Toxins-Antitoxins: Plasmid Maintenance, Programmed Cell Death, and Cell Cycle Arrest

See allHide authors and affiliations

Science  12 Sep 2003:
Vol. 301, Issue 5639, pp. 1496-1499
DOI: 10.1126/science.1088157


Antibiotic resistance, virulence, and other plasmids in bacteria use toxin-antitoxin gene pairs to ensure their persistence during host replication. The toxin-antitoxin system eliminates plasmid-free cells that emerge as a result of segregation or replication defects and contributes to intra- and interspecies plasmid dissemination. Chromosomal homologs of toxin-antitoxin genes are widely distributed in pathogenic and other bacteria and induce reversible cell cycle arrest or programmed cell death in response to starvation or other adverse conditions. The dissection of the interaction of the toxins with intracellular targets and the elucidation of the tertiary structures of toxin-antitoxin complexes have provided exciting insights into toxin-antitoxin behavior.

The horizontal gene pool in bacteria is composed of an array of accessory mobile genetic elements that profoundly influence genome plasticity, organization, and evolution (1). Plasmids are extrachromosomal, autonomously replicating members of this pool important for bacterial adaptability and persistence because they provide functions that might not be encoded by the chromosome (2). Although plasmids can be highly beneficial to host bacteria by allowing them to persist in otherwise hostile ecological niches, pathogenic properties are also endowed by plasmids. The clinical failure of antibiotics in recent years is in part linked to the rapid dispersal of plasmid-borne resistance genes in bacterial populations. Virulence determinants in bacterial pathogens may also be plasmid-located and similarly dispersed. A topical example of the contribution of plasmids to bacterial diversity is represented by the foodborne pathogen Bacillus cereus, the anthrax agent B. anthracis, and the insect pathogen and commercial pesticide B. thuringiensis; these are very closely related species but, from a human perspective, have quite distinct biological properties. These differences reflect, at least in part, differences in these species' plasmid contents (3).

Although there is considerable heterogeneity in the physical and genetic properties of plasmids, large, low-copy-number plasmids associated with antibiotic resistance and pathogenicity tend to include core regions required for replication, segregational stability, and conjugative transfer. The segregational stability of these and other plasmids in bacterial populations is achieved by the activity of plasmid-specified partitioning proteins that direct plasmid copies to new daughter cells at cell division. Plasmid-directed events resulting in selective killing or growth impairment of cells that have failed to acquire a plasmid copy were identified in the 1980s (4, 5). These mechanisms confer an advantage on plasmid-retaining cells by reducing the competitiveness of their plasmid-free counterparts, thereby ensuring the retention of the plasmid in the population. There has been a recent resurgence of interest in these bacterial cell–poisoning systems because of new insights that have been acquired into these events, but also because of enhanced appreciation of their widespread distribution both on plasmids of medical importance and on bacterial chromosomes (611). These toxin-antitoxin (TA) mechanisms, also known as postsegregational cell killing and addiction systems, attack cells from within. This is in contrast to the action of colicins or antibiotics that are secreted by bacteria into their environment as inhibitors of neighboring microorganisms. The toxin component produced by TA cassettes is designed to maim bacterial cells, which raises the exciting possibility that these factors might be exploited as novel antibacterial agents in the treatment of infectious diseases. Restriction-modification enzyme pairs can be either plasmid- or chromosomally encoded and are also now viewed as multifunctional TA systems that can promote segregational stability, as well as providing protection against invading DNAs and directing genome rearrangements. The restriction enzyme is analogous to the toxin; the modification methylase is equivalent to the antitoxin (12).

Toxin-Antitoxins and Plasmid Maintenance

TA cassettes have a characteristic organization in which the gene for the antitoxin component precedes the toxin gene (Fig. 1); the two loci often overlap, reflecting a common autoregulatory mechanism exerted by both components. Although most TA modules conform to this arrangement, there are examples of TA cassettes in which the gene order is reversed, where the antitoxin alone exerts the regulatory effect or where the product of a third gene is implicated (6, 7, 13).

Fig. 1.

