Biotechnological and Medical Exploitations of Toxin-Antitoxin Genes and Their Components
We review here the state of the art on the application of toxin–antitoxin pairs in biotechnology and medicine, touching on technologies that range from simple selection of recombinant DNA in the laboratory, to complex and ambitious therapeutic strategies that may become routine in the future.
19.1 TA Pairs as Selection Markers in Bacteria
19.1.1 Selection of Recombinant DNA Clones by Toxin Inactivation
Recombinant DNA technologies have become of paramount importance in most biology laboratories today. One inconvenience of DNA cloning is the low frequency with which DNA inserts are successfully introduced into target plasmid molecules in ligation reactions. This means that most target DNA plasmids may be circularized back during ligation without carrying an insert, and this “empty” vector population can generate a predominant background of false positive clones in subsequent selection steps. Distinguishing bacterial colonies carrying the pursued recombinant DNA from those carrying only the empty vector is costly and time-consuming. TA systems have been exploited to eliminate false positives from this mixture of clones, circumventing the problem above by only allowing the growth of bacterial cells that receive insert-bearing plasmids during transformation.
The first of these systems was developed using the CcdA and CcdB TA pair from plasmid F (Bernard 1995, 1996). These proteins contribute to plasmid F stability by killing newborn daughter cells that have not inherited a plasmid copy at cell division (Ogura and Hiraga 1983; Jaffe et al. 1985). CcdB toxin kills cells by inhibiting Gyrase A, one of the subunits of E. coli’s topoisomerase II, and this poisoning activity is neutralized by CcdA (Miki et al. 1984). The ccdB gene was fused to the multiple cloning site (MCS) of plasmid pUC18 (Vieira and Messing 1982), creating a fusion gene encoding for the N-terminal fragment of beta-galactosidase and CcdB itself (Bernard et al. 1994). The resulting CcdB variant produced from this vector is sufficiently toxic to kill bacteria in the absence of CcdA. However, if ccdB is inactivated by insertion of a foreign DNA fragment in the MCS above, the recombinant plasmid cannot produce CcdB and no longer interferes with host viability. This positive selection tool for recombinant clones is highly efficient and simplifies cloning procedures, as only cells containing recombinant plasmids give rise to colonies. Similar technologies were implemented later on using toxin Kid from plasmid R1 (Gabant et al. 2000) and toxin ParE from plasmid RK2 (Kim et al. 2004). In all cases, empty vectors can be amplified using E. coli strains that express the corresponding antitoxin partner (Kis or ParD) from chromosomal locations, or that produce a CcdB-insensitive GyrA mutant.
The technology based on CcdB was initially marketed by Delphi Genetics SA, a spin-off company of the Université Libre de Bruxelles (ULB), and subsequently licensed to Invitrogen Inc. The latter company has included the technology in a large portfolio of positive selection vectors with different copy numbers, host range, and/or transfer properties, making it a standard tool in many molecular biology laboratories worldwide. For instance the technology is included in Invitrogen’s Gateway vectors, which exploit site-specific recombination to shuttle DNA inserts from donor vectors into all sorts of recipient vectors rapidly and efficiently. In this technology positive recombinants are selected because site-specific recombination replaces the ccdB gene in recipient vectors by the gene of interest transferred from donor vectors (Walhout et al. 2000).
19.1.2 Selection of DNA Recombinant Clones by Antitoxin Reconstitution
A new generation of positive selection vectors based on CcdB has recently been generated by Delphi Genetics SA to produce the StabyCloning™ system. In this case, the ccdB gene is inserted in the chromosome of E. coli cells, and a truncated inactive version of the antitoxin gene is present in the cloning vector. Insert DNAs that include in their 5′ end the 14 base pairs missing from the antitoxin gene can reconstitute the inactive ccdA gene if they are integrated in the vector in the right orientation. This selection strategy is well suited for cloning blunt-ended PCR products produced using 5′ oligonucleotides that include the missing base pairs above, an approach that has been used recently for the generation of microsatellite libraries (Tzika et al. 2008). One drawback of this strategy is that it requires modification of the strain used for selection. An advantage is that there is no need to include antibiotic resistance genes in the DNA vector, as reconstituted ccdA functions as the selective marker for plasmid maintenance. This eliminates some of the pitfalls associated with the use of some antibiotics during selection, like growth of satellite plasmid-free colonies.
