OUP user menu

Toxin–antitoxin systems are ubiquitous and versatile modulators of prokaryotic cell fate

Christopher F. Schuster, Ralph Bertram
DOI: http://dx.doi.org/10.1111/1574-6968.12074 73-85 First published online: 1 March 2013


Toxin–antitoxin (TA) systems are small genetic elements found on plasmids or chromosomes of countless bacteria, archaea, and possibly also unicellular fungi. Under normal growth conditions, the activity of the toxin protein or its translation is counteracted by an antitoxin protein or noncoding RNA. Five types of TA systems have been proposed that differ markedly in their genetic architectures and modes of activity control. Subtle regulatory properties, frequently responsive to environmental cues, impact the behavior of TA systems. Typically, stress conditions result in the degradation or depletion of the antitoxin. Unleashed toxin proteins impede or alter cellular processes including translation, DNA replication, or ATP or cell wall synthesis. TA toxin activity can then result in cell death or in the formation of drug-tolerant persister cells. The versatile properties of TA systems have also been exploited in biotechnology and may aid in combating infectious diseases.

  • toxin
  • antitoxin
  • persisters
  • stress response
  • physiology
  • applications
  • review


The first toxin–antitoxin (TA) system was discovered in the Escherichia coli F plasmid. It was noticed that the bicistronic locus ccdAB (for ‘coupled cell division’) ensured faithful plasmid partitioning upon cell division. Plasmid-free cells were eradicated by the action of the stable CcdB protein, the toxic effect of which was abrogated by antitoxin CcdA (Ogura & Hiraga, 1983; Jaffé et al., 1985). A similar outcome for plasmid-free cells was subsequently deduced from the small hok–sok system of the R1 plasmid, which, however, employed a different mode of action, regulation, and genetic architecture, reflected by the fact that Hok was a toxic protein but sok a small noncoding RNA antitoxin (Gerdes et al., 1986b). According to the chemical nature of the antitoxin, CcdAB and hok–sok are regarded as canonical type II or type I TA systems, respectively. Later on, numerous related genetic elements associated with the phenomenon of plasmid addiction were identified (Hayes & Van Melderen, 2011). It is common to all known TA systems that they encode a stable toxin protein that can cause cell death or cell arrest and an RNA or protein antitoxin that can preclude toxicity (Gerdes et al., 2005). Under a variety of stress conditions (Christensen et al., 2001, 2003; Vogel et al., 2004), the antitoxin is either proteolysed or its transcription is changed. This unleashes the toxin, which thereupon exerts an inhibitory effect on a specific and frequently essential bacterial target.

Over the years, TA systems have also been identified in the chromosomes of countless bacteria (Makarova et al., 2009; Fozo et al., 2010; Blower et al., 2012b), archaea (Grønlund & Gerdes, 1999), and possibly also in yeast (Satwika et al., 2012) and other fungi (Yamaguchi et al., 2011). It appears that some bacteria are devoid of TA systems, whereas others harbor several dozens of them (Pandey & Gerdes, 2005). Chromosomal location of TA systems implies functions extending beyond plasmid addiction (Magnuson, 2007). In particular, the revelation of causal relationships between TA systems, bacterial growth impediment, and multidrug tolerance of persister cells has recently sparked great interest in the field (Lewis, 2010; Gerdes & Maisonneuve, 2012). TA systems with novel modes of action have been revealed in the last years (Mutschler et al., 2011; Tan et al., 2011; Holberger et al., 2012; Masuda et al., 2012a, b; Wang et al., 2012). We here provide current insights into TA systems regarding different types and categories, regulatory properties, targets of toxicity, physiological roles, and applications in bacterial genetics and biotechnology.

Types of TA systems

Five types of TA systems have been proposed to date (Fig. 1 and Table 1). All of them comprise a toxin protein and an antitoxin that can be either a small noncoding RNA (sRNA) in type I and type III TA systems or a low molecular weight protein in TA types II, IV, and V. The genetic architecture and the mode of toxin activity control by the antitoxin differ markedly between these types as outlined in the following.

