Polymorphic toxin systemsComprehensive characterization of trafficking modes, processing, mechanisms of action, immunity and ecology using comparative genomics

If bacteria live in dense colonies competing for space and nutrients, then natural selection should favor weapons. If those weapons can be swapped, reshuffled, and upgraded like interchangeable parts, then the arms race never ends. That is exactly what bacteria have built — a modular arsenal of over 150 distinct toxin domains, delivered by at least eight different secretory systems. Zhang and colleagues mapped the entire system from first principles using comparative genomics. The ecological logic comes first. Bacteria wage war not just against other species but also against their own kind. In biofilms, in the densely packed microfilms of the human oral cavity, cells compete for identical resources with genetically near-identical neighbors. That pressure strongly selects for kin versus non-kin discrimination, cooperation, cheating, and direct antagonism. Polymorphic toxin systems are the molecular machinery that resolves those conflicts. What makes these toxins "polymorphic" is their architecture. Each complete toxin follows a tripartite design: N-terminal domains handle trafficking, central filamentous repeats present the toxin outward, and an extreme C-terminal domain carries the actual killing activity. The conserved scaffold remains constant. The C-terminal payload is endlessly swappable, generated by recombination with standalone toxin-coding cassettes. Genomic loci typically show a consistent toxin-immunity gene order — the toxin gene is upstream, and its cognate immunity gene is immediately downstream. They often retain strings of additional immunity genes at the three-prime end, interpreted as remnants of successive cassette-replacement events. One recombination at the tip swaps the weapon and simultaneously imports a matching shield. Now, getting a toxin out of a cell and into a neighboring cell is its own engineering problem, and Zhang and colleagues found that bacteria have solved it in at least eight distinct ways. The general Type II secretion pathway is the most prevalent export route overall. Beyond that, there is Type V two-partner secretion, where toxins are extruded through outer-membrane porin partners; Type VI secretion, which uses a phage-tail-derived injection syringe to punch directly into adjacent cells; Type VII, also called ESX, driven by a FtsK-type ATPase and carrying small WXG-motif substrates; the Photorhabdus virulence cassette pathway, or PVC, which shares the phage-tail injection mechanism with Type VI and couples it to AAA plus ATPases; a PrsW-dependent route in which an eight-helix transmembrane peptidase spans the membrane and processes toxins simultaneously; a MuF phage-capsid-like system that appears to package toxin polypeptides into capsid-like structures for delivery; and a TcdB/TcaC-like export pathway. Type VII, Type VI, and Type V account for approximately twelve, eleven, and ten percent of complete toxins, respectively. Most of these pathways require one additional step that Zhang and colleagues emphasize as central to the entire system: autoproteolytic maturation. A toxin polypeptide is often cleaved from its carrier before it becomes active. The paper identifies six classes of processing peptidases associated with these systems — HINT, ZU5, PrsW, caspase-like, papain-like, and a novel metallopeptidase tied to the PVC system. Their positions are telling. ZU5 domains typically sit at the N-terminal base of filamentous stalks, likely cleaving during extrusion. HINT and caspase-like peptidases sit immediately upstream of the C-terminal toxin domain, releasing the active payload upon delivery. The PVC metallopeptidase appears functionally equivalent to HINT but mutually exclusive with it — different injection contexts demand different solutions. With the delivery machinery established, the catalog of what gets delivered is staggering. There are over 150 distinct toxin domains, spanning 23 distinct clades of peptidases alone. They are organized by what they destroy: nucleases are the single largest class, targeting both DNA and RNA from multiple structural folds. The BECR fold — named for Barnase, EndoU, Colicin E5 or D, and RelE — supplies one of the most extensive radiations of RNA-cleaving toxins across the entire study, with seven previously unrecognized clades defined. Many act by metal-dependent or metal-independent phosphoester hydrolysis; some produce two-prime, three-prime cyclic phosphate RNA termini that block further processing by the target cell. Deaminases appear in multiple clades, including two highly divergent new ones — one in giant Wolbachia proteins apparently directed at host cells, another in classical polymorphic toxins from proteobacteria and actinobacteria used in intraspecific conflict. ADP-ribosyltransferases come in two canonical clades. The R-S-E clade includes toxins related to cholera and pertussis toxins and appears in both polymorphic toxins and in host-directed effectors from Pseudomonas syringae and Legionella. The H-Y-E clade includes PARP-related toxin families. Alongside