19.1.3 Non-Antibiotic Markers for Selection of Plasmid-Bearing Cells
Antibiotics are commonly used during bacterial fermentation, and the vast majority of expression vectors contain antibiotic resistance genes to enable counter-selection of plasmid-free cells and avoid that they dominate the culture. The use of these genes raises safety concerns that are increasingly highlighted by regulatory authorities worldwide (Vandermeulen et al. 2011). Thus, the expectation is that a “zero tolerance” toward antibiotic-based selection and production systems will be the general consensus in the near future (Peubez et al. 2010). Given the tremendous impact that this decision would have on both academic and industrial settings, other approaches for selection of plasmid-containing cells are required.
TA pairs found in plasmids function as stability systems and therefore they offer an attractive alternative to the use of antibiotic resistance genes for plasmid selection. The introduction of TA pair hok-sok (Gerdes et al. 1986) or parDE (Roberts et al. 1993) in a high-copy number DNA vector was shown to reduce the plasmid loss frequency from 1.4 × 10−2 cells per generation to 2.4 × 10−8 and 2 × 10−5 cells per generation, respectively, in E. coli. Placing both toxin–antitoxin pairs on the same plasmid reduced loss frequency even further, down to 5.4 × 10−11 per cell and generation. This level of stability enhancement increases from 3 to 43 h the time required to accumulate 10 % of plasmid free cells in exponentially growing cultures, without affecting cell growth rates, or their efficiency producing recombinant proteins. This stability enhancement is well suited for batch and fed-batch fermentations but not for continuous operations because TA pairs only delay, but do not prevent, the takeover of a culture by plasmid-free cells. For instance, in the case above only 10 % of cells in the culture still retain the plasmid after 90 h of growth. In view of the stabilizing synergy displayed by hoksok and parDE it would be interesting to test whether other TA combinations, or the addition of other TA systems to the plasmid above, would enhance its stability even further.
Physical separation of toxin and antitoxin partners has also been used to achieve plasmid stabilization in the absence of antibiotics (Szpirer and Milinkovitch 2005). As with the StabyCloning technology, in this strategy toxin CcdB is expressed from the cell’s chromosome while antitoxin CcdA is expressed from the bacterial plasmid. Plasmid maintenance, and therefore CcdA expression, ensures neutralization of CcdB expressed in host cells. Cells losing the plasmid cannot produce the antitoxin, and continuous production of CcdB eliminates the plasmid-free cell. This approach makes it very difficult for the latter cells to escape cell death and, when combined with stabilization systems found in high-copy number plasmids, like the cer locus of ColE1 (Summers and Sherratt 1984), provides an excellent stabilization enhancement of plasmids in bacteria. One pitfall of this approach is that it requires a modification of the host strain, but the system improves recombinant protein expression and/or plasmid recovery from those modified cells and therefore provides a safer and more efficient manufacturing alternative for the production of DNA and protein vaccines (Peubez et al. 2010).
19.2 Artificial Activation of TA Pairs as Antibiotic Strategies
The discovery of antibiotics is a major medical achievement of the past century. The availability of penicillin, streptomycin, and sulfonamides in the late 1940s, and of chloramphenicol, erythromycin, and tetracycline in the following decade raised the expectation that deaths associated to bacterial infections would rapidly become a problem of the past. However, such anticipated hopes were shadowed by the overwhelming adaptability of microorganisms, and the rate at which they developed resistance to antibiotics (Davies 2007). The problem persists today despite a considerable expansion of our antibiotic weaponry, and both the number of resistant microorganisms and the breadth of resistance in single bacteria are unprecedented and mounting (Levy and Marshall 2004). The consequences of this are dramatic; 1.7 million U.S. citizens acquired bacterial infections in hospitals in 2002, causing 99,000 deaths and health care costs between $5 and 10 billion, with similar statistics reported from the EU, and reaching a disproportionate scale in developing countries (Peleg and Hooper 2010; Okeke et al. 2005).
Most bacteria become insensitive to antibiotics through the acquisition of resistance determinants that accumulate in plasmids called R-factors (Alekshun and Levy 2007; Nikaido 2009). R-factors are mobile genetic elements that can promote their conjugative transfer between bacteria of different genera. This, and their ability to exchange genetic material with co-existing plasmids and host cell chromosomes, explains the facility with which antibiotic resistance evolves and spreads in both Gram-positive and -negative bacteria (Martinez and Baquero 2002; Carattoli 2009; Norman et al. 2009; Jensen et al. 2010).