View this table:
Table 1

The five currently known TA system types

Chemical nature of
ToxinAntitoxinMode of action or target of the toxinToxin inactivationExamplesReferences
Type IProteinRNAHydrophobic proteins that interfere with membranes,
possibly also endoribonucleases
mRNA (T) — RNA (AT) interactionHok/Sok, TisAB/IstR SymERGerdes et al. (1986a), Vogel et al. (2004), Kawano et al. (2007)
Type IIProteinProteinGyrase inhibitors
Ribosome (in)-dependent endoribonucleases
Ribosome association inhibition
EF-TU phosphorylation
UNAG phosphorylation
Protein (T) — Protein (AT) interactionCcdAB
MazEF, RelBE, VapBC
Ogura & Hiraga (1983), Bernard & Couturier (1992), Ceglowski et al. (1993), Lehnherr et al. (1993), Black et al. (1994), Aizenman et al. (1996), Gotfredsen & Gerdes (1998), Christensen et al. (2003), Pedersen et al. (2003), Zhang et al. (2004), Zielenkiewicz & Ceglowski (2005), Liu et al. (2008), Schumacher et al. (2009),Mutschler et al. (2011), Winther & Gerdes (2011)
Type IIIProteinRNAEndoribonucleaseProtein (T) — RNA (AT) interactionToxINFineran et al. (2009), Blower et al. (2011b)
Type IVProteinProteinDestabilization and inhibition of MreB and FtsZ polymerizationProtein (AT) — target structure interactionCbtATan et al. (2011), Masuda et al. (2012b)
Type VProteinProteinHydrophobic proteins that interfere with membranesCleavage of mRNA (T) by Protein (AT)GhoSTWang et al. (2012)
  • Currently only one known representative of this type. T, Toxin; AT, antitoxin.

Schematic representations of the currently known TA types. Toxin genes and proteins are depicted in red color, antitoxins (AT) in blue, DNA as sinus curves. (a) Type I: the toxin is usually a hydrophobic peptide whose mRNA transcript is intercepted by the RNA antitoxin (the overlapping antisense regions can be differently distributed). (b) Type II: toxin and antitoxin are proteins, and the toxicity is prevented by formation of a TA complex. (c) Type III: the RNA antitoxin consists of 5.5 repeats (short, blue arrows), is processed, and inhibits toxicity by directly binding the toxic protein. (d) Type IV: the toxin protein inhibits the target molecules (orange box), whereas the antitoxin molecules counteract these effects. (e) Type V: the mRNA of the small toxin-encoding ORF is cleaved by the antitoxin.

Type I

In a typical type I TA system (Fig. 1a), a protein-coding toxin gene lies adjacent to an antitoxin sRNA gene of opposite orientation, often resulting in overlapping transcripts (Gerdes et al., 1997; Gerdes & Wagner, 2007; Fozo et al., 2008a; Durand et al., 2012). Type I toxins are usually small (about 20–65 amino acids) and exhibit hydrophobic properties (Fozo et al., 2010). Most of them are assumed to act by disrupting the membrane potential (Gerdes et al., 1986b; Fozo et al., 2008b), by a mechanism not yet fully understood. Downregulation of toxin activity occurs post-transcriptionally as the sRNA antitoxin scavenges the toxin mRNA by an antisense mechanism. Concomitant inhibition of translation is usually accomplished by the degradation of RNA duplexes and/or masking of the ribosome binding site (Vogel et al., 2004; Darfeuille et al., 2007; Fozo et al., 2008b).

