Toxin-antitoxin systems are genetic modules widespread among bacteria and archaea that pair a growth-inhibiting toxin with a cognate antitoxin that keeps it in check. In type III systems, the antitoxin is not a protein but an RNA molecule that directly binds and neutralizes the toxin. Under normal conditions the pair sits in a dormant embrace, but when a phage attacks, the antitoxin is overwhelmed or neutralized, freeing the toxin. The unleashed toxin acts as an endoribonuclease, cleaving both bacterial and phage RNA and thereby halting viral replication in an act of altruistic suicide that protects the wider bacterial population from infection.

Type III toxin-antitoxin systems have traditionally been divided into three families based on toxin sequence homology: toxIN, cptIN, and tenpIN. While the ribonucleoprotein complexes of the ToxIN and CptIN systems had already been structurally characterized, the tenpIN family remained the least understood member of the trio, with no structure available since its first functional characterization in the bacterium Photorhabdus luminescens more than a decade ago. Recent bioinformatic surveys have changed the picture dramatically, identifying more than 700 unique tenpN sequences across prokaryotic genomes and viruses, and revealing that these systems are far more widespread than previously appreciated, including within the ESKAPE pathogens responsible for most antibiotic-resistant infections.

The research team selected a putative tenpIN system whose toxin sequence is 100 percent identical across several enterobacterial pathogens, making it an attractive representative for structural study. The genetic architecture follows the canonical type III layout: a shared promoter drives an antitoxin cassette of 102 nucleotides, followed by a Rho-independent terminator and a toxin gene encoding a 153-amino-acid protein. To confirm the system’s function, the researchers expressed the TenpN toxin alone in E. coli cells and observed significant growth inhibition, a bacteriostatic effect that was fully rescued when the tenpI antitoxin sequence was co-expressed. These assays established that the conserved pathogenic sequence encodes a genuine, working type III toxin-antitoxin module.

A striking feature of this system emerged from close inspection of the antitoxin sequence. Standard tandem-repeat-finding software could not detect any repeats at all, because unlike the near-identical repeats of toxIN and cptIN systems, the tenpI antitoxin contains two non-identical repeats, termed Rp1 and Rp2, which differ at multiple nucleotide positions, and Rp2 is three nucleotides shorter than Rp1. Despite this sequence divergence, both repeats were predicted to fold into H-type pseudoknots, the hallmark tertiary structure of type III antitoxin RNAs. When each repeat was cloned separately and tested, both Rp1 and Rp2 independently rescued bacteria from TenpN-mediated growth inhibition, and colony-forming unit assays showed that the full-length antitoxin, Rp1, and Rp2 all counteracted toxicity to a similar extent.

To visualize the complex, the team co-expressed and co-purified the toxin and antitoxin from E. coli using affinity, ion-exchange, and size-exclusion chromatography. The purified complex eluted at an apparent molecular weight of roughly 72 kilodaltons, consistent with a heterotetramer of two toxin proteins and two antitoxin RNA molecules, while free TenpN eluted as an 18-kilodalton monomer. Sequencing of the RNA recovered from the purified complex confirmed that both Rp1 and Rp2 were incorporated. Circular dichroism thermal melting experiments showed that both repeats unfold with a melting temperature of about 56 degrees Celsius, demonstrating that the non-identical repeats are equally thermally stable, and in vitro assembly experiments confirmed that TenpN forms complexes with either repeat on its own.

The cryo-EM analysis itself demanded careful sample preparation. Negative-stain transmission electron microscopy identified the most homogeneous particle fractions, and cryo-EM data collected on a 200-kilovolt Talos Arctica microscope with a K2 direct electron detector yielded 2,855 movies. From more than a million automatically picked particles, iterative two-dimensional and three-dimensional classification converged on a well-resolved reconstruction built from 327,240 particles. The final map displayed clear C2 symmetry, and symmetric and asymmetric reconstructions proved nearly identical, indicating no significant structural asymmetry within the assembly. The resulting structure, determined in the complex’s native state without antibodies or tags, measures approximately 80 by 60 angstroms and shows the protein and RNA components arranged alternately in a closed cyclic ring with a hole at its center.

