Cross-Regulations between Bacterial Toxin–Antitoxin Systems: Evidence of an Interconnected Regulatory Network?

TA systems are genetic modules in bacteria that encode both a stable toxic protein and a labile antitoxin which neutralizes the toxin’s activity.
They are classified into six types depending on the antitoxin’s nature and mode of action, with toxins that are always proteins and antitoxins that are either small RNAs or proteins.
The coexistence of both homologous and nonhomologous TA systems in bacterial genomes suggests that cross-regulation between them is very likely. Cross-regulations can influence RNA levels or the toxicity of other TA systems.
Type II antitoxins are global regulators of gene expression via their DNA-binding domains.
TA systems can act individually or within an extended gene network containing cross-regulations between homologous and nonhomologous TA systems among various types.

Toxin–antitoxin (TA) systems are ubiquitous among bacteria and include stable toxins whose toxicity can be counteracted by RNA or protein antitoxins. They are involved in multiple functions that range from stability maintenance for mobile genetic elements to stress adaptation. Bacterial chromosomes frequently have multiple homologues of TA system loci, and it is unclear why there are so many of them. In this review we focus on cross-regulations between TA systems, which occur between both homologous and nonhomologous systems, from similar or distinct types, whether encoded from plasmids or chromosomes. In addition to being able to modulate RNA expression levels, cross-regulations between these systems can also influence their toxicity. This suggests the idea that they are involved in an interconnected regulatory network.
Camille Riffaud, Marie-Laure Pinel-Marie, Brice Felden  Published:June 12, 2020

Toxin-Antitoxin Systems Are Abundant in Bacterial Genomes

Toxin-antitoxin (TA) systems are widespread genetic modules that encode both a stable toxic protein, whose overexpression can lead to growth arrest or cell death, and a corresponding labile antitoxin, which neutralizes the toxin’s activity during bacterial growth [1]. TA systems are classified into six types depending on the antitoxin’s nature and mode of action. While all the toxins are proteins, the antitoxins can be either noncoding RNAs (in type I and III systems) or low-molecular-weight proteins (in types II, IV, V, and VI) (Box 1). One exception is the DarTG system, a type II/IV TA hybrid system in which the DarG antitoxin interacts with and inhibits expression of the DarT toxin, but also acts on the target of DarT [,]. Antitoxins are unstable and are more susceptible than toxins to degradation by proteases, leaving toxins to interfere with cellular housekeeping processes such as replication, translation, and cell division [,]. The first discovery of TA systems was in the 1980s on Escherichia coli plasmids. These systems prevent plasmid loss during cell division via postsegregational killing [,]. In this process, if the daughter cell does not receive a TA system plasmid, the labile antitoxin is degraded and the stable toxin expresses its toxicity, killing the cells that lack the plasmid. Meanwhile, the bacteria that do inherit the plasmid are unaffected by toxin activation because of de novo and continuous antitoxin synthesis. Since their initial discovery, homologues of these plasmid-encoded TA systems have been identified in the chromosomes of many bacteria, including human pathogens such as Mycobacterium tuberculosis, Staphylococcus aureus, and enterobacteria. The biological role of chromosomal TA systems is still elusive even though significant scientific advances have been made []. In addition to maintaining the stability of genetic elements, TA systems are also involved in many other functions, including growth adaptation as a response to environmental stresses or persister cell formation [,,,,,]. However, contradictory papers have been published, and the involvement of TA systems in persister cell formation is now questioned [,,,,].

