hinton agar plates supplemented with ticarcillin (100 mg/L) and rifampicin (100 mg/L). E. coli BL21 (DE3) (Novagen, Fontenay-sous-Bois, France) was used for overproduction of NDM-1 and NDM-9 enzymes.

MICs of β-lactams (ticarcillin, piperacillin-tazobactam, ceftazidime, cefepime, imipenem, meropenem), ciprofloxacin, trimethoprim-sulfamethoxazole, amikacin, tigecycline, minocycline and colistin were determined by the broth microdilution method in cation-adjusted Mueller-Hinton broth (MHB), by using customized microplates (ThermoFisher, Illkirch-Graffenstaden, France). Cefiderocol MICs were determined using an Iron-Depleted and Cation-Adjusted Mueller-Hinton Broth (ID-CAMHB) and a titrated powder of cefiderocol (MedChem Express) according to the recommendations of the Clinical and Laboratory Standards Institute (2020). MIC values were interpreted according to the current EUCAST or CLSI (for minocycline and tigecycline) clinical breakpoints, except for cefiderocol, for which the EUCAST pharmacokinetic-pharmacodynamic (PK-PD) breakpoint (S ≤ 2; R > 2 mg/L) was applied [Clinical and Laboratory Standards Institute, 2020; The European Committee on Antimicrobial Susceptibility Testing (EUCAST), 2021].

Total DNA of strains 13A297 and 19A2597 was purified (NucleoSpin Microbial DNA, Macherey-Nagel, Germany), and then quantified spectrophotometrically (NanoDrop, Thermo Fisher Scientific, France). Preparation of DNA libraries and whole-genome sequencing experiments were outsourced to services using Illumina platforms generating paired-end reads (P2M platform, Institut Pasteur, Paris, France). In addition, long-reads sequencing using the Oxford Nanopore technology was performed for the strain 13A297 (MinION, R9.4.1 flow cell). Short and long reads were assembled as hybrid using SPAdes v.3.15.5 and Trycycler v.0.5.0. Quality control of reads, assembly, and genome annotation were carried out using FastQC v.0.11.9, QUAST v.5.0.2, and RASTtk on the PATRIC 3.6.9 platform. Antimicrobial resistance genes (AMR) were identified by the National Reference Center for Antimicrobial Resistance (NRC-AR) pipelines and databases. The sequence type (ST) of isolates was determined according to the MLST schemes available at PubMLST.¹

From genomic DNA extracted using the QIAamp DNA mini kit (Qiagen, Courtaboeuf, France), the bla_NDM-1 and bla_NDM-9 genes were amplified with the specific primers NDM_fwd (5’-CAAGCTTGGTACCGAGCTCGTGGCTTTTGAAACTGTCGC-3′) and NDM_rev (5’-ACTGGCGGCCGTTACTAGTGTCGAGGTCAGGATAGGGG-3′) and subsequently cloned into the shuttle vector pABEC using the NEBuilder Hifi assembly kit according to the manufacturer’s recommendations (New England Biolabs, Evry, France) (Vuillemenot et al., 2022).

For the production of NDM-1 and NDM-9 proteins, the complete coding sequences of the β-lactamase genes were amplified by PCR using the specific primers NdeI-NDM_29-270 (5′-AAAAACATATGGGTGAAATCCGCCCGACG-3′) and NDM-XhoI_{delta STOP} (5′-AAAAACTCGAGGCGCAGCTTGTCGGCCAT-3′). The amplicons and the plasmid pET41b were digested with the restriction enzymes NdeI and XhoI before ligation. The recombinant plasmids pET41b-NDM-1 and pET41b-NDM-9 were transferred by transformation into E. coli BL21 (DE3).

