Toxin-antitoxin systems: Biology, identification, and application.

Toxin-antitoxin (TA) systems are small genetic elements composed of a toxin gene and its cognate antitoxin. The toxins of all known TA systems are proteins while the antitoxins are either proteins or non-coding RNAs. Based on the molecular nature of the antitoxin and its mode of interaction with the toxin the TA modules are currently grouped into five classes. In general, the toxin is more stable than the antitoxin but the latter is expressed to a higher level. If supply of the antitoxin stops, for instance under special growth conditions or by plasmid loss in case of plasmid encoded TA systems, the antitoxin is rapidly degraded and can no longer counteract the toxin. Consequently, the toxin becomes activated and can act on its cellular targets. Typically, TA toxins act on crucial cellular processes including translation, replication, cytoskeleton formation, membrane integrity, and cell wall biosynthesis. TA systems and their components are also versatile tools for a multitude of purposes in basic research and biotechnology. Currently, TA systems are frequently used for selection in cloning and for single protein expression in living bacterial cells. Since several TA toxins exhibit activity in yeast and mammalian cells they may be useful for applications in eukaryotic systems. TA modules are also considered as promising targets for the development of antibacterial drugs and their potential to combat viral infection may aid in controlling infectious diseases.

Introduction

The first TA system, ccdA/ccdB, was described 30 years ago as a module enhancing the stability of the F plasmid by post-segregational killing of plasmid-free daughter cells. Subsequently, further plasmid encoded TA systems were identified by their ability to enhance plasmid stability but also a number of TA systems were detected by homology searches on bacterial chromosomes. Intriguingly, bacteria were identified that contain dozens of putative TA systems per genome, for instance Mycobacterium tuberculosis, and Nitrosomonas europaea. While the function of TA systems on plasmids was obvious, their presence on chromosomes remained enigmatic for a long time and it is only now that some of the proposed functions have been confirmed by experimental evidence. Exhaustive homology searches in sequence databases and novel bioinformatic approaches allowed the identification of numerous TA systems and today more than 10000 (putative) TA modules are known.,- This nearly ubiquitous presence has increased the interest in TA systems considerably. Moreover, they represent convenient systems to address basic scientific questions including gene regulation, stress response and persister cell formation and are versatile tools for biotechnological applications.

In this review we provide a brief description of how TA systems are composed at the molecular level, highlight novel findings about the action of TA toxins on their cellular targets and discuss functions proposed for TA systems. In addition, approaches for identification and characterization of putative TA systems are described. A special focus of this review is also the application of TA systems in biotechnology and basic research. Finally, the potential of TA systems for combating pathogens is discussed.

Biology of TA Modules

The toxins of all characterized bacterial TA systems are proteins, while the antitoxins are either proteins or small RNAs (sRNAs). In general, the toxin is more stable than the antitoxin. Currently, TA systems are assigned to five classes (I–V) according to their genetic structure and regulation. In type I and III TA modules the antitoxins are small noncoding RNAs, while the antitoxins of the other TA classes are small proteins.

Type I

Type I antitoxins are unstable antisense sRNAs. In most cases expression of the toxin is downregulated by base-pairing of the antitoxin sRNA with the stable mRNA of the toxin. This interaction was shown to prevent binding to the ribosome and thereby arrests translation of the toxin mRNA. An example for a type I TA system regulated by this mode is the symR/symE module of Escherichia coli (Fig. 1A). Additional levels of regulation involve the control of symE expression by LexA, a SOS-response regulated transcriptional repressor, and the degradation of SymE by the protease Lon. Further examples of TA systems regulated by inhibition of toxin mRNA translation are tisB/istR-1 and ibs/sib, both encoded by the chromosome of E. coli, fstI/rnaII from plasmid pAD1 and hok/sok encoded by plasmid R1 (Fig. 1B). In case of the latter regulation is more complex. The RNA antitoxin sok is expressed from a strong promoter but the transcript has a very short half-life of approximately 0.5 min. In contrast, the hok mRNA is expressed from a very weak promoter but has a half-life of approximately 20 min. The hok transcript shows extensive secondary structure and the 3′ end folds back to the 5′ end. The folded full-length hok mRNA is neither accessible for the ribosome nor the Sok-RNA antitoxin. Processing by RNase II removes a part of the 3′ terminus causing a major structural rearrangement including the 5′ part of the hok mRNA. This allows translation but also binding of the Sok-RNA. However, the Sok antisense-RNA does not show complementarity to the Shine-Dalgarno sequence of hok. Interestingly, the hok/sok locus contains in addition to the hok toxin and sok antitoxin a third gene called mok for modulator of killing, which overlaps with almost the entire hok gene. Analysis of point mutations revealed that prevention of mok translation abolished efficient expression of hok, indicating that the hok and mok open reading frames are translationally coupled and that the Sok-RNA regulates hok translation indirectly by preventing translation of mok. Finally, the hok mRNA/Sok-RNA hybrids are cleaved by RNase III, which is the initial step for decay of the hok mRNA but not essential for regulation. In contrast, for some type I modules including txpA/ratA (Fig. 1C), bsrG/sr4, and yonT/as-yonT of Bacillus subtilis cleavage of double-stranded RNA regions created by binding of the antitoxin RNA to the toxin mRNA by RNase III is crucial for regulation of toxin expression.

