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

The new cell line, named Calendula Root 1, or CR1, was born from a deceptively simple procedure. Seeds of Calendula officinalis were surface-sterilized and germinated on agar plates, and the emerging rootlets were sliced into small fragments and placed on callus induction medium containing the auxin 2,4-dichlorophenoxyacetic acid. Within days, a soft white callus appeared on the root pieces, and just thirty-five days after callus formation was first observed, the researchers transferred the tissue into liquid medium to start a suspension culture. The resulting cells, mostly round and measuring between fifty and three hundred micrometers, grew quickly and homogeneously, reaching packed cell volumes of fifty to sixty percent within a week when passaged weekly, figures very similar to those of BY2 cultures grown under comparable conditions.

One of the most consequential findings of the study came from an analysis of what the cells leave behind in their growth medium. When recombinant proteins are secreted into the culture supernatant, purification becomes dramatically simpler, but secreted proteins are vulnerable to the proteases that plant cells release into the medium. In BY2 cultures, these extracellular proteases have long been a thorn in the side of protein producers. Using gelatin zymography, a technique in which gelatin is embedded in a polyacrylamide gel so that proteolytic activity appears as clear bands of degraded substrate, the team found that the CR1 supernatant showed only minimal degradation at a single position of roughly eighty kilodaltons. The BY2 supernatant analyzed in the same experiment, by contrast, displayed five distinct degradation bands between fifty and sixty-eight kilodaltons plus a weaker signal at one hundred kilodaltons. This stark difference positions CR1 as a potentially far friendlier host for secreted products.

To test the production capacity of the new line, the researchers engineered a transformation construct carrying two independent expression cassettes, both driven by the cauliflower mosaic virus 35S promoter. The first encoded DsRed, the tetrameric red fluorescent protein from the coral Discosoma, targeted to the endoplasmic reticulum. The second encoded brazzein, a small, intensely sweet protein of just 6.5 kilodaltons from the West African plant Pentadiplandra brazzeana, directed to the apoplast, the space between the cell membrane and the cell wall from which proteins can reach the culture medium. DsRed served as a visible marker, allowing the team to pick fluorescent calli with a simple cold light source and filters, while brazzein represented a genuine industrial target with no commercial antibody available for detection.

Transformation was carried out using Agrobacterium tumefaciens strain GV3101, with cocultivation performed in the dark on solid medium. Ten days after plating on selection medium containing kanamycin, the first seventy-two fluorescent calli were picked, followed by another eighty-four five days later. Notably, the researchers chose not to optimize the codon sequences of either gene, reasoning that the available codon usage data for Calendula officinalis was sparse and that analysis of the closely related sunflower revealed no truly rare codons. Some calli turned visibly pink to the naked eye, indicating very high DsRed accumulation, and these were used to initiate suspension cultures that grew homogeneously within two weeks, although growth rates varied considerably between transformation events.

Quantifying brazzein posed a methodological challenge that the team solved with an elegant two-step approach. First, they purified brazzein from the supernatant of transformation event 74 using cation exchange chromatography and ultrafiltration, creating an in-house reference standard. Top-down liquid chromatography with tandem mass spectrometry confirmed the integrity of this standard, detecting the sevenfold-charged ion at a mass-to-charge ratio of 925.69 and deconvoluting the signals to a molecular weight of 6473 daltons, exactly the correct average mass of brazzein. The standard was then quantified by tryptic digest and comparison with a commercial isotope-labeled peptide, yielding a concentration of 0.21 milligrams per milliliter. Armed with this reference, the team used top-down LC-MS/MS with external calibration to measure brazzein in the supernatants of seven transformation events.

The results were striking. Brazzein concentrations in the culture supernatants ranged from 4.1 to 24.5 micrograms per milliliter, with an average of 10.5 micrograms per milliliter. Expressed per gram of fresh cell weight, the best line produced 259 micrograms of brazzein after fifteen days, roughly two hundred times more than the 1.2 micrograms per gram previously reported for brazzein production in transgenic carrot cells. Crucially, the brazzein produced in CR1 was untagged and secreted into the medium, an important distinction because attachments to either terminus of the molecule can alter or even completely abolish its sweetness. While yeast fermentation still holds the record for brazzein titers, with reports exceeding 300 milligrams per liter, the authors emphasize that the purpose of their work was to showcase the platform rather than to maximize this particular product.

The DsRed figures were even more dramatic. The best transformation event, line 155, accumulated 1322 micrograms of DsRed per gram of fresh weight, equivalent to 34.9 percent of total soluble protein, a figure the authors note clearly exceeds production rates reported for other recombinant proteins in the literature. Even the average across the seven analyzed lines, 533 micrograms per gram, was more than twice the roughly 200 milligrams per kilogram of biomass achieved with cytosol-targeted DsRed in transiently transformed BY2 plant cell packs. Because stable transformants are generally assumed to produce less recombinant protein than transiently transformed cells, the true capacity of CR1 may be even greater than these numbers suggest. Only small amounts of DsRed, between zero and 19.9 percent, were found in the supernatants, indicating that the cultures consisted largely of intact cells at harvest.

The study also revealed intriguing biology that will require further investigation. Transformation events that produced the most DsRed tended to grow more slowly, with line 155 being the most prominent example, though the authors did not explore the causal relationship in this dataset. Moreover, the molar production ratio of DsRed to brazzein varied widely between lines, from 1:2 in event 74 to 1:0.24 in event 16, showing that high DsRed output is not a reliable predictor of output from the second cassette in the tandem construct. The researchers attribute this asymmetry to position effects and to the multiple or partial T-DNA insertions that commonly occur during Agrobacterium-mediated transformation. Encouragingly, thirty-three months after transformation, all analyzed calli and dozens of others were still alive and expressing DsRed, suggesting robust transgene stability.

The authors acknowledge one cosmetic quirk: like BY2 cultures, which accumulate the pigment phytomelanin, aging CR1 cultures turn brownish-black, a trait known in marigold and other members of the Asteraceae family. Timely subculturing kept the browning from interfering with the work. Taken together, the findings establish CR1 as a rapidly growing, easily transformed cell culture with remarkably low protease activity in its medium and exceptional recombinant protein productivity in both the cytosolic-to-ER and secretory routes. Beyond the raw numbers, the choice of marigold carries strategic weight: as a well-known therapeutic herb with a positive public image, Calendula may sidestep the customer acceptance concerns that can complicate the use of tobacco-derived lines for food-related products, while its rich secondary metabolism offers untapped possibilities for producing high-value plant natural products in bioreactors.

Subject of Research: Establishment of a marigold (Calendula officinalis) suspension cell culture for recombinant protein production

Article Title: Marigold suspension cell culture as production system for recombinant DsRed and brazzein

Article References: Buntru, M., Croon, A., Müller, J., Hanke, M., & Schinkel, H. (2026). Marigold suspension cell culture as production system for recombinant DsRed and brazzein. Plant Cell Reports, 45(10), Article 317. https://doi.org/10.1007/s00299-026-03982-8

Image Credits: AI Generated

DOI: 10.1007/s00299-026-03982-8

Keywords: plant cell culture, Calendula officinalis, recombinant proteins, DsRed, brazzein, BY2 cells, Agrobacterium tumefaciens, LC-MS/MS, protease activity, suspension culture, plant biotechnology, Fraunhofer IME