Bacterial genomes included in the analysis were manually selected from the PubMLST database¹ until January 2022. In detail, we selected 39 A. baumannii complete genomes assigned to international clonal lineages ICI, ICII, and ICIII, which corresponded to Pasteur ST1, ST2, and ST3, respectively, and to additional epidemic clonal lineages assigned to Pasteur ST10, ST25, ST32, ST52, ST78, and ST79 (Gaiarsa et al., 2019). Furthermore, we collected from the National Center for Biotechnology Information (NCBI) high-quality complete genomes and, when not available, scaffolded sequences of all species into the Acinetobacter genus. The “parameters” to consider the genomes of “high quality” were: N50 ≥ 10,000 bp; number of contigs ≤1,000; identification of the correct species through rMLST (Jolley et al., 2012). The Acinetobacter selected genomes were typed by A. baumannii Pasteur MLST scheme (Diancourt et al., 2010) and rMLST (Jolley et al., 2012) using the BIGSdb software available at https://pubmlst.org/organisms/acinetobacter-baumannii/ (Jolley et al., 2018). The characteristics of the genomes were included in Supplementary Table S1 and available at https://pubmlst.org/bigsdb?db=pubmlst_abaumannii_isolates&page=project&project_id=8.

The allelic profiles of Pasteur and ribosomal MLST schemes were extracted from all the genomes, and then the sequences were aligned using Muscle (Edgar, 2004) to generate the neighbor-joining trees using the BIGSdb software (Jolley et al., 2018). The core genome single-nucleotide polymorphisms (SNPs) were detected using the tools of PARSNP v1.1.2 (Treangen et al., 2014), and the SNP alignment was performed considering the ascertainment bias using the Lewis correction (Lewis, 2001). In detail, each genome was aligned to the reference genome NC_010611.1 of ACICU, and the alignments were then concatenated using Muscle (Edgar, 2004). The maximum-likelihood phylogenies of 574 genomes belonging to the A. baumannii group (A. baumannii, A. pittii, A. seifertii, A. lactucae, and A. nosocomialis) were performed using the concatenated alleles of the Pasteur and ribosomal MLST schemes and a reference phylogeny using genome-wide data (a core genome of 372 high-quality genes and an alignment of 17,072 SNPs). The phylogenies of Pasteur MLST, ribosomal MLST, and core SNP alignments were inferred through the GTR-GAMMA model at 100 bootstrap replicates using RAxML v.8 (Stamatakis, 2014). GTR’s GAMMA model was used for its ability to optimize the transition/transversion speed ratio and the α parameter of the gamma rate heterogeneity distribution (Stamatakis, 2014). The phylogenetic trees and annotations were visualized using the iTol v6 software.²

The statistical analyses were performed using the Robinson-Fould (R-F) and Matching clusters (M-C) topology-based tests employing TreeCmp (Bogdanowicz et al., 2012). The M-C test calculates the number of topological changes that must be made to transform a tree into a reference tree. The R-F test counts the number of splits that are unique to one of the two trees. In both cases, the two analyzed trees are identical if the value is zero. The likelihood-based Shimodaira-Hasegawa (SH) test (Shimodaira and Hasegawa, 1999) was performed with RAXML (Stamatakis, 2014). In this test, a null hypothesis assumes that two compared trees are both a correct interpretation of an alignment. The tested hypothesis is that one or more trees better represent the data. p-values lower than 0.05 indicate that the two trees are significantly different.

The antimicrobial resistance genes were detected using abricate³ based on the ResFinder 4.0 database (Bortolaia et al., 2020). The presence of the gene and percentage of identity were indicated in the matrix in Supplementary Table S2. The threshold identity of 80%, the minimum length of 80% matches, and the coverage value of more than 90% were selected for each gene. The following acquired resistance genes were analyzed: class A, B, C, D β-lactamase, folate pathway inhibitors, rifampicin, aminoglycoside, chloramphenicol, colistin, macrolide, quaternary ammonium salts, and tetracycline resistance genes (Supplementary Table S2). The multiple alignment and phylogeny of class D β-lactamase genes were performed using Clustal W (Thompson et al., 1994) and RAXML v.8 (Stamatakis, 2014), respectively. The genomes of Acinetobacter genus were classified as MDR if they carried antimicrobial-resistant genes for at least three of the nine classes of antimicrobials considered by Magiorakos et al. (2012).