Figure 1. Types of TA systems. (A) The symR/symE module of E. coli as an example for a type I system regulated by interference of toxin mRNA translation. SD, Shine-Dalgarno sequence. (B) Regulation of the type I system hok/sok of plasmid R1. (C) The ratA/tpxA module from Bacillus subtilis represents a type I system where toxin mRNA degradation is promoted. (D) The relB/relE two module type II system from E. coli. (E) The ω-ε-ζ three module type II systems from Streptococcus pyogenes plasmid pSM19035. (F) The toxI/N type III system from the Erwinia carotovora plasmid pECA1039. (G) The yeeU/yeeV type IV system of E. coli. (H) The ghoS/ghoT type V system of E. coli. In this and all subsequent figures the toxin and its encoding gene are shown in orange while the antitoxin and its encoding gene are shown in green.

Type I TA modules were found in plasmids and on chromosomes. Interestingly, the majority of characterized plasmid encoded sRNA antitoxins have extensive complementarity with the toxin mRNA, while many sRNAs of chromosomally encoded type I TA systems act by base pairing with targets that have limited complementarity.

Type II

The type II TA systems are the best studied class of TA modules. Within this class both the toxin and the antitoxin are small proteins. The antitoxin forms a protein-protein complex with the toxin resulting in its neutralization (Fig. 1D). While the toxin protein is stable the antitoxin is rapidly degraded by proteases of the Clp family- or by Lon.- The operon of type II TA modules typically comprises two small open reading frames where the upstream gene encodes the antitoxin. However, exceptions of this conserved gene organization are known, for instance the higB/higA TA module, where the toxin gene higB is located upstream of the antitoxin gene higA. Typically, transcription of the TA operon is autoregulated by binding of the antitoxin or by the toxin–antitoxin complex to the promoter. Depending on the stoichiometric ratio of the antitoxin to the toxin several types of complexes may be formed with distinct affinities to the promoter. For instance, in presence of an excess of RelB over RelE dimers of RelB (RelB2) and the 2:1 complex RelB2RelE are formed, which both inhibit the relB/relE promoter (Fig. 1D). The RelB2RelE has a stronger inhibitory effect on the relB/relE promoter than RelB2 and thus RelE acts as a transcriptional co-repressor., In contrast, in the presence of an excess of RelE the 2:2 complex (RelB2RelE2) is formed, which cannot bind the promoter and, consequently, transcription is activated. This mode of regulation is frequently called conditional cooperativity and is believed to be important for stabilization of the antitoxin level in rapidly growing cells to minimise random induction of relB/relE. Also other TA systems including phd-doc from the E.coli bacteriophage P1 and vapB/vapC of Salmonella enterica are regulated by conditional cooperativity.

Within this group of TA modules also three-component TA systems were reported. One example for this extraordinary genetic organization is the ω-ε-ζ TA module encoded by the Streptococcus pyogenes plasmid pSM19035 (Fig. 1E). The cytotoxic effect of the ζ protein is counteracted by the dimeric ε2 antitoxin, which forms a stable ζ2ε2 heterotetramer complex.- In contrast to other type II TA systems neither the ζ toxin nor the ε antitoxin or the ζ2ε2 complex regulate their own expression. Instead, the activity of promoter Pω is regulated by dimeric ω2, a global regulator of transcription., In addition, basal levels of ε and ζ are expressed from the constitutive but very weak promoter Pε., Also the paaR-paaA-parE TA module encoded by E. coli O157:H7 represents a three-component system where the upstream located parR is necessary for transcriptional control of the TA module. However, in contrast to ω-ε-ζ, the ParE-PaaA complex is involved in the regulation of its own transcription. The prototype of the third group of the three-component type II TA systems is the pasA/pasB/pasC module of plasmid pTF-FC2 from Thiobacillus ferrooxidans, where autoregulation is achieved by the antitoxin PasA and its complex with the toxin PasB. The third component, PasC, is not involved in regulation of expression but enhances toxin–antitoxin complex formation.

Originally type II TA systems were grouped into 8–14 families based on sequence similarity and gene structure, and it was assumed that each toxin family is associated with a specific antitoxin family. However, it has now become evident that many hybrid systems exist where a TA locus contains a toxin of one class and an antitoxin of another class. The functionality of a number of such hybrid systems has already been confirmed.- Thus, it was recently suggested to classify toxin and antitoxin families independently and 13 type II toxin superfamilies and 20 antitoxin superfamilies have been proposed., In addition, 4 superfamilies of ‘solitary’ toxins were identified, which showed inhibition of growth if overexpressed in E. coli but experimental proof of antitoxin activity of open reading frames flanking these toxins failed. However, it cannot be excluded that the antitoxin activity might be provided by a small RNA similar to type I and III systems. With respect to that it is interesting to note that toxin/antitoxin shuffling may also occur between the different types of TA systems since the toxin ToxN of the type III TA module toxI/toxN shows 3D structure similarity to the type II toxin MazF.