Type I toxin and antitoxin genes are independently transcribed from their own promoters. Their antitoxins are unstable cis- or trans-encoded antisense sRNAs that interact with toxin-encoding mRNAs by pairing, thereby inhibiting toxin mRNA translation and/or inducing its degradation (Figure IA). Under stress conditions, the RNA antitoxin pool is reduced by specific RNases which, in turn, induce toxin translation. Most type I toxins are amphipathic peptides with a predicted α-helical transmembrane domain, and they induce pore formation (TisB, HokB) or nucleoid condensation (DinQ, Fst, LdrD, BsrG) that cause defects in membrane potential and/or cell division. Some toxins (SymE and RalR) act within the cytosol and can cleave nucleic acids [].
Type II genes are in the same operon and are cotranscribed []. These antitoxins have two domains: a structured N-terminal one with DNA-binding motifs for binding the TA promoter and thus repressing transcription; and a C-terminal domain involved in toxin neutralization by protein–protein interactions. Antitoxins are susceptible to cellular proteases, allowing the activity of stable toxins (Figure IB). These toxins function by inhibiting various things: DNA gyrase action (CcdB, ParE); EF-Tu activity during translation (Doc); tRNA synthetase aminoacylation (HipA); or synthesis of tRNA (VapC, AtaT, TacT) or peptidoglycan precursor (PezT). However, most type II toxins are RNases, and are either ribosome-independent (MazF, Kid, PemK, MqsR) or ribosome-dependent (RelE, YoeB, HigB) [].
Type III toxin and antitoxin genes are cotranscribed and are in the same operon. A short, imperfect palindromic repeat between the genes forms a transcription terminator which controls the relative levels of toxin and antitoxin RNA. These toxins, such as ToxN from the ToxIN system, are all RNases that process the RNA antitoxin precursor into smaller RNAs and cleave host RNAs. ToxI antitoxin forms a secondary pseudoknot structure that binds to and blocks the toxin’s active site (Figure IC) [,]. This system induces an abortive phage infection by degrading phage RNAs, promoting the suicide of the phage-infected cells.
In type IV systems, the antitoxins and toxins do not interact. E. coli YeeV and CptA toxins prevent cell division by binding to MreB and FtsZ cytoskeletal proteins, inhibiting their polymerization. YeeU and CptB antitoxins counterbalance this toxic activity by stabilizing MreB and FtsZ cytoskeletal filaments (Figure ID) [].
The only type V module so far reported is the ghoST system in E. coli and Shigella spp. Here, the GhoS antitoxin is an RNase that inhibits GhoT toxin translation by mRNA cleavage during growth (Figure IE) []. Under stress conditions, ghoS mRNA is degraded by type II MqsR, resulting in the translation of GhoT, a hydrophobic peptide that disrupts cell membranes in the same way as do type I toxins.
The most recent detection of a type VI TA system was in C. crescentus []. Here, the SocA antitoxin is a proteolytic ‘adaptor’ protein that inhibits SocB toxicity by promoting its degradation by the ClpXP protease (Figure IF). SocB interferes with replication elongation by binding to the DnaN β clamp [].
Toxins (T) are red and antitoxins (A) are blue. SD points to the Shine–Dalgarno sequence.

The number of expressed TA loci vary extensively between species (Table 1). The abundance of both homologous and nonhomologous TA systems in bacterial genomes and, for some of them, their sequence identity, sequence homology, or similar mechanisms and functions, suggest that redundancy, and/or cross-regulations between them, is very likely to occur and interfere with bacterial physiology []. In this review we discuss the cross-regulations that have been reported between type I or II TA systems in various bacteria, and also those between TA systems of different types. Particular attention is given to cross-regulations that influence the RNA levels of other TA pairs, and to how these regulations affect the overall toxicity of the systems.

Cross-Regulations between Type I TA Systems

Many type I TA homologues were identified when either looking for small RNAs or subsequent to sequence comparisons when searching for conserved toxin amino acid residues and antitoxin nucleotide sequences (Table 1). Within the E. coli O157:H7 genome, 26 loci of type I TA systems (including 14 hok and 7 ibs genes) were found [], and recently seven more from TxpA and Fst families were identified in the S. aureus MW2 genome []. Below, we describe the putative cross-regulations between homologous type I TA systems in both Gram-negative and Gram-positive bacteria.

Cross-Regulations between Type I TA Systems in Gram-Negative Bacteria

The ibs/Sib type I TA systems, named ibsA to-E/SibA to-E, are expressed from five homologues on the E. coli K12 genome (Figure 1A and Table 1) []. Ibs peptides contain 18 or 19 amino acids and are located at the E. coli membrane. Their overexpression is lethal to E. coli because the cell envelope is disrupted. This toxicity can be repressed by the Sib RNA antitoxins through base-pairing with ibs toxins open reading frames []. Sib RNA antitoxins contain two domains, TRD1 and TRD2, that interact with ibs mRNA []. Only the SibA and SibE TRD2 sequences are identical, and a sibA-deleted strain is viable even when ibsA toxin expression is induced, which suggests that SibE might prevent toxicity by recognizing ibsA mRNA (Figure 1A) []. Additional experiments should be performed to highlight any direct interactions between SibE and ibsA mRNA and to explore the impact of these interactions on ibsA mRNA levels and toxicity. Thus, for that system, the current evidence for cross-regulations is weak.