Type II

In type II TA systems, both toxin and antitoxin are small proteins (Hayes & Van Melderen, 2011; Wang & Wood, 2011; Yamaguchi & Inouye, 2011; Yamaguchi et al., 2011; Fig. 1b). Canonically, the antitoxin gene precedes and frequently overlaps the toxin gene in a bicistronic operon, but this order can also be reversed (Tian et al., 1996). Binding of the usually flexible antitoxin protein results in steric changes in the toxin or blocking of critical sites for its action. In addition to toxin inactivation, type II TA antitoxins are typically also transcriptional repressors that autoregulate their own operons. Rarely, the dual antitoxin function of transcriptional regulation and toxin inhibition is allocated to two separate proteins in tripartite TA systems (Smith & Rawlings, 1997; Zielenkiewicz & Ceglowski, 2005; Hallez et al., 2010). To activate the toxin (e.g. on stress, see below), the antitoxin component is degraded by the Clp (Lehnherr & Yarmolinsky, 1995; Aizenman et al., 1996; Donegan et al., 2010) or the Lon protease (Van Melderen et al., 1994, 1996; Smith & Rawlings, 1998; Christensen et al., 2001; Hansen et al., 2012). Characteristics like the coupled arrangement of two small open reading frames (ORFs), conserved domains, and sequences facilitate type II TA systems' prediction in silico (Sevin & Barloy-Hubler, 2007; Makarova et al., 2009; Leplae et al., 2011; Shao et al., 2011). The type II TA systems have been grouped into different classes according to their cognate toxin and antitoxin pairs (Gerdes et al., 2005). Recent reports, however, cast doubts on the notion of strictly pairwise T–A interaction and rather corroborate the ‘mix and match’ principle which suggests that type II TA toxins can also interact with antitoxins from different classes (Guglielmini & Van Melderen, 2011; Leplae et al., 2011). It thus seems reasonable to categorize type II TA toxins and antitoxins independently of each other and based on a bioinformatical approach. Leplae et al. (2011) came up with a list of 12 toxin and 20 antitoxin superfamilies. Crystal structures of a number of type II TA single components or complexes have been revealed, including MazEF (Kamada et al., 2003), YefM/YoeB (Kamada & Hanaoka, 2005), RelBE (Takagi et al., 2005), MqsRA (Brown et al., 2009), or VapBC2 (Maté et al., 2012). According to structural properties, six different classes of type II TA toxins have been proposed, namely Kid, RelE, Doc, VapC, HipA, and ζ (Blower et al., 2011a). To exert transcriptional autoregulation, type II TA antitoxins act alone or in complex with suitable toxin proteins. As found recently, the outcome of transcriptional repression or derepression is dictated by the stoichiometry between toxin and antitoxin proteins, termed conditional cooperativity (Overgaard et al., 2008; Garcia-Pino et al., 2010; Maisonneuve et al., 2011). For example, the relEB locus is repressed by RelB antitoxin dimers or the ternary RelB2–RelE complex, whereas the heterotetramer RelB2–RelE2 is incapable of relEB promoter binding (Overgaard et al., 2008; Boggild et al., 2012). Recently, the regulation dynamics of the relEB system, in which the antitoxin can be degraded in the either free or toxin-bound state, was assessed mathematically (Cataudella et al., 2012). The underlying models indicate that individual cells binarily switch between a high or low toxin level state. The switch back to low toxin levels can occur rapidly due to conditional cooperativity, which also ensures that toxicity is not accidentally unleashed. Another recent study addressed the regulatory network of the E. coli hipBA locus involved in persister cell formation (Koh & Dunlop, 2012). According to the mathematical model, the entry into a persister state is caused by stochastic fluctuations that could modulate the frequency of persister formation.

Type III

The ToxIN locus from Pectobacterium atrosepticum, the so far only validated type III TA system (Fig. 1c), consists of the endoribonuclease ToxN and the RNA antitoxin ToxI (Blower et al., 2009, 2011b; Fineran et al., 2009). ToxI consists of five complete plus one incomplete nearly identical 36-nt long repeats and is processed to inhibit the toxin by protein–RNA interaction. This TA system is important in phage defense (Fineran et al., 2009) by employing a mechanism similar to apoptosis in eukaryotes. Infection of a bacterial cell by a phage activates ToxN, induces cell death, and thus prevents the spreading of the virus in the bacterial population (abortive infection). New experiments, however, show that mutants of a Myoviridae virus themselves encode the antitoxin toxI to overcome this line of defense (Blower et al., 2012a). It was shown that the operon is negatively autoregulated, and it was hypothesized that on phage attack, the TA levels get out of balance and thus lead to toxicity (Blower et al., 2009). Recent in silico analyses suggest three different type III TA families distributed over a range of different prokaryotic phyla (Blower et al., 2012b).