Type III

Similar to type I systems the antitoxin of type III modules is a sRNA. However, the mode of interaction is different. So far the only characterized member of this class is the toxI/toxN TA module of pasmid pECA1039 from Pectobacterium carotovoum, which was originally described as a protection system of bacteria against bacteriophage infection. The toxN gene is preceded by a short inverted repeat and by a tandem array of direct repeats (Fig. 1F). The inverted repeat serves as a terminator for regulating the amount of antitoxic sRNA and toxin mRNA. The ToxN protein has RNase activity and cleaves the toxI/toxN transcript at the direct repeats to release the active 36 nt RNA antitoxin. In contrast to type I TA modules the RNA antitoxin neutralises the toxin protein directly by binding. Studies showed that the protein-RNA interaction results in the formation of a heterohexameric triangular assembly of three ToxN proteins, which are interspersed by three 36 nt ToxI sRNA pseudoknots and that this binding reaction has an exquisite molecular specificity. In a phylogenetic study a total of 125 putative type III systems were identified and assigned to 3 families, toxIN, cptIN, and tenpIN. The majority of the type III systems is encoded by bacterial chromosomes but approximately 15% of the toxIN and the tenpIN modules are encoded by plasmids and one toxIN module is encoded by a prophage. The functionality of selected cptIN and tenpIN systems was confirmed in E. coli by assessing the toxicity of the proposed toxin gene and the ability of the cognate antitoxic repeats to inhibit the lethal effects.

Type IV

A type IV TA system designation was proposed for the yeeU/yeeV (also named ctbA/ctbB) TA module of E. coli. The functional analysis of this TA module revealed that the toxin YeeV interacts with MreB and FtsZ and thereby interferes with their polymerization and therefore with cytoskeleton assembly (Fig. 1G). The YeeU antitoxin protein counteracts YeeV by stabilizing MreB and FtsZ polymers. A similar mode of action was also reported for cptA/cptB (ygfX/ygfY), another TA module of E. coli. While the toxin and antitoxin of all other TA classes interact either at the RNA or the protein level, the toxin and antitoxin of this TA class do not directly interact.

Type V

Recently, the ghoS/ghoT TA module of E. coli was designated as a type V TA system (Fig. 1H). The antitoxin protein GhoS has a sequence specific endoribonuclease activity for the cleavage of the GhoT toxin mRNA and thereby prevents the translation of the toxin.

Cellular targets of the toxin compound

Toxin proteins of TA modules target a wide range of cellular processes and structures (Table 1). Most characterized toxins act as translational inhibitors and a multitude of mechanisms has been identified how they impact on protein biosynthesis. Many toxins function as mRNA interferases, either in a ribosome-dependent or independent manner. Typical examples for toxins cleaving free mRNAs are MazF, Kid, ChpBK, MqsR, and HicA. While HicA has no specific consensus recognition motif, Kid shows a preference for UA(A/C) and ChpBK, MqsR, and MazF cleave specifically at UAC, GCU, and ACA sites, respectively. The recognition motif among MazF homologs varies and consist of 3, 5, or even 7 nucleotides. MazF can also target the 3′ terminus of the 16S rRNA within the 30S ribosomal subunit to remove 43 nucleotides containing the anti-Shine-Dalgarno sequence required for initiation of translation. Interestingly, these modified ribosomes do not recognize canonical mRNAs containing Shine-Dalgarno sequences but were shown to translate so called leaderless mRNAs generated by MazF by cleaving ACA sites at or closely upstream of the AUG start codon of some specific mRNAs. Recently, it was shown that mycobacterial MazF-mt6 cleaves mRNA at the sequence UUCCU and that it targets an evolutionarily conserved region of the 23S rRNA in the ribosomal A site, which inhibits translation and destabilizes the association of the 30S and 50S ribosomal subunits. Structural similarities with MazF indicate an action as endoribonuclease also for ToxI, the toxin of the type III TA system toxI/toxN, however, the targeted RNA species remains elusive. RelE is the best characterized example of a RNA interferase cleaving mRNAs in a ribosome-dependent manner. RelE cleaves the mRNAs in the A site of the ribosome and was reported to show some sequence preferences. Also the type II TA toxin VapC exhibits RNase activity but, in contrast to the toxins mentioned above, shows specificity for the tRNAfMet in vivo. Another way in which TA modules can interfere with mRNA synthesis is by modifying the ribosomal activity via direct interaction with different subunits of the ribosome. Direct association of the toxin Doc with the 30S ribosomal subunit stopped elongation of translation resulting in a translational arrest. Similarly, the RatA toxin of the ratA/ratB type II system was shown to bind to the 50S ribosomal subunit, which prevents its association with the 30S ribosomal subunit and consequently the formation of 70S ribosomes. In contrast, the toxin HipA, a protein kinase, mediates inhibition of translation by phosphorylation of the elongation factor EF-Tu, preventing its interaction with tRNA.