The clusters of Pasteur sequence types (STs) were determined using the eBURST algorithm as described previously (Feil et al., 2004). Minimum spanning trees of STs were built with Phyloviz using the goeBURST algorithm (Ribeiro-Gonçalves et al., 2016). Minimum-spanning trees were generated from the seven alleles of each MLST scheme, and species were assigned based on clustering with reference STs. Additionally, SNP analysis by PARSNP was visualized using Gingr (Treangen et al., 2014), which provided an interactive display of multi-alignment variants and a phylogenetic tree estimated from the core genome alignment. Then, the values of the maximum unique matches and the data were evaluated to study the cluster phylogeny of SNPs (Treangen et al., 2014).

The database consisted of 837 genomes assigned to 72 distinct species in the Acinetobacter genus, which were identified at genus and species levels using the ribosomal MLST scheme (Jolley et al., 2012) and showed a genome size range of 2.85–4.85 Mpb (Supplementary Table S1). The Pasteur MLST scheme (Diancourt et al., 2010) was able to assign an allelic profile and an ST to all 837 genomes belonging to 72 species in the Acinetobacter genus; while, the ribosomal MLST scheme assigned an ST to 806 out of 837 genomes because from 1 to 20 alleles could not be identified in the genomes belonging to A. baumannii, A. nosocomialis, A. bereziniae, A. pittii, A. radioresistens, and A. seifertii. In addition, the rMLST scheme identified paralogues in 11 out of 53 loci paralogous in 57 Acinetobacter spp. (Supplementary Table S1). The genome-wide distances of the whole database analyzed using the minimum spanning tree (MST) with the Pasteur MLST scheme showed that the most abundant species were A. pittii (n = 282), A. nosocomialis (n = 175), A. seifertii (n = 61), A. baumannii (n = 39), and A. lactucae (n = 14), while 263 genomes were assigned to other 67 species of Acinetobacter genus (Figure 1). The 574 genomes assigned to the above five species have been considered the most clinically relevant species and were included in the A. baumannii group (Cosgaya et al., 2017). Among genomes within the A. baumannii group, the genetically closest species were A. pittii with A. lactucae or A. seifertii, showing 5 and 6 locus variants (LVs) genome-wide distance, respectively. The A. nosocomialis genomes assigned to ST782 and to the most frequent ST279 showed 4 and 5 LVs genome-wide distances with respect to CC2 A. baumannii genomes, respectively (Figure 1).

The core genome-SNP, Pasteur MLST, and rMLST maximum-likelihood phylogenies of 574 genomes belonging to the A. baumannii group (A. baumannii, A. pittii, A. seifertii, A. lactucae, and A. nosocomialis) showed similar inter- and intra-species distributions of branch lengths and nodes (Figure 2). In particular, the branch lengths of SNPs and rMLST phylogenies had values between ~10⁻⁶ and ~ 10⁻² (Figures 2A,C), while the branch lengths of Pasteur MLST phylogeny were between ~10⁻⁶ and ~ 10⁻³ (Figure 2B). The core genome SNPs and Pasteur MLST phylogenies showed bootstrap values greater than 50 for ancestral nodes of all species belonging to the A. baumannii group (Figures 2A,B). In addition, rMLST phylogeny assigned bootstrap values greater than 50 to the ancestral nodes of A. baumannii, A. seifertii, and A. lactucae species, while bootstrap values of 36 for A. pittii ancestral node and more uneven values ranging from 21 to 94 were assigned across A. nosocomialis genomes (Figure 2C). The above data demonstrated that all three phylogenies identified A. baumannii, A. pittii, A. seifertii, A. lactucae, and A. nosocomialis as distinct species in the A. baumannii group and a strong genomic similarity between A. lactucae and A. pittii species. Moreover, all three maximum-likelihood phylogenies showed high genomic heterogeneity among A. nosocomialis genomes (Figure 2). The statistical comparison between core genome SNP phylogeny and Pasteur MLST or rMLST phylogenies using the R-F and M-C tests showed similar statistical values and concordance among phylogenies (Table 1).

To evaluate the discriminatory power of the core genome SNP phylogeny and the Pasteur MLST phylogeny, we analyzed and compared the clusters identified by the two phylogenies. Clustering of core SNP phylogeny identified three clusters (1–3) in A. pittii genomes, one cluster (4) in A. seifertii genomes, and no clusters in A. nosocomialis genomes (Figure 3A). Cluster analysis of Pasteur MLST phylogeny showed four clusters in A. pittii having ST63, ST119, ST207, and ST396 as ancestral STs; two clusters in A. nosocomialis with ST410 and ST279 as ancestral STs; and one cluster in A. seifertii with ancestral ST553. Acinetobacter baumannii genomes showed three clades assigned to ST1, ST2, and ST10; and A. lactucae genomes counted 14 singletons and no clades (Figure 3B). Interestingly, A. pittii clusters 1, 2, and 3 of the core SNP phylogeny corresponded to A. pittii ST396, ST119, and ST207 clusters in the Pasteur MLST phylogeny, respectively; A. seifertii cluster 4 of the core SNP phylogeny corresponded to A. seifertii ST553 cluster of the Pasteur MLST phylogeny (Figure 3).