(A) Putative cross-regulations between the five homologues of the ibs/Sib system in Escherichia coli. An RNA complex forms and creates putative crosstalk between the ibsA toxin mRNA and the noncognate SibE RNA antitoxin, allowing E. coli growth. (B) No cr
The zor/Orz system is another type I TA system which occurs in multiple homologues on the E. coli chromosome (Table 1). It is expressed from two gene pairs, zorO/OrzO and zorP/OrzP. An overproduction of ZorO is toxic for E. coli, but its toxicity is neutralized in an RNase III-dependent manner by the OrzO RNA antitoxin []. OrzO and OrzP antitoxins are structural homologues, with OrzP having a 15-nucleotide (nt) domain at the 5′ end that is predicted to interact with zorO mRNA, while the OrzO has an 18-nt sequence. However, toxicity induced by ZorO during growth is only rescued by OrzO, demonstrating the tight specificity of the zorO/OrzO system. A 3-nt variable region ‘V1’ was identified in OrzO and OrzP, and since an OrzP V1 mutant represses zorO expression and toxicity, V1 was been shown to be responsible for the OrzO RNA antitoxin’s specificity for the zorO target mRNA [
]. So, in fact, just mutating the three nucleotides located in the mRNA toxin interaction domain is sufficient to confer specificity to two antitoxin homologues.
Five homologues of the aapA/IsoA TA systems are found in Helicobacter pylori strains (Table 1) [,]. Overexpression of the 30-amino acid peptide encoded by aapA1 induces cell death, and aapA1 toxin translation is prevented by mRNA folding, RNA processing, and by the IsoA1 RNA antitoxin []. Despite the fact that IsoA1 and IsoA3 have similar secondary structures in solution, footprint assays did not detect interactions between aapA1 and IsoA3 RNA, which suggests that toxin neutralization is enabled by the tight specificity between aapA and IsoA RNA [].
As for the hok/Sok and ldr/Rdl TA systems, all that is known is that they are abundant in enterobacterial genomes (Table 1). For hok/Sok homologues, many are nonfunctional due to mutations, insertions, or large rearrangements in E. coli [,,]. Moreover, it was demonstrated that sokE and sokX loci can be expressed without hok transcripts, suggesting that these antitoxins could possess additional roles beyond toxicity neutralization [,]. However, nobody has yet looked for cross-regulations between the multiple homologues, so this could warrant investigation [,,].

Cross-Regulations between Type I TA Systems in Gram-Positive Bacteria

Two type I TA systems, sprA/SprAAS and sprG/SprF, are found in two and four homologues, respectively, in the S. aureus genome (Table 1) []. The overexpression of toxic membrane peptides encoded by sprA and sprG RNA triggers S. aureus stasis or death [,,,]. Expression of SprA toxins is prevented in trans by SprAAScis-RNA antitoxins when they pair at the internal ribosome binding site (RBS) and thus inhibit toxin translation [,]. By contrast, SprF antitoxins prevent toxin translation by interacting in cis with their overlapping regions [,]. Despite high sequence identities, growth kinetics between noncognate pairs demonstrate a lack of cross-regulation between the two sprA/SprAAS TA homologues and between the four sprG/SprF TA homologues (Figure 1B,C) [,]. However, sprG3-SprF1 and sprG2-SprF4 RNA complexes were detected in vitro, which might suggest that cross-regulation could occur in vivo between the sprG3/SprF1 and sprG2/SprF4 RNA pairs. RNA level variations during S. aureus growth were observed by northern blot analysis in strains overexpressing SprF1 (which decreased sprG2 and sprG3 RNA levels); sprG2 (decreased SprF1, SprF2, and SprF3); sprG2/SprF2 (increased sprG1); or SprF4 (decreased sprG1 and sprG2) []. The sprG4 toxin is not expressed in S. aureus HG003 because sprG4 lacks an upstream promoter sequence, suggesting that SprF4’s primary function is to interact with and regulate sprG2 mRNA expression. Together, these results show that in each sprG/SprF TA system, SprG toxicity control is specific to the cognate antitoxin, but that cross-regulations between the different sprG/SprF TA systems can modulate the RNA levels of homologues (Figure 1C). Further experiments are needed to investigate the physiological relevance of this sort of network in the S. aureus sprG/SprF TA pairs.
So, in all type I TA systems found in multiple homologues, cross-regulations have been documented only between homologous systems, with no studies yet published on nonhomologous cross-regulation. In Gram-negative bacteria, toxicity-affecting cross-regulations were only observed between the zorO/OrzO and zorP/OrzP systems (in site-directed mutagenesis assays), and they were suggested for ibsA/SibE. A lack of cross-regulation of toxicity in Gram-positive bacteria does not, however, exclude possible regulation at the post-transcriptional level, as demonstrated to exist in the sprG/SprF TA systems.

