Genomic Analysis of a Strain Collection Containing Multidrug-, Extensively Drug-, Pandrug-, and Carbapenem-Resistant Modern Clinical Isolates of Acinetobacter baumannii

ABSTRACT In this study, we characterize a new collection that comprises multidrug-resistant (MDR), extensively drug-resistant (XDR), pandrug-resistant (PDR), and carbapenem-resistant modern clinical isolates of Acinetobacter baumannii collected from hospitals through national microbiological surveillance in Belgium. Bacterial isolates (n = 43) were subjected to whole-genome sequencing (WGS), combining Illumina (MiSeq) and Nanopore (MinION) technologies, from which high-quality genomes (chromosome and plasmids) were de novo assembled. Antimicrobial susceptibility testing was performed along with genome analyses, which identified intrinsic and acquired resistance determinants along with their genetic environments and vehicles. Furthermore, the bacterial isolates were compared to the most prevalent A. baumannii sequence type 2 (ST2) (Pasteur scheme) genomes available from the BIGSdb database. Of the 43 strains, 40 carried determinants of resistance to carbapenems; blaOXA-23 (n = 29) was the most abundant acquired antimicrobial resistance gene, with 39 isolates encoding at least two different types of OXA enzymes. According to the Pasteur scheme, the majority of the isolates were globally disseminated clones of ST2 (n = 25), while less frequent sequence types included ST636 (n = 6), ST1 (n = 4), ST85 and ST78 (n = 2 each), and ST604, ST215, ST158, and ST10 (n = 1 each). Using the Oxford typing scheme, we identified 22 STs, including two novel types (ST2454 and ST2455). While the majority (26/29) of blaOXA-23 genes were chromosomally carried, all blaOXA-72 genes were plasmid borne. Our results show the presence of high-risk clones of A. baumannii within Belgian health care facilities with frequent occurrences of genes encoding carbapenemases, highlighting the crucial need for constant surveillance.


MDR, XDR, and PDR A. baumannii Modern Clinical Isolates
Antimicrobial Agents and Chemotherapy which belongs to the rare ST604, which was first identified in Egypt (16). Isolates representing other less frequently detected STs (one isolate per ST) are of ST215 (AB231-VUB) and ST158 (AB32-VUB). Two isolates of ST85 (AB177-VUB and AB186-VUB) and two isolates of ST78 (AB21-VUB and AB40-VUB) did not cluster with any of the major branches ST2, ST1, and ST636 in the phylogenetic tree ( Fig. 1), showing their distinct genetic backgrounds.
Concerning the geographical repartition of the different ST identified in our collection, we have detected six isolates of ST636 which have been described to cause outbreaks within hospital settings in Serbia and Colombia (17,18). A. baumannii ST215 has been common in Thailand since 2010 (19), while GES-producing A. baumannii ST158 caused an outbreak in a Tunisian neonatal unit and was linked to a GES-producing clone from the Middle East; it has also been identified in Denmark (20). ST78 (AB21-VUB and AB40-VUB) was recently detected in Russia as an uncommon clone known as "Italian clone." Indeed, it was reported from several Italian hospitals in 2010, and since then, it has been detected from other Mediterranean countries, the United States, Germany, Kuwait, and French Guiana, pointing toward successful global dissemination (21). ST85 is represented by two isolates (AB177-VUB and AB186-VUB), yet only AB177-VUB possesses both bla NDM-1 and bla OXA-94 . A. baumannii ST85 possessing the bla NDM-1 and bla OXA-94 genes was previously detected in France, Algeria, Turkey, Syria, Tunisia, and, recently, also Spain (22). AB186-VUB possesses bla OXA-94 but not bla NDM-1 , pointing toward geographical unrelatedness of AB177-VUB and AB186-VUB.
The comparison of our 43 A. baumannii isolates from this study with 603 wholegenome sequences of A. baumannii ST2 obtained from BIGSdb showed great variety and a distribution of A. baumannii ST2 across the world (Fig. 7 and Fig. S1 and S2). The relatedness of the isolates was assessed based on single-nucleotide polymorphisms (SNPs) in coding regions, with threshold for a clonal isolate set for #10 as described before (23,24). The complete overview of SNP distances can be found in the SNP matrix in Table S2. Only two isolates (AB189-VUB and AB222-VUB) met this criterion of the relatedness. Isolate AB189-VUB can be clonally linked to 31 genomes of A. baumannii from the United States (n = 28) and France (n = 1) and of unknown origin (n = 2) (Table  S2). On the other hand, strain AB222-VUB is clonally related to 36 genomes of A. baumannii ST2, from the United States (n = 33) and France (n = 1) and of unknown origin (n = 2) (Table S2). While the majority of clonally related strains were the same for both AB189-VUB and AB222-VUB, two isolates from the United States were specific for AB189-VUB and nine isolates from the United States were specific for AB222-VUB. The complete overview on the origin of publicly available A. baumannii ST2 (Pasteur) from BIGSdb can be found in Table S3.