Table 1. Cellular targets of selected well characterized TA-toxins

Toxin	Antitoxin/molecular species	Type	Toxin activity	Cellular process	Ref.
Hok	Sok / RNA	I	Integrates into the inner cell membrane	ATP synthesis	59
TisB	IstR-1 / RNA	I	Integrates into the inner cell membrane	ATP synthesis	60
SymE	SymR / RNA	I	mRNA cleavage	Translation	9
CcdB	CcdA / Protein	II	Inhibition of DNA gyrase	Replication	61
ParE	ParD / Protein	II	Inhibition of DNA gyrase	Replication	62
MazF	MazE / Protein	II	Ribosome-independent mRNA cleavage and cleavage of 16S rRNA	Translation	63,64
MazF-mt6	MazE-mt6 / Protein	II	Ribosome-independent mRNA cleavage and cleavage of 23S rRNA	Translation	65
Kid	Kis / Protein	II	Ribosome-independent mRNA cleavage	Translation	66
HicA	HicB / Protein	II	Ribosome-independent mRNA cleavage	Translation	67
RelE	RelB / Protein	II	Cleavage of ribosome-bound mRNA	Translation	68
VapC	VapB / Protein	II	Cleavage of tRNAfMet	Translation	69
Doc	Phd / Protein	II	Binds to the 30S ribosomal subunit	Translation	70
RatA	RatB / Protein	II	Binds to the 50S ribosomal subunit	Translation	71
HipA	HipB / Protein	II	Phosphorylation of EF-Tu	Translation	72
ζ	ε / Protein	II	Phosphorylation ofUDP-N-acetylglucosamine	Peptidoglycan synthesis	73
ToxN	ToxI / RNA	III	RNA cleavage?	Translation?	54
YeeV	YeeU / Protein	IV	Inhibition of FtsZ and MreB polymerization	Cytoskeleton	56
CptA	CptB / Protein	IV	Inhibition of FtsZ and MreB polymerization	Cytoskeleton	57
GhoT	GhoS / Protein	V	Integrates into the inner cell membrane	ATP synthesis	58

Also the toxin ζ of the ω-ε-ζ type II module shows kinase activity but its cellular target is distinct from HipA: ζ phosphorylates the peptidoglycan precursor UDP-N-acetylglucosamine (UNAG), which inhibits MurA, the enzyme catalyzing the initial step in peptidoglycan synthesis, and consequently impairs formation of the bacterial cell wall.

In addition to the cell wall the inner membrane may also be targeted by TA toxins. Most type I systems encode small, hydrophobic proteins that seem to function like phage holins introducing pores into the inner cell membrane, which leads to loss of the membrane potential and consequently impairs ATP synthesis. A similar mechanism is also anticipated for GhoT, the toxin of the type V system ghoS/ghoT. GhoT is a small highly hydrophobic protein with two predicted transmembrane domains and its induction causes formation of lysed cells with damaged membranes, which are also referred to as ghost cells.

The cytoskeleton is targeted by the type IV toxin YeeV, which interacts with FtsZ and inhibits its polymerization and GTPase activity. In addition, YeeV was also shown to disrupt MreB polymers, which are important for cell division and maintenance of the cell shape. Similar results were reported for the CptA toxin.

TA toxins can also inhibit DNA replication. The type II toxins ParE and CcdB were shown to inhibit GyrA,, a subunit of an essential type II topoisomerase, which relaxes positive supercoils that arise by unwinding the DNA double strand by helicase during replication. Gyrase poisoning induces DNA double strand breaks, activates the SOS response and efficiently triggers programmed cell death. However, it must be emphasized that most TA toxins including HipA, StbE, and RelE do not cause cell death but induce a static condition in which the cells are still viable but unable to proliferate.

Functions of TA systems

While the role of TA modules located on plasmids is quite clear, the widespread appearance of TA modules in chromosomes remained enigmatic for a long time. Plasmid encoded TA modules are important for plasmid stabilization. If the plasmid bearing a TA module is not transmitted to a daughter cell the unstable antitoxin is degraded while the stable toxin remains and acts on its cellular targets to kill or to inhibit growth of the plasmid-free cells (Fig. 2A). This effect is called post-segregational killing or addiction. In addition, plasmid encoded TA systems are also important for mediation of exclusion of co-existent compatible plasmids. Conjugation can create cells containing two plasmids of the same incompatibility group, which cannot be stably maintained in the same host. Loss of the plasmid possessing the TA module will kill the daughter cell while loss of the other plasmid leaves the cell unaffected (Fig. 2B). Over several rounds of conjugation and subsequent exclusion the plasmid containing a TA module can outcompete the second plasmid from the bacterial population.