Model building combined computational prediction with experimental density. An AlphaFold2 model of TenpN was fitted and refined into the map, while the RNA had to be constructed de novo, fragment by fragment, using tools including simRNA, RhoFold+, Coot, Chimera, and Phenix. The Rp1 model fit the density better than Rp2, reflected in map-to-model correlation coefficients of 0.66 versus 0.52, likely because Rp2 lacks three nucleotides, so Rp1 was used to represent both RNA positions. The refined structure shows TenpN as a well-folded protein of six alpha-helices and five beta-strands, with a four-stranded antiparallel beta-sheet core and a long kinked helix that closely mirrors features previously seen in ToxN, where the kinked helix helps hold the antitoxin RNA in place. The tenpI RNA folds into a central H-type pseudoknot with two stems and two loops, its single-stranded 5-prime and 3-prime ends binding along an electropositive groove on the toxin surface in a head-to-tail arrangement that generates the cyclic tetramer.

Comparative analysis placed TenpN firmly within the type III toxin lineage. Despite sharing only about 13 percent sequence identity with ToxN and CptN, TenpN adopts a conserved overall fold, and DALI structural comparison showed it resembles ToxN, with a backbone root-mean-square deviation of 5.4 angstroms, more closely than CptN, at 10.1 angstroms. Sequence alignment highlighted conserved residues in the region forming the catalytic center, prompting the researchers to create alanine mutants of TenpN residues Pro53, Leu54, Arg55, and Ser56. Growth assays revealed that P53A and L54A mutants remained fully toxic, while R55A and S56A lost their ability to inhibit bacterial growth, identifying Arg55 and Ser56 as functionally critical residues that likely constitute the putative active site of the TenpN endoribonuclease, analogous to the validated active-site triads of ToxN and CptN.

The structural comparison also revealed a fascinating chimeric character of the TenpIN system: the toxin resembles ToxN, yet the antitoxin RNA is architecturally closer to cptI. Both tenpI and cptI feature a single-nucleotide loop that twists to coaxially stack the second stem beneath the first, and a long A-rich second loop that runs behind the first stem like a third strand, potentially forming A-minor interactions, whereas toxI RNA has shorter stems, more balanced loops, and a distinctive UUU triplet at the helical junction. These RNA size differences appear to dictate assembly stoichiometry: the larger cptI and tenpI RNAs form heterotetramers with their toxins, while the compact toxI assembles a heterohexamer. Because the single-stranded ends of tenpI and cptI converge on the same side of the active site, the tetrameric arrangement follows naturally. Beyond resolving an evolutionary puzzle, the structure opens a practical avenue: small molecules or peptides designed to disrupt the TenpIN complex could free the toxin inside pathogenic bacteria, offering a fundamentally new antibacterial strategy at a time when resistance to conventional antibiotics continues to spread.

Subject of Research: Cryo-EM structure of the TenpIN type III toxin-antitoxin ribonucleoprotein complex in pathogenic bacteria

Article Title: Cryo-EM reveals the quaternary architecture of TenpIN type III toxin-antitoxin RNP complex common to pathogenic bacteria

Article References: Nadig, K., Padmanaban, S., Noor, S., Sandhya, S., Manikandan, P., Dutta, S., & Singh, M. (2026). Cryo-EM reveals the quaternary architecture of TenpIN type III toxin-antitoxin RNP complex common to pathogenic bacteria. PLOS Pathogens, 22(10), e1014690. https://doi.org/10.1371/journal.ppat.1014690

Image Credits: AI Generated

DOI: 10.1371/journal.ppat.1014690

Keywords: cryo-EM, toxin-antitoxin systems, TenpIN, bacteriophage defense, RNA pseudoknot, endoribonuclease, pathogenic bacteria, ESKAPE pathogens, structural biology, antimicrobial targets, ribonucleoprotein complex, bacterial immunity

The choice of target reflects a deliberate strategy to move beyond the two viral proteins that dominate the current antiviral arsenal. Approved and emergency-authorized drugs against COVID-19 largely fall into two mechanistic classes: inhibitors of the main protease Mpro, such as nirmatrelvir, and agents that attack the RNA-dependent RNA polymerase, including remdesivir and molnupiravir. While these drugs remain valuable, the continued evolution of SARS-CoV-2 has produced variants that partially escape vaccine-induced and infection-induced immunity and reduce the effectiveness of some monoclonal antibodies. Diversifying the antiviral toolkit by targeting additional conserved viral enzymes has therefore become a pressing priority for the field.