Cross-Regulations between Type II TA Systems

Type II TA systems are also abundant in bacterial genomes (Table 1). In M. tuberculosis, 79 systems were identified after research in the toxin–antitoxin database ( and through a homologue search of the M. tuberculosis genomes for toxin and antitoxin protein sequences from each of eight major type II families (CcdBA, HigBA, HipBA, MazEF, ParDE, RelBE, VapBC, and Doc/PhD). These type II systems include 50 from the VapBC family and 10 from the MazEF family, likely the result of extensive gene duplication events (Table 1) [,,,]. Cross-regulations between type II TA systems are more documented than for the type I systems as type II antitoxins and toxins possess global regulatory functions. Type II antitoxins possess a DNA-binding domain and act as transcriptional regulators via their interactions with TA gene promoters. Most type II toxins are endoribonucleases (RNases) that inhibit bacterial translation by cleaving mRNAs when loaded onto translating ribosomes (in the cases of RelE, MazE, Kid, and MqsR), or by targeting tRNAs (VapC, TacT, and AtaT) [,,]. In this section we describe the cross-regulations between homologous or nonhomologous type II TA systems. In fact, cross-regulations between type II TA systems can occur both between systems from the same replicon (whether encoded by plasmids or chromosomes) and between systems that have different origins.

Cross-Regulations between Type II TA Systems in Gram-Negative Bacteria

Cross-Regulations between Homologous Type II TA Systems

Cross-regulations can occur between homologues where one is encoded by chromosomes and the other by plasmids. Investigations were conducted on the CcdAB TA system expressed from the F plasmid and the chromosome of the pathogenic E. coli O157:H7 strain, and it was shown that the plasmid CcdB toxin is not neutralized by the chromosome-encoded CcdA antitoxin [,]. Other examples of studies between differently-encoded systems in E. coli include the par locus located on the pR100 plasmid with the chromosome-encoded MazEF TA system, but they turned out to be specific [,]. In E. coli, overproduction of the MazE antitoxin (a homologue of chromosomal Kis) rescues the toxicity induced by Kid toxin overexpression in the presence of a mutated version of the Kis antitoxin []. This cross-regulation between MazE and Kid probably occurs due to structural similarities between the antitoxins as a result of a protein–protein complex formation [].
The first study of cross-regulations between homologous chromosome-encoded type II TA systems in Gram-negative bacteria was described in Caulobacter crescentus, an α-proteobacterium. There are four homologues of the ParDE and RelBE TA systems (Table 1) []. ParD and RelB proteins act as antitoxins against ParE and RelE toxicity. Overexpression of the ParE1 toxin inhibits cell division, causing filamentation, but cell growth is unaffected. There is no cross-regulation between the ParDE and RelBE homologues [
Another example of specificity between homologous type II TA systems was reported in E. coli O157:H7 for the two paaR–paaA–parE systems which contain the PaaR regulator, the PaaA antitoxin protein, and the ParE bactericidal toxin [,]. ‘Plate toxicity’ assays prove that no cross-regulation occurs between these two homologues, as each PaaA antitoxin specifically counteracts the toxicity of its associated ParE toxin [].
At the amino acid sequence level, Polom et al. investigated the specificity of TA interactions between plasmid-encoded Axe–Txe and chromosome-encoded YefM–YoeB homologues. No cross-interactions were detected between the wild-type proteins. However, a single N83Y amino acid substitution which converts a Txe-specific residue to a YoeB-specific one will allow functional interaction between the mutated Txe toxin and the YefM antitoxin, resulting in reversal of the toxicity []. Using the same approach in Haemophilus influenzae, Walling and Butler showed that VapC1 and VapC2 toxicity is neutralized by a VapB1 T47W mutated antitoxin that has an additional tryptophan residue which is involved in the VapB2–VapC2 interaction []. These results indicate that type II antitoxin specificity can be decreased by a single mutation at an essential residue.

