Letter to the Editor: Prophages Encode Antibiotic Resistance Genes in Acinetobacter baumannii

Dear Editor:

Multidrug-resistant Acinetobacter baumannii is one of the most prominent human pathogens, and it is an important cause of infections in intensive care units in hospitals all over the world. ¹ Similar to other Gram-negative bacteria, in this species conjugation seems to be the major process dispersing antibiotic resistance genes (ARGs), ² but natural competence has also been reported. ³ Phages have not been described as relevant vehicles of ARGs in this species despite the fact that transduction is a well-known factor driving horizontal gene transfer among bacteria. ⁴ However, a recent study claimed that an ARG could have been transferred by transduction in A. baumannii. ⁵ Furthermore, another very recent study has shown that chromosomal ARGs can be transferred between A. baumannii strains due to prophages. ⁶ In this study we demonstrate that a previously understudied prophage diversity encodes many ARGs in A. baumannii.

To explore the presence of prophages in A. baumannii, we created a database of 133 genomes (Supplementary Table S1) representing the breadth of diversity within this species. The data set has genomes from the major international clones and covers >50 sequence types (STs). We searched for prophage signals using VirSorter ⁷ ; to increase the proportion of correct phage predictions only VirSorter categories 1 and 2 were considered (see methods in Supplementary Data) and we limited predicted phages to contigs longer than 10 kb. Remarkably, 1,529 prophages were identified in the 133 genomes, with a mode of 8 prophages per genome. Furthermore, every single genome had phage signals. Notably, most of the prophages are bigger than 10 kb and many even bigger than 30 kb (Supplementary Table S1), implying that many of these prophages maybe functional. We then constructed a maximum likelihood phylogeny based on core genes (Supplementary Fig. S1) as in a previous study ⁸ and mapped the prophages on this tree. Supplementary Fig. S1 shows that prophages are distributed across the tree. Moreover, phages seem to be particularly abundant in strains from a recently emerged clade (ST758) found in Latin America.^9,10 These findings suggest that prophages are commonly found in different lineages of A. baumannii.

We used the Prokaryotic Virus Orthologous Groups database to classify these prophages. We found that 1,427 phages (93.3%) were classified into the Caudovirales order (family classification is described in Supplementary Data). Remarkably, there were hardly any phages with close matches to sequences in the GenBank: the best match had just 11% coverage alignment and 10% of nucleotide identity. We then focused on the abundance of ARGs encoded in the predicted prophages. We clustered prophages using CD-HIT to avoid overestimation of ARGs (Supplementary Data) and the inferred clusters were searched in the Comprehensive Antibiotic Resistance Database using a very strict sequence similarity threshold (≥95% coverage and ≥98% nucleotide identity) to confidently identify ARGs. We found nine different drug families encoded in 723 genes (Fig. 1A). The three major contributors were the multidrug efflux resistance-nodulation-division transporter family (see red bars in Fig. 1), the small multidrug resistance efflux pump AbeS (yellow bar), and oxacillinases (OXA) from the OXA-23-like and OXA-51-like gene families (some blue bars). This analysis clearly shows that prophage-encoded ARGs are a commonplace in A. baumannii.

OPEN IN VIEWER

FIG. 1.

ARGs found in prophages and induction assays. (A) ARGs present in prophages; the different types of ARGs are shown on the x-axis, whereas the y-axis gives the number of prophages having that particular gene. The color coding of the bar gives the different antibiotic resistance classes and is explained in the key. (B) Induction plots; the 11 strains that were positive for the induction assays plus the control (strain ATCC17978). The strains in bold also did form lytic plaques. ARGs, antibiotic resistance genes. Color images are available online.

Finally, we determined if these prophages could enter a lytic cycle. To do that we conducted induction assays with mitomycin C (Supplementary Data) on 30 A. baumannii strains belonging to different STs (Supplementary Table S2). In 11 strains (Fig. 1B) we detected cases of induction, most of these strains belong to the ST758 (Supplementary Table S2). However, there were strains from other STs (ST231 and ST369) that also showed induction. Notably, in 5 (45%) of the 11 strains where we saw induction were able to form lytic plaques (Supplementary Fig. S2 and Supplementary Table S2). Of note, 9 of the 11 strains positive for the induction assay had whole genome sequences and 56% (5 strains) of these strains had phage-encoded ARGs (Supplementary Table S3). Thus, it seems that phages with lytic potential are present in several lineages of this species and that some phages encode ARGs. However, whether the phages with lytic potential are precisely those encoding ARG remains to be established.

In summary, our study shows that a sizeable amount of uncharacterized prophage diversity exists in A. baumannii and that some prophages may enter a lytic phase. Remarkably, many of the prophages encode ARGs, yet at this point we cannot ascertain if they have a lytic potential, nor that they contribute to ARG spread. For the past years there has been a hot debate about whether or not phages can encode ARGs^11,12; in this regard, our study supports the notion that prophage encode ARGs.^11–13 Thus, lysogenization could be an important player in HGT mediated by phages.