those, ADP-ribosyl cyclase domains generate cyclic ADP-ribose from NAD and could perturb calcium homeostasis in target cells. RelA/SpoT-like nucleotidyltransferases produce the bacterial stress signal ppGpp; when expressed without regulation in a target cell, they shut down growth and protein synthesis. Glycosyltransferases related to Clostridioides difficile toxin B modify small GTPases, disrupting host signaling. Three distinct lipid-modifying toxin classes damage membranes directly — a glycerophosphoryldiester phosphodiesterase, a CDP-alcohol phosphatidyltransferase, and a phospholipase A2. And then there are pore-forming toxins: small two-helix hydrophobic domains predicted to span the membrane and kill by disrupting its integrity entirely. Two novel findings stand out for their conceptual reach. The BECR-fold RNases, already the most radiated toxin class, turn out to use at least two distinct metal-binding strategies across their clades. That level of mechanistic diversification is unusual even within a single structural fold. The repurposing of JAB-domain proteins and ParB-related domains as toxins is unexpected: JAB domains normally function as deubiquitinating peptidases, but in this context, Zhang and colleagues predict they act as nucleases. The paper also notes repeatedly that many of these toxin chemistries are shared with host-directed effectors from Legionella, Pseudomonas, and Xanthomonas — creating a direct mechanistic bridge between intraspecific bacterial warfare and the virulence factors implicated in human disease. Every weapon, of course, demands a shield. Zhang and colleagues recovered over 90 families of immunity proteins, and those families vary enormously in breadth. Some neutralize a single toxin type, while others cover up to 27 distinct toxin domains within a single family. Two superfamilies dominate the defensive side. The SUKH superfamily provides immunity across 18 distinct nuclease families spanning eight structural folds, three families of deaminases, AMP and UMP transferring enzymes, an aldo-keto reductase, and two mechanistically distinct peptidases. The SuFu superfamily — expanded here to include a second highly divergent clade — covers six families of HNH-fold nucleases, a ParB domain, peptidases from two unrelated folds, and a glycerophosphodiester phosphodiesterase. Both superfamilies achieve this breadth by deploying variable binding interfaces instead of fixed active-site contacts. Organizationally, bacteria concentrate their defenses into what the authors call polyimmunity loci — tandem arrays of immunity genes, either homogeneous repeats of one family or heterogeneous mixtures of many. One example from Bacteroides sp. D22 encodes 19 different immunity proteins drawn from 13 distinct superfamilies, six of them distinct SUKH versions. Polyimmunity proteins compact multiple immunity domains into a single polypeptide; some contain up to ten distinct immunity domains in one chain. About thirteen percent of organisms with polyimmunity loci carry more than one. The evolutionary logic proposed by Zhang and colleagues is direct: these arrays function as preemptive backups, allowing cells to neutralize toxins from non-kin strains and rapidly cover new weapons acquired by lateral transfer. Zooming to the phylogenetic scale, these systems span all major bacterial lineages and appear in a small number of archaea. Ecologically, the study links richer arsenals to social niches — the oral cavity and soil bacilli are named explicitly — and suggests that polyimmunity systems could enable cheating behavior in cooperative bacterial communities, a hypothesis the paper notes is testable by competition experiments. The most far-reaching claim in the paper is that many of these domains have crossed kingdom boundaries. SUKH, SuFu, ZU5, and HINT derivatives have been transferred into eukaryotes. The cargo-binding domain unique to animal myosin VI is evolutionarily related to a bacterial immunity protein, with statistical support reaching a p-value below one in ten million in iteration four of JACKHMMER searches. Zhang and colleagues name endosymbiotic alphaproteobacteria — Wolbachia, Rickettsia, Odyssella — and chlamydiae as plausible donors during early eukaryotic evolution and propose that these acquisitions may have contributed to the origin of deubiquitinating peptidases, ZU5-containing apoptosis components, and poly-ADP-ribosylation machinery. The speculation is brief in the paper, and properly so, but it is grounded in specific sequence evidence. What Zhang and colleagues deliver, in the end, is the first comprehensive map of a bacterial arms race: over 250 toxin and immunity domains catalogued, testable predictions about active sites and catalytic mechanisms attached to each, and a resource that makes targeted biochemical follow-up possible. The survey is not microbiological completionism for its own sake. It is a record of how conflict sculpts genomes over evolutionary time — and, if the eukaryotic transfer evidence holds, a partial account of how some of our own cellular machinery was forged in that same conflict. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.