Phylogeographical Analyses and Antibiotic Resistance Genes of Acinetobacter johnsonii Highlight Its Clinical Relevance

Acinetobacter johnsonii has been severely understudied and its population structure and the presence of antibiotic resistance genes (ARGs) are very much uncertain. Our phylogeographical analysis shows that intercontinental transmission has occurred frequently and that different lineages are circulating within single countries; notably, clinical and nonclinical strains are not well differentiated from one another. Importantly, in this species recombination is a significant source of single nucleotide polymorphisms.

genome (UBA3112) belonged to A. johnsonii since they shared ANI values higher than the 95% (the cutoff value for species demarcation) when they were compared. Of note, UBA3112 and UBA8888 were not included in downstream analyses because they did not have high-quality genomes according to CheckM (13) (see also the footnote for  Table S1).
A pangenome analysis through Roary (14) yielded a total of 13,531 groups of homologous genes (GHGs), most of them (89%) within the accessory genome (see Table S2). The strict core genome consisted of 1,538 GHGs and the majority (ϳ67%) was found in 15% or fewer of the genomes (see Table S2). Moreover, this is an open pangenome (see Fig. S1), since the number of GHGs kept growing as more genomes were considered without tailing off; we therefore did not fully sample the gene repertoire of this species. Then, to evaluate the level of synteny, we conducted a genome alignment considering five genomes (one from each of the clusters identified in the population structure analysis [see below]) using progressiveMauve (15). Figure S2 shows that a significant number of inversions and large-scale changes occurred within these genomes, indicating that this species has undergone a considerable amount of genome rearrangement.
To establish the population structure and the evolutionary relationships of these isolates, maximum-likelihood phylogeny using PhyML (16) (model GTRϩRϩI) and population structure analyses via hierBAPS (17) (in Rstudio, with K ϭ 20) were conducted on the core genome alignment, which had 161,087 segregating sites and a nucleotide diversity of 0.0295. We found five genetically differentiated clusters (colored labels in Fig. 1), and four seemed to be real populations, since they were monophyletic groups according to the phylogeny (blue, maroon, green, and purple labels in the figure); in contrast, cluster 2 appeared to be an exclusion group (red labels, Fig. 1). Some of the real clusters had isolates from different continents. For instance, cluster 1 (blue labels) had isolates from South America (Chile and Argentina), Africa (Morocco), Asia (China), and Australia, whereas cluster 3 (maroon labels) contained isolates from Asia (Japan), Europe (Germany), and North America (USA). In addition, different lineages can be circulating in the same country. For instance, isolates from China were found in three of the clusters; this pattern also applies for the Japanese and U.S. strains. Remarkably, we noted that in cluster 2 the clinical isolates (XBB1, Aj2199, and UCO-489) grouped together with environmental isolates such as JH7 (recovered from mine tailings), WCHAJo010049 (collected from sewage), or 18QD2AZ57W (sampled from pig feces).
Thus, these analyses reveal a clear population structure in this species, where some clusters are composed of isolates from distant geographic regions, showing that intercontinental transmission has occurred frequently. Furthermore, different lineages circulate within single countries, implying that several introduction events have happened in the same country. Importantly, there seems to be no clear delimitation between clinical and nonclinical isolates. We used Gubbins (18) to assess the impact of homologous recombination. Clearly, recombination is of paramount importance since the average per-branch recombination/mutation ratio was 4.64, implying that recombination is introducing almost five times more single nucleotide polymorphisms than does mutation.
Finally, we conducted an in silico prediction of ARGs by conducting BLAST searches (similarity criteria, Ն80% identity and Ն70% coverage) of the A. johnsonii proteomes against the Comprehensive Antibiotic Resistance Database (19). Notably, all the strains, even the environmental ones, had at least two ARGs (see Fig. 2); for instance, isolates C6 and LXL_C1 both had oxacillinases and multidrug efflux resistance-nodulationdivision (RND) transporter genes. We found resistant determinants for several drug classes in many isolates (see Fig. 2, drug class). We also looked for mutations conferring resistance to fluoroquinolones via ResFinder (20), but we did not find any. In agreement with previous studies (3, 4), we found some ␤-lactamase genes (bla NDM-1 , bla PER-1 , bla PER-2 , and bla OXA-58 ). In addition to some clinical isolates, two sewage strains (Acsw19 and WCHAJo010049) and a strain collected from pig feces (18QD2AZ57W) had the largest amount of ARGs. In this regard, Tang et al. determined that strain Acsw19 has 12 ARGs in plasmids and in the chromosome (10). Considering the OXA ␤-lactamases, we found several families: OXA-211-like, OXA-58-like, and OXA-23-like. However, the most abundant-OXA-281, OXA-334 and OXA-373-belong to the OXA-211-like family, which was described rather recently in non-baumannii Acinetobacter spp. Remarkably, many ARGs have undergone horizontal gene transfer (HGT) since 81% of them had identical sequences in other bacteria from clinically relevant genera such as Salmonella, Klebsiella, Vibrio, etc. (see Fig. 2 and Table S3). As a case in point, the carbapenemase NDM-1 was present in four isolates (see gray rectangles in Fig. 1) on noncontiguous branches of the tree, implying independent acquisitions of this gene, and identical sequences of this gene were found in many genera other than Acinetobacter (see Table S3). Taken together, these results show that many strains, both clinical and nonclinical, had ARGs with signals of HGT and thus could function as a reservoir of ARGs for other bacteria.
In conclusion, we highlight the clinical relevance of this species, since environmental