Schematic representation of cell death induced by plasmid-located type II TA modules. The toxin (red) and antitoxin (blue) proteins form a tight complex that negates the harmful activity of the toxin. The antitoxin is degraded by a protease (green) more rapidly than the toxin, but the latter is continually sequestered by fresh antitoxin. As long as the plasmid is maintained, the cell tolerates the presence of the TA complex (right). If a missegregation event or replication defect produces a plasmid-free cell (left), the degraded antitoxin cannot be replenished so that the liberated toxin attacks an intracellular target to cause death or growth restriction of the plasmid-free cell. The targeting of DNA by the toxin is illustrative only.

The toxin gene encodes a stable protein, whereas the antitoxin is either a labile protein or an untranslated, antisense RNA species. The toxin is neutralized by inhibition of toxin translation when the antitoxin is an RNA (type I), or by avid binding of the partner antitoxin when the latter is a protein (type II). If a plasmid-free variant is produced, owing to a replication error or other defect in plasmid maintenance, the new cell will still inherit the TA complex. The antitoxin component is degraded more rapidly by host enzymes as compared with the toxin but is not replenished because of the absence of the plasmid. The toxin is released from the TA complex and interacts with an essential host target to cause death or growth restriction of the plasmid-free cell (Fig. 1).

The best-characterized type I module is the hok-sok locus of plasmid R1 of Escherichia coli, the analysis of which has not only provided important insights into TA mechanisms but also into RNA-RNA interactions and posttranscriptional control in bacteria (14). The hok toxin gene specifies a transcript that is translationally inactive and unable to bind the sok countertranscript RNA owing to extensive secondary structure features. The inactive hok mRNA is converted gradually to a translationally active derivative by 3′-end processing. However, translation of this truncated transcript is inhibited irreversibly by binding of the sok transcript that is complementary to the 5′ end of the hok mRNA. The duplex RNA is degraded by an RNase III–dependent process. Thus, the translational activation of the hok gene is subtly controlled at multiple levels. The absence of a renewable source of the sok antitoxin RNA in a plasmid-free cell and the differential decay patterns of the sok and hok RNAs result in the accumulation of the truncated, active version of the hok mRNA. The latter is translated into the 52–amino acid, toxic Hok protein. Hok exerts its poisonous effect by cell membrane depolarization in a manner analogous to holin proteins produced by some bacteriophages before cell lysis (14).

The majority of TA loci that have been identified on plasmids are of type II (Table 1). Here, the toxin protein is tightly sequestered by the antitoxin so that free toxin is unavailable within the cell. However, the antitoxin is more susceptible to degradation by host proteases than the toxin, so that if a plasmid-free cell arises the toxin becomes available to target an essential intracellular host factor to induce cell death or severe growth impairment. The Clp and Lon proteases have been implicated in antitoxin degradation in different TA complexes (7), although other proteins can modulate this proteolysis (15).

Table 1.

Major type II TA loci identified on plasmids. Prototypical TA proteins only are listed; many other plasmids and chromosomes produce homologs of these proteins. aa, amino acid; ND, not determined.

Plasmid Bacterium Toxin (aa) Antitoxin (aa) Toxin target
F E. coli CcdB (101)CcdA (72)DNA gyrase
P1 E. coli Doc (126)Phd (73)Translation?
RK2 E. coli ParE (103)ParD (83)DNA gyrase
R1 E. coli Kid (110)Kis (84)DnaB
pTF-FC2 A. ferrooxidans PasB (90)PasA (74)ND
pSM19035 S. pyogenes ζ (287)ϵ (90)ND
Rts1 E. coli HigB (92)HigA (104)ND
P307 E. coli RelE (95)RelB (83)ND
pMYSH6000 S. flexneri MvpT (133)MvpA (75)ND
pRUM E. faecium Txe (85)Axe (89)ND