We analyzed the presence of antimicrobial resistance genes in the 837 genomes assigned to 72 distinct Acinetobacter species. The ResFinder software identified 188 genes encoding for resistance to 12 antimicrobial categories (Supplementary Table S2). The presence of class A, B, C, and D β-lactamases in the neighbor-joining tree of the 837 genomes assigned to 72 distinct Acinetobacter species is shown in Figure 4. The class A β-lactamase genes bla_{CARB-8,-2,-14,-16} were found in 1 A. nosocomialis, 4 A. pittii, 1 A. towneri, and 1 A. bereziniae genomes; the class A β-lactamase gene bla_SCO-1 was found in 1 A. radioresistens genome. In addition, class A β-lactamase genes bla_{TEM-1A,-1B,-1D} were found in 9 of 39 A. baumannii genomes; bla_PER-1,-2 ESBL genes were found in 1 A. baumannii, 2 A. nosocomialis, 1 A. pittii, and 1 A. radioresistens genomes; bla_VEB-1,-7 ESBL genes were found in 1 A. baumannii and 2 A. pittii genomes. The class B metallo-β-lactamase (MBL) genes bla_GIM-1, bla_{IMP-1,-4,-14,-19,-34} bla_NDM-1,-16, and bla_VIM-2,-4 were identified in 12 of 72 species of Acinetobacter genus. Among species, 34 of 282 A. pittii genomes showed at least one MBL gene. The bla_NDM-1 was the most frequent MBL gene and was found in A. nosocomialis, A. pittii, A. lactucae, A. junii, A. bereziniae, A. cumulans, A. wuhouensis, A. sichuanensis, A. rongchengensis, A. indicus, and A. variabilis genomes. The bla_ADC-25 class C β-lactamase was found in the five species belonging to A. baumannii group and A. calcoaceticus genomes, but not in other Acinetobacter spp. (Figure 4; Supplementary Table S2). The class D oxacillinase, which is a naturally occurring enzyme in several Acinetobacter species, was found in 503 isolates assigned to 35 Acinetobacter species (Figure 4; Supplementary Table S2). In all, 94 class D β-lactamase genes belonging to 11 distinct bla_OXA family genes (bla_OXA-211, bla_OXA-134, bla_OXA-214, bla_{OXA 294}, bla_OXA-51, bla_OXA-213, bla_OXA-274, bla_OXA-286, bla_OXA-58, bla_OXA-40, and bla_OXA-23) were identified (Supplementary Figure S1). Among the five species of A. baumannii group, A. baumannii, A. pittii, and A. lactucae genomes showed intrinsic class D β-lactamase belonging to bla_OXA-51 and bla_OXA-213 family genes, respectively, while A. nosocomialis and A. seifertii genomes did not show any intrinsic class D β-lactamase. Similarly, intrinsic class D β-lactamases were identified in other Acinetobacter species, such as bla_OXA-134-like in A. lwoffii and A. schindleri, bla_OXA-211 family in A. johnsonii, bla_OXA-213 family in A. calcoaceticus, bla_OXA-214 family in A. haemolyticus, bla_OXA-228 family in A. bereziniae, bla_OXA-286 family in A. viviani, A. disperses, and A. courvalini genomes, and bla_OXA-294 family gene in A. proteoliticus, A. gyllenbergii, and A. colistinresistens (Figure 4). Also, the presence of class D carbapenemase genes belonging to bla_OXA-23, bla_OXA-40, or bla_OXA-58 family genes were found in 16 of 39 A. baumannii genomes, 27 of 282 A. pittii genomes, 8 of 145 A. nosocomialis genomes, 2 of 2 A. seifertii genomes, 9 of 9 A. radioresistens genomes, 5 of 6 A. cumulans genomes, and 3 of 11 A. colistinresistens genomes (Figure 4). Moreover, 46 of 72 Acinetobacter species showed at least one antimicrobial resistance gene (Supplementary Table S2), and 91 isolates assigned to 17 species of the Acinetobacter genus showed at least three genes encoding for resistance to three different antimicrobial classes and were classified as MDR isolates (Magiorakos et al., 2012) (Supplementary Table S2; Figure 4).