Type IV

Very recently, a type IV TA system was proposed (Fig. 1d; Masuda et al., 2012b). It includes the toxin protein CbtA that inhibits polymerization of the pivotal cell shape and division proteins MreB and FtsZ and the antitoxin CbeA (Tan et al., 2011). In a novel mode of TA antitoxin action, CbeA stabilizes MreB and FtsZ polymers and thereby counteracts not only the toxin CbtA but also other MreB/FtsZ polymerization inhibitors. The inhibitory effect of CbeA onto CbtA activity is thus achieved by shielding of the toxin target rather than by interaction with the toxin itself. CbtA and CbeA thus show functional analogy to bacterial DNA restriction–modification systems (Dziewit et al., 2011). It is debatable, if this new type of TA system classification is justified, as the proposed mechanism that involves antagonizing but not directly interacting factors does not fulfill the criteria of established TA systems.

Type V

In the E. coli GhoST type V TA system, the activity of the ghoT toxin gene is regulated post-transcriptionally by the antitoxin GhoS (Fig. 1e; Wang et al., 2012). GhoS is a sequence-specific endoribonuclease primed for ghoT mRNA degradation. This unique mode of toxin inactivation is the principal distinctive criterion for the type V TA system, which is otherwise genetically similar to type II TA loci. Also, in contrast to type II TA systems, GhoS is stable and not a transcriptional regulator of its own operon. Overexpression of the small hydrophobic toxin GhoT causes cell membrane damage and ultimately lysis resulting in so-called ghost cells. While E. coli GhoST is the only characterized type V TA system to date, orthologues of GhoS are found in Serratia and Erwinia (own blast searches) and possibly other organisms.

Targets of TA toxins

TA toxins impede vital cellular functions by corrupting a variety of target structures and processes (Fig. 2 and Table 1). The small hydrophobic type I TA system toxins generally appear membrane associated, resulting in proton-motive force perturbations and impediment of ATP synthesis (Fozo et al., 2008a, 2010). A similar mode of action is assumed for the type V TA toxin GhoT (Wang et al., 2012). An exception among the type I TA toxins is SymE, which probably acts as an RNase (Kawano et al., 2007), as do many of the type II TA toxins: endoribonucleases of the RelE family (Christensen & Gerdes, 2003), such as YafQ (Motiejūnait≐ et al., 2007) and YoeB (Christensen et al., 2004), target mRNAs associated with the ribosomal A site. Exceptionally, the RelE-related MqsR (YgiU) toxin (Christensen-Dalsgaard et al., 2010) can cut mRNA in a ribosome-independent, sequence-specific manner. This also pertains to the phylogenetically widespread MazF toxin (Zhang et al., 2003). Its RNA cleavage specificity differs between homologues, which has intriguing implications for gene regulation. The toxin ToxN of the type III TA system ToxIN is structurally related to the endoribonuclease toxins Kid and MazF, and preliminary experiments suggest that ToxN can also cleave RNA (Blower et al., 2012b). Initiator tRNAfMet molecules are cut by toxin VapC, which may result in translation initiation at elongator codons of otherwise silent genes (Winther & Gerdes, 2011). Inhibition of protein synthesis by the HipA toxin is realized by phosphorylation of the elongation factor EF-Tu, which blocks its interaction with aminoacyl tRNA (Schumacher et al., 2009). The E. coli TA toxins Doc and RatA [also known as YfjG, not to be confused with the type I antitoxin RatA of Bacillus subtilis (Silvaggi et al., 2005)] both target bacterial ribosomes. RatA binds to the ribosomal 50S complex and thereby prevents association with the 30S subunit (Zhang & Inouye, 2011), whereas Doc interacts with the 30S complex to inhibit translation elongation (Liu et al., 2008). Also, topoisomerase function is affected by TA toxins. Gyrase is corrupted both by CcdB (Bernard & Couturier, 1992) and ParE (Jiang et al., 2002) via direct interaction with the GyrA subunit. The two toxins, however, apparently employ two distinct mechanisms, as ParE requires ATP and the binding sites of CcdB and ParE to GyrA are most likely different (Yuan et al., 2010). Remarkably, although deficient in RNase activity, toxin CcdB is structurally related to the endoribonucleases MazF and Kid. Indeed, structure–function delineations are not straightforward in TA systems, as also exemplified by the Kid and RelE families. Despite structural differences, they share the same RNase T1-like mechanism, suggesting convergent development (Blower et al., 2011a). Evolutionary and structural relationships among TA systems have been reviewed (Arbing et al., 2010). As discovered only recently, TA toxins may also target the bacterial cell envelope. To this end, the ζ toxin of the ε/ζ type II TA system phosphorylates the cell wall precursor uridine diphosphate-N-acetylglucosamine (UNAG) at the N-acetylglucosamine 3′-hydroxyl group and therefore no longer functions as a peptidoglycan precursor (Mutschler & Meinhart, 2011; Mutschler et al., 2011). As described above, the cell shape proteins MreB and FtsZ are targeted by toxins CbtA and CptA (Tan et al., 2011; Masuda et al., 2012a). Finally, protease activity is assumed as the causative toxic function of toxin IetS from a tumor-inducing Agrobacterium plasmid (Yamamoto et al., 2009), but this requires further characterization.