The emergence and spread of pathogenic bacteria that have become resistant to multiple antibiotics through lateral gene transfer have increased the need for novel antimicrobials. Interestingly, TA pairs are ubiquitous among plasmids conferring antibiotic resistance and chromosomes of pathogenic bacteria. In the latter case TA pairs tend to be clustered and linked to mobile genetic elements suggesting that, as described for antibiotic resistance genes, TA pairs are mobile elements that move frequently within and between plasmids and chromosomes (Pandey and Gerdes 2005; Guglielmini and Van Melderen 2011; Leplae et al. 2011; Shao et al. 2011). These observations, the apparent absence of TA pairs from eukaryotic organisms, and the fact that these toxins inhibit bacterial growth and/or induce bacterial cell death (Yamaguchi and Inouye 2011) has increased efforts directed at inducing artificial activation of toxins as an antibiotic strategy.
19.2.1 Indirect Approaches for Toxin Activation
22.214.171.124 Inhibitors of Toxin–Antitoxin Production
The report above pointed out that sequence-specific DNA binders that inhibit transcription of target genes have been recently designed (Raskatov et al. 2012) and that molecules like this, able to bind promoters at TA operons, may be designed to repress transcription at these loci. We feel that this strategy may not be suitable in the latter case as sequence-specific DNA binders are designed to inhibit transcription of target genes by competing with transcriptional activators (Raskatov et al. 2012). Most TA proteins function as transcriptional auto-repressors, so inhibiting transcription of these operons would require ‘stapling’ TA complexes to their DNA binding sites, rather than inhibiting their interaction with them. Intuitively, developing ‘TA-DNA staplers’ might be a very challenging task.
An alternative would be to design antisense (as) RNAs that provoke the specific degradation of the mRNAs encoding type II TA pairs, or that outcompetes antitoxin RNAs in the case of type I TA pairs. The suitability of the latter approach has already been validated experimentally and has proved to be an efficient way of inducing toxin activation and cell killing in E. coli cells bearing plasmids that encode the hok-sok TA locus (Gerdes et al. 1986; Faridani et al. 2006). Delivering asRNA can be achieved using liposomes or self-penetrating peptide nucleic acids (Faridani et al. 2006; Meng et al. 2012; Bai et al. 2012). Of note, these delivery systems can also translocate mRNA through mammalian cell membranes, so the effects in human cells of any asRNAs designed to activate toxin expression should be carefully evaluated before hand, to discard that it silences the expression of essential genes in humans.
A pitfall of any approach directed at inhibiting the expression of type II TA pairs is that they would only induce the activation of toxin molecules that were produced in cells before treatment. The concentration of these proteins is kept relatively low in cells due to feedback loop regulatory mechanisms, like transcriptional auto-repression (Yamaguchi and Inouye 2011). Thus, permanent inhibition of toxin and antitoxin production is likely to induce only transient and mild toxicity. This may not be enough to kill cells, as many of these toxins only kill bacterial cells if activation is sustained long enough (Amitai et al. 2004; Lioy et al. 2006, 2012).
126.96.36.199 Increased Degradation of Antitoxins
Activation of antitoxin-degrading proteases (i.e. ClpP and Lon) is more likely to trigger sustained toxin activation, as constant antitoxin degradation would impede transcriptional auto-repression by TA complexes, overcoming the limitation above. Compounds that stimulate the activity of ClpP have been developed (Sass et al. 2011; Leung et al. 2011) and it would be interesting to test whether they induce activation of toxins degraded by this protease. In any case, it should be noted that this strategy is likely to induce indiscriminate killing of bacterial cells, independently of whether they contain TA pairs or not, as suggested by results obtained with Lon and ClpP (Goff and Goldberg 1987; Christensen et al. 2004; Sass et al. 2011). Furthermore, human cells encode homologues of these proteases and the possibility that activators may lead to undesired side effects in treated patients should be discarded appropriately.
19.2.2 Direct Approaches for Toxin Activation
Direct approaches constitute a more straightforward approach to induce toxin activation, where a drug directly targets toxin and/or antitoxin proteins to either disrupt preformed complexes or to prevent their formation in the first instance. So far, only disruption of preformed complexes has been explored. In one work, hepta- and octapeptides representing fragments of a region in Bacillus anthracis antitoxin PemI that were predicted to interact with its toxin partner, PemK, were examined for their ability to disrupt PemK-PemI complexes (Agarwal et al. 2010). Some of these peptides managed to disrupt the TA complex with considerable efficiency. However, they also reduced the endoribonucleolytic activity of PemK, which is responsible for its toxic effect in cells. Most probably this is due to the fact that screened peptides were derived from the stretches of the antitoxin known to interact with the toxin. Although these results are encouraging and constitute a valuable proof of principle, partial activation of the toxin may not be enough to achieve the pursued antibiotic effect. It would be desirable to find new molecules able to disrupt TA complexes completely by targeting the antitoxin component, as these should not interfere with the activity of the toxin. Peptides like these may be derived from PemK regions known to interact with PemI, a possibility that is worth testing.
