Ethical approval for the study was granted by the Ethics and Research Committee of the Obafemi Awolowo University Teaching Hospitals Complex, Ife, Nigeria with protocol number ERC/2017/06/13. Participants provided written informed consent before voluntarily participating in the study. Patient confidentiality and anonymity were maintained. Only de-identified patient metadata with no traceability to patients was collected and analyzed.

A prospective observational study was conducted to determine the colonization and transmission of CRAB among all new patients admitted into the ICU of the Obafemi Awolowo University Teaching Hospitals Complex (OAUTHC), Ife, Osun State, Nigeria between August 2017 and June 2018. A total of 108 patients were recruited. Rectal swabs were collected from each patient within 48 h of ICU admission and, thereafter, weekly until exit using the protocol for active surveillance cultures by the US Centres for Disease Control and Prevention (18). Acquisition rate was defined as the percentage of patients who acquired CRAB that was absent on admission. CRAB was isolated from the rectal swab samples and preliminarily identified using standard microbiological procedures, and then cryopreserved at −80°C. We also verified all A. baumannii isolates from human specimens and associated metadata retrospectively submitted to the then-new Nigerian antimicrobial resistance (AMR) surveillance system by OAUTHC, Ife, Osun State and from the University College Hospital (UCH), Ibadan, located in the adjacent Oyo State. The identities of all presumptive Acinetobacter species were verified by culture on CHROMagar™ Acinetobacter media with CHROMagar MDR Supplement CR102 (CHROMagar, Paris, France) and identification on a VITEK 2 automated system (bioMérieux, Inc., Marcy-l’Étoile, France) following manufacturer instructions using GN ID (reference number: 21341) cards. Antimicrobial susceptibility testing was also carried out on the Vitek 2 automated system using the AST N281 (reference number: 414531) cards. Antibiotics tested were ticarcillin/clavulanic acid, piperacillin/tazobactam, ceftazidime, cefoperazone/sulbactam, cefepime, doripenem, imipenem, meropenem, gentamicin, ciprofloxacin, levofloxacin, minocycline, tigecycline and trimethoprim/sulfamethoxazole. Minimum inhibitory concentration values were interpreted according to the Clinical Laboratory Standards Institute guidelines (19).

Genomic DNA extraction was carried out using the FastDNA Spin Kit for Soil (MP Biomedicals, Irvine, CA, United States) with protocols modified for bacterial genomic DNA extraction. The Acinetobacter isolates were grown overnight in Tryptone Soy Broth (Oxoid, Basingstoke, United Kingdom) and centrifuged to obtain the cell pellets at 6,000 revolutions per minute for 5 min. The pellet was then used for extracting the DNA following the manufacturer’s instructions. Quantification of extracted DNA was done using the Qubit™ dsDNA BR Assay Kit (Invitrogen, Waltham, MA, United States). Library preparation was done using the Covaris LC220 for fragmentation, and the NEBNext Ultra II FS DNA library kit for Illumina with 384-unique indexes (New England Biolabs, Ipswich, MA, United States). The double-stranded DNA libraries (avg. 500 bp) were sequenced using the HiSeq X10 with 150 bp paired-end chemistry (Illumina, San Diego, CA, United States).

All sequence analyses, except where otherwise stated, were carried out as described in the Global Health Research Unit (GHRU) protocol¹ using publicly available Nextflow pipelines. De novo assembly, species identification, and quality control were carried out using the De novo assembly pipeline with default parameters. Quality checks included total bases between 3,340,530–4,776,219 megabase pairs (Mbp), contamination levels <5%, N50 scores >25000, and number of contigs <300; assemblies that “failed” any of these checks were excluded from the downstream analyses.

To determine evolutionary relationships among the isolates, single nucleotide polymorphism (SNP) phylogeny analysis was conducted using the SNP phylogeny pipeline with default parameters. Briefly, Bactinspector² was used to select the closest reference to the A. baumannii sequences (Genbank accession: GCA_000830055.1). Reads were mapped to the reference, and variants were called, filtered and concatenated into pseudogenomes. Afterward, the pseudogenomes were aligned and a maximum likelihood tree was constructed from the aligned pseudogenomes with the GTR + G model and 1,000 bootstraps. To determine the clonality of isolates within suspected outbreak clades, we re-selected a reference sequence (NZ_CP016298.1) more closely related to the strains of interest, computed pseudogenome alignments as previously described and calculated the pair-wise SNP distances between the strains based on the aligned pseudogenomes using FastaDist³.