Two liters of brain heart infusion broth medium containing 30 mg/L kanamycin were inoculated with 2 mL overnight cultures of E. coli strains BL21 (DE3) pET41b-NDM-1 and of E. coli BL21 (DE3) pET41b-NDM-9. Production of recombinant proteins harboring the His tag was induced overnight at 22°C with 0.2 mM IPTG. Then, the cultures were centrifuged at 6,000 g for 15 min and the pellets were resuspended with the binding buffer (10 mM imidazole, 25 mM sodium phosphate pH 7.4 and 300 mM NaCl). Bacterial cells were lysed by sonication and the pellet was removed by two consecutive centrifugation steps at 10,000 g for 1 h at 4°C; the supernatant was then centrifuged at 96,000 g for 1 h at 4°C. The soluble fractions were filtered, passed through a HisTrap^TM HP column (GE Healthcare) and proteins were eluted with the elution buffer (500 mM imidazole, 25 mM sodium phosphate pH 7.4 and 300 mM NaCl). Finally, gel filtration was performed with 50 mM sodium phosphate buffer pH 7 and 150 mM NaCl with a Superdex 75 column (GE Healthcare). The fractions were pooled according to purity by SDS–PAGE, dialyzed in 100 mM sodium phosphate pH 7, 50 μM ZnSO₄, and concentrated using Vivaspin columns. The concentration was determined by measuring the OD at 280 nm and with the extinction coefficients obtained from the ProtParamtool (Swiss Institute of Bioinformatics online resource portal).

Kinetic parameters were determined using purified NDM-1 and NDM-9 β-lactamases in 20 mM Hepes pH 7.5 supplemented with 50 μM ZnSO₄. The k_cat and K_m values were determined by analyzing β-lactam hydrolysis under initial-rate conditions with an ULTROSPEC 2000 UV spectrophotometer with the SWIFT II software (GE Healthcare, Velizy-Villacoublay, France) and the Eadie-Hofstee linearization of the Michaelis–Menten equation.

Three-dimensional structure of NDM-9 was generated by in silico mutagenesis from the structure of NDM-1 (PDB code 4RL0) (Feng et al., 2014) using UCSF Chimera (version 1.14) (Pettersen et al., 2004). Molecular docking of cefiderocol (open form) into the active sites of both NDM-1 and NDM-9 was performed using CCDC Gold (part of CSD 2021 software) (Verdonk et al., 2003) with the GoldScore scoring function and the binding site defined as a sphere with 20 Å radius around Zn301.

The whole genome sequences generated in the study have been submitted to the Genbank nucleotide sequence database under the accession number PRJNA989410.

Acinetobacter baumannii 13A297 strain was resistant to all β-lactams, including cefiderocol (MIC >32 mg/L), aminoglycosides, fluoroquinolones, trimethoprim-sulfamethoxazole and remained susceptible to colistin and tetracyclines (Table 1). WGS sequencing of A. baumannii 13A297 revealed a genome of 4,229,816 bp with a mean coverage of 280 X. Analysis of WGS data revealed that this isolate possessed the naturally-occurring bla_OXA-64 and bla_ADC-26 genes. Its resistance profile to β-lactams was the result of acquisition of the carbapenemase NDM-9, that differs from NDM-1 by a single amino-acid substitution (Glu152Lys). NDM-9 was first reported in a clinical Klebsiella pneumoniae isolate from China in 2013 (Wang et al., 2014). This enzyme was also identified in a mcr-1-harboring Escherichia coli strain, in three environmental Klebsiella variicola isolates in South Korea (Yao et al., 2016; Di et al., 2017) and in a K. pneumoniae strain isolated from wastewater in Switzerland in 2015 (Nüesch-Inderbinen et al., 2018). Since then, NDM-9 was sporadically detected in clinical, environmental or animal strains belonging to various Enterobacterales species including Salmonella sp. in Asia and Europe (Italy), but also in Africa (Tunisia) and in French Polynesia (Falcone et al., 2020; Li et al., 2021; Oueslati et al., 2021; Song et al., 2022; Sun et al., 2023; Zouaoui et al., 2023). MLST assigned A. baumannii 13A297 strain to ST^Pas25, and ST^Ox2280, according to the MLST Pasteur and Oxford schemes, respectively. In addition to the globally disseminated multidrug-resistant ST^Pas1 and ST^Pas2, ST^Pas25, which belongs to international clone 7 (IC7), has emerged worldwide as epidemic, endemic or sporadic cases in the last two decades (Sahl et al., 2015; Rodríguez et al., 2018; Rafei et al., 2022). A hallmark of this clone is its evolution toward multidrug resistance and a high capacity for biofilm formation (Rafei et al., 2022). In addition, it has been associated with high mortality rates among infected patients (da Silva et al., 2018).

