Toxin-Antitoxin (TA) genetic modules encode two-component systems documented in virtually all the prokaryotic kingdom. The toxin part of such a system is responsible for reversibly or irreversibly halting bacterial cell growth while the antitoxin moiety neutralizes the toxin. So far, six classes of TA systems have been described based on the molecular mechanism of toxin neutralization by the antitoxin, but type II TA systems are most well studied in terms of toxins' target, toxin-antitoxin interactions and biological functions. TA systems were first identified as plasmid "addiction" modules involved in post-segregational killing as a way to ensure long-term plasmid maintenance in bacterial populations. However, the discovery that TA modules are widespread on bacterial chromosomes pushed to revisit the initial assumption that their biological function would be confined to mobile DNA. Since then, a broader range of biological functions of TA systems in defense against bacteriophages and in promoting persistent infections has been described.
Bacterial persisters are antibiotic-tolerant cells that form as phenotypic variants among a population of genetically sensitive cells and contribute to the recalcitrance of chronic and relapsing infections. The genetic basis of this phenomenon has remained elusive and is hotly debated particularly regarding the role of TA systems. In this Ph.D. thesis, I therefore critically investigated the contribution of type II TA systems in the formation or survival of persister cells of Escherichia coli strains. Indeed, Papers I and II disprove the previously dominant model that type II TA systems are the primary drivers of persistence in E. coli K-12 and show that PasTI, previously thought to be a TA system important for persistence of uropathogenic E. coli (UPECs), is not a TA system.
In Paper I, we re-evaluated the well-known but controversial model of persister formation in the commonly used E. coli K-12 laboratory strain that had linked a specific set of 10 type II TA systems to the induction of a persister state under unstressed conditions. This model proposed a cascade of molecular events starting from (p)ppGpp-dependent Lon protease activation and ending up with the release of free toxins responsible for the onset of the dormancy underlying persistence. While it had generally been widely accepted, inconsistent results of different laboratories cast doubt about some of the results. We therefore critically revisited previous work on this model in order to explain and reconcile the experimental divergences. We showed that results obtained with traditional methods for the detection of persisters were distorted by insufficiently controlled experimental parameters. In addition, the inadvertent lysogenization of many important mutant strains with bacteriophage ø80 caused many of the phenotypes that had supported the model before. We therefore reconstructed these mutants and adopted a more controlled antibiotic killing assay protocol to follow the dynamics of persister formation/survival throughout the phases of bacterial growth. While we readily supported role of (p)ppGpp and Lon in antibiotic tolerance, we were unable to confirm a role of type II TA systems in E. coli persistence under any tested experimental condition.
In Paper II, we were intrigued by previous work suggesting that persister formation of UPEC – a major pathogen causing chronic and relapsing infections – would be driven by a largely uncharacterized TA module, pasTI. We therefore studied the role of pasTI in persister formation of pathogenic as well as standard laboratory E. coli strains. Surprisingly, we found that pasTI is not a TA system. Instead, PasT is the previously overlooked bacterial homolog of Coq10, a mitochondrial protein known as a key factor for ubiquinone-dependent respiration in yeast and humans. Consistently, we show that E. coli mutants lacking pasT have respiratory defects that cause pleiotropic phenotypes including impaired persister formation. Remarkably, the respiratory defects and impaired antibiotic tolerance of E. coli pasT mutants could be complemented by ectopic expression of human coq10. Our work thus unites PasT/Coq10 as a single family of proteins supporting cellular respiration that is conserved from bacteria to humans. In addition, our work sheds new light on previously recognized links of persister formation to a tight balance of cellular energy metabolism.
Papers I and II lie within recent debates revolving around the relevance of TA systems as mediators of the dormancy underlying persistence and untangle knots crucial for the further conceptual and practical developments of the field. On one side, we stressed the importance of defining a common future ground for robust methodology and careful genetics, which would promote progress by making research of different groups more comparable and transparent. On the other side, our work caused a paradigm shift in the field by disproving the link between type II TA systems and persistence in E. coli K-12, which had been dominant for almost a full decade. However, our research does not dispute that some TA systems might play roles in persistence of E. coli and other bacteria as was shown, e.g., for some type I TA systems even in E. coli K-12. Furthermore, our work does not unravel the molecular basis of how exactly persister formation depends on PasT, (p)ppGpp, and Lon, though it seems unlikely that these factors act together in some kind of a defined, linear genetic pathway similar to the previously dominant model. Finally, understanding the molecular function of PasT (and, by homology, human/yeast Coq10) would be of great importance for our knowledge of ubiquinone-dependent respiration and have great implications for biomedical applications in humans.
Another aspect of this Ph.D. project addressed in Papers III-IV and Manuscript V is the comprehensive analysis of the diversity and evolutionary dynamics of type II (and IV) TA modules in clinically relevant contexts. In short, we developed and/or applied bioinformatics tools to search for toxin-antitoxin combinations and their profiles of distribution in clinical isolates of enteropathogenic E. coli (EPEC), Klebsiella pneumoniae as well as within the entire European Nucleotide Archive (ENA) database.
In Paper III, we introduced SLING (Search for LINked Genes) as a new tool for the flexible batch annotation of gene pairs, e.g., operons, encoding functionally dependent proteins. Briefly, SLING uses the Hidden Markov model (HMM) profiles of a protein of interest to retrieve corresponding coding sequences (CDSs) within a genome collection. It then scans the neighboring regions of the selected CDSs for the presence of paired CDSs according to customized genetic structural requirements. We exploited SLING to predict TA operons and to study their diversity in terms of pairing and distribution in clinical isolates of EPEC and K. pneumoniae – two pathogens with emergent multidrug resistant and hypervirulent variants – in Paper III and IV, respectively. Overall, our results unraveled highly dynamic and plastic scenarios where toxins vary in their distribution patterns (e.g., species-specific, sporadic), in the range of paired antitoxins for each toxin, and in the operon structures of toxin-antitoxin genes. In Paper IV specifically, we experimentally confirmed the toxicity of ca. two thirds of newly predicted toxins and the neutralization activity of ca. two thirds of predicted antitoxins, some of which were novel. Finally, we revealed different degrees of association of the predicted K. pneumoniae toxin groups with resistance genes, plasmid replicons and virulence factors.
In Manuscript V, we aimed at providing novel insight into the distribution of TA systems with Gp49 domain-containing toxins and their association with resistance genes and plasmid replicons. Gp49 domain-containing toxins belong to the HigB toxin family within the RelE superfamily, i.e., are ribosome-dependent mRNA endonucleases. However, they are poorly studied and their biological relevance as well as the functional differences with the HigB toxin family are unclear. We investigated the distribution of representative Gp49 domain-containing TA systems across thousands of genomes and went beyond the species level by exploiting the recently developed tool BIGSI (Bitsliced Genomic Signature Index), which indexes the entire ENA sequence content. We shed more light on the evolutionary dynamics of Gp49 domain-containing toxins by revealing distinctive host range properties of the 24H-67P subfamily, which is uniquely broadly distributed and not restricted to a single genus. Deeper analyses revealed that the remarkable mobility of this Gp49 subfamily was likely due to residence on antibiotic resistance plasmids which could indicate a relevance in clinical settings.
Papers III-IV and Manuscript V turn our attention to the need for a more comprehensive and systematic way of studying TA systems. TA systems are known to be part of the mobilome, and therefore the investigation of their distribution, diversity and evolutionary histories might give important functional insight. To the best of our knowledge, we are the first to exploit BIGSI for this purpose and we anticipate that this kind of far-reaching approach can be extended to other TA systems.