DNA Metabolism Proteins as Targets

Surprisingly few intracellular targets of plasmid-encoded toxin proteins have been elucidated definitively, which probably reflects the difficulties associated with analyzing proteins that are inherently poisonous to their bacterial hosts (Table 1). Among type II modules, the target of the CcdB toxin of the F plasmid was the first to be identified and has been most extensively characterized (16). CcdB is a topoisomerase II DNA gyrase poison that interacts with the catalytic GyrA subunit of gyrase and entraps a cleavage complex between gyrase and DNA. This behavior is analogous to the inhibitory action of quinolone drugs on gyrase, although quinolones and CcdB target different GyrA domains. The labile CcdA antitoxin can inhibit and reverse the interaction between CcdB and gyrase. However, polymerases are unable to traverse the trapped CcdB-gyrase-DNA complex that leads to the induction of stress responses. CcdB also associates with gyrase to produce a complex that is impaired in supercoiling. The concerted effects of CcdB as a gyrase poison and as an inhibitor of gyrase-mediated DNA supercoiling are lethal for E. coli (16). The ParE toxin of the ParDE complex of plasmid RK2 is unrelated evolutionarily to the CcdB toxin. Nevertheless, ParE recently was shown to modulate the activity of DNA gyrase in a manner strikingly similar to that of CcdB (17). It is intriguing that a pair of toxins without any apparent sequence homology have evolved to target the same essential host factor.

DnaB is the primary replicative DNA helicase in E. coli. The R1 plasmid of E. coli expresses a type II complex, Kis-Kid, in addition to the hok-sok type I system. The Kid toxin specifically inhibits DnaB-dependent DNA replication (18), probably either by entirely blocking assembly of the replication complex or by interfering with DnaB-mediated protein interactions that are necessary for the formation of a productive complex (19). Despite their very low sequence similarity and different intracellular targets, the Kid and CcdB toxins have remarkably similar tertiary structures. Nevertheless, the surfaces of Kid and CcdB that have been implicated in contacting their respective targets are on entirely opposite faces of the proteins (9).

DNA gyrase and DnaB, proteins required for DNA metabolism, are the only known intracellular targets of plasmid-encoded protein toxins. Although too few targets have been defined (Table 1) to assert whether this will reflect a general pattern among TA proteins, it is logical that a fundamental process such as DNA replication is disrupted by TA action, because this is likely to impede proliferation of a plasmid-free cell most effectively. Analogously, translation, another basic macromolecular process, is a target of chromosomal TA systems, as discussed below.

Plasmid-Based Toxin-Antitoxins in Pathogenic Bacteria

TAs of E. coli plasmids have been studied most intently because of their tractability and because this bacterium is readily manipulated genetically. Nevertheless, database mining and some experimentation have revealed that TA genes are probably widely dispersed in a gamut of different bacteria, including many pathogens (711). The Gram-negative bacterium Shigella causes an acute dysentery and is responsible for hundreds of thousands of deaths annually worldwide. Isolates of one of the four main Shigella species, Shigella flexneri, typically harbor a large plasmid that specifies several virulence factors required for intestinal epithelial cell invasion, one of the initial steps in pathogenesis. This plasmid contains a type II TA module, mvpAT, that has homologs in a variety of other disease-causing bacteria (20).

Unrelated TA modules have been identified on plasmids of Gram-positive pathogens. Streptococcus pyogenes is a serious and reemerging human pathogen that can induce a range of conditions from sore throat to toxic shock–like syndrome, acute rheumatic fever, and necrotizing fasciitis. The pSM19035 plasmid was identified more than 20 years ago in S. pyogenes but only recently has been shown to harbor a type II TA cassette (21). The ζ toxin protein of this plasmid is unusually large, two to three times the size of any other known toxin (Table 1), and forms a heterotetrameric complex with the ϵ antitoxin. The formation of heteromultimeric species appears to be a characteristic of TA complexes. The tertiary structure of the ϵ2ζ2 complex reveals that the ζ toxin is probably a phosphotransferase with Walker-type adenosine triphosphate (ATP) binding motifs (11). The ATP binding sites are protected within the heterotetramer by the N-terminal α-helical arms of the ϵ antitoxin subunits. This suggests that the AT-Pase activity of ζ is important for its toxicity but that this becomes apparent only when the ATP binding site is uncloaked in the absence of the antitoxin. This is supported by the finding that mutations in the ATP binding site abrogate toxicity. However, the exact role of phosphotransferase activity in toxicity is unknown, because the intracellular target for the ζ poison has yet to be determined (11).

