Plasmid maintenance in many bacteria is attributed to the presence of toxin–antitoxin loci on the plasmids. These loci consist of two genes: one encodes a stable toxic protein, and the second an unstable antitoxin. If the plasmid is lost from the cell upon division, the unstable antitoxin is degraded, and the stable toxin is able to kill the cell. This phenomena, referred to as ‘post-segregational killing’ or ‘plasmid addiction’ has been described for plasmids in both Gram negative and Gram positive bacteria. The toxin–antitoxin loci are categorized into two broad classes based on the type of antitoxin: the antitoxin of type I systems is a small RNA (sRNA) which base pairs with the toxin mRNA to prevent protein synthesis, whereas the antitoxin of the type II systems is a protein that binds to and inhibits the toxin protein. Generally, type I toxins are small (under 60 amino acids in length), highly hydrophobic proteins, while type II toxins are slightly larger (~100 amino acids) and less hydrophobic. The best-studied type I toxin–antitoxin systems include the hok
locus of plasmid R1, and the par
locus of plasmid pAD1 of Enterococcus faecalis
Although the toxin–antitoxin loci were initially described on plasmids, recent studies have shown that many of these gene pairs are also present on bacterial chromosomes. The type II toxin–antitoxin systems, in which the antitoxin is a protein, have been documented in diverse bacteria with many genomes carrying dozens of distinct toxin–antitoxin pairs (3
). The type II toxins have been shown to degrade RNA or inhibit cellular enzymes such as DNA gyrase (4
). The physiological role(s) of the type II systems remains a subject of debate; proposed functions include stress survival, protection of the bacteria against foreign DNA, and stabilization of chromosomal regions (6
Several studies have shown that type I toxin–antitoxin systems, in which the antitoxin is an sRNA, are also present on some bacterial chromosomes. The hok-sok
locus from plasmid R1 is encoded in the genomes of several enteric bacteria (8
). In some strains, the sequences of these loci have degenerated and appear to be non-functional whereas in other cases, the systems are intact. Similarly, the par
locus from plasmid pAD1 is present on the chromosomes of E. faecalis
, Lactobacillus casei
and a Staphylococcus saprophyticus
). Additional type I toxin–antitoxin loci were found serendipitously on bacterial chromosomes (1
). These include the ldr-rdl
loci of Escherichia coli
and the txpA-ratA
locus of Bacillus subtilis
. Interestingly, for these loci, there was no reported homology to known plasmid sequences. However, as for the plasmid-encoded systems, overproduction of the corresponding protein leads to cell death, and this toxicity is repressed by an antisense sRNA regulator. The exact biochemical activities of the small, hydrophobic toxin proteins are not known, although similarity to phage holin proteins has been noted, and overexpression of the proteins is associated with membrane depolarization and increased membrane permeability (1
). As for the chromosomally-encoded type II toxin–antitoxin loci, the physiological function(s) of the chromosomally-encoded type I toxin–antitoxin systems remains unclear.
As mentioned above, type II toxin–antitoxin loci are broadly distributed among diverse bacteria. We hypothesized that type I systems are also widespread. To test this, we sought to identify homologs of the known type I toxins. Our computational approach identified many more putative toxins than have been previously reported. We experimentally validated a homolog of the par locus encoded in the chromosome of Streptococcus pneumoniae, the first report of a type I toxin–antitoxin system in this pathogen.
In addition to documenting the distribution of known type I systems in bacteria, we sought to identify new type I loci. Given the hydrophobicity and short length of type I toxins, and the difficulties in predicting sRNAs computationally, we developed search parameters based upon the characteristics of the known type I toxin–antitoxin systems. For example, given that the ibs-sib and ldr-rdl loci of E. coli are duplicated in the same intergenic region, we hypothesized that a short open reading frame (ORF) encoding a protein with a putative transmembrane domain and repeated in tandem could be a component of a type I toxin–antitoxin system. We also searched for amino acid sequences containing specific features derived from the analysis of known toxins, such as polar C-terminal residues. Finally, because the known antitoxin sRNAs form complex secondary structures, we developed a computational approach based upon the RNA folding energy profile of a putative type I locus to identify the location of the antisense sRNAs. Through these multiple approaches, we identified three new type I toxin–antitoxin loci which were experimentally validated. Our searches greatly expand the number of type I toxin–antitoxin systems known to be encoded in bacterial genomes.