Cross-Regulations between Nonhomologous Type II TA Systems

Smith et al. studied cross-regulations between the CcdAB TA system produced by the F plasmid and the parD locus from the R1 plasmid in E. coli, a system which contains the Kis antitoxin and the Kid toxin []. CcdB and Kid toxins have different targets, with CcdB targeting DNA gyrase and Kid targeting RNA via an RNase action. Despite the different cellular targets, these toxins both have a similar structural fold, leading the authors to conclude that cross-regulations are possible. Kid toxicity is rescued by the Kis cognate antitoxin and increased by the CcdA antitoxin, while conversely, the overexpression of Kis antitoxin inhibits CcdB toxicity, although less efficiently than CcdA. By testing in vitro the RNase activity of Kid onto CopT RNA in the presence of CcdA antitoxin, Smith et al. showed an unexpected and unexplained stimulation of the RNase activity of Kid by CcdA. Direct interactions between toxins and antitoxins from nonhomologous complexes were also demonstrated, suggesting that cross-regulations between the ccd and parD systems are possible (Figure 2A ) [].

The red arrows indicate events related to the presence of cross-regulations. (A) Cross-regulations between the nonhomologous, plasmid-encoded CcdAB and Kis/Kid TA systems in Escherichia coli. The complex identified between the Kis antitoxin and CcdB toxin

Cross-Regulations between Different Types of TA System

Cross-regulation between different types of TA system was initially discovered between the MqsR/MqsA type II and GhoT/GhoS type V systems. In its active form, the GhoT toxin is located at the bacterial membrane and induces a ghost-cell phenotype. This is rescued by the GhoS antitoxin when it specifically cleaves and inactivates the ghoT mRNA []. The MqsA antitoxin represses stationary-phase sigma factor RpoS and influences various stress responses [], while 5′-GCU sites are targeted by MqsR RNase mRNA cleavages [,]. MqsR regulation of ghoST mRNA stability was investigated, and ghoT mRNA is enriched in the presence of the MqsR toxin. On the other hand, introducing a cleavage site recognized by MqsR into the ghoT mRNA will reduce GhoT toxicity and cell death, because MqsR degrades ghoT mRNA (Figure 3A ) []. Only 14 mRNA transcripts in E. coli lack an MqsR-induced 5′-GCU cleavage site, and the ghoS mRNA is an MqsR target [,]. During stress, such as the presence of an antibiotic, the MqsR toxin degrades the ghoS antitoxin, allowing GhoT toxin translation and ghost-cell formation. This is the first description of a TA system being regulated by another type of TA system (Figure 3A) [].

The red arrows indicate events related to the presence of cross-regulations. PghoST, PmqsRA, PmazEF1-P2, PtxpA, and PratA are TA promoters. T, toxin; A, antitoxin. (A) Cross-regulations between the ghoST type V TA system and the mqsRA type II system in Es
Cross-regulation between different types of TA system was also observed in V583, a clinical isolate of Enterococcus faecalis. A locus encoding two adjacent TA modules was discovered in its genome. One module is related to the type I TxpA family; the other is related to the type II MazEF family []. In E. coli, this newly predicted MazEF-like system turned out to be a type II TA system, as MazF is toxic when RNA degradation is triggered, but this effect is counteracted by the MazE antitoxin. In fact, the MazEF complex presents dual regulatory activity, both repressing mazEF operon expression and activating RatA RNA antitoxin transcription []. Indeed, overexpression of mazEF in trans induces a ~2.5-fold increase in mazEF and RatA RNA levels, in comparison with a control strain with the empty vector. A common sequence encompassing the –10 boxes, the transcription start sites (TSSs) and the 5′ transcribed region, was identified in the PmazEF and PratA promoters (called drm1 for PmazEF and drm2 for PratA). In a double mutant generated to abolish transcription initiation of txpA and ratA by mutagenesis of the –10 boxes (‘–10txpARatA’), the RatA RNA level was increased as compared with an isogenic strain, suggesting that the interaction of MazEF complex with the drm2 sequence activates RatA transcription []. This is a case of cross-regulation between a type II antitoxin and a type I TA module (Figure 3B).
To summarize, in this section we reported on two examples of cross-regulation between different types of TA system: between type II and V, and between type I and II. In the second example, from E. faecalis, cross-regulation is favored by the genomic proximity of the two TA systems. Cross-regulations between three homologous type IV TA systems were also described in E. coli []. Given the abundance of these systems in bacterial genomes, cross-regulations between different types of TA system remain a topic worthy of further exploration, and these examples may present just the tip of the iceberg.

