A type VI secretion ADP-ribosyltransferase Rhs toxin selectively modifies charged tRNAThr(GGU) anticodon loops to block ACC codon decoding Authors: Maïalène Chabalier, Jonas B Desjardins, Julius Martinkus, Eric Cascales Published: 2026 ABSTRACT AbstractThe type VI secretion system (T6SS) is a sophisticated machine that uses a contractile mechanism to deliver antibacterial toxins into neighboring cells, hence participating to interbacterial competition and pathogenesis. Here, we characterize the T6SS-associated rearrangement hotspot RhsF polymorphic toxin from Chromobacterium violaceum. Toxicity assays showed that the RhsF C-terminal extension, TreThr, is a cytoplasmic-acting antibacterial toxin, which is neutralized by the downstream immunity protein. Structural modeling and mutagenesis demonstrated that TreThr belongs to the ADP-ribosyltransferase (ART) family, adopting a canonical ART fold. Using coupled in vitro transcription-translation assays, we found that TreThr inhibits protein synthesis in a Nicotinamide Adenine Dinucleotide (NAD)-dependent manner, through isoacceptor-specific ADP-ribosylation of transfer RNAs (tRNAs). Biotin–NAD labeling experiments and mapping further revealed that TreThr modifies the GGU anticodon of charged thrT and thrV tRNA isoacceptors and prevents decoding of threonine ACC codons. Introduction Bacteria live in densely populated environments in which access to space and nutrients is limited [1–4]. For fitness and niche occupation, they deploy specialized and efficient secretion machines to antagonize neighboring cells [5]. Among these machineries, the type VI secretion system (T6SS) is a contractile injection system (CIS) widely used by Gram-negative bacteria to deliver antibacterial and anti-eukaryotic effectors into target cells, hence participating to interbacterial competition and pathogenesis [6–9]. As other CISs, the T6SS fires an inner tube, made of Hcp hexamers, topped by a spike made of a VgrG-PAAR complex, together with associated toxic proteins [7–12]. These effector proteins can be either independent polypeptide that binds to the tube or spike complex directly or through the help of adaptors, or domains fused to tube and spike components [13–16]. The latter class corresponds to polymorphic toxins i.e. multidomain modular proteins that comprise an N-terminal trafficking region, which determines the route of secretion, and a variable toxic C-terminal extension [17]. A specific class of T6SS-associated polymorphic toxins, called Rearrangement hotspot (Rhs) proteins, share an additional central shell-like YD-repeat region that forms a β-barrel cocoon encapsulating the variable toxic C-terminal domain [18–22]. The cage is flanked by conserved autoproteolytic cleavage motifs that mediate internal processing required for maturation and toxin release [18, 23]. The T6SS is capable of delivering a remarkable diversity of effectors spanning a large repertoire of biochemical activities [9, 13–16]. These effectors usually target essential cellular macromolecules including membranes, nucleic acids, proteins, peptidoglycan and protein synthesis or cell division machineries, as well as critical molecules or cofactors such as ATP or NAD [9, 16]. In recent years, several T6SS toxins with ADP-ribosyltransferase (ART) activity have been characterized [24–28]. ART enzymes cleave the NAD and transfer the ADP-ribose moiety to cellular targets [29–31]. The Serratia proteomaculans Tre1 and Xenorhabdus bovienii TreX ART effectors ADP-ribosylate the FtsZ ring-forming component of the divisome, hence blocking cell division [24, 28]. The Photorhabdus laumondii Tre23 and Salmonella enterica Typhimurium TreTu toxins inhibit translation through ADP-ribosylation of 23S ribosomal RNA and of the elongation factor EF-Tu, respectively [25, 26]. In Pseudomonas aeruginosa, RhsP2 ADP-ribosylates structured noncoding RNAs, including transfer RNAs (tRNAs) and RNase P [27]. These studies established that T6SS-associated ART effectors represent a versatile class of antibacterial effectors that inhibit bacterial growth by targeting various essential processes, highlighting the remarkable plasticity of ART-mediated intoxication mechanisms. Based on the composition of their catalytic triads, ARTs are evolutionarily clustered into three major types: H-H-h (where h denotes any hydrophobic residue), H-Y-[QED], and R-[ST]-E [29], although atypical variants such as R-H-h ARTs have also been reported [26]. Here, we expand the repertoire of T6SS-delivered ART toxins targeting the protein synthesis machinery. We show that the Chromobacterium violaceum vgrG3 island, which encodes the VgrG3 spike protein playing a major role in T6SS function [32], encodes a putative polymorphic toxin of the Rhs class, RhsF. We show that the RhsF C-terminal domain, which is toxic when produced in the Escherichia coli cytoplasm, belongs to the ART family, whereas the small downstream protein CV_1430 acts as the cognate immunity protein that neutralizes its toxicity. Structural modeling using AlphaFold3 and mutagenesis further show that the