The fact that these isolates were detected in Belgium points toward their persistence and successful global dissemination, especially in the case of the clones of AB189-VUB and AB222-VUB, which were detected in the genomes of isolates from the United States and France. However, specific routes of transmission cannot be established in this study, and certain bias of the sequencing capacity of each country is also present. Since none of the isolates harbored the mcr gene, we have explored the genetic background of the isolates for mutations in the two-component lipid A-encoding system pmrAB. Only isolate AB173-VUB harbored a substitution in pmrB T235I while pmrB T235N was described to provide resistance to colistin (25), possibly providing the same colistinresistant phenotype. We have also examined interruption of the Lpx pathway as a possible cause of colistin resistance; however, the genes lpxACD were intact, suggesting that this mechanism was not present in the studied set of isolates. However, other factors such as outer membrane asymmetry or efflux pumps might be involved (26). Recent data from 30 European countries showed that 4% of the tested CRAb isolates were resistant to colistin, with the majority originating in southern Europe (Greece and Italy) (4).
Despite a limited number of isolates in this study, our findings provide important epidemiological data for Belgium, since most of the data related to MDR A. baumannii and CRAb in Belgium were published before 2010 (27)(28)(29).
The data described here provide an insight in the genotype and phenotype of MDR, XDR, and PDR A. baumannii from Belgian hospitals. Carriage of determinants of resistance to carbapenems on mobile genetic elements such as plasmids enables horizontal gene transfer, for which several A. baumannii isolates are naturally competent, and further spread of carbapenem resistance. Our study demonstrates the wide distribution of internationally disseminated MDR, XDR, and PDR clones of A. baumannii in Belgian health care facilities and also shows their detection throughout several years in America, especially in the United States. These strains pose a serious health issue to patients, especially those admitted to high-risk wards such as the intensive care units, and have the potential to cause nosocomial infections and difficult-to-control outbreaks.

MATERIALS AND METHODS
Bacterial isolates. A collection of 43 nonredundant clinical A. baumannii isolates (Fig. 1) collected across Belgium was provided by the National Reference Center (NRC) Laboratory for Antibiotic-Resistant Gram-Negative Bacilli (CHU UCL Namur, Yvoir, Belgium), which acquired these isolates to confirm and characterize carbapenem resistance mechanisms. All isolates were confirmed as A. baumannii by matrixassisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometry (MALDI Biotyper; Bruker Daltonics).

MDR, XDR, and PDR A. baumannii Modern Clinical Isolates
Antimicrobial Agents and Chemotherapy instructions (Erba Lachema, Brno, Czech Republic) in triplicates. The results were evaluated according to the CLSI (30,31). In order to evaluate multidrug-resistant (MDR), extensively drug-resistant (XDR), and pandrug-resistant (PDR) phenotypes (32), susceptibility to tetracycline was tested too, using Etest (bioMérieux). The susceptibility to tetracycline was tested only in isolates carrying the tet(B) gene, for which resistance to tetracycline would alter the phenotype from XDR to PDR. Two isolates (AB14-VUB and AB189-VUB) not carrying genes conferring resistance to tetracycline were included as a negative controls.
Whole-genome sequencing. A total of 43 clinical isolates were subjected to whole-genome sequencing (WGS) using short-read (Illumina) and long-read sequencing (Nanopore) and de novo assembly of the draft genomes. The subcultured isolates were used for DNA extraction and following independent sequencing and bioinformatical analyses. Seeing the high genomic dynamics of A. baumannii bacteria, we followed the nomenclature in the field (33) by renaming the subcultured strains by adding "-VUB," although these strains are a priori identical or very similar.
