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

Acinetobacter johnsonii has not been studied as much as A. baumannii, and few studies have been carried out to examine this species. A. johnsonii has been found in aquatic sources, human skin, and animals (1, 2). However, it also causes severe human infections (3–6), highlighting its clinical importance. For instance, Turton et al. showed that 1.7% of 690 nonduplicate Acinetobacter isolates associated with bacteremia were A. johnsonii (6). Moreover, Cleland et al. identified A. johnsonii as a relevant pathogen involved in chronic rhinosinusitis (7), and some studies have described A. johnsonii isolates carrying antibiotic resistance genes (ARGs) (3, 8, 9). For example, different carbapenemase genes, such as bla_NDM-1 and bla_OXA-58, have been identified in A. johnsonii (3, 9). The first description of bla_NDM-1-positive A. johnsonii occurred in two isolates recovered from sewage in China in 2010 (9), and another isolate (also collected from sewage) was found to coproduce the plasmid-encoded carbapenemases NDM-1, OXA-58, and PER-1 (3). Interestingly, NDM-1 has also been found in phages not only in A. johnsonii (10) but also in A. baumannii (11). Hence, this species could be a potential reservoir of ARGs against last-line antibiotics, which is particularly worrying since these genes can be transferred to other clinically relevant microorganisms.

Population genomics studies are needed to achieve a better understanding of the phylogeny and the ARGs within A. johnsonii. Although two previous studies conducted some comparative genomics analyses of A. johnsonii (3, 4), these only considered a very small number of genomes. Thus, our aim was to characterize the phylogeography and ARGs in A. johnsonii using all the genomes available to date. The lack of information on A. johnsonii is clear; as of 16 March 2020, there were only 31 genomes on the National Center for Biotechnology Information database. We downloaded these genomes (see Table S1 in the supplemental material) and corroborated that they were A. johnsonii by conducting an average nucleotide identity (ANI) analysis via OrthoANI (12). All but one 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 (∼67%) 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 GTR+R+I) and population structure analyses via hierBAPS (17) (in Rstudio, with K = 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, ≥80% identity and ≥70% 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-nodulation-division (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 β-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 β-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 and clinical strains are intermingled with one another, and all the strains show ARGs. Further (genomic and functional) studies of clinical and nonclinical strains are needed to fully understand the clinical nature of this species.

S.C.-R. and V.M.-E. are very grateful to Alfredo José Hernández Álvarez and Victor Manuel del Moral Chávez for installing some of the programs used for this study on our servers.

This study was supported by National Fund for the Scientific and Technological Development (FONDECYT) of Chile (FONDECYT-Iniciación, grant number 11190602) to A.O.-C. and by the ANID Millennium Science Initiative/ Millennium Initiative for Collaborative Research on Bacterial Resistance, MICROB-R, NCN17_081, to A.O.-C. and G.G.-R. This study was also partially supported by the Programa de Apoyo a Proyectos de Investigación e Innovación Tecnológica (grant IN206019) and the Consejo Nacional de Ciencia y Tecnología Ciencia Básica 2016 (grant 284276) to S.C.-R.

All authors report no potential conflicts of interest.