Figure 2. Functions of plasmid encoded TA systems. (A) Stabilization of plasmids by post segregational killing. (B) Exclusion of co-existent compatible plasmids.

Some TA systems present on chromosomes may fulfil a similar function and mediate stabilization of superintegrons. This type of genetic element encodes for proteins with adaptive functions like resistance, virulence and metabolic activities and has been found to frequently contain TA modules. In the genome of Vibrio fischeri a stabilizing role of a superintegron was suggested for the type II TA module ccdA/ccdB and in a superintegron of V. cholera putative TA modules with similarities to parA/parD, higA/higB, and phd/doc systems were found. In addition, the introduction of relB/relE and parD/parE into the chromosome of E. coli resulted in the stabilization of the flanking DNA regions.

Chromosomal TA modules may also play a role in protection against invading DNAs such as plasmids and phages. Bacteria have evolved multiple phage resistance mechanisms, including abortive infection, during which the bacteriophage-infected cells commit suicide to prevent spreading of phages in the bacterial population. The type I TA module hok/sok of plasmid R1 was shown to exclude T4 pages in E. coli and the chromosomal mazE/mazF type II TA module induced abortive infection upon P1 bacteriophage attack. A further example is the toxI/toxN type III TA module isolated from Erwinia carotovora. Upon infection of bacterial cells with phages the toxin ToxN is activated and kills the cells. Recent findings showed that some phages evolved sequences encoding mimics of the bacterial antitoxin toxI that can neutralise bacterial ToxN and thus evade host cell defense systems.,

Several studies indicate a role of TA systems in the formation of bacterial persisters upon stress conditions. When bacterial populations are exposed to bactericidal factors some cells enter a dormant state in which they are recalcitrant to the unfavorable conditions. Persister cell formation in E. coli after ampicillin treatment was suggested to be regulated by chromosomal TA modules. The toxin HipA of the hipA/hipB TA module was identified as an important determinant for persister cell formation.- In addition, analysis of cellular fractions enriched for persisters revealed that these cells show increased levels of TA mRNAs., A recent study illustrates the role of several mRNase toxins in the formation of persister cells and suggests that the different TA loci encoding mRNases cumulatively contribute to this phenomenon. Since all these mRNase toxins are degraded by Lon, whose activity is regulated by environmental factors, it was suggested that the level of persister cells can be modulated by environmental conditions in a Lon-dependent way. With respect to that it is interesting to note that cross-activation between different TA systems has been described recently., The observation that TA toxin activation induces perciceter cell formation demonstrated that TA systems can enhance survival of bacteria under unfavorable conditions. However, other studies indicate that at least some TA systems mediate programmed cell death. For instance, a number of publications suggest that MazF promotes cell death in response to nutritional stress, phage infection, treatment with antibiotics, or extracellular signal peptides- while other studies report that MazF induces a reversible bacteriostatic condition, and contributes to persister cell formation. Thus it remains controversially discussed whether MazF is bactericidal or induces a bacteriostatic stage.,

Other possible functions of chromosomally encoded TA systems include regulation of biofilm formation and action as global regulators and were recently reviewed by Wang and Wood. Although a number of questions regarding the function of chromosomally encoded TA systems remain, their role is no longer as enigmatic as it was a few years ago.

Methods for Identification of TA Systems and Confirmation of Their Activity

While the first described TA systems were discovered by their plasmid stabilizing effect, novel TA systems are now usually identified by BLAST, TBLASTN, and PSI-BLAST database searches using sequences of characterized TA modules.,, To remove false positives additional filters were developed that make use of certain typical properties of TA modules, mainly the close physical association of the toxin and the antitoxin gene and the sizes of the predicted genes., Also a number of sequence independent search algorithms were established. For instance, in a search for pairs of small open reading frames with sizes of 65 to 135 amino acids each and a spacing of less than 150 bp, 18 pairs of putative TA systems were identified in the genome of E. coli MG1655. Subsequent assays for toxicity revealed that 6 genes inhibited cellular growth upon overexpression. One of them, yeeV, was recently shown to inhibit cytoskeleton polymerization and represents the toxic compound of the first type IV TA system described. For identification of type I TA modules Fozo et al. used an algorithm based on searches for small ORFs, transmembrane domain predictions and enrichment of polar or aromatic amino acids at the C-terminus. In addition, also the tendency of some type I TA modules to be tandemly duplicated in bacterial genomes was applied for identification of novel TA modules. These approaches allowed the identification of functional type I modules that could not be detected by extensive PSI-BLAST and TBLASTN searches performed in the same study.