Targets of known TA systems as validated today. A schematic bacterial cell, including inner membrane and peptidoglycan layer, is depicted. Other components of the bacterial cell envelope are omitted for simplicity. (1) Membrane disruption (type V and most type I toxins), (2) inhibition of cell wall synthesis by UNAG phosphorylation (ζ-toxin), (3) cleavage (scissors symbol) of free mRNAs by endoribonucleases (e.g. MazF, PemK, ToxN), (4) cleavage of rRNA, leading to ‘leaderless’ ribosomes (MazF), (5) ribosome-dependent cleavage of mRNA (RelE), (6) protein synthesis inhibition at 30S ribosome subunit (Doc), (7) phosphorylation of EF-Tu (HipA), (8) inhibition of ribosome association (RatA), (9) cleavage of tRNAfMet (VapC), (10) gyrase inhibition (CcdB, ParE), (11) destabilization of MreB and FtsZ polymerization (CbtA), (?) possibly serine protease function (IetS).

TA systems could play a multitude of roles in physiology

Stabilization of episomal elements by plasmid-encoded TA systems appears evident as in the case of ccdB (Ogura & Hiraga, 1983; Gerdes et al., 1985; Gerdes & Wagner, 2007), whereas defining roles of chromosomally encoded systems is less straightforward (Tripathi et al., 2012).