RhsF ART domain, named TreThr, belongs to the R-S/T-E ART clade. We then demonstrate that TreThr inhibits protein synthesis in a NAD-dependent manner. We show that TreThr is a highly specific tRNA-directed toxin that selectively ADP-ribosylates the threonine tRNAThr(GGU) isoacceptors thrT and thrV at their anticodon therefore preventing decoding of ACC threonine codons. Materials and methods Bacterial strains, media, and growth conditions Escherichia coli DH5α and W3110 were used for cloning and heterologous toxicity assays, respectively. Cells were routinely grown in lysogeny broth (LB) with agitation at 37°C. When required, media were supplemented with chloramphenicol (40 μg ml−1) for maintenance of pBAD33 plasmid derivatives. Expression from pBAD33 arabinose-inducible vectors [33] was induced by addition of 0.2% of L-arabinose and repressed by addition of 1% of glucose. Plasmid construction DNA fragments corresponding to the TreThr R1403A and E1497A domains (amino-acids 1393–1513 of RhsF) and to the wild-type TreThr domain and the full-length RhsFI immunity protein, were synthesized and cloned into pBAD33 by Twist Bioscience. The pBAD33 vector encoding wild-type TreThr was not obtained by the company, likely due to its toxicity, and was engineered by Quick-change mutagenesis by re-introducing a glutamate codon at codon 1497 in the pBAD33-TreThr-E1497A vector, using PrimeSTAR Max Polymerase (Takata) and oligonucleotides GCTAAGCGGTTTGCCACTAAAGACCAAGAAGTATTAATAAGAAATGGCATTCCTAAGGGTTCGTACGAGG and CCTCGTACGAACCCTTAGGAATGCCATTTCTTATTAATACTTCTTGGTCTTTAGTGGCAAACCGCTTAGC (mutagenized codon italicized). All constructs were verified by DNA sequencing (Eurofins genomics). Heterologous toxicity and immunity rescue assays Toxicity and rescue assays were performed as previously described [25, 26, 34, 35] with minor modifications. Briefly, E. coli W3110 cells were transformed with the pBAD33 derivative plasmids encoding wild-type or R1403A or E1497A TreThr, or both TreThr and RhsFI. Transformants, selected on LB agar plates supplemented with chloramphenicol and 1% of glucose, were grown to an optical density at λ = 600 nm (A600) of ∼ 0.6 in LB supplemented with chloramphenicol and glucose, washed twice in LB, normalized to an A600 of 0.5 and serially diluted. Ten-microliter drops were spotted onto LB agar plates containing chloramphenicol and either 1% of glucose for repression or 0.2% of L-arabinose for induction. LB plates were incubated overnight at 37°C before imaging. The data shown are representative of at least three independent experiments. In vitro coupled transcription-translation (IVT) assays To examine the effect of TreThr on transcription and protein synthesis, IVT assays were performed as previously described [25, 26] using the PURExpress® in vitro protein synthesis system (New England Biolabs). Briefly, 5-μl reactions were assembled in the presence of a 100 ng of DNA template encoding Strep-tagged GFP, or wild-type and variant TreThr, obtained using oligonucleotides GCGAATTAATACGACTCACTATAGGGCTTAAGTATAAGGAGGAAAAAATATGAGTAAAGGAGAAGAACTTTTCAC and AAACCCCTCCGTTTAGAGAGGGGTTATGCTAGTTATTATTTTTCGAACTGCGGGTGGCTCCATTTGTATAGTTCATCCATGCCA, or GCGAATTAATACGACTCACTATAGGGCTTAAGTATAAGGAGGAAAAAATATGACGGGGAAACAGTTCACAGGGACG and AAACCCCTCCGTTTAGAGAGGGGTTATGCTAGTTATTATTTTTCGAACTGCGGGTGGCTCCACTCTATAACCTCGTACGAACCCTTAGGAATGCC, respectively, in the presence or absence of 0.1 mM of NAD+. Reactions were incubated at 37°C for 3 h, mixed with 2 × Laemmli loading buffer, and analysed by sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS–PAGE), transfer onto nitrocellulose membranes, and immunoblotting using the anti-Streptag antibody (Classic, Bio-Rad). To test the impact of TreThr on transcription/translation of the GFP, a second coupled IVT reaction was performed with the DNA template coding for the GFP reporter protein, in the presence or absence of 0.1 mM of NAD+ and/or 0.5 μM of IVT-produced wild-type or mutant TreThr. To test the codon specificity of TreThr-mediated translation inhibition, similar IVT reactions were performed using synthetic DNA templates. The AllCodons construct contained all sense codons represented in triplicate, followed by a C-terminal FLAG tag. The sequence of AllCodons was defined by Artificial Intelligence with ChatGPT, using the following criteria: each codon in consecutive triplicates with the exception of codons TAG, TGA, and TAA, starting with ATG, randomly organized, with a GC content comparable all along the sequence, with limited palindromes and hairpins, without long GC or AT repeats. The AllCodons-ACC construct was identical except that the ACC threonine codon was omitted. The synthetic genes, synthesized by GeneArtTM (Invitrogen), were amplified using oligonucleotides GCGAATTAATACGACTCACTATAGGGCTTAAGTATAAGG and AAACCCCTCCGTTTAGAGAGGGGTTATGC and used as DNA templates in the IVT reaction in the presence or absence of 0.1 mM of NAD+ and 0.5 μM of IVT-produced TreThr or GFP. After incubation 3 h at 37°C, translation products were analyzed by SDS–PAGE and immunoblotting using anti-FLAG antibodies (Clone M2, Sigma–Aldrich). The data shown are representative of three independent experiments. Biotin-NAD ADP-ribosylation assays To identify the targets modified by TreThr, IVT reactions using the PURExpress® system were supplemented 0.1 mM of 6-biotin-17-NAD+ (Biolog Life Science), a NAD+ derivative in which the ADP-ribose is decorated with a biotin, and the TreThr-encoding DNA template, as previously described [25, 26]. Reactions were incubated for 2 h at 37°C. For testing the role of ribosomes, IVT reactions were performed using the PURExpress® ΔRibosome kit (New England Biolabs) supplemented, or not, 0.6 μM of IVT-produced TreThr, and incubated for 20 min at 37°C. For protein analysis, reactions were mixed with 2 × Laemmli loading buffer, separated by SDS–PAGE, transferred to nitrocellulose membranes and biotin-labeled species were detected with alkaline phosphatase streptavidin conjugate (Invitrogen). For RNA analysis, nucleic acids were extracted from the reaction after addition of 15 μl of water, by mixing with 20 μl of phenol and 20 μl of chloroform. After vortexing for 30 s and centrifugation, the aqueous phase was recolted and RNA species were then separated on 1.5%-agarose Tris–acetate–EDTA gels. The gel was fixed with 1 × Saline-Sodium Citrate (SSC) buffer supplemented with 0.05 M of NaOH for 20 min, washed in 10 × SSC, and then transferred overnight by capillarity to nylon membrane (Hybond-N+, Amersham) in 10 × SSC. The membrane was then washed, air-dried, and nucleic acids were fixed by exposure to ultraviolet light for 30 s, before immunoblotting with alkaline phosphatase streptavidin conjugate. For testing the impact on free tRNAs, purified tRNAs from the PURExpress® Δ(aa, tRNA) were mixed in the presence or absence of 0.1 mM of 6-biotin-17-NAD+ and/or 0.6 μM of IVT-produced TreThr for 20 min at 37°C, prior to nucleic acid extraction and analyses. For testing the impact of tRNA aminoacylation, purified tRNAs were mixed with amino-acids (a home-made mixture of all amino-acids, using each purified amino-acid (Sigma–Aldrich), or of a mixture lacking threonine), 0.1 mM of ATP and E. coli aminoacyl–tRNA synthetase (Sigma–Aldrich), in the presence of 0.1 mM of 6-biotin-17-NAD+ and/or 0.6 μM of IVT-produced TreThr and incubated for 20 min at 37°C prior to extraction and analyses. For testing the impact on charged tRNAs, purified tRNAs were incubated with amino-acids, ATP and E. coli aminoacyl–tRNA synthetase for 20 min at 37°C. Aminoacyl–tRNAs were extracted using phenol and chloroform, and incubated with 0.1 mM of 6-biotin-17-NAD+ and 0.6 μM of IVT-produced TreThr for 20 min at 37°C prior to extraction and analyses. The data shown are representative of three independent experiments. Aminoacylation of purified tRNAs was assessed by acid urea polyacrylamide gel electrophoresis as described [36], with modifications. Purified tRNAs were incubated under the same aminoacylation conditions used for the ADP-ribosylation assays, with ATP and purified E. coli aminoacyl–tRNA synthetases in the presence of either the amino-acid mixture lacking threonine, threonine alone, or the indicated combinations. Following incubation, the reactions were mixed with 300 μl of 300 mM sodium acetate (pH 5), 10 mM EDTA, and tRNAs were extracted using phenol and chloroform, precipitated by the addition of 750 μl of ice-cold ethanol and 20 ng of glycogen (Roche), incubation overnight at −80°C and centrifugation at 13 000 revolutions per minute (rpm) for 30 min, washed with 500 μl of 70% ethanol, resuspended in 5 μl of 100 mM sodium acetate (pH 5), 8M urea 8M, 10% glycerol, 0.05% bromophenol blue, separated by 6%-acrylamide acid urea–Poly-Acrylamide Gel Electrophoresis (PAGE) [6% acrylamide (acrylamide:bis-acrylamide 19:1), 100 mM sodium acetate (pH 5), 8M urea], and visualized by ethidium bromide (EtBr) staining after gel fixation in 10% acetic acid. Enrichment, cloning, and sequencing of biotinylated RNA species The protocol for the identification of the sites of ADP-ribosylation was adapted from [25]. Three independent 20-μl IVT reactions were performed for 20 min at 37°C in the presence of 0.1 mM of 6-biotin-17-NAD+ and 0.6 μM of IVT-produced TreThr. After dilution into a final volume of 300 μl with 200 mM Tris–HCl (pH 9.5) and incubation for 1 h at 37°C to deacylate tRNAs, RNAs were extracted by phenol-chloroform. The aqueous phase was precipitated by the addition of 700 μl of isopropanol and 20 ng of glycogen. After overnight incubation at −80°C and centrifugation at 13 000 rpm for 30 min, the pellet was washed with 1 ml of 70% ethanol and resuspended in Binding buffer [5 mM Tris–HCl (pH 7.5), 0.5 mM EDTA, 1 mM NaCl]. For experiments aimed at mapping the modified region, RNA preparations were fragmented by sonication for 20 s (Branson Sonifier 450 equipped with a microtip, duty cycle 50%). Biotinylated RNAs or RNA fragments were captured using 100 μl of streptavidin-coated magnetic beads (DynabeadsTM M-270, Invitrogen) equilibrated in Binding buffer. After extensive washes with Binding buffer, beads were washed with T4 RNA