Comprehensive comparative-genomic analysis of type 2 toxin-antitoxin systems and related mobile stress response systems in prokaryotes

Abstract

Background: The prokaryotic toxin-antitoxin systems (TAS, also referred to as TA loci) are widespread, mobile two-gene modules that can be viewed as selfish genetic elements because they evolved mechanisms to become addictive for replicons and cells in which they reside, but also possess "normal" cellular functions in various forms of stress response and management of prokaryotic population. Several distinct TAS of type 1, where the toxin is a protein and the antitoxin is an antisense RNA, and numerous, unrelated TAS of type 2, in which both the toxin and the antitoxin are proteins, have been experimentally characterized, and it is suspected that many more remain to be identified.

Results: We report a comprehensive comparative-genomic analysis of Type 2 toxin-antitoxin systems in prokaryotes. Using sensitive methods for distant sequence similarity search, genome context analysis and a new approach for the identification of mobile two-component systems, we identified numerous, previously unnoticed protein families that are homologous to toxins and antitoxins of known type 2 TAS. In addition, we predict 12 new families of toxins and 13 families of antitoxins, and also, predict a TAS or TAS-like activity for several gene modules that were not previously suspected to function in that capacity. In particular, we present indications that the two-gene module that encodes a minimal nucleotidyl transferase and the accompanying HEPN protein, and is extremely abundant in many archaea and bacteria, especially, thermophiles might comprise a novel TAS. We present a survey of previously known and newly predicted TAS in 750 complete genomes of archaea and bacteria, quantitatively demonstrate the exceptional mobility of the TAS, and explore the network of toxin-antitoxin pairings that combines plasticity with selectivity.

Conclusion: The defining properties of the TAS, namely, the typically small size of the toxin and antitoxin genes, fast evolution, and extensive horizontal mobility, make the task of comprehensive identification of these systems particularly challenging. However, these same properties can be exploited to develop context-based computational approaches which, combined with exhaustive analysis of subtle sequence similarities were employed in this work to substantially expand the current collection of TAS by predicting both previously unnoticed, derived versions of known toxins and antitoxins, and putative novel TAS-like systems. In a broader context, the TAS belong to the resistome domain of the prokaryotic mobilome which includes partially selfish, addictive gene cassettes involved in various aspects of stress response and organized under the same general principles as the TAS. The "selfish altruism", or "responsible selfishness", of TAS-like systems appears to be a defining feature of the resistome and an important characteristic of the entire prokaryotic pan-genome given that in the prokaryotic world the mobilome and the "stable" chromosomes form a dynamic continuum.

Reviewers: This paper was reviewed by Kenn Gerdes (nominated by Arcady Mushegian), Daniel Haft, Arcady Mushegian, and Andrei Osterman. For full reviews, go to the Reviewers' Reports section.

Figure 1

Two computational strategies for the identification of TAS.

Figure 2

Predicted new families of toxins. A. Multiple alignment of COG4679 family (RelE interferase supefamily). B. Multiple alignment of SMa0917 family (PemK/MazF interferase superfamily). C. Multiple alignment of COG2929 family (RelE interferase superfamily) representative. The sequences are denoted by Gene Identification (GI) numbers from the GenBank database and abbreviated species names. Species name abbreviations (generally consisting of 3 first letters of genus name and 4 first letters of species) for all alignments are given in Additional file 13. The positions of the first and the last residues of the aligned region in the corresponding protein are indicated for each sequence. The numbers within the alignment represent poorly conserved inserts that are not shown. The coloring is based on the consensus shown underneath the alignment; h indicates hydrophobic residues (ACFILMVWY), p indicates polar residues (STEDKRNQH), s indicates small residues (AGSVC) and a indicates aromatic residues (WYFH). The secondary structure elements are shown according to structural data if the structure is available or predicted using the PSIPRED program [106]; E indicates β-strand and H indicates α-helix.

Figure 3

Distribution of TAS across bacterial and archaeal taxa. Black: TAS absent in the taxon while random expectation is significantly non-zero. Dark gray: TAS absent in the taxon with random expectation not significantly different from zero. Blue: TAS is significantly underrepresented in a taxon with more than twofold difference from random expectation. Cyan: TAS is significantly underrepresented in a taxon with less than twofold difference from random expectation. Light gray: abundance of a TAS in a taxon does not significantly differ from random expectation. Orange: TAS is significantly overrepresented in a taxon with less than twofold difference from random expectation. Red: TAS is significantly overrepresented in a taxon with more than twofold difference from random expectation. The random expectation estimate is based on the total number of TAS of the given type and the total number of protein-coding genes in the given taxon. The statistical significance was estimated using the χ² test (critical χ² value of 3.84 for 1 degree of freedom and p-value of 0.05).

Figure 4

Predicted new families of antitoxins. A. Multiple alignment of COG2442 family domain, a predicted DNA-binding antitoxin protein of winged HTH motif superfamily. B. Multiple alignment of COG2886 family of predicted antitoxins containing the HTH domain. C. Multiple alignment of COG2880 family, an AbrB superfamily representative. Designations are the same as in Figure 2.

Figure 5

Multiple alignments of distinct predicted antitoxin families containing ribbon-helix-helix (RHH) domains. A. RHH domain of COG1753 family. B. RHH domain in MJ1172 family. C. RHH domain in COG5304/COG3514. D. RHH domain fused to HEPN domain (paREP 1 subfamily). Designations are as in Figure 2.

Figure 6

Multiple alignment of the conserved cores of two distinct families of HEPN domains. The distinct subfamilies of HEPN domains are indicated by brackets on the right. Designations are as in Figure 2.

Figure 7

Relative abundance of HEPN_T, HEPN_M and MNT domains in thermophiles and mesophiles. A. The total number of HEPN-MNT pairs in hyperthermophiles and thermophiles ("Thermo"), mesophiles and psychrophiles ("Meso") and all ("Both") genomes. B. The number of HEPN_T, HEPN_M and MNT genes in selected genomes. Font color indicates the temperature preference: red – hyperthermophiles; gold – thermophiles; green – mesophiles; blue – psychrophiles. Asterisks indicate Archaea.

Figure 8

The relationship between the number of detected TA pairs and genome size.

Figure 9

Graph of relationships between different families of toxins and antitoxins. Known (black) and predicted (magenta) toxins (red circles) and antitoxins (blue circles) and their operon organizations. Lines connect genes with 5 or more two-component operons found; thickness of a line is proportional to the frequency of the respective operon.

Figure 10

TAS in selected genomes. Red dots show the approximate position of TAS genes on the circular chromosomes.

Figure 11

Fractions of solo and two-gene operon occurrences for each family of toxins and antitoxins. Red, fraction of solo genes; blue, fraction of genes in (predicted) operons.