Multi-locus sequence types (MLST) of the strains were determined in silico as described in the GHRU protocol based on the Oxford and Pasteur MLST schemes (20, 21). The goeBURST software⁴ was used to assign the predicted sequence types (STs) to IC groups based on locus similarities to known international clones (22). The acquired antimicrobial resistance determinants harbored by each of the isolates were also predicted in silico as described in the aforementioned GHRU protocol using the Ariba software and the National Center for Biotechnology Information’s (NCBI) Bacterial Antimicrobial Resistance Reference Gene Database⁵. Only predicted genes tagged by the ariba software as “yes” or “yes_nonunique” were regarded as present in the genome.

Patient clinical data was entered into WHONET 5.6⁶ and cleaned on Microsoft Excel^® (Microsoft Corporation, Redmond, WA, United States) spreadsheet. Data was then analyzed using Statistical Package for the Social Science (SPSS^®, IBM Corp., Armonk, NY, United States) version 20. A descriptive statistical analysis was carried out to summarize the demographic data and results were presented as frequency distribution, percentages, mean, and standard deviation. Risk factors for CRAB-colonization/infection were assessed using bivariate analysis with Chi-square and multivariate analysis with logistic regression. P-values <0.05 were considered statistically significant.

Rectal swab samples were obtained from 108 patients admitted to the ICU at OAUTHC between August 2017 and June 2018. Carbapenem-resistant A. baumannii (CRAB) was recovered from 20 (18.5%) patients. The acquisition rate was 8.3% (8/96), while 12 (11.1%) patients were positive for CRAB within 48 h of admission. Patients that acquired CRAB had seven times the odds of subsequent bloodstream infection (OR = 7.41; 95% CI 2.39–22.92). The mortality rate among CRAB-colonized patients was 50% and the odds of death was almost two times higher among the CRAB-colonized patients compared to patients with no CRAB colonization (OR = 1.84, 95%; CI = 0.69–4.90).

A total of 36 bacterial genomes were analyzed; 20 were isolated from the ICU at OAUTHC between 2017 and 2018 and were part of the suspected rectal colonization outbreak, while the remaining 16 were retrospective isolates submitted to Nigeria’s AMR surveillance system (2017 – 2019) by OAUTHC and UCH. The isolates were identified as A. baumannii (34), A. nosocomialis (one), and A. pittii (one) based on their whole-genome sequences.

The 34 A. baumannii strains segregated into clear evolutionarily distinct lineages, with isolates in each lineage belonging to identical STs. Half (17/34; 50.0%) of the isolates belonged to the two major clades observed (clade 1 and clade 2) and three different STs: ST1114 and ST1841 (which we later found to be the same ST but were artifactually different due to a duplicated gdhB gene in the strains) in clade 1, and ST1089 in clade 2 (Figure 1).

All nine clade 1 (IC2) strains and five of the eight clade 2 strains were from the OAUTHC ICU. Based on phylogenetic relatedness, timeline of isolation and other epidemiological data, we hypothesize that the clade 1 isolates were part of an outbreak in the ICU, while the others, including those in clade 2, were endemic circulating strains isolated during the same period. After the isolation of the first clade 1 strain on 2nd January 2018, seven more identical strains belonging to this clade were isolated within 3 weeks (between 17th March 2018 and 9th April 2018) of each other from patients in the same ICU. Conversely, and further supporting our hypothesis of repeated introduction, the five clade 2 strains from the ICU were isolated over 42 weeks. The intra-clade SNP distances between the isolates ranged from 0–213 SNPs. Also, one of the clade 2 strains was isolated at another tertiary facility in a different state in southwest Nigeria; the date of isolation was not available for this as well as most of the other non-ICU isolates.

Clade 1 isolates all had an identical AMR and virulence gene composition and resolved into two sub-clades – clade 1A (two isolates) and clade 1B (seven isolates). Each sub-clade had a maximum within-clone distance of two SNPs. The SNP distances between the two sub-clades were ∼48 SNPs. A previous study reported a threshold of ∼2.5 core-genome SNPs for distinguishing outbreak and non-outbreak A. baumannii strains (24). We conclude that clade 1B isolates were part of a definite outbreak; however, as there were only 2 clade 1A isolates, it could not be determined if these isolates were also part of an independent and concurrent outbreak in the ICU during the study period.