In a different work a library of 6-, 14-, and 17-aminoacid long peptides was screened in search of molecules that could disrupt the interaction between toxin ζ and antitoxin ε encoded by plasmid pSM19035 from Streptococcus pyogenes. Disruption of toxin ζ-antitoxin ε complexes was accomplished using two different mixtures of 17-aminoacid peptides, although the effect was not observed when single peptides in those mixtures were used instead (Lioy et al. 2010). These results pointed out that complex disruption might have been achieved through the combined action of different peptides with weak disrupting activity. Whether these results could be exploited to generate single disruptor molecules remains to be established. Furthermore, the work was performed using an inactive derivative of toxin ζ, and did not include an analysis of whether these peptides interacted with the toxin or with the antitoxin. Thus, further work will be required to determine if these peptides disrupt the TA complex without inhibiting toxin ζ.
19.2.3 Selecting the Right Strategy
Although still too early to predict the therapeutic value of toxin activation in clinical settings, the results above suggest that development of TA disrupting agents is technically possible. Yet, drug discovery and development is a long and expensive process, and selecting the right approach will be important to speed up progress toward new antibiotics based on TA disruptors. An ideal TA complex-disruptor should be small, to facilitate its pharmacological formulation. This may prove difficult to accomplish, given the high affinity with which most antitoxin interact with their partners and the extension and physical–chemical nature of these interactions (Kamada et al. 2003; Koga et al. 2011; Overgaard et al. 2009; Chopra et al. 2011; Fiebig et al. 2010; Yang et al. 2010; Zhu et al. 2010). Lead compound identification should rely on screenings designed to select for disruptors that interact with antitoxins, so that positive hits do not interfere with the toxin’s activity. It will be essential to select the right TA targets, a decision that should be based on data concerning their clinical relevance (i.e. their prevalence in antibiotic resistance clinical isolates) (Moritz and Hergenrother 2007; Williams et al. 2011), their functionality, and evidence of their bactericidal effect. Furthermore, the mode of action of some toxins enables cells to develop resistance against them (Bernard and Couturier 1992). Some toxins induce persistence, a dormant, reversible state that protects cells from the action of antibiotics (Wang and Wood 2011; Maisonneuve et al. 2011), or promote the survival of cells under stress (Amitai et al. 2004; Lioy et al. 2006, 2012). Toxins known to induce any of these phenotypes should also be discarded as targets.
TA pairs encoded in plasmids are thought to function as post-segregational killing (PSK) systems (Van Melderen and Saavedra De Bast 2009) and, intuitively, they appear to be ideal candidates for the approaches above. However, it has been shown that the kis-kid TA pair encoded in plasmid R1 functions as a rescue system, and that its activation increases plasmid copy numbers and enforces retention of antibiotic resistance in host cells, without killing them (Pimentel et al. 2005; de la Cueva-Mendez and Pimentel 2007). It is possible that other plasmid-encoded TA pairs may function as rescue systems. If so, their artificial activation is likely to keep cells alive and at the same time increase quantitatively their antibiotic resistance. Moreover, the simultaneous increase in plasmid copy numbers and sustained TA disruption should also lead to higher production of both the toxin and the antitoxin, which could titrate out the complex disruptor. Therefore, TA pairs like these should also be discarded as potential targets for disruptor discovery programs.
19.3 TA Pairs as Facilitator of Protein Structural Analysis
Deciphering the three-dimensional structure of proteins and their complexes is in many cases essential to understand biological function, and constitutes a major requirement for the rational design of pharmacological agents directed against them. Unfortunately, protein structure determination is often a long and expensive process, with a success rate below 5 %. More than 75 % attempts fail because target proteins cannot be prepared in sufficient amounts or suitable forms for structural analysis, and 70 % of the remaining cases do not succeed at producing good crystals or high quality NMR spectra (Terwilliger et al. 2009). Membrane proteins are particularly vulnerable to the limitations above, as evidenced by the fact that they constitute less than 1 % of the high-resolution protein structures solved to date (Watts 2005). This hampers drug development because membrane proteins comprise one-third of all cellular polypeptides and are involved in 85 % of cell signaling pathways, making them very attractive drug targets in the pharmaceutical industry.