Concluding Remarks

In this review we present recent data on TA systems which act individually or within a gene network (see Outstanding Questions). This network contains cross-regulations between homologous and nonhomologous TA systems that may or may not be of the same types and origins. These subtle cross-regulations are specific to each bacterial species and must play a fundamental role in species such as M. tuberculosis, E. coli, and S. aureus where a high number of TA systems are found. This implies the existence of sophisticated networks in bacteria between all the TA systems to control RNA levels and/or toxicity (Figure 4, Key Figure). However, in cases where crosstalks were described, it is either under overexpression conditions and/or with mutated antitoxins. Therefore, the biological relevance of all these interactions is unclear, especially because the roles of TA systems in bacterial physiology is debated [,]. However, TA diversity, abundance, and cross-interaction were all selected during continuous host–pathogen warfare and/or partnerships in order to increase fitness. A deeper understanding of the many and varying functions of TA systems will require better characterization of their cross-interactions, as well as the identification of the functions of each TA module or RNA in the pair. Type II antitoxins are global regulators of gene expression via their DNA-binding domains, responding to a plethora of environmental stimuli (Figure 4). Type II toxins are endoribonucleases inducing depletion of type II antitoxins by translation inhibition, but they also lead to the accumulation of toxin-encoding mRNA fragments upon cleavage (Figure 4). This toxin production constitutes a possible positive feedback loop which contributes to a complex regulatory network that controls bacterial growth. Moreover, small RNAs are widely found in bacteria, regulating many RNA and/or protein targets. This raises an interesting question. RNA antitoxins from type I and III TA systems counteract their respective toxins, but what if they also have sRNA regulatory functions? Cross-regulations can arise between an antitoxin and several toxins from different systems [,,,,], and they can both improve toxicity regulation and, conversely, titrate noncognate antitoxins. Instead of individual entities we should be considering TA systems as members of intricate and connected gene networks. In the near future, we predict a much better understanding of the many important effects of TA systems in bacterial physiology as they are being actively investigated worldwide.

Outstanding questions

  • Are cross-regulations between homologous and nonhomologous TA systems exceptions or the rule?
  • To what extent are TA systems interconnected?
  • Are there cross-regulations involving type III, IV, and VI TA systems?
  • Can RNA antitoxins and type II antitoxins be considered as global regulators with targets other than their toxins?
  • How do cross-regulations between bacterial TA systems affect their functions?

Red arrows indicate cross-regulations between TA systems. Type I TA systems (blue) can interact with their homologues, sometimes via RNA complex formation, to regulate the RNA levels of other type I TA systems. However, type II TA systems (green) appear t

Acknowledgments and references

We are thankful to Juliana Berland for help with editing the manuscript. C.R. is the recipient of a fellowship funded by the French Ministère de l’Enseignement Supérieur et de la Recherche (grant MENRT) and the School of Pharmacy and Medical Sciences of the ' Université de Rennes 1 '. This work was funded by the Agence Nationale pour la Recherche (grant number ANR-15-CE12-0003-01 'sRNA-Fit' to B.F.); by the Fondation pour la Recherche Médicale (grant number DBF20160635724 'Bactéries et champignons face aux antibiotiques et antifongiques' to B.F.); and by the Institut National de la Santé et de la Recherche Médicale.

See references on this link.