Ligase reaction buffer (New England Biolabs). Ligation of enriched biotinylated RNA fragments to a linker was performed overnight at 25°C by incubating the RNA bound to the streptavidin beads with 1 μM of Universal microRNA Cloning Linker (NEB), 20 units of T4 RNA Ligase 1 (NEB, M0204S), 40 units of Murine RNase inhibitor (NEB), 10% PEG8000 in T4 RNA Ligase Reaction buffer in a 50-μl volume. After heat-inactivation at 65°C for 15 min, beads were washed twice with binding buffer and once with Reverse Transcription buffer (Thermo Fisher). Reverse transcription was assembled in 25 μl and performed as recommended by the manufacturer (Thermo Fisher). RNAs bound to beads were mixed with dNTPs and reverse transcription primer (RT-uni-NEB-Hind, NEB, AGCCAAGCTTATTGATGGTGCCTACAG) and incubated at 65°C for 5 min for primer annealing. After incubation on ice for 5 min, the Reverse Transcriptase (SuperScript IV, Thermo Fisher), its reaction buffer and 5 mM of Dithiothreitol (DTT) were added. Reverse transcription was allowed at 55°C for 20 min and heat-inactivated for 10 min at 80°C. The beads were then washed three times with Binding buffer, resuspended to a final volume of 25 μl. Complementary DNA (cDNA) was extracted by resuspending the beads in RNase H buffer and RNA strands were digested with RNase H (NEB) for 2 h at 37°C. RNaseH was heat-inactivated at 65°C for 20 min, and the cDNA elution was used for ligation of the second Linker (phosphorylated ss-linker oligonucleotide, pACTCTCTACTGTTTCTCCAT) as described before, in the presence of 0.5 mM of ATP. Five-μl of ligation reactions were mixed in 25-μl PCR reactions with oligonucleotides annealing to the linkers [F-sslinker-Eco (GGAGGAATTCATGGAGAAACAGTAGAGAGT) and R-NEB-Hind (AGCCAAGCTTATTGATGGTG)] using the PrimeSTAR Max DNA polymerase (Takata). Half-μl of the PCR reaction were used for cloning into pBAD18 using Gibson assembly. The pBAD18 plasmid and PCRs of enriched tRNA or tRNA fragment were amplified using oligonucleotides ACTCTCTACTGTTTCTCCATGAATTCGCTAGCCCAAAAAA and ACTGTAGGCACCATCAATAAGCTTGGCTGTTTTGGCGG. Amplified fragments were then assembled using NEBuilder® HiFi DNA Assembly Master Mix (New England Biolabs) and transformed into CaCl2-competent DH5α E. coli cells. Extracted plasmids were sequenced using oligonucleotide TCCATAAGATTAGCGGATCC (Eurofins Genomics) and sequences of the cloned fragments were aligned to the E. coli genome using the NCBI Blastn server. Sequence analyses and structural predictions Conserved motifs delimiting the Rhs N-terminal region, the central Rhs shell, and the C-terminal toxin domain were identified by sequence inspection and comparison with previously characterized Rhs proteins. Catalytic residues were inferred from comparison with structurally related ARTs and by inspection of the predicted catalytic cleft. Structural predictions for the RhsF protein and TreThr were generated using AlphaFold3 [37], and the complexes of TreThr with NAD, RhsFI, or candidate tRNAs were generated using the multimer mode of AlphaFold3. Models were visualized in ChimeraX [38]. Structural similarity searches were performed using DALI [39]. Results A C. violaceum vgrG island encodes an Rhs toxin–immunity module The C. violaceum ATCC12472 genome harbors a gene cluster encoding a functional T6SS, as well as four additional vgrG islands. Among the VgrG spike proteins encoded within these islands, VgrG3 was previously shown to play a preponderant role for T6SS sheath assembly and T6SS function [32]. The vgrG3 island shares a genetic organization characteristic of T6SS-associated effector modules, with the vgrG3 gene (CV_1432) followed by four genes encoding a full-length Rhs protein, RhsF (CV_1431), a small ORF (RhsFI, CV_1430), a truncated Rhs protein (RhsG, CV_1429) and a small Open-Reading Frame (ORF)(RhsGI, CV_1428) (Fig. 1A). This organization suggests that, as demonstrated for S. enterica Typhimurium rhs genes [40], RhsF and its potential cognate immunity protein, RhsFI, may serve as rearrangement hot spots for the repositioning of the truncated Rhs fragment and its diversification. Organization of the C. violaceum vgrG3 island and domain architecture of the RhsF protein. (A) Schematic representation of the vgrG3 island genetic organization from C. violaceum ATCC 12472. The locus comprises the vgrG3 gene (CV_1432, pink), a rhs gene (CV_1431, blue) encoding the full-length RhsF protein, CV_1430 (purple) encoding its putative cognate immunity protein, RhsFI, and two short genes: CV_1429 (flesh) and CV_1428 (red), encoding putative truncated Rhs protein (RhsG), and immunity protein (RhsGI). Genes are represented as arrows and are drawn to scale. (B, C) Schematic representation of 1513-aminoacids RhsF protein (B) and ribbon representation of its AlphaFold3 structural model (C), comprised of an N-terminal FIX domain (N-ter, green) predicted to adopt a α-helical fold similar to that of Acinetobacter baumannii Tse15, a central Rhs core predicted to form the shell (orange) with the two lid domains (cyan and dark blue), and a C-terminal toxin extension (light