New enabling technologies are required to overcome the problem above, and TA pairs are being used to implement some of them. For instance, solid-state (ss) in cell NMR facilitates the study of drug interactions with membrane proteins, and even the structural analysis of large functionally active complexes with atomic resolution. Importantly, in-cell NMR does not require sample purification and when applied in solid-state, the hydrophobic nature of membrane proteins is not an issue that affects sample preparation. However, in-cell NMR is not a very sensitive technique, and it requires that target proteins are over-expressed by cells, and that they incorporate isotopic labels both uniformly and exclusively, to distinguish them from other cellular proteins (Reckel et al. 2005). The recent development of a condensed single protein production (cSPP) system, which exploits E. coli’s toxin MazF, constitutes an important step in this direction. cSPP is a highly effective method for high yield production of recombinant proteins that also enables their exclusive isotopic labeling in vivo (Suzuki et al. 2005).
19.3.1 Condensed Single Protein Production
cSPP relies on two separate technologies, both devised by Masayori Inouye’s laboratory. The first one is an expression system (pCOLD) that achieves formidable yields of recombinant protein production in conditions that increase solubility and stability. pCOLD is based on the observation that, when exposed to low temperature, E. coli induces the expression of a handful proteins that help cells to adapt to cold-shock (Jones et al. 1987). CspA (cold-shock protein A) stands out among these proteins, and constitutes up to 2 % of total protein content in E. coli cells grown at low temperature (Goldstein et al. 1990). The expression of CspA is regulated post-transcriptionally. cspA mRNA is produced both at 37 and 15 ºC, but its 5′ and 3′ untranslated regions (UTR) function as a thermosensor (Giuliodori et al. 2010). These UTR regions fold differently at each one of these temperatures, and the structures produced have opposite consequences for the stability and translatability of the mRNA. Folding at 37 ºC hides the mRNA’s Shine-Dalgarno (SD) sequence and initiation codon from ribosomes, and also turns the mRNA very unstable. In contrast, folding at 15 ºC exposes the SD sequence and initiation codon in a structure that is avidly recognized by ribosomes, increasing translation rates, and also augments the stability of the transcript 100-fold (Xia et al. 2002).
These features have been exploited to generate a series of expression vectors (pCOLD) that are induced at low temperature, enabling production of large amounts of soluble proteins in E. coli cells by simply growing them at 15 ºC (Qing et al. 2004). In these conditions, most of the cell’s translation machinery is devoted to produce the recombinant protein, which enables its preferential isotopic labeling. Using pCOLD it is possible to perform a basic NMR characterization of recombinant proteins directly from cleared cell lysates, thus skipping the need to purify them to homogeneity. However, although NMR spectra obtained this way provide valuable information about protein structural integrity they do not allow solving their 3D structures with atomic resolution.
The latter requires exclusive (rather than predominant) isotopic labeling of target polypeptides, and background protein synthesis in cells under cold-shock is sufficiently high to impede this happening. To sort this out, Inouye’s laboratory has cleverly exploited the toxin component of the mazEF TA pair in E. coli. MazF is an mRNA interferase that cleaves RNA at single-stranded ACA sites, and that is neutralized by its antitoxin partner MazE (Zhang et al. 2003). ACA sites are abundant in genes and therefore MazF depletes cells from mRNAs, inducing a dramatic inhibition of protein synthesis in E. coli cells. Interestingly, MazF arrests cells in a “quasi-dormant” state that sustains ATP regeneration, transcription, and translation (of ACA-less mRNAs), and that can be prolonged for days. ACA sites can be eliminated silently from genes, and work in Inouye’s laboratory demonstrated that concomitant expression of MazF and a target gene engineered to encode an ACA-less mRNA results in sustained and high-level target expression, in the virtual absence of background cellular protein synthesis (Suzuki et al. 2005). Inouye’s group also showed that these features are maintained in E. coli cells grown at low temperature, making it possible to combine the effects of MazF and pCOLD vectors to achieve very high yield production of recombinant proteins from ACA-less mRNAs (20–30 % of total protein) and their almost exclusive isotopic labeling (90 % of the incorporated isotope) in vivo. Furthermore the system allows production of target proteins in the absence of cell growth, which has two important advantages. On the one hand, it may be used to express and characterize the structure of some toxic proteins. On the other hand, it allows up to 150-fold condensation of cultures at the expensive stage of isotopic labeling, reducing costs as much as 99 % without compromising yields, which still remain in the range of milligrams per milliliter of culture.