The significance of a number of physiological roles of TA systems is under constant debate (Table 2; Magnuson, 2007). A plethora of studies provide compelling evidence that chromosomal TA systems are major players in physiological downshifts of prokaryotes. Whether such changes are reversible or initiate a programmed cell death pathway is controversial. The essential roles of TA systems in the bacterial persister phenotype have, however, been well established (Lewis, 2010; Gerdes & Maisonneuve, 2012). Persisters are phenotypic variants of bacterial or unicellular fungal cells that are much less sensitive to antibiotics than most other cells in an isogenic population (Lewis, 2010). It has been conjectured that biofilms constitute reservoirs for persisters and accordingly, MazEF and MqsRA among other TA systems, are important in biofilm formation (Wang & Wood, 2011). As reported recently, deleting the mqsRA locus alone (Kim & Wood, 2010), or 10 different TA systems simultaneously, markedly decreased the antibiotic tolerance of E. coli cultures. Accordingly, when the lon protease was overproduced in wild-type E. coli, persister levels increased, possibly due to the degradation of antitoxin molecules and concomitant toxin activation (Maisonneuve et al., 2011). A causal relationship between the persister phenotype and TA systems has been established in case of the type I TA TisB toxin, which is part of the SOS regulon (Dörr et al., 2010). TA toxins active as RNases appear to regulate genes post-transcriptionally. For example, MazF of Staphylococcus aureus controls expression of virulence factors and other genes by selective mRNA cleavage. Unusual abundances of MazF cleavage sites indicate respective target genes in staphylococci (Zhu et al., 2009; Schuster et al., 2013). As described recently, the global influence of TA systems can extend beyond the modulation of the bacterial transcriptome. Selective RNA cleavage of E. coli MazF results in the production of specialized ribosomes with modified 16S rRNA gene, capable of specifically translating leaderless mRNAs (Vesper et al., 2011; Moll & Engelberg-Kulka, 2012). Type III (Blower et al., 2009), type I (Pecota & Wood, 1996), and possibly also type II (Hazan & Engelberg-Kulka, 2004) TA systems have also been proposed to serve as an antiphage infection mechanism. Affected cells could selectively be eradicated to ensure survival of the uninfected subpopulation. For Myxoxoccocus xanthus, the involvement of an orphan mazF gene in programmed cell death (PCD) during fruiting body formation was reported (Nariya & Inouye, 2008). Accordingly, also E. coli mazEF has been suggested to be responsible for PCD (Aizenman et al., 1996; Amitai et al., 2004; Engelberg-Kulka et al., 2006), but this claim has been disputed (Pandey & Gerdes, 2005; Tsilibaris et al., 2007). The ability to produce proteins over days in the single protein production system (Suzuki et al., 2005; described later in detail) and the translation of leaderless mRNAs by MazF-modified ribosomes (Vesper et al., 2011) also argue against this view. Besides plasmids, also other mobile genetic elements (MGEs) could be stabilized by TA systems, as in the case of the integrative and conjugative element SXT from Vibrio cholerae (Wozniak & Waldor, 2009). According to the controversial anti-addiction model (Saavedra De Bast et al., 2008), chromosomal TA systems could counteract toxicity from TA components encoded on MGEs, rendering the host resistant to postsegregational killing. Eventually, some TA systems might just be selfish genes devoid of physiological roles. This could be particularly true for evolutionarily late acquired systems that have not (yet) been implemented into the physiology framework of the cell. The vast differences between individual TA systems suggest assessing their physiological role(s) on a case to case basis, bearing in mind that more than one function might be exerted.

View this table:
Table 2

Validated and hypothesized roles of TA systems

Gene regulationmazEF
Genes or proteins can be regulated by toxin and/or antitoxin components. Specific cleavage of transcripts by toxins with RNase activityZhu et al. (2008), Kim et al. (2010)
Growth control/stress responsemazEF
Stress conditions such as nutritional stresses can trigger TA loci. Adaptation to changing environmentsChristensen et al. (2001), Hazan et al. (2004), Wang et al. (2011)
Programmed cell arrestrelEB
Toxin-induced reversible cell growth arrest for long-time survivalPedersen et al. (2002)
Persister cellsmultiple, mqsAR
Deletion of single or multiple TA systems, or specific mutations (e.g. hipA7) results in increased overall drug tolerance of a cultureMoyed & Bertrand (1983), Kim & Wood (2010), Maisonneuve et al. (2011)
Programmed cell deathmazEFAltruistic PCD of individuals to confer a fitness advantage to the population (e.g. nutrient supply). Under debateAmitai et al. (2004), Engelberg-Kulka et al. (2005), Hu et al. (2010)
Phage protectiontoxIN
Abortive infection of infected cells to prevent phage spread in the populationPecota & Wood (1996), Blower et al. (2009)
Stabilization of MGEshok/sok
Addiction system of MGEs including pathogenicity islands, plasmids, or integrative conjugative elementsGerdes et al. (1986b), Wozniak & Waldor (2009)
Anti-addiction elementsccdABCells harboring a chromosomal TA system can be protected against postsegregational killing by plasmid-encoded TA lociSaavedra De Bast et al. (2008)
Selfish genesUnknownTA systems as genetic commensals within a genome
  • Data according to Magnuson (2007), modified and extended.