An experimental approach for identification of type I TA modules involves the cloning or microarray-based hunt for small untranslated RNAs and the subsequent search of small ORFs in close proximity. The shoB/ohsC, symE/symR, and tpxA/ratA modules could be identified by this strategy. In a recent study novel functional type II TA systems were discovered by shotgun cloning. This approach is based on the fact that a functional toxin can only be cloned together with its cognate antitoxin (Fig. 3A). During whole-genome shotgun Sanger sequencing the microbial genome is randomly fragmented, ligated into a vector and transformed in E. coli. If a toxin is detached from its cognate antitoxin growth of the E. coli clone is arrested and, consequently, such clones are not obtained. Searches in more than 300 bacterial and archaeal genomes for genes, which are absent from clones (the putative toxin) unless the adjacent gene (the putative antitoxin) is present, were performed using publicly available data from genome sequencing projects. After application of statistical tests and additional selection criteria 8 putative, previously unidentified families of TA systems were identified and 6 of them could be experimentally evaluated. Subsequent experiments revealed that sanaTA from Shewanella sp and rlegTA from Rhizobium leguminosarum, two representatives of the newly identified TA systems, could mediate partial resistance of E. coli against infection with T7 phages. Approaches similar to “shotgun cloning” may also be suitable for detection of other types of TA systems. Important drawbacks of this method are that TA modules with very small antitoxin genes are difficult to detect (the probability of fragmentation of such a gene is too low) and that only TA systems active in E. coli can be obtained. In addition, microbial genome data are currently mainly obtained by cloning-independent next generation sequencing techniques and such data are unsuitable for analysis with the “shotgun cloning” approach.

Figure 3. Experimental approaches for identification of TA systems. (A) Shotgun cloning: the genome to be investigated is randomly fragmented, cloned and transformed into E. coli. Clones comprising the toxin but not the antitoxin do not proliferate and are absent in the assembly. (B) Plasmid stabilization: the fragmented DNA is cloned in a vector that can normally replicate in wild type hosts but that is highly unstable in polA- strains. After several rounds of replica plating under non selective conditions only colonies with an insert mediating plasmid stability still contain plasmids at a high frequency.

A versatile although rarely used method for isolation of TA systems uses their ability to stabilize plasmids (Fig. 3B). The genome or plasmid to be investigated is randomly fragmented and ligated into pALA136, a vector containing the ColE1 origin and the P1 replicon. Since this plasmid replicates in wild-type E. coli by the ColE1 origin to a moderate copy number such a gene library can be prepared with ease. The library is then transformed into a polA deficient host where the ColE1 origin is non-functional and replication switches to a low copy number under control of the P1 replicon. The transformants are subsequently replica-plated several times under non-selective conditions and finally maintenance of the plasmid is investigated by plating on an antibiotic-containing selective medium. This approach allowed the isolation of the first representatives of the stbD/stbE TA systems, which belong to the relE/parE superfamily and were later shown to inhibit translation.

To confirm the functionality of a putative TA system a number of criteria have been established. The toxin should arrest growth of its host if overexpressed in the absence of the antitoxin. Co-expression of the antitoxin should revert this effect. Such studies are often performed with a two plasmid system where one contains an IPTG-inducible lac promoter and the second an inducible ara promoter.,,, Such experiments are also frequently used to determine the so called “point of no return”, a delay time after which cells cannot be rescued anymore even if the antitoxin is expressed. However, it must be mentioned that the two plasmid system is artificial, especially with respect to the expression level and thus the results of such “point of no return” experiments should be interpreted with care. Another commonly used assay for functionality of TA systems is the stabilization of a plasmid showing a high loss rate.,,

Morphological changes of the bacterial cell after toxin induction can provide first evidence for the targeted process. Cell filamentation may be indicative for inhibition of DNA replication, bulge formation was observed in cells with impaired cell wall biosynthesis and a lemon to spherical cell shape was reported for toxins targeting the cytoskeleton. However, such morphological changes are only indicative and require confirmation by additional methods. Pulse-chase experiments using radio-labeled precursors like 3H-thymidine, 3H-uracile, and 35S-methionine (or 3H-leucine) whose incorporation is specific for DNA, RNA, and protein biosynthesis, respectively, are frequently used to investigate whether the toxin effects one of these processes.,,,,, Targeting of the cell membrane can be investigated with several staining techniques and evaluation by fluorescence microscopy., Membrane staining techniques but also the quantification of the ATP level and the activity of NAD(P)H-dependent oxidoreductases are useful for distinguishing living and dead cells. It is likely that the application of such methods will increase in the investigation of TA systems since there is a general debate whether TA systems are bacteriostatic or bactericidal.