Enterococcus spp. have emerged in recent decades as important human pathogens that are the aetiological agents of a variety of noscomial infections, including surgical wound, urinary tract, blood-stream and cardiovascular infections. Importantly, enterococci are often resistant to a wide spectrum of antibiotics, making treatment of enterococcal infections increasingly difficult. Plasmid pAD1 of Enterococcus faecalis is a conjugative virulence plasmid that has similarities to many plasmids identified from clinical isolates of this bacterium. The par locus on pAD1 is a TA module that produces a 33-amino-acid toxin and an antitoxin that is a countertranscript RNA species (22). This is the only known type I cassette in Gram-positive bacteria, and it shares mechanistic similarities, but also important contrasts, with hok-sok-like TAs of Gram negatives. Plasmid pRUM is a multidrug resistance plasmid recently identified in a clinical isolate of E. faecium that also harbors a larger, vancomycin-resistance plasmid. A novel type II TA cassette (axe-txe) on pRUM was suggested by bioinformatics and was subsesquently shown to be functional both in its native host and in evolutionarily divergent bacteria (10).

Francisella tularensis is a zoonotic agent and a current cause of concern because of its potential as a bioterror weapon. A small cryptic plasmid of a related Francisella species harbors homologs of axe-txe (23), as do many bacterial chromosomes (10). Although experimental evidence has been provided only recently for TAs in plasmids of pathogenic bacteria, further searches for these systems in other important pathogens are likely to reveal a variety of different, possibly novel, TA cassettes.

Chromosomal Toxin-Antitoxins: Programmed Cell Death or Cell Cycle Arrest?

TA modules are a parasitic device which ensures that cells that have missegregated a plasmid do not survive. Bacterial chromosomes also harbor TA cassettes (cTAs) that are homologous to those identified on plasmids but apparently fulfill a different function to related elements.

Programmed cell death (PCD, or apoptosis) is a normal physiological process that occurs during development and tissue turnover in multicellular, eukaryotic organisms. However, various pathological conditions, including tumor formation, autoimmune disease, and neurodegenerative disorders, involve aberrant PCD mediated by either suppression or up-regulation of critical molecular components of the PCD apparatus. As unicellular organisms, bacteria were not considered to undergo PCD; however, bacteria in natural environments exist as multicellular colonies or as biofilms displaying coordinated multicellular processes (24). Colonial bacteria also maintain discrete, ordered spatial structures. The concept that bacteria might also possess PCD mechanisms involved in regulating their multicellular organization has emerged relatively recently (25, 26). Homologs of the eukaryotic PCD apparatus are present in bacteria (27), and components of this eukaryotic machinery can also elicit bacterial cell death (28, 29). Moreover, genome sequence data suggests that the E. coli genome, for example, harbors several cTA modules that are either PCD genes or mediators of reversible cell cycle arrest (Fig. 2). PCD in bacteria might allow surviving cells to scavenge nutrients from dead siblings or might prevent the systemic spread of bacteriophages within a population, for example (6). By acting as cell cycle arrest factors, cTA proteins alternatively might allow cells to enter a dormant or semidormant state as a protection against severe nutrient limitation and then to revive when environmental conditions become more conducive (30).

Fig. 2.

Locations of known type I (red) and type II (yellow) toxin-antitoxin modules on the E. coli genome. Asterisks denote genes that are inactive or relics.

Several type I cTAs in E. coli K12 are related to hok-sok but apparently are inactive or relics, although functional hok-sok modules are probably present in wild-type strains of E. coli (31). Long direct repeat (LDR) sequences recently identified on the E. coli chromosome appear to be active analogs of type I modules (32).

The K12 genome includes a number of known type II cTA genes that either enhance segregational stability and/or exhibit toxic-antitoxic behavior when inserted ectopically into multicopy plasmids: chpBIK, mazEF, relBE, and yefM-yoeB (Fig. 2). The ecnAB locus also has many of the hallmarks of a type II module, but an important difference is that the EcnB bacteriolytic toxin is actively synthesized in response to osmotic conditions during stationary phase, rather than being released from a sequestered state as the result of proteolytic degradation of an unstable antitoxin (33).

The cTA locus mazEF was the first identified in E. coli (6). MazE isolates the MazF toxin. However, transcriptional or translational inhibition of mazEF blocks de novo synthesis of the unstable antitoxin so that uncomplexed MazF can be directed to its intracellular target. This target is unknown, although expression of mazEF interferes with translation and/or DNA replication (30) (Fig. 3). Blockage of mazEF expression is triggered by 3′,5′ guanosine bispyrophosphate (ppGpp), an amino acid starvation signal in E. coli. Antibiotics that are general inhibitors of transcription or translation exert a similar effect, as does thymine deprivation (34). The MazEF proteins show significant sequence identity with both the ChpBIK cTA proteins and with the Kis-Kid proteins of plasmid R1. The homology between the MazF and Kid toxins is mirrored by close tertiary structure similarities (35).