TA systems can respond to external stimuli and can be embedded into larger transcriptional networks

Besides autoregulatory functions mostly operative in type II representatives, a number of TA systems are also controlled by pleiotropic or specific regulators and/or environmental cues. For example, in S. aureus, the activities of the alternative sigma factor σB and the mazEF system are tightly intertwined (Senn et al., 2005; Donegan & Cheung, 2009). In E. coli, the SOS response was shown to influence the istR-tisAB or yafNO TA systems (Vogel et al., 2004; Singletary et al., 2009; Wagner & Unoson, 2012). Corruption of topoisomerase by ciprofloxacin or the CcdB toxin leads to DNA damage and in turn cleavage of the SOS master regulator LexA that also controls tisB expression (Dörr et al., 2010; Tripathi et al., 2012). A connection between amino acid and carbon limitation was established for TA systems such as mazEF, relBE, yafNO, higBA, mqsRA, or hicAB (Aizenman et al., 1996; Christensen et al., 2001; Lemos et al., 2005; Jørgensen et al., 2009; Christensen-Dalsgaard et al., 2010) also highlighting a role of the stringent response alarmone ppGpp on certain TA systems (Aizenman et al., 1996; Korch et al., 2003). The regulation of GhoST by the subordinate TA component MqsR indicates hierarchical networks of TA system control (Wang et al., 2012). Notably, the fact that the most prevalent modes of TA systems' control act post-transcriptionally indicates an advantage over transcriptional regulation to ensure rapid adaptation to stress. Particularly, free-living bacteria frequently encounter environmental changes that could have favored the accumulation of TA systems acting as stress response loci in these organisms (Pandey & Gerdes, 2005). Accordingly, TA systems are absent or are scarcely found in strongly host-associated prokaryotes, whereas these genetic elements can constitute up to 2.5% of the genome size of free-living bacteria (Pandey & Gerdes, 2005; Leplae et al., 2011).

Applications of TA systems

Exploiting the properties of TA systems has led to a number of applications for various purposes (Table 3). A straightforward use, akin to the natural function, is the stabilization of episomal elements. High-yield production of plasmid-encoded proteins or other compounds can be a physiologic burden for the producing cells, leading to plasmid loss and concomitant reduction in yields (Friehs, 2004). Antibiotic markers, which can occasionally be inappropriate, have been replaced by TA systems to stabilize plasmids in a bacterial culture (Kroll et al., 2010). Likewise, the loss of plasmids equipped with TA systems can be delayed, with beneficial consequences for gene therapy (Vandermeulen et al., 2011). On the other hand, yeast strains could be provided with a ‘self-destruct unit’ consisting of TA system components to ensure their containment in the laboratory. For this purpose, cells could carry a toxin gene under a glucose-repressible promoter, constitutively express the antitoxin and be cultivated in glucose-containing broth. It was rationalized that cells accidentally escaping from the fermentation vessel would be eradicated by the toxin activated by low glucose levels (Kristoffersen et al., 2000). In molecular cloning, the ccdB gene has been employed for positive selection of recombinant plasmids. Successful insertion of DNA into the multiple cloning site disrupts ccdB, leading to a growth advantage of cells with modified plasmids and therefore reduces screening time for desired clones (Bernard et al., 1994). The single protein production (SPP) process exploits the E. coli MazF toxin, which cleaves RNA at ACA sites (Suzuki et al., 2005), stochastically found in virtually every native transcript. Induction of mazF results in degradation of most mRNA species, whereas an artificially ACA-free mRNA is unaffected in expression and translation. This technique facilitates protein labeling with 13C and 15N isotopes for NMR studies or the production of toxic proteins. Escherichia coli MazF has also been employed to combat viral infection of human cells. mazF controlled by an HIV-dependent promoter conferred resistance of CD4+ T-lymphoid cells to HIV, presumably by selective viral RNA decay upon infection (Chono et al., 2011). Specific eradication of herpes C virus (HPC)-infected cells was achieved by proteolytic activation of MazF (Shapira et al., 2012). To this end, mazF and mazE were translationally fused to an HPC protease-cleavable linker. MazF unleashed after viral attack is then supposed to cleave viral RNAs and thus inhibit infection. Another interesting use of TA systems in eukaryotes was exemplified by Slanchev et al. (2005). The kid toxin was expressed in primordial germ cells of male zebrafish embryos, while somatic cells were protected by kis antitoxin expression, producing sterile male fish. These approaches provide promising future perspectives for the eradication of a defined set of cells, including treatment of viral infections and possibly cancer.