blue) encapsulated within the shell. Domain boundaries and residue numbers are shown. pLDDT and PAE of the RhsF AlphaFold3 model are presented in Supplementary Fig. S1. A structural alignment of the RhsF N-terminal domain with the N-terminal clade domain of A. baumannii Tse15 is presented in Supplementary Fig. S2. Domain annotation and AlphaFold3 structural prediction of the 1513-residue RhsF polymorphic toxin showed that it presents the canonical tripartite architecture of Rhs proteins [18–21]: (i) an α-helical N-terminal region [amino-acids (aa) 1–337] usually involved in Rhs trafficking, (ii) a central Rhs core (aa 338–1394) predicted to form a β-rich cocoon-like fold typical of Rhs proteins, and (iii) a C-terminal extension (aa 1395–1513) predicted to be encapsulated into the Rhs shell and likely corresponding to the toxic module (Fig. 1B and C, and Supplementary Fig. S1). The N-terminal α-helical domain of RhsF is predicted to adopt a fold closely related to the clade domain recently described in the A. baumannii Tse15 effector (Supplementary Fig. S2) [22], corresponding to a FIX domain [41]. FIX domains were proposed to mediate association with their cognate VgrG spike proteins, notably through contacts with C-terminal extensions containing an Ig-like domain followed by a helical region [22, 41]. Such a C-terminal extension is carried by VgrG3, suggesting that RhsF uses a similar strategy for docking to the T6SS puncturing apparatus. The predicted domain boundaries suggest that RhsF undergoes processing through autoproteolytic cleavage of each side of the Rhs shell. The central core is demarcated of the N-terminal region by the sequence PLLPIGE that bears conserved residues of N-terminal processing motif reported for other Rhs effectors [18, 23]. At the distal end of the shell, the C-terminal toxin domain is separated from the Rhs core by the canonical DPIGL-X17-DPLGL signature [18, 23]. These features indicate that, like other Rhs toxins, RhsF is likely synthesized as a multidomain precursor that binds to the VgrG spike and that is processed to release its effector domain into the target cell. The C-terminal extension of RhsF, TreThr, is a cytoplasmic-acting antibacterial toxin neutralized by a cognate immunity protein To determine whether the C-terminal extension of RhsF, named hereafter TreThr, is toxic, its sequence was cloned into a pBAD33 vector under the arabinose-inducible pBAD promoter. Production of the TreThr isolated domain in the cytoplasm of E. coli K-12 W3110 caused a severe growth defect under inducing conditions (Fig. 2A). Genes encoding Rhs toxins are usually organized in operon with genes encoding their cognate immunity proteins. In agreement with this canonical organization, co-expression of the rhsFI gene located downstream of rhsF fully restored viability (Fig. 2A). These results demonstrate that the RhsF and RhsFI form an effector-immunity pair. TreThr is a cytoplasmic-acting toxin neutralized by RhsFI and shares a canonical ART fold with a catalytic pocket comprised of R1403 and E1497. (A) Toxicity and rescue assays in the heterologous host E. coli. Serial dilutions (10−1–10−6) of E. coli W3110 cells carrying the empty vector, the plasmid encoding TreThr, or both TreThr and RhsFI were spotted onto LB-agar plates supplemented with 1% of glucose (repression conditions, left panel) or 0.2% of L-arabinose (induction conditions, right panel). (B) Ribbon representation of the Alphafold3 model of the TreThr toxin domain (pTM score = 0.87). TreThr adopts the typical ART architecture with six β-strands organized as a split β-sheet in the order 4-5-2/1-3-6 (strand numbers indicated). N, N-terminus; C, C-terminus. pLDDT and PAE of the TreThr AlphaFold3 model are presented in Supplementary Fig. S3. (C) Close-up view of the predicted catalytic cleft highlighting residues Arg1403 (red) and Glu1497 (yellow). (D) Ribbon representation of the AlphaFold3 model of TreThr bound to NAD (red), showing positioning of the cofactor within the central cleft formed by the split β-sheet (pTM score = 0.85, ipTM score = 0.74). pLDDT and PAE of the TreThr-NAD AlphaFold3 model are presented in Supplementary Fig. S4. (E) Ribbon representation of the AlphaFold3 model of TreThr in complex with RhsFI (purple), showing occlusion of the entrance of the TreThr catalytic pocket by the N-terminal α-helix of RhsFI (pTM score = 0.92, ipTM score = 0.92). pLDDT and PAE of the TreThr-RhsFI AlphaFold3 model are presented in Supplementary Fig. S5. (F) Toxicity assay in the heterologous host E. coli. Serial dilutions (10−1–10−6) of E. coli W3110 cells carrying the empty vector, the plasmid encoding wild-type TreThr, or its R1403A or E1497A variants were spotted onto LB-agar plates supplemented with 1% of glucose (repression conditions, left panel) or 0.2% of L-arabinose (induction conditions, right panel). TreThr adopts a canonical ART fold with a functional catalytic pocket involving Arg1403 and Glu1497 To gain insight into the biochemical function of the toxin domain, we analyzed TreThr using structure-guided approaches. AlphaFold3 