19.3.2 SPP and in-Cell NMR
The system has been used for the structural analysis of prokaryotic and eukaryotic membrane proteins, which could be incorporated into E. coli membranes and produced reasonably well-resolved NMR spectra using whole cells (Mao et al. 2009), and even better ones when performed on outer membrane fractions from these samples (Vaiphei et al. 2011). Furthermore, as the SPP system allows protein production in the absence of cell growth, cerulenin, an inhibitor of phospholipid biosynthesis (Heath et al. 2001) could be used without killing cells to eliminate the typical spurious signals in ssNMR spectra that arise by non-specific isotopic labeling of phospholipids when 13C-glucose is used as precursor, further improving NMR spectra (Mao et al. 2011).
19.3.3 Improvement of cSPP
The cSPP system is being subject of clever improvement. For instance, co-induction of MazF and the target protein results in 20 % of the latter protein being produced without isotopic enrichment, which reduces the amount of final sample that can be observed by NMR. This problem was eliminated by either inducing transcription of both proteins independently (Schneider et al. 2009) or by using tryptophan-or histidine-free MazF variants to perform cSPP in E. coli auxotrophs for one of those amino acids. This trick enables exclusive production of the mRNA interferase in media lacking tryptophan or histidine, and subsequent synthesis of target proteins in the presence of these aminoacids (Vaiphei et al. 2011). Using this approach more than 98 % of the target protein is labeled in the final sample. More recently, the system has also been modified to produce soluble, isotopically labeled, human proteins in E. coli that accumulate in the periplasmic space of E. coli, from where they can be recovered easily for their subsequent NMR analysis (Mao et al. 2010).
The most critical step in the cSPP system is the preparation of an ACA-less gene for target proteins. Although this can be achieved by oligonucleotide-directed site-specific mutagenesis ACA sites can be very abundant in genes, making the approach cumbersome and time-consuming. The problem can be circumvented by chemically synthesizing the entire gene. An attractive option would be to generate cSPP systems based on mRNA interferases that are more specific than E coli’s MazF. For instance, Kid, an homologue of MazF encoded in plasmid R1, cleaves mRNA at single-stranded UUACU sites, and also inhibits protein synthesis and cell growth without killing E. coli cells (Pimentel et al. 2005). Similarly, some MazF homologues in S. aureus, B. subtilis, M. xanthus, M. tuberculosis, and H. walsbyi, cleave mRNA at specific pentanucleotide (even heptanucleotide) sequences (Yamaguchi and Inouye 2011). It would be interesting to test if these mRNA interferases can be used to develop cSPP systems that reduce (even eliminate) the hassle of mutating cleavage sites from target genes.
A limitation of in-cell NMR spectroscopy in bacteria is its inability to study post-translational protein modifications or to analyze interactions between eukaryotic targets and their ligands in more physiological environments. To overcome this limitation development of a eukaryotic SPP system would be ideal. Interestingly, mRNA interferases Kid and MazF are functional in eukaryotic cells (de la Cueva-Mendez et al. 2003; Shimazu et al. 2007). Both proteins induce apoptosis in human cells, but the observation that MazF still shuts off protein synthesis completely in NBK/BIK- or BAK-deficient human cells, without killing them, provides an interesting clue of how a human cell-based SPP system could be generated.
19.4 Some TA Pairs are Functional in Eukaryotic Cells
As pointed out above, several groups have analyzed the effects of TA pairs in eukaryotic cells. Expression of RelE was shown to inhibit growth in Saccharomyces cerevisiae, an effect that was counteracted by co-expression of antitoxin RelB (Kristoffersen et al. 2000). Expression of RelE is toxic also in human cells, where it induces apoptosis, although in this species the neutralizing effect of RelB was not evaluated (Yamamoto et al. 2002). The results above led to the proposal that relE and relB genes could be used to restrict growth of genetically modified yeast strains to controlled environments. For instance, toxin RelE may be expressed constitutively in cells, independently of environmental conditions, while production of neutralizing amounts of antitoxin RelB would only be allowed in controlled environments, and not in cells that may escape from these.
Similar results were obtained later on using toxin MazF and antitoxin MazE, the chromosomal homologues of Kid and Kis in E. coli, in human cells (Shimazu et al. 2007). This work showed that the mRNA interferase activity of MazF (and presumably also of Kid) is responsible for its lethal effect in human cells. Furthermore, it described that induction of apoptosis by MazF in human cells requires pro-apoptotic protein BAK and its upstream regulator NBK/BIK. Cells deficient in BAK do not induce apoptosis when exposed to MazF, although the protein still inhibits protein synthesis in these cells, completely. As mentioned earlier this offers the attractive possibility of developing an SPP system in mammalian cells, which would have great value for the structural characterization of human proteins by in-cell NMR.
