TASmania: A bacterial Toxin-Antitoxin Systems database

Abstract

Bacterial Toxin-Antitoxin systems (TAS) are involved in key biological functions including plasmid maintenance, defense against phages, persistence and virulence. They are found in nearly all phyla and classified into 6 different types based on the mode of inactivation of the toxin, with the type II TAS being the best characterized so far. We have herein developed a new in silico discovery pipeline named TASmania, which mines the >41K assemblies of the EnsemblBacteria database for known and uncharacterized protein components of type I to IV TAS loci. Our pipeline annotates the proteins based on a list of curated HMMs, which leads to >2.106 loci candidates, including orphan toxins and antitoxins, and organises the candidates in pseudo-operon structures in order to identify new TAS candidates based on a guilt-by-association strategy. In addition, we classify the two-component TAS with an unsupervised method on top of the pseudo-operon (pop) gene structures, leading to 1567 "popTA" models offering a more robust classification of the TAs families. These results give valuable clues in understanding the toxin/antitoxin modular structures and the TAS phylum specificities. Preliminary in vivo work confirmed six putative new hits in Mycobacterium tuberculosis as promising candidates. The TASmania database is available on the following server https://shiny.bioinformatics.unibe.ch/apps/tasmania/.

Fig 1. Overview of the pipeline to build the TASmania database.

The different steps include: downloading EnsemblBacteria, updating the InterPro annotation, selecting the proteins matching an arbitrary list of reference TAS IPR, building the corresponding HMM profiles and scanning the proteomes. In parallel, we structure target genomes into pseudo-operons and include phylum information. Finally, we add extra value to TASmania by clustering the HMM profiles into larger families for TA combinations analysis.

Fig 2. Unique proteins length distribution of TASmania putative hits.

(A) Antitoxins length distribution (in amino acids). (B) Toxins length distribution (in amino acids). Blue and red vertical lines correspond to default thresholds used by TAfinder.

Fig 3. Pseudo-operon types distribution.

(A) All hits from the TASmania (only the 20 most frequent pseudo-operons structures are shown). (B). Canonical hits only (two-genes T->A or A->T modules) highlighting the higher abundance of the A->T module type versus the T->A type.

Fig 4. Comparison of TASmania and TAfinder hits.

Using M.tuberculosis as a proof-of-principle, a list of manually curated, new and promising TASmania-specific hits is shown in Table 1, compared to the results obtained by TAfinder on the same genomes. (A) Mycobacterium tuberculosis H37Rv. (B) Mycobacterium smegmatis HMC2 155. (C) Caulobacter crescentus CB15. (D) Staphylococcus aureus NCTC8325. These TASmania-specific TA hits correspond mostly to: i) type I or type IV systems; ii) orphan loci; iii) guilt-by-association “x” loci iv) unusual combinations (“TT”, “AA”). This confirms that our strategy of not filtering out any unusual TAS operon structures or protein lengths allows us to be more discovery-orientated. Including the guilt-by-association “x” cognates is also useful when looking for uncharacterized TAS families.

Fig 5. Expression of putative toxins in M.smegmatis.

M.smegmatis strain MC²155 was freshly transformed with pLAM12-based constructs expressing the putative toxin encoding genes of M.tuberculosis identified in this work, namely Rv0078A, Rv0207c, Rv0229c, Rv0269c, Rv0366c, Rv0569, Rv2016, Rv2165c, Rv2514c, Rv3641c and Rv3662c. Transformants were plated on LB agar supplemented with kanamycin, without (-) or with 0.2% acetamide inducer (+). Plates were incubated 3 days at 37°C.

Fig 6. Six putative TA of M.tuberculosis validated by rescue test in M.smegmatis.

M.smegmatis strain MC²155 was freshly transformed with pLAM12-based constructs expressing the putative toxic genes of M.tuberculosis (Rv0078A, Rv0207c, Rv0269c, Rv0366c, Rv2016 and Rv2514c) either alone or as an operon together with their respective putative antitoxin genes, namely Rv0078B/Rv0078A, Rv0208c/Rv0207c, Rv0269c/Rv0268c, Rv0367c/Rv0366c, Rv2016/Rv2017 and Rv2515c/Rv2514c. Transformants were plated on LB agar supplemented with kanamycin and acetamide inducer (0.2%), except for Rv0366c and Rv0367c/Rv0366c, which shows suppression by the putative antitoxin only in the absence of acetamide. Plates were incubated for three days at 37°C.

Fig 7. Examples of co-occurrence of toxin and antitoxin clusters within two-genes pseudo-operons (popTAs).

The color key correspond to percentages (%), given in each cell. (A) Antitoxin clusters in A->T orientation, and their relation to toxin clusters. For instance, the modular A74 antitoxin cluster has three main cognates the T4, T65 and T78 toxin clusters: A74.T4 (31.06% of A74 popTAs) (nearest Pfam PhdYeFM_antitox.YafQ_toxin), A74.T65 (44.41% of A74 popTAs) (nearest Pfam PhdYeFM_antitox.PIN) and A74.T78 (13.59% of A74 popTAs) (nearest Pfam PhdYeFM_antitox.ParE_toxin). (B) Antitoxin clusters in T->A orientation, and their relation to toxin clusters. For instance, the bi-directional A12 antitoxin cluster’s main toxin cognate is T102, as in T102.A12 (29.15% of A12 popTAs) (nearest Pfam HicB_lk_antitox.HicA_toxin). A restrictive antitoxin cluster is also highlighted with A124 co-occurring mainly with T34 as in T34.A124 (99.88% of A124 popTAs) (nearest Pfam BrnT_toxin.BrnA_antitoxin). (C) Toxin clusters in A->T orientation, and their relation to antitoxin clusters. The restrictive T60 toxin cluster and its association with A46 in A46.T60 (98.79% of T60 popTAs) (nearest Pfam CcdA.CcdB) is given as example. (D) Toxin clusters in T->A orientation, and their relation to antitoxin clusters. T152 is also a quite restrictive toxin cluster that mostly has A23 as the main antitoxin cognate, as in T152.A23 (nearest Pfam HigB-like_toxin.HTH_3). The complete co-occurrence is shown in S3 Fig and described in S5 Table.

Fig 8. popTAs across phyla.

The most abundant popTAs in relative numbers are specific to each phylum.

Fig 9. A9 cluster as example of cluster modularity.

(A) Multiple sequence alignment of A9 antitoxin cluster (nearest Pfam PhdYeFM_antitox) proteins that are associated with T6 (nearest Pfam YoeB_toxin) or T9 (nearest Pfam ParE_toxin) toxin clusters. (B) HMM profile from antitoxin A9 cluster proteins, in A9.T6 and A9.T9 popTA. (C) Multiple sequence alignment of T6 or T9 toxin clusters proteins associated with A9 antitoxin cluster. (D) HMM profile from toxin T6 and T9 clusters proteins, in A9.T6 and A9.T9 popTA. Note: for clarity, only a subset of sequences are drawn. The magenta bars and stars highlight the conserved residues and regions.