modeling predicted that TreThr adopts a typical ART fold, with six β-strands arranged as a split β-sheet with the characteristic 4-5-2/1-3-6 topology and forming a cleft (Fig. 2B and Supplementary Fig. S3). This core fold is similar to that observed in previously characterized T6SS-associated ART effectors, including P. laumondii Tre23, S. enterica TreTu, P. aeruginosa Tse6 and RhsP2, and X. bovienii TreX [24–28]. Indeed, structural comparisons of the TreThr AlphaFold3 model against the Protein Data Bank (PDB) using the DALI server [39] returned the X. bovienii TreX (PDB 8S2M, Root mean square deviation (RMSD) 2.3 Å, Z-score 11.8), and Streptomyces coelicolor Guanine-specific ART (PDB 5ZJ5, RMSD 2.5 Å, Z-score 10.9), as the closest structural homologues. In canonical ARTs, catalytic residues are located on strands 1, 2, and 5, which form a cavity accommodating the NAD molecule. Based on the composition of these catalytic triads, ARTs are evolutionary clustered into the H-H-h (h being any hydrophobic residue), H-Y-[QED] and R-[ST]-E clades [29]. In TreThr, these positions are occupied by Arg1403, Ser1445, or Thr1446, and Glu1497 (Fig. 2C), suggesting an R-(S/T)-E-type ART active site. This assignment is reinforced by the AlphaFold3 structural model of TreThr bound to NAD, in which the cofactor is accommodated within the cleft formed, a configuration in which the split β-sheet cradled the nicotinamide-ribose (Fig. 2D and Supplementary Fig. S4). In addition, modeling of TreThr in complex with RhsFI shows that the C-terminal α-helix of the immunity protein sterically occupies the entrance of the TreThr catalytic cavity (Fig. 2E and Supplementary Fig. S5), suggesting a mechanism of neutralization based on direct active-site occlusion. This predicted arrangement strongly resembles the inhibition modes reported for other antibacterial ART toxin–immunity pairs, in which the immunity partner blocks access to the catalytic pocket and thereby prevents cofactor engagement. In agreement with all these observations, substitutions of the Arg1403 and Glu1497 residues with alanine (yielding the R1403A and E1497A variants, respectively) abolished TreThr toxicity (Fig. 2F). Taken together, the structural predictions and mutational data identify TreThr as an active ART toxin whose activity depends on residues Arg1403 and Glu1497 residues. TreThr inhibits translation in a NAD-dependent manner Several T6SS-associated ART toxins characterized so far have been shown to inhibit bacterial growth by interfering with translation, although they do so by modifying distinct substrates, including 23S ribosomal RNA, structured noncoding RNAs, or the elongation factor EF-Tu. To test whether TreThr affects protein synthesis, we used an in vitro coupled transcription–translation (IVT) assay to produce TreThr itself. Fig. 3A shows that TreThr was synthesized when NAD+ was omitted from the reaction. However, similar to the ART toxin TreTu, TreThr did not accumulate in the presence of NAD+. Under the same NAD+-containing conditions, the catalytic variants R1403A and E1497A were readily produced (Fig. 3A), indicating that TreThr inhibits the coupled transcription/translation system in a catalytic and NAD+-dependent manner. TreThr inhibits translation in a NAD+-dependent manner. (A) Coupled in vitro transcription-translation assay monitoring production of Strep-tagged TreThr (TreThrST) or its catalytic variants R1403A and E1497A, in the presence or absence of NAD+. Two controls were included: strep-tagged GFP (GFPST), and the S. enterica Typhimurium TreTu ART toxin that inhibits protein synthesis by modifying the EF-Tu elongation factor [26]. After 3 h of incubation, proteins were separated by SDS–PAGE, transferred onto nitrocellulose and immunodetected using anti-Streptag antibody. Molecular weight standards (Molecular Weight (MW), in kDa) are indicated on the left. (B) Coupled in vitro transcription-translation assay monitoring production of Strep-tagged GFP (GFPST) in the presence or absence of TreThr (TreThrST) or its catalytic variants R1403A and E1497A, in the presence or absence of NAD+. After 3 h of incubation, proteins were separated by SDS–PAGE, transferred onto nitrocellulose and immunodetected using anti-Streptag antibody. Molecular weight standards (MW, in kDa) are indicated on the left. To directly assess whether TreThr inhibits transcription or of translation, we next monitored GFP production in IVT reactions supplemented with TreThr and NAD+. The GFP reporter was synthesized in the absence of TreThr or of NAD+, as well as in the presence of the inactive TreThr variants R1403A and E1497A (Fig. 3B). In contrast, GFP synthesis was abolished when the IVT reaction was supplemented with both wild-type TreThr and NAD+ (Fig. 3B). These results demonstrate that TreThr inhibits transcription and/or protein synthesis and that this inhibition depends strictly on its NAD+-dependent ART activity. TreThr ADP-ribosylates the thrT and thrV tRNAs, hence preventing decoding of threonine ACC codons To identify the target(s) of TreThr, we supplemented the IVT