Nsp15, the uridine-specific endoribonuclease also known as NendoU, has attracted growing attention as exactly such a complementary target. The enzyme is conserved across nidoviruses and cleaves both single- and double-stranded RNA at uridine residues, generating 2′,3′-cyclic phosphodiester and 5′-hydroxyl termini. Beyond its role in viral RNA processing, Nsp15 helps the virus evade innate immune sensors by degrading viral double-stranded RNA intermediates that would otherwise trigger antiviral signaling. Genetic and biochemical studies indicate that this activity contributes to optimal viral replication and pathogenesis, which makes disabling it an attractive proposition for small-molecule drug design.

The design of compound 5 drew on structural work showing that tipiracil, an approved thymidine phosphorylase inhibitor, binds to the uridine recognition site of SARS-CoV-2 Nsp15 and suppresses its endoribonuclease activity. Although tipiracil alone does not fully block viral replication at clinically relevant concentrations, it provides a validated structural starting point. The Brazilian team, building on prior QSAR-guided virtual screening and on literature describing phenylthio-uracils as potent non-nucleoside inhibitors of HIV-1 reverse transcriptase, added a 2-aminophenylthio group at the C-6 position of the uracil core and a bromine atom at C-5. The rationale was that the bulky aromatic thioether would extend into a hydrophobic subpocket adjacent to the uracil-binding site, adding van der Waals and possible π-stacking contacts beyond those available to tipiracil’s pyrrolidinylmethyl substituent, while the electron-withdrawing bromine would tune the electronic character and hydrogen-bonding capacity of the pyrimidinedione core.

Synthesis proceeded in three steps from commercially available 2,4,6-trichloropyrimidine. Controlled hydrolysis under basic conditions selectively yielded 6-chlorouracil in 76 percent yield. Bromination at C-5 with elemental bromine then furnished 5-bromo-6-chlorouracil, albeit in a modest 31 percent yield. The final step, a nucleophilic aromatic substitution of 2-aminothiophenol onto the C-6 chloride in the presence of triethylamine under nitrogen, delivered compound 5 in 59 percent yield, for an overall yield of roughly 14 percent. Nuclear magnetic resonance spectroscopy and high-resolution mass spectrometry confirmed the structure, with the observed molecular ion at m/z 313.9590 matching the calculated value of 313.9593 for the protonated molecule. The authors note that the route uses inexpensive reagents and is amenable to scale-up and diversification, for example by varying the arylthio group at C-6.

Computational docking against the crystal structure of Nsp15 deposited under Protein Data Bank entry 9HH5, resolved at 2.08 angstroms, supported the design hypothesis. Compound 5 achieved a total interaction energy of −65.4 kilocalories per mole, compared with −56.4 kilocalories per mole for tipiracil, with the advantage coming primarily from stronger van der Waals contacts (−54.4 versus −49.1 kilocalories per mole) and hydrogen bonding (−10.9 versus −7.3 kilocalories per mole). At the residue level, the compound engaged His235, Gly248, Trp333, Glu340, Thr341 and Tyr343, residues previously implicated in catalysis and uracil recognition. Trp333 in particular contributed a substantial van der Waals interaction of −12.5 kilocalories per mole plus a hydrogen bond of −3.5 kilocalories per mole with the uracil moiety. Visualization of the best-ranked poses confirmed that both ligands occupy the uracil-binding pocket in similar orientations, with the aminophenylthio substituent of compound 5 reaching into the adjacent hydrophobic subpocket. Docking runs were repeated in triplicate and poses were cross-checked by an independent evaluator.