View this table:
Table 3

Established and conceivable applications of TA system components

ApplicationTA system(s)DescriptionReferences
Vector stabilizationhok/sok
Replacement of antibiotic markers in production strains and for gene therapyPecota et al. (1997), Kroll et al. (2010), Vandermeulen et al. (2011)
Containment controlrelBEA glucose-repressed relE toxin gene could ensure elimination of production strains outside a laboratory or a fermenter, where low glucose conditions are encounteredKristoffersen et al. (2000)
CounterselectionccdBDisruption of plasmid-encoded ccdB upon insertion of a DNA fragment allows for positive selectionBernard et al. (1994)
Single protein productionmazFSimultaneous expression of mazF and a gene of interest lacking MazF restriction sites results in large relative amounts of target proteinSuzuki et al. (2005)
HIV treatmentmazFMammalian cells carrying mazF under the control of an HIV-controlled promoter are resistant to HIV infection, supposedly because of MazF-dependent HIV eradicationChono et al. (2011)
HCV treatmentmazEFMammalian cells carrying HCV protease-cleavable translational fusions of toxin and antitoxin are eliminated upon HCV infection due to toxin activationShapira et al. (2012)
Selective killing of eukaryotic cell typeskid/kisCertain cells in a multicellular organism are killed off by the toxin, while others are protected by expression of the antitoxin. Used to select for sterile zebrafish males, lacking germ cellsde la Cueva-Mendez et al. (2003), Slanchev et al. (2005)
Antibacterial treatmentpemIK
Artificial activation of TA toxins in bacteria by changing the toxin–antitoxin ratio or specifically disrupting or preventing complex formation (Fig. 3)Agarwal et al. (2010), Lioy et al. (2010)

TA systems might be suitable as antibacterial targets

TA systems have been suggested as targets for antibacterial treatment (Williams & Hergenrother, 2012). Intuitively, the inherent antibacterial activity of TA toxins could be exploited to kill cells directly or to support antibiotic treatments. To this end, TA toxins could be activated by preventing formation or breaking up TA complexes, or by degradation or interference with production of antitoxins (Fig. 3; Williams & Hergenrother, 2012).

TA systems as possible antibacterial targets. (Left) In the complexed form, the toxin (red circular sector) is inhibited by the antitoxin (blue triangle). Two possible ways of toxin activation are depicted. (a) Binding of a suitable ligand (green box) to the TA interface or to allosteric sites of either protein could preclude TA complexes. (b) TA ratios could be shifted to increase free toxin. Above: stimulation of antitoxin degradation by Clp or Lon; below: selective up- or downregulation of toxin or antitoxin expression, respectively. This figure is drawn exemplarily for type II TA systems, but similar strategies are also adaptable to the other types.

Although appealing, these strategies are associated with some pitfalls and drawbacks. First, TA systems suitable for attack should be widespread and active over a number of strains (Williams et al., 2011). An inhomogeneous distribution of TA systems among a consortium of infective strains could give those lacking these elements a selective growth advantage. Second, influencing TA systems might raise persister cells that are even more difficult to tackle during infection (Gerdes & Maisonneuve, 2012). Despite some promising results (Agarwal et al., 2010; Lioy et al., 2010; Chopra et al., 2011), compounds that efficiently activate TA toxins and are suitable for use in patients are yet to be identified.

Closing remarks

To date, five different TA types based on the mode of action and nature of the antitoxin have been defined, and multiple toxin targets and toxicity mechanisms have been elucidated. Knowledge today tremendously exceeds the notion of plasmid stabilization that was proposed when the TA field was in its infancy. In the future, it will be interesting to pinpoint more specific purposes of TA systems in different microorganisms, to reveal their regulation and impact on the modulation of single cells or on a population scale. The use of TA systems in various fields of biology is just emerging.


We are grateful to Dr Volkmar Braun for critically reading the manuscript. Work in the authors' laboratory was supported by the Deutsche Forschungsgemeinschaft through grants BE4038/2 and BE4038/5 of the priority programmes 1316 (‘host adapted metabolism of bacterial pathogens’) and 1617 (‘phenotypic heterogeneity and sociobiology of bacterial populations’).


View Abstract