For the short-read sequencing, the genomic DNA was extracted using the phenol-chloroform method. Stationary-phase bacteria (2 mL) at an optical density at 600 nm (OD 600 ) of 4 were centrifuged for 1 min at 12,000 Â g and resuspended in 200 mL of breaking buffer (2% Triton X-100, 1% SDS, The supernatant was transferred into a new Eppendorf tube and 400 mL of phenol-chloroform was added. After centrifugation, the aqueous layer was transferred to a new recipient tube and 1 mL of 100% ethanol was added, mixed, and centrifuged for 3 min at 12,000 Â g. The supernatant was then removed, the pellet was resuspended with 400 mL of TE buffer, and 30 mL of 1-mg/mL RNase was added. After incubation for 15 min at 37°C, 10 mL of 4 M ammonium acetate was mixed, then 1 mL of ethanol 100% was added. After centrifugation (5 min at 12,000 Â g), the pellet was resuspended in 100 mL of TE buffer and the final DNA concentration was determined by spectrophotometry. The sequencing libraries were prepared using Nextera XT and subjected to 2 Â 250-bp paired-end sequencing on MiSeq (Illumina) using V3_600 kit. The fastq files were generated and demultiplexed using bcl2fastq (Illumina).
The DNA for long-read MinION (Oxford Nanopore Technologies [ONT]) sequencing was extracted using Genomic-tip 100/G (Qiagen, Hilden, Germany). The long-read sequencing libraries were prepared using a 1D ligation barcoding kit (SQK-LSK109 and EXP-NBD104; ONT, Oxford, UK). Samples were quality controlled using Qubit (double-stranded DNA [dsDNA] broad range (BR) chemistry; Thermo Fisher Scientific) and Fragment Analyzer (Agilent Technologies; using a DNF-464 kit). The average size of the fragments was 45 to 70 kb. Samples were equimolarly pooled and 12 samples were run per sequencing run which was always 2Â reloaded. MinION flow cells had a minimum of 1,200 sequenceable pores at the start, and initial loading was approximately 35 fmol followed by 2 reloads each after 24 h of sequencing. The sequencing was performed on a MinION Mk1b instrument (ONT) using R9.4.1 (FLO-MIN106) flow cells.
Genotypic characterization. The assembled draft genomes were subjected to multilocus sequence typing (MLST) using mlst (https://github.com/tseemann/mlst) employing the PubMLST database (37) based on the Pasteur and Oxford schemes. Two isolates (AB21-VUB and AB179-VUB) were of a novel ST Oxford and were deposited to PubMLST database and assigned a new ST. The resistance genes were detected using ABRicate (https://github.com/tseemann/abricate) employing ResFinder (38) with a 95% threshold for both identity and query coverage. The point mutations were characterized using the BLAST algorithm and Geneious R9 (Biomatters, New Zealand). The genetic environment was assessed using Mobile Element Finder by the Center for Genomic Epidemiology (39).
Phylogenetic analysis. The maximum likelihood tree depicting the relatedness of the isolates was constructed from assembled draft genomes using precited open reading frames obtained by Prokka (40) as an input for the core genome alignment created using Roary (41). RAxML (42) was used for calculation of the phylogenetic tree using general time reversible with optimization of substitution rates under the GAMMA model of rate heterogeneity method supported by 500 bootstraps. The phylogenetic tree was visualized and completed with metadata in iTOL (43).
Comparison with publicly available genomes. The most clinically relevant isolates belonging to worldwide-spread ST2 (Pasteur scheme) were compared to genomes of A. baumannii ST2 available in BIGSdb (44). The search was performed on 14 June 2022 and resulted in 607 hits for A. baumannii ST2 containing sequencing data. Out of 607 entries, 4 were excluded after not passing the ST verification using in silico MLST by mlst (https://github.com/tseemann/mlst). All 43 isolates from our study were involved. The annotation and core genome alignment were performed using Prokka and Roary as described above. SNPs were extracted from the core genome alignment using snp-sites (https://github .com/sanger-pathogens/snp-sites), and the phylogenetic tree was constructed using RAxML under the GTRGAMMA model supported by 100 bootstraps. The minimum spanning tree (MST) was visualized using GrapeTree (45). The relatedness of the isolates was assessed based on core genome SNP count obtained using snp-dists (https://github.com/tseemann/snp-dists) with a cutoff value of #10 for clonal relationship as described before (23,24).
Data availability. The draft and complete assemblies with the short-and long-read sequencing reads were deposited in GenBank under BioProject PRJNA734485, PRJNA701627, and PRJNA798866.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. This study was supported by the Flanders Institute for Biotechnology (VIB). This project received funding from the European Union's Horizon 2020 research and innovation program under Marie Sklodowska-Curie grant agreement no. 748032.
We have no conflicts of interest to declare.