reaction with 6-biotin-17-NAD+, an NAD+ derivative in which a biotin moiety is ligated to the adenine base, and analyzed incorporation of the labeled ADP-ribose. Examination of the protein fraction of the IVT reaction revealed no major change in protein biotinylation profiles in the presence or absence of TreThr (Supplementary Fig. S6). Although this result does not formally exclude low-abundance protein substrates, it suggested that protein modification is not the dominant biochemical signature associated with TreThr activity in the IVT system. By contrast, analysis of RNA extracted from the same reactions revealed a TreThr-dependent biotin signal in the low-molecular-weight RNA fraction corresponding to tRNAs (Fig. 4A), suggesting that TreThr primarily ADP-ribosylates one or more tRNA species. TreThr ADP-ribosylates aminoacylated tRNAs. (A) RNA extracted from of in vitro transcription-translation (IVT) reactions performed in the presence of the TreThr-encoding DNA template in the presence or absence of 6-biotin-17-NAD+ (NAD-biot) were stained with EtBr (upper panel) or detected using streptavidin-AP conjugate (lower panel). The positions of tRNAs, 16S and 23S ribosomal RNAs are indicated. The asteriks indicate the TreThr mRNA that is detected upon prolonged incubation time of the reaction and/or prolonged detection with the streptavidin-AP conjugate. (B) Purified tRNAs incubated or not with TreThr and 6-biotin-17-NAD+ (NAD-biot) were stained with Ethidium Bromide (EtBr, upper panel) or detected using streptavidin-AP conjugate (lower panel). The complete IVT reaction was used as positive control. (C) RNA extracted from of in vitro transcription-translation (IVT) reactions containing or not ribosomes (ribos), and supplemented with TreThr in the presence or absence of 6-biotin-17-NAD+ (NAD-biot) were stained with EtBr (upper panel) or detected using streptavidin-AP conjugate (lower panel). The positions of tRNAs, 16S and 23S ribosomal RNAs are indicated. (D) Purified tRNAs were incubated with ATP, TreThr and 6-biotin-17-NAD+ in the presence or absence of aminoacids and/or of aminoacyl–tRNA synthetases (aaRSs) were extracted and stained with EtBr (upper panel) or detected using streptavidin-AP conjugate (lower panel). Charged tRNAs correspond to purified tRNAs, pre-incubated with ATP, aminoacids and aminoacyl–tRNA synthetases prior to extraction and incubation with TreThr and 6-biotin-17-NAD+. We next asked whether TreThr could directly modify isolated tRNAs. Figure 4B shows that no detectable labeling was observed when purified tRNAs were incubated with TreThr and biotin-NAD+. This result suggested that a specific biochemical or conformational state of tRNA, absent from the purified preparation, might be required for productive recognition by the toxin. It also raised the possibility that additional factors present in the coupled transcription–translation reaction contribute to substrate formation or presentation. To test whether active translation might be required to generate the susceptible tRNA state, we extracted RNAs from IVT reactions lacking ribosomes. Under these conditions, TreThr modified tRNA efficiently (Fig. 4C), demonstrating that active protein synthesis and ribosome engagement are not required. The relevant tRNA state must therefore arise upstream of ribosomal decoding. Because the ribosome-free reaction contains amino-acids and aminoacyl–tRNA synthetases, we next examined the role of these components. Strikingly, tRNA modification was observed only in the presence of both amino-acids and aminoacyl–tRNA synthetases (Fig. 4D), suggesting that tRNAs are targeted by TreThr either after aminoacylation or while associated with aminoacyl–tRNA synthetases. To test whether tRNAs were efficiently aminoacylated under these experimental conditions, reaction products were analyzed by acid-urea–PAGE, which preserves aminoacyl–tRNA ester bonds and separates charged from uncharged tRNAs based on their differential electrophoretic mobility [36]. In the presence of aminoacyl–tRNA synthetases and amino-acids, most of the tRNAs displayed the mobility shift expected for aminoacylated species (Supplementary Fig. S7). To gain further insight into the tRNA state, we purified charged tRNAs from aminoacylation reactions and incubated them with TreThr. Aminoacyl–tRNAs were readily modified (Fig. 4D, last lane), supporting the conclusion that TreThr acts on charged tRNAs. To determine whether TreThr targets all charged tRNA or instead selectively modifies specific tRNA isoacceptors, total RNAs from IVT reactions performed in the presence of TreThr and biotin-NAD were extracted, enriched on streptavidin beads, ligated, and cloned. Sequencing of 22 independent full-length enriched tRNAs exclusively identified only thrT and thrV, the two threonine tRNAThr isoacceptors bearing the GGU anticodon (Fig. 5A). To further map the modified region, TreThr-biotinylated tRNAs were fragmented by sonication prior to enrichment and cloning. Sequencing of 15 additional independent fragments again exclusively identified thrT and thrV tRNA