Fig. 3.

Schematic representation of the known interactions of toxins of type I and II TA systems with intracellular targets. The plasmid-encoded Kid, CcdB, and ParE toxins target factors involved in DNA replication and metabolism. The chromosomally encoded RelE toxin inhibits translation. The MazF and Doc toxins are also thought to affect translation, the latter perhaps via an interaction with MazEF. MazF might also inhibit DNA replication. The plasmid-encoded Hok and chromosomally specified EcnB toxins disrupt membrane organization.

Like mazEF, the cTA locus relBE is activated in response to amino acid starvation and elevated levels of ppGpp. RelE is an inhibitor of translation and induces a bacteriostatic response that is fully reversible if the RelB antitoxin is subsequently produced (30, 36). It has recently been shown that RelE cleaves mRNAs within the ribosome but not free transcripts; i.e., cleavage is dependent on translation (37, 38). The cleaved mRNAs cannot be processed further, which results in the accumulation of stalled ribosomes on damaged mRNAs and translation inhibition. thereby blocking cell cycle progression (37, 38). Transfer-messenger RNA (tmRNA) releases trapped ribosomes from defective mRNAs and tags polypeptides from these ribosomes for proteolysis. Replenishment of the pool of tmRNA is likely to permit recycling of amino acids from the RelE-stalled ribosomes and to promote translational restart (38). As with RelBE, the detrimental effect of the MazF toxin also can be fully reversed when its cognate antitoxin is subsequently produced. In view of this, it has been proposed that both RelBE and MazEF are modulators of the physiological response to poor nutritional conditions rather than bona fide PCD mechanisms (30). It remains to be elucidated whether other cTA modules in E. coli (Fig. 2) evoke PCD or cell cycle arrest or have another role.

Perspectives and Questions

Postsegregational cell killing by TA modules is a highly effective strategy that plasmids deploy to ensure their persistence within bacterial populations. In contrast, chromosomally-located TA genes, homologous to those on plasmids, tune the physiology of the cell in response to external cues, inducing either reversible bacteriostasis or cell death (6, 7, 30). The mechanisms of action of type II plasmid-based TAs and cTAs are clearly similar; however, the stimuli that prompt antitoxin decay are different.

The metabolic state of the cell triggers cTA activity. Because most bacteria apparently harbor multiple cTA genes (Fig. 2), it will be intriguing to assess whether these are activated simultaneously in response to one or more inputs or separately in response to different signals. It is not yet known whether RelBE and MazEF are switched on by common or dissimilar nutritional deficits (6, 30). By contrast, the relBE-like dinJ-yafQ cassette in E. coli (Fig. 2) has been hypothesized to be a DNA-damage-responsive element (7). Whether other cTA modules respond to nutritional alarms or to other stimuli is uncertain, but it is clear that the acquisition by bacteria of multiple cTA-based mechanisms has been a widespread strategem for cell survival under stressful conditions.

The toxin components of TAs are active not only in the hosts from which the elements originate, but also in diverse bacteria (8, 10, 38) and even in eukaryotes (3941), possibly reflecting an evolutionary relationship between TAs and PCD in eucaryotes (42). TA systems probably aid the persistence of plasmids in a diverse range of bacteria by poisoning broadly conserved factors. It is not possible to estimate whether TA modules first developed on chromosomes and were then recruited by plasmids for their own purposes or vice versa. However, lateral transfer, perhaps mediated by plasmids or transposons, appears to have contributed to the dissemination of TA genes (7, 10).

The roles and targets of TA systems can help elucidate important cellular processes. For example, details of DNA gyrase function have been revealed by studies of its interaction with the CcdB toxin (16). Similarly, the association of the RelE toxin with ribosomes has provided new glimpses into the bacterial translation apparatus (37, 38). Augmented by the potential application of TAs as biocontainment and antibiosis agents, continued investigation of TAs will provide further fascinating insights into plasmid maintenance and evolution, as well as programmed cell death and cell cycle arrest phenomena in bacteria.

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article