In silico toxicology profiling using two independent platforms, Deep-PK and ProTox 3.0, suggested a generally favorable preliminary safety profile. Both compounds were predicted to be inactive or low-risk for Ames mutagenicity and carcinogenicity, and neither was flagged for significant eye irritation or blockade of the hERG potassium channel, a common cause of cardiac toxicity. Notably, tipiracil was predicted to carry a high probability of respiratory toxicity, whereas compound 5 was assigned a low probability for that endpoint. The authors caution that such machine-learning predictions do not replace experimental safety testing, but they argue that this kind of early computational triage can help minimize late-stage attrition in antiviral drug development, where rapid progression from hit to lead is critical.

The decisive test came in cell culture. Working with a Brazilian clinical isolate of SARS-CoV-2 in Vero E6 cells, the team used MTT-based viability assays to measure both cytotoxicity and antiviral protection. In uninfected cells treated with compound 5 at 20 milligrams per milliliter, viability was 72.6 percent, marginally below the commonly accepted 80 percent non-cytotoxic threshold and indicating low to moderate cytotoxicity at this high concentration. In infected cells, however, the picture was striking: treatment raised viability to 68.1 percent, roughly three times the 22 percent seen in infected vehicle controls and far above the 17 percent in PBS controls. Chloroquine diphosphate at 20 micromolar, the positive control, achieved 82 percent viability under the same conditions. The protective effect of compound 5 therefore approached, though did not match, that of the reference antiviral, consistent with an antiviral mechanism potentially linked to interference with Nsp15 function, although direct enzymatic inhibition was not measured in this study.

The authors are careful to delineate the limits of the work and the steps needed before the scaffold can advance. Antiviral activity was assessed at a single concentration in Vero E6 cells, a line widely used for coronavirus propagation but lacking certain components of the human innate immune response, so validation in human-derived systems such as Calu-3 cells, A549-ACE2 cells or primary airway epithelial cells will be essential. Quantitative enzymatic IC50 values against Nsp15, dose-response curves and selectivity indices remain to be determined, as do preliminary ADME studies and in vivo safety and efficacy testing. The field context, however, is encouraging: recent first-in-class thiazolidinedione and rhodanine Nsp15 inhibitors such as KCO237 have achieved sub-micromolar enzymatic potency, confirming that the uridine active site is druggable and that structurally diverse chemotypes are needed. With its accessible synthesis, favorable predicted binding to key catalytic residues, low-risk toxicological profile and demonstrable cell protection, compound 5 offers other laboratories a convenient, reproducible template for structure-activity relationship exploration, spanning modifications of the aryl ring, the amino group and the C-5 halogen, as new variants of the virus continue to emerge.

Subject of Research: Design and antiviral evaluation of a thio-aminophenyl-bromopyrimidinone inhibitor of the SARS-CoV-2 endoribonuclease Nsp15

Article Title: Design, synthesis and antiviral evaluation of a new thio-aminophenyl-bromopyrimidinone as an inhibitor of the SARS-CoV-2 viral enzyme Nsp15

Article References: Collin, H. P., Delevati, M. A., dos Santos, P. R., da Silva da Fonseca, G., Tasso, L., Reisdorfer, F., Gonzalez, L. C., Machado, P., Gonçalves, R. S. B., & Moura, S. (2026). Design, synthesis and antiviral evaluation of a new thio-aminophenyl-bromopyrimidinone as an inhibitor of the SARS-CoV-2 viral enzyme Nsp15. Results in Chemistry, 31, Article 103895. https://doi.org/10.1016/j.rechem.2026.103895

Image Credits: AI Generated

DOI: 10.1016/j.rechem.2026.103895

Keywords: SARS-CoV-2, Nsp15, endoribonuclease, uracil derivatives, tipiracil, antiviral drug discovery, molecular docking, Vero E6 cells, bromopyrimidinone, COVID-19, drug design, in silico toxicology