Fusion Platform Recovers Previously Unproducible Bacterial Toxins

Bacterial polymorphic toxins are compact, modular proteins that can attack essential processes in competing microbes, making them potentially valuable leads for new antibacterial and antifungal medicines. Yet their promise has been constrained by a basic biotechnology problem: the toxins are so potent that the bacteria used to manufacture them can be damaged before enough protein accumulates for study. Researchers have now developed a modular production system that temporarily neutralizes each toxin by linking it to its natural immunity partner, then releases the active protein for purification. The approach produced several toxins that conventional recombinant-expression methods could not yield and revealed that one previously uncharacterized family member acts as a nuclease.

The work addresses an urgent need for antimicrobial agents with new mechanisms of action. Drug-resistant bacterial pathogens were linked to an estimated 4.95 million deaths worldwide in 2019, including 1.27 million directly attributable to antimicrobial resistance, while invasive fungal diseases cause millions of infections and deaths each year. The shrinking antimicrobial pipeline and the limited number of drug classes available against fungi have increased interest in naturally occurring microbial weapons. Polymorphic toxin systems are particularly attractive because their C-terminal toxin domains can interfere with nucleic-acid metabolism, protein synthesis or cell-wall construction, providing mechanisms that differ from many conventional antibiotics.

Producing these toxins in Escherichia coli is unexpectedly difficult. Even minute amounts of active protein can poison the host or inhibit the machinery required for further protein synthesis. In the study, the researchers tested tightly regulated arabinose- and IPTG-inducible systems in several commonly used E. coli expression strains, co-expression of cognate immunity proteins from the same or separate plasmids, and bacterial cell-free synthesis. None produced detectable amounts of the purified wild-type toxins PT7, PT8 or PT9. PT1 was more difficult still: its expression construct could not even be propagated reliably, and colonies recovered from transformation attempts contained frameshift mutations in the toxin sequence.

The failure of separate co-expressed toxin and immunity proteins suggested that the molecules became dangerous before they could bind each other. Multicopy plasmids can generate toxins faster than immunity proteins can neutralize them, and a catalytic toxin may damage its host after only a small number of molecules becomes active. Cell-free production also failed, indicating that host-cell death alone does not explain the barrier. PT7 and PT8 were later shown to possess both DNA- and RNA-degrading activities, raising the possibility that they directly attack messenger RNA, ribosomal RNA or other components needed to sustain the cell-free translation reaction.

The successful strategy converts the natural toxin–immunity interaction into a temporary intramolecular safeguard. Each toxin is fused to its cognate polymorphic immunity protein through a flexible linker containing a cleavage site for tobacco etch virus protease. Because the two domains are translated as one polypeptide, the immunity protein can engage its toxin immediately, minimizing the interval during which an unbound, active toxin could damage the host. A stretch of glycine, glycine, glycine and serine repeats was used to connect the domains. Since experimental structures of the selected toxin–immunity complexes were unavailable, researchers used deep-learning structure prediction to model their relative orientation and chose the shortest linker that preserved a plausible arrangement.

This design restored bacterial growth to levels comparable to those seen with catalytically inactive toxin mutants and permitted productive expression of fusion proteins containing PT7, PT8 and PT9. Subsequent purification exploited distinct affinity tags on the two domains. The intact fusion was captured through a Strep tag, cleaved with tobacco etch virus protease, denatured in urea to separate the toxin from its immunity partner, and refolded while immobilized on a nickel-affinity column through its histidine tag. Size-exclusion chromatography then yielded purified, predominantly monomeric proteins. An unrelated green fluorescent protein fused to the toxins did not rescue transformation efficiency, supporting toxin-specific neutralization rather than a general effect of adding a soluble fusion partner.

The researchers subsequently shortened the original four-stage, three-day purification procedure by arranging for tobacco etch virus protease to cleave the fusion inside the bacterial cell. An engineered host strain expressed the protease under arabinose control, while the toxin fusion was induced from a T7 promoter. Once released, the toxin and immunity protein remained associated, preserving neutralization during harvest and lysis. Direct application of the soluble lysate to a nickel-affinity column, followed by on-column refolding and size-exclusion chromatography, reduced the workflow to two principal steps. For PT9, this optimized method increased recovery from approximately 1–2 milligrams per litre of culture to about 4 milligrams per litre, a two- to fourfold improvement.

Computational redesign extended the method beyond toxins whose natural immunity partners worked effectively. PT37, an uncharacterized toxin from E. coli, could not be detected when expressed alone, and a fusion containing its predicted natural immunity protein restored growth but still failed to accumulate as recombinant protein. Using the ProteinMPNN sequence-design system, the researchers generated a construct in which both the immunity protein and linker were redesigned while the toxin sequence remained fixed. The redesigned fusion produced substantially more transformants, improved the rescue of the toxic phenotype and supported detectable expression, allowing PT37 to be purified. Experiments did not isolate whether greater binding affinity, improved protein stability or other design effects were responsible, so the mechanism remains open.

Purified PT7 and PT8 retained their expected enzymatic activities after denaturation and refolding. Both degraded plasmid DNA and mammalian ribosomal RNA in vitro, identifying PT8 as a previously unrecognized nuclease and confirming that the platform can recover biologically active material. PT9 showed no detectable activity against either substrate, so its folding was assessed independently by reconstituting it with its cognate immunity protein and determining the complex structure by X-ray crystallography. The structure was solved at 1.5-ångström resolution, and 61 C-alpha atoms of the experimental PT9 model closely matched the computational prediction, with a root-mean-square deviation of 0.67 ångström.

The results establish transient intramolecular neutralization as a potentially general strategy for making intrinsically toxic proteins that interfere with conventional recombinant production. It could support structural biology, biochemical screening and the development of antimicrobial proteins whose mechanisms remain poorly understood, while also offering a route to engineered neutralizing partners for toxins without suitable natural immunity proteins. Important limitations remain. PT1 could not be rescued because damage occurred during cloning, and the study did not test whether purified toxins can enter bacterial cells and kill them when applied outside the cell. Those delivery questions will be crucial if the proteins are to progress from research reagents toward antimicrobial therapeutics.

Subject of Research: A modular platform for producing highly toxic bacterial polymorphic toxins through temporary toxin–immunity neutralization.

Article Title: Modular Platform for Recombinant Production of Highly Toxic Bacterial Proteins

Article References: Fraenkel, R., Cahana, I., Sivan, T., Fisher, T., Nadav, H., Bruchim, S., Weisz, J., Deouell, N., Cheskis, S., Shalom, M., Imbert, L., Oppenheimer‐Shaanan, Y., Levy, A., & Tzarum, N. (2026). Modular Platform for Recombinant Production of Highly Toxic Bacterial Proteins. Microbial Biotechnology, 19(10), Article e70453. https://doi.org/10.1111/1751-7915.70453

Image Credits: AI Generated

DOI: 10.1111/1751-7915.70453

Keywords: bacterial toxins, recombinant proteins, toxin immunity, protein engineering, antimicrobial resistance, nuclease activity, protein structure prediction, microbial biotechnology, Modular, Platform, Recombinant, Production