As the suspected outbreak isolates within clade 1 were close to identical (≤2 SNPs), we further investigated the reason for the different ST assignments of the isolates within this clade. Based on the Oxford MLST scheme, ST1114 and ST1841 strains are single locus variants with allelic differences in the gdhB gene. We retrieved the sequences of these two alleles from the pubMLST database and searched them against the contigs of all clade 1 isolates using BLAST. This revealed two copies of the gdhB gene within all nine clade 1 isolates, with each copy sharing 100% identity with either allele 3 or 189. This demonstrated that all clade 1 strains belonged to the same ST and supported our outbreak hypothesis. With repeated runs, however, our MLST pipeline consistently detected allele 189 among ST1841 isolates and allele 3 among the ST1114 isolates.

We characterized the strains based on MLST data to identify the lineages causing A. baumannii infections in Nigeria. Nine of the 34 (20.6%) A. baumannii isolates, and the single A. pittii isolate, had novel MLST allelic profiles (Figure 2). Upon submission of these isolates and profiles to the PubMLST database, we found that two of the STs had also been detected in other studies – one in China (ST2417) and the other in Ghana (ST2146) – around the same period and submitted to the database. The other eight novel ST profiles were submitted to the PubMLST database with submission ID BIGSdb_20210908105012_023242_52326 and have been assigned STs (Supplementary Table 1). The nine outbreak A. baumannii isolates (ST1114 and ST1841), as well as one of the novel ST strains (G20500013; ST2456), belonged to IC2, while three of the strains belonged to IC1 (ST231, ST441, and the novel ST2451 strain). Other international clones of A. baumannii were also present among our strains. However, the majority (19/34; 55.9%) of the A. baumannii isolates were singletons, non-major international clones or novel. The eight ST1089 strains clustered with other IC9 sequence types in the PubMLST database, while the two ST229 strains clustered with other IC7 sequence types. Both ST229 isolates were phylogenetically identical and had identical AMR and virulence genes but were isolated from different locations, one from UCH and the other from OAUTHC.

All 34 A. baumannii isolates harbored variants of the intrinsically encoded blaOXA-51-family carbapenemase gene (Figure 3 and Supplementary Table 2). In addition to these intrinsic blaOXA-51 family genes, all but one of the A. baumannii isolates carried other carbapenem resistance genes. Seventeen (50.0%) isolates, including all IC1, IC2 (clade 1) and IC7 strains harbored variants of the potent blaOXA-23-like carbapenemase gene; four of these strains also harbored the blaNDM-1 gene. blaNDM-1 was present in 15 (44.1%) strains, including all eight ST1089 (clade 2) strains and two of the novel ST strains (ST2450 and ST2456). This ST2450 strain was highly resistant and was the only strain that carried the blaOXA-58 carbapenemase gene. Two other strains carried the blaOXA-420 variant of the blaOXA-58 family. Phenotypically, thirty-one (91.2%) strains were resistant to at least one of the carbapenem antibiotics tested; 31 strains were resistant to meropenem, 30 were resistant to doripenem, and 28 were resistant to imipenem (Figure 4 and Supplementary Table 3).

Other beta-lactam resistance genes were found in most of the isolates. Several variants of the AmpC gene, blaADC, which is an intrinsic chromosomally encoded gene that confers cephalosporin resistance, were detected in 28 (82.4%) of the isolates. The remaining six isolates carried copies of the gene that were flagged by the ariba prediction software as being either “fragmented” or “interrupted,” indicating either a non-coding variant or a gene structure significantly different from that of the gene in the NCBI AMR database, respectively. Furthermore, all ten IC2 isolates, as well as the ST441 strain, harbored the broad-spectrum beta-lactamase gene, blaTEM-84. All Acinetobacter isolates harbored genes conferring streptomycin resistance. Amikacin, gentamicin and kanamycin resistance genes were also detected in 32 (94.1%) of the A. baumannii isolates. Thirty-one (91.2%) A. baumannii isolates also harbored at least one sulfonamide resistance gene; 15 isolates carried both sul1 and sul2. At least one of the tetracycline efflux transporter genes, tetA, tetB, and tet39 was detected in 19 (55.9%) of the A. baumannii isolates. The resistance rate among A. baumannii strains in this study was very high, with >88% of the strains being resistant or intermediate to 10 of the 14 antibiotics tested, including the carbapenems.

The outbreak strains in clade 1 all harbored multiple resistance genes. All of them harbored the blaOXA-23 gene and were all phenotypically resistant to doripenem, imipenem, and meropenem. They also harbored genes mediating resistance to streptomycin, beta-lactams/cephalosporins, sulfonamides, tetracycline, and other aminoglycosides, including amikacin, gentamicin and kanamycin, and were only sensitive to tigecycline amongst all the antibiotics tested.