19.5 Selection of High Transgene-Expressing Mammalian Cell Pools
The strength and stability of transgene expression in stably transfected mammalian cells depends on their chromosomal integration site. Integration often occurs randomly and the identification of cells that have integrated the transgene in chromosomal locations favoring sustained and high expression levels is time-consuming. Kid and Kis have been exploited to carry out effortless selection of clones with the characteristics above. For this, cells are stably transfected with a plasmid from which Kid expression can be transactivated with doxycycline. These cells are then transfected with a second plasmid from which the protein of interest and Kis are co-expressed using a bicistronic transcriptional unit. Strict co-expression of the transgene with antitoxin Kis in cells that express Kid allows enrichment of cells expressing high levels of the transgene. Progressive increase of Kid expression in these cells using doxycycline selects for cells that increase transgene and Kis expression appropriately. Extending this procedure for several weeks allows the selection of cells that express the transgene 120-fold higher than the initial population, in the absence of antibiotic selection (Nehlsen et al. 2010).
19.6 Selective Cell Killing in Multi-Cellular Organisms
Toxicity of Kid and protection by Kis occur in both somatic and embryonic cells. Experiments where single blastomeres of two-cell embryos of X. laevis were microinjected with Kid, or Kis, or both Kid and Kis proteins showed that the toxin inhibits development of the injected half of the embryo, and that this does not happen when Kis alone or both Kid and Kis are injected instead (de la Cueva-Mendez et al. 2003). These observations provided the grounds to analyze whether kis and kid could be exploited to kill specific cells in animals without harming any other cell in the organism. Targeted cell ablation is commonly used as a tool for studying the role of a particular cell line in the multicellular context of the entire organism. Genetic methods for expressing toxic molecules under the control of tissue-specific promoters allow high specificity of cell targeting (Arase et al. 1999; Roman et al. 2001). Nevertheless, in many cases, even minute amounts of the toxin expressed outside of the targeted cells can compromise the viability of the organism. Obstacles of a similar nature often prevent the use of targeted cell ablation for medical purposes, such as in cancer treatment. It had been claimed that Kid and Kis could be used to refine the specificity of the toxic action to a well-defined cell population by protecting non-targeted cells using minimal background levels of Kis expression, but not enough to protect targeted cells from toxicity (de la Cueva-Mendez et al. 2003).
19.7 Antiviral Therapies
MazF and MazE have also been used to implement novel antiviral strategies against Hepatitis C virus (HVC) and Human Immunodeficiency Virus (HIV). NS3 serine protease and PR aspartyl proteases are essential for HCV and HIV replication, respectively, and represent prime targets for developing antiviral therapies. In a recent work, the C-terminal 41-residue fragment of antitoxin MazE was fused to the N-terminal end of toxin MazF using linkers bearing a specific protease cleavage site for either HIV PR (HIV-1 protease) or NS3 protease (HCV protease). Incubation of these ‘MazF-zymoxins’ with the corresponding proteases induced proteolytic cleavage and release of the MazE peptide and enabled cleavage of single-stranded RNA in vitro (Park et al. 2012). Similar observations were made by a different group, which also fused MazF to a neutralizing fragment of MazE via an NS3-cleavable linker. Expression of this fusion protein was well tolerated in naive healthy cells but killed HCV-infected cells and cells expressing NS3 (Shapira et al. 2012).
Activation of mazF transcription directed by the long terminal repeat (LTR) of HIV-1 has also been used to develop an anti-viral therapy (Chono et al. 2011a). Transcriptional activation at the LTR requires the Tat protein, which is produced by HIV-1 infection. CD4+ T cells carrying mazF downstream of HIV1-LTR in their chromosome were infected with HIV-1. Tat expressed in these cells upon infection induced MazF expression and this resulted in degradation of ACA-rich HIV mRNA, and prevented viral replication. Moreover, the same result was observed when monkey primary CD4+ T cells carrying LTR-mazF were infected with simian/human immunodeficiency virus (SHIV). Strikingly, Tat-induced expression of MazF did not result in cleavage of cellular mRNA, and neither killed nor inhibited growth of CD4+ T cells. This result is somehow surprising considering the potent endoribonucleolytic and cytotoxic effect described for MazF in mammalian cells (Shimazu et al. 2007) and therefore deserves further investigation.