Application of TA Systems

TA systems as tools for biotechnology and molecular biology

A major challenge in DNA cloning is the low frequency of insertion of fragments into linearized vectors. A number of vectors for positive selection of inserts contain a toxin gene, typically ccdB from the F-plasmid, that is inactivated upon insertion of foreign DNA,, allowing only insert-containing clones to grow (Fig. 4A). The ccdA/ccdB module was also used in the StabyCloningTM system, where the vector contains a truncated version of the ccdA antitoxin. Attachment of a 14 bp sequence to the 5′ end of the DNA fragment to be cloned, for instance by including it in the PCR primer, restores the active antitoxin, which is capable to counteract the toxin that has been introduced into the genome of the host cell. Thus, only cells containing a vector with an insert in the desired orientation can form colonies (Fig. 4B). The use of antibodies is not necessary for this system. The commonly used GatewayTM system is based on the recombination system of phage λ. The attB1 and attB2 sites are attached by PCR to the 5′ and 3′ end of the DNA to be cloned, respectively. Subsequently, an in vitro recombination reaction is performed using the PCR fragment, the vector containing the attP1 and attP2 sites, the λ integrase INT and the integration host factor (IHF) from E. coli. To distinguish between vectors containing the desired fragment and empty vectors the ccdB toxin gene is placed between the attP1 and attP2 sites providing powerful selection against empty vectors (Fig. 4C). In the subsequent cloning steps the same strategy is used to distinguish between empty and insert-containing vectors.

Figure 4. Application of TA systems for DNA cloning. (A) Insertion of the gene of interest destroys the toxin gene and allows the bacteria to growth. (B) Principle of the StabyCloningTM system. (C) Principle of the selection used in the Gateway cloningTM system.

Plasmid instability can be a severe problem for the production of proteins in microorganisms. Cells that had lost the expression plasmid have usually a significantly higher growth rate than construct-containing cells and can rapidly overgrow the latter. Thus antibiotic selection pressure must be maintained during the whole fermentation process in order to obtain high yields. This practice is costly, poses a risk of contaminating the product with antibiotics and requires special waste treatment to avoid the release of antibiotics into the environment. Moreover, in dense cultures it might be difficult to maintain selective pressure because the antibiotics are rapidly inactivated by resistant cells. As an alternative strategy, TA systems can be used that kill their hosts upon plasmid loss. Supplementing an existing expression vector with one or several TA systems can significantly enhance its stability. Importantly, the obtained constructs have the same requirements as the original expression vectors and can be used with standard host strains allowing for maximal flexibility. However, this approach can only delay but not completely prevent the appearance of plasmid-free cells. This disadvantage can be circumvented by integrating the toxin gene into the genome while placing the antitoxin gene on the expression vector. This separated-component-stabilization allows stabilization without the use of antibiotics, increases the yield of recombinant protein and does not require modification of the culture conditions. However, special host strains are necessary. Recently, a system was developed that makes use of the ACA-specificity of MazF, the toxic component of the E. coli mazE/mazF module, for expression of a single protein in living E. coli cells. The gene of interest is engineered to encode an ACA-less mRNA and expressed in E. coli together with MazF. The latter induces bacteriostasis by cleaving mRNAs at ACA sites. Upon toxin induction protein synthesis is shut down except for the engineered gene. Despite growth arrest, the cells retain essential metabolic and biosynthetic activities for energy metabolism, transcription, translation and biosynthesis of nucleotides and amino acids for several days allowing production of the target protein in high yields. In addition, bacteriostasis offers the opportunity to work with much denser cultures than traditional methods. This system is ideal for production of stable isotope-labeled proteins as required for NMR studies, since the amount of costly stable isotope-labeled precursors can be kept at a minimum.

In eukaryotic cells gene silencing due to epigenetic effects is a major drawback in the generation of stably expressing cell lines. Moreover, transfection of animal cell lines leads to a heterogeneous population of clones expressing the transgene at different levels. Since high expression often reduces the growth rate, cells with low or no expression are selected over time. Recently, a novel technique was described circumventing these problems: the transgene is tightly co-expressed with the Kis antitoxin in cells that expresses the Kid toxin, a ribonuclease digesting free mRNA. This system provides a positive selection for cells expressing the transgene and the antitoxin at high levels since they can overcome Kid toxicity. Indeed, a significant enrichment of cells with strong transgene/antitoxin expression was observed over time. The Kid toxin was also suggested as a tool for selectable elimination of cells in tissue cultures or even whole organisms. Other TA toxins with proven activity in mammalian cell lines include RelE and MazF and might be useful for similar techniques.

In summary, TA systems and their components are currently successfully applied in DNA cloning and protein expression in microorganisms but may also be useful tools for manipulation of eukaryotic cells.