Analysis of Nanopore sequencing data showed that the bla_NDM-9 gene was located on a full-length Tn125 composite transposon, carried by a 166,254-bp long plasmid named pNDM-9. A 3-bp target site duplication (TTT) identified on both sides of Tn125 was indicative of a transposition event. Interestingly, the genetic environment of bla_NDM-9 in A. baumannii 13A297 corresponded to the original Tn125 and differed from that identified in Enterobacterales. Indeed, IS26 was associated with bla_NDM-9 in all but one of the plasmids identified in Enterobacterales, highlighting that IS26 may play a key role in the transfer of the bla_NDM-9 gene between Enterobacterales (Li et al., 2021). Plasmid pNDM-9 encoded 165 open reading frames and possessed a GC content of 39.1%. It exhibited significant similarity (99.9%) to 75% of the sequence of plasmid pIOMTU433 (accession number: AP014650) (Figure 1). Plasmid pIOMTU433 is a 189,354 bp long plasmid identified in an A. baumannii clinical strain isolated in Nepal in 2013. Interestingly, pIOMTU433 did not carry a bla_NDM-like gene but a bla_PER-7 ESBL-encoding gene. In addition to bla_NDM-9, pNDM-9 also carried other resistance genes, such as sul2, aph(3″)-Ib and aph(6)-Id, which confer resistance to sulfonamides and aminoglycosides, respectively. Furthermore, the pNDM-9 plasmid shared 99.9% nucleotide identity with 90% of the sequence of the bla_NDM-1 positive plasmid pPM194122 (150,385 bp-long, accession number CP050426), which was sequenced from an A. baumannii strain from India (Figure 1). Consistent with the identification of an incomplete tra operon, attempts to transfer the pNDM-9 plasmid by mating-out assays were unsuccessful.

To clarify the impact of NDM-9 on cefiderocol susceptibility and to compare it to that of NDM-1, both corresponding bla_NDM-1 and bla_NDM-9 genes were cloned and expressed in E. coli and A. baumannii reference strains. NDM expression from the plasmid pABEC in E. coli DH5α conferred resistance to ceftazidime, imipenem, meropenem, and cefiderocol for NDM-9 (8 mg/L) and a decrease in susceptibility to cefiderocol (at least 64-fold) for NDM-1 (2 mg/L), using the pharmacokinetic-pharmacodynamic (PK-PD) cut-off of 2 mg/L (S ≤ 2 mg/L; R > 2 mg/L) (Table 1). Interestingly, DH5α cells transformed with bla_NDM-9-carrying vector pABEC-NDM-9 were 4-fold more resistant to cefiderocol and at least 2-fold to meropenem than those containing pABEC-NDM-1. In agreement with this result, the transfer of the vector pABEC-NDM-9 into the A. baumannii CIP70.10 strain led to a higher level of resistance to cefiderocol (MIC >32 mg/L) than that of pABEC-NDM-1 (MIC = 4 mg/L) (Table 1).

NDM-9 was previously shown to possess slightly increased hydrolytic activity toward meropenem, imipenem and cefotaxime as compared to NDM-1 (Wang et al., 2014). To gain insight into the structural features that may explain the different levels of resistance of NDM-1 and NDM-9 to cefiderocol, both enzymes were purified and steady-state kinetic parameters of ceftazidime and cefiderocol were determined. NDM-9 exhibited a 7-fold higher hydrolytic efficiency (k_cat/K_m) toward ceftazidime, which is linked to a higher turnover rate (k_cat) of this antibiotic by NDM-9 (Table 2). Although weak, a slight hydrolysis of cefiderocol by NDM-1 was evidenced. As compared to NDM-1, NDM-9 possessed a 3-fold higher affinity for cefiderocol, which was associated with a 2-fold increase in its catalytic efficiency (Table 2).

Molecular modeling revealed stabilizing interactions between cefiderocol and residues Lys211, Lys216, Asn220 of NDM-9, and a strong ionic interaction of Lys152 with both Asp223 and the carboxylate group of R1 substituent in cefiderocol (Figure 2). In NDM-1, this latter favorable interaction is replaced by a repulsive interaction with Glu152, leading to a lower affinity for cefiderocol. Altogether, these results explain the higher cefiderocol MIC and catalytic efficiency observed for NDM-9 as compared to NDM-1.