In a later work, MazF-transduced CD4+ T cells were infused autologously in monkeys (Chono et al. 2011b). These cells persisted for very long time in animals, and they did not develop antibodies against MazF, suggesting that the approach may constitute a suitable therapeutic approach against HIV. Supporting this, cells harvested from treated monkey more than 6 months post-infusion still suppressed the replication of SHIV ex vivo. All this supports that a MazF-based CD4+ T cell therapy may help the immune system to maintain a stable condition in HIV infected patients. As mentioned above understanding why mRNAs escape from MazF cleavage in these cells requires further investigation. This may happen because MazF is not expressed strongly enough upon infection. Accordingly, the mRNA interferase could not be detected in cells transduced with HIV in the experiments above. However, detection was possible when these cells were transfected with retroviral vector overexpressing Tat. A multiplicity of infection (MOI) of 0.01 was used to transduce HIV in the experiments above. This may allow enough MazF expression to degrade HIV RNA, but not cellular mRNAs. Once HIV RNA is degraded, no more Tat (and therefore no more MazF) is produced, and the system is reset. It will be important to establish whether viral amplification taking place in cells of infected patients would lead to higher MOI than those tested in the work above, and whether this would increase MazF expression in cells to an extent that induces their death, If so, the population of infused mazF-CD4+ T cells may decrease more rapidly than expected, which would be detrimental from a therapeutic point of view.
Work in Dr. de la Cueva-Méndez’s laboratory is supported by Fundación Pública Andaluza Progreso y Salud, which depends on the Consejeria de Salud y Bienestar Social of the Junta de Andalucia.
- Arase, K., Saijo, K., Watanabe, H., Konno, A., Arase, H., & Saito, T. (1999). Ablation of a specific cell population by the replacement of a uniquely expressed gene with a toxin gene. Proceedings of the National Academy of Sciences of the United States of America, 96, 9264–9268.PubMedCrossRefGoogle Scholar
- Chopra, N., Agarwal, S., Verma, S., Bhatnagar, S., & Bhatnagar, R. (2011). Modeling of the structure and interactions of the B. anthracis antitoxin, MoxX: Deletion mutant studies highlight its modular structure and repressor function. Journal of Computer-Aided Molecular Design, 25, 275–291.PubMedCrossRefGoogle Scholar
- Christensen, S. K., Maenhaut-Michel, G., Mine, N., Gottesman, S., Gerdes, K., & Van Melderen, L. (2004). Overproduction of the Lon protease triggers inhibition of translation in Escherichia coli: Involvement of the yefM-yoeB toxin–antitoxin system. Molecular Microbiology, 51, 1705–1717.PubMedCrossRefGoogle Scholar
- Kim, H. G., Hwang, H. J., Kim, M. S., Lee, D. Y., Chung, S. K., Lee, J. M., et al. (2004). pTOC-KR: A positive selection cloning vector based on the ParE toxin. BioTechniques, 36(60–62), 64.Google Scholar
- Koprunner, M., Thisse, C., Thisse, B., & Raz, E. (2001). A zebrafish nanos-related gene is essential for the development of primordial germ cells. Genes & Development, 15, 2877–2885.Google Scholar
- Raskatov, J. A., Meier, J. L., Puckett, J. W., Yang, F., Ramakrishnan, P., & Dervan, P. B. (2012). Modulation of NF-kappaB-dependent gene transcription using programmable DNA minor groove binders. Proceedings of the National Academy of Sciences of the United States of America, 109, 1023–1028.PubMedCrossRefGoogle Scholar
- Sass, P., Josten, M., Famulla, K., Schiffer, G., Sahl, H. G., Hamoen, L., et al. (2011). Antibiotic acyldepsipeptides activate ClpP peptidase to degrade the cell division protein FtsZ. Proceedings of the National Academy of Sciences of the United States of America, 108, 17474–17479.PubMedCrossRefGoogle Scholar
- Schneider, W. M., Inouye, M., Montelione, G. T., & Roth, M. J. (2009). Independently inducible system of gene expression for condensed single protein production (cSPP) suitable for high efficiency isotope enrichment. Journal of Structural and Functional Genomics, 10, 219–225.PubMedCrossRefGoogle Scholar
- Van Melderen, L., & Saavedra De Bast, M. (2009). Bacterial toxin–antitoxin systems: More than selfish entities? PLoS Genetics, 5(3), .Google Scholar
- Williams, J. J., Halvorsen, E. M., Dwyer, E. M., DiFazio, R. M., & Hergenrother, P. J. (2011). Toxin–antitoxin (TA) systems are prevalent and transcribed in clinical isolates of Pseudomonas aeruginosa and methicillin-resistant Staphylococcus aureus. FEMS Microbiology Letters, 322, 41–50.PubMedCrossRefGoogle Scholar