Strategies to use TA systems for pathogen control

Since TA systems are ubiquitously present in bacterial genomes and have a considerable potency to repress growth or even kill cells they have been proposed as potential targets for development of antibacterial drugs.,- Because of their proteinaceous nature type II systems seem to be the most feasible targets for artificial activation of the toxin compound. The most straight forward approach for activation of the toxin would be a drug that disrupts TA complexes (Fig. 5A) or prevents their formation (Fig. 5B), which would directly activate the toxin. Alternatively, activation of the cellular proteases would lead to enhanced degradation of the proteolytically highly sensitive antitoxin and thereby activate the toxin (Fig. 5C). Similarly, repression of transcription of the TA operon would prevent de novo synthesis of the labile antitoxin (Fig. 5D). However, transcription of TA systems is usually autorepressed by binding of the antitoxin or the TA complex to its own promoter, which might counteract the latter strategy. In contrast, a drug interfering with autoregulation would increase the TA complex pool. After removal of the drug the excess of TA complexes would tightly repress de novo synthesis of the labile antitoxin and thus a critical level of free active toxin might be generated by proteolytic cleavage (Fig. 5E). Finally, plasmid-located TA systems might be artificially activated by inducing plasmid loss, for instance by interfering with the replication system of the plasmid (Fig. 5F). However, there are considerable difficulties in the application of these strategies. The targeted TA system must be widespread in strains of the pathogenic bacterial species to ensure reliable efficacy. In addition, the activation of TA systems bears the danger of persister cell induction, clearly an undesirable effect. Thus the TA systems to be targeted must be carefully selected. So far several peptides were reported to induce TA systems in vitro.,, One example called EDF (extracellular death factor) was reported to induce death of E. coli also upon extracellular treatment by activating MazF.-, EDF is a NNWNN pentapeptide, which is likely derived proteolytically from glucose-6-phosphate dehydrogenase in a ClpXP-dependent manner and subsequently secreted into the medium. Treatment of exponentially growing E. coli cells with culture supernatant of EDF producing bacteria or with synthetic EDF activates MazF and its homolog ChpBK by overcoming the inhibitory effects of the antitoxins MazE and ChpBI, respectively, and induces thereby cell death. Recently, it was shown that also the culture supernatants of other bacterial species contain peptides that can trigger MazF activation. Although these data are promising, further studies are required to evaluate the potential of EDF or homologous peptides as a novel class of antibiotics.

Figure 5. Strategies for artificial activation of TA systems. (A) Disruption of TA complexes. (B) Prevention of complex formation. (C) Activation of cellular proteases, for instance Lon or Clp. (D) Inhibition of TA transcription. (E) Overexpression of the TA system and subsequent removal of the activating drug. (F) Induction of plasmid loss (for plasmid encoded TA systems).

Another potential use of TA systems may be in gene therapy against viral infections. Recently, a retroviral vector was developed containing the E. coli mazF gene under control of the TAR promoter from HIV-1 (Fig. 6A). The HIV infection cycle starts with expression of the viral Tat protein, which binds to the TAR sequence to induce transcription of the entire HIV-1 genome. Infection of CD4+ cells containing the TAR-mazF construct with HIV-1 induced the expression of MazF, which in turn efficiently cleaved the viral mRNA and thus prevented replication of HIV-1 in such cells in vitro. An alternative approach was recently investigated for removal of Hepatitis C virus (HCV) infected cells. An expression cassette was constructed encoding a polypeptide incorporating MazF and a fragment of MazE (MazEp) linked via a stretch containing a NS3 cleavage site (Fig. 6B). NS3 is a viral serine protease essential for processing of polyproteins encoded by HCV. Cells containing the mazF-linker-mazEp construct can grow since MazF toxicity is neutralized by the covalently attached MazEp. In contrast, in cells infected with HCV the viral NS3 protease activates MazF proteolytically, which efficiently shuts down protein biosynthesis and eradicates infected cells. This strategy may also be useful for other viruses, since many of them, for instance HIV, depend on viral proteases.

Figure 6. TA systems as antiviral tools. (A) CD4+ cells were transfected with a construct containing mazF under control of TAR. After infection of CD4+ cells with HIV-1 the viral TAT protein is produced, which binds to the TAR sequence and triggers expression of MazF. The active MazF protein cleaves RNA including HIV-1 and prevents thereby its replication. (B) Cells were transformed with a construct containing a part of mazE (mazEp), a linker and mazF as a polyprotein. The polyprotein remains inactive until the hepatitis C virus (HCV) encoded protease NS3 cleaves the linker. The released active MazF protein cleaves RNA and triggers cell death.

Conclusion

Our knowledge about TA systems and their functions has greatly increased since their discovery 30 y ago. However, it must be emphasized that still many scientific questions remain. For instance, in 2012 two novel types of TA systems were reported. There is also a general debate whether TA systems are bactericidal or bacteriostatic and it is becoming evident that particularly chromosome encoded TA systems may have a multitude of physiological functions including protection against phage infections, induction of persister cell formation, general stress response, regulation of biofilm formation and action as global regulators. This raises also the question for host specificity of TA systems, a topic that has been little addressed as yet. Currently, TA systems and their components are used in DNA cloning and protein expression in bacterial cells while their application in eukaryotes is just emerging. Since several biotechnological applications require TA systems with specific properties, for instance certain RNA cleaving sites, there is clearly a demand for the development of novel high-throughput methods for identification and particularly for biochemical characterization of TA systems. TA systems may also offer an exciting opportunity for development of novel strategies to control pathogens. However, additional research including in vivo studies are required to fully assess their potential.