sequences, all encompassing bases 35 and 36, corresponding to the second and third positions of the GGU anticodon (Fig. 5A and B). These results therefore identify the anticodon of the tRNAThr(GGU) thrT and thrV isoacceptors as the target region of TreThr-mediated ADP-ribosylation. TreThr ADP-ribosylates the tRNAs thrT and thrV anticodons, preventing decoding of threonine ACC codons. (A) Identification and mapping of TreThr-modified tRNAs. Biotinylated RNAs were fragmented and enriched from in vitro reactions containing TreThr and 6-biotin-17-NAD+, cloned and sequenced. The different secondary structures of tRNA are shown with a color code (acceptor stem, green; D-loop, pink; anticodon loop, blue; anticodon, red; variable loop, orange; TψC loop, yellow). The fragments of the thrT and thrV tRNAs identified are represented by the black horizontal lines, that all encompass bases 35 and 36 (underlined in red) corresponding to the second and third bases of the anticodon. (B) Schematic secondary-structure representation of the tRNAThr(GGU) thrT [same color code as in panel (A)] highlighting the bases targeted by TreThr in the anticodon loop. (C) Purified tRNAs incubated with ATP, TreThr and 6-biotin-17-NAD+ in the presence or absence of a mixture aminoacids lacking threonine and/or of threonine, and/or of aminoacyl–tRNA synthetases (aaRSs) were extracted and stained with EtBr (upper panel) or detected using streptavidin-AP conjugate (lower panel). (D) Coupled in vitro transcription-translation assay monitoring production of a FLAG-tagged synthetic construct containing all codons in triplicate (AllCodons) or containing all codons except codon ACC (AllCodons-ACC) in the presence or absence of GFP or of TreThr, and of NAD+. After 3 h of incubation, proteins were separated by SDS–PAGE, transferred onto nitrocellulose and immunodetected using anti-FLAG antibody. The position of the artificial proteins is indicated by the arrow. Molecular weight standards (MW, in kDa) are indicated on the left. (E) Distribution of AlphaFold3 ipTM scores obtained for multimer models generated between TreThr, NAD and the 86 E. coli tRNAs. The highest-scoring models include thrT and thrV (red circles). (F) Ribbon representation of the AlphaFold3 model of TreThr in complex with NAD and the thrT tRNA (ipTM score = 0.72). The close-up view highlights the relative positions of the TreThr catalytic cleft, of NAD (yellow) and of the three bases of the thrT anticodon (red letters). pLDDT and PAE of the TreThr-NAD-thrT AlphaFold3 model are presented in Supplementary Fig. S9. To independently test the specificity of TreThr toward threonine tRNAs, the complete pool of purified tRNAs was incubated with aminoacyl–tRNA synthetases and ATP in the presence of either all amino acids except threonine, or threonine alone. Thus, the aminoacylation state of tRNAThr species was selectively controlled by the presence or absence of threonine. Acid-urea–PAGE analysis confirmed the presence of aminoacylated tRNA species when threonine was provided as the sole amino-acid (Supplementary Fig. S7), consistent with aminoacylation of tRNAThr by the threonyl–tRNA synthetases present in the aminoacyl–tRNA synthetase mixture. As shown in Fig. 5C, the absence of threonine in the amino-acid mixture prevented TreThr-dependent tRNA labeling, whereas threonine alone was sufficient to support modification. To test more directly whether translation inhibition results specifically from targeting tRNAThr(GGU), we performed IVT reactions using synthetic reporter genes. Because the thrT and thrV tRNAs decode ACC, we assessed the impact of TreThr on translation of FLAG-tagged synthetic ORFs containing or lacking threonine ACC codons: the ‘AllCodons’ template corresponded to a gene containing all codons in triplicate, whereas the ‘AllCodons-ACC’ gene lacked the ACC threonine codon (Supplementary Fig. S8). While the presence of GFP in the IVT reaction did not interfere with the synthesis of these two proteins, translation of the ‘AllCodons’ construct was abolished in the presence of TreThr and NAD+, whereas the ‘AllCodons-ACC’ construct was translated (Fig. 5D). To complement the biochemical data, we generated AlphaFold3 multimer models between TreThr, NAD, and each of the 86 E. coli tRNAs. Fig. 5E shows that the predicted complexes spanned a broad range of ipTM confidence scores, but thrT and thrV ranked among the highest-scoring candidates, with ipTM values of 0.72 and 0.71, respectively. Although these structural models remain predictive, inspection of the TreThr-NAD-thrT model showed that the anticodon loop is positioned directly adjacent to the TreThr catalytic cleft, with the GGU anticodon oriented toward the bound cofactor and the catalytic pocket. This arrangement is consistent with a geometry compatible with ADP-ribose transfer to one of the anticodon nucleotides (Fig. 5F and Supplementary Fig. S9). Because aminoacylated elongator tRNAs are normally bound by EF-Tu before delivery to the ribosome, we also generated an AlphaFold3 model containing TreThr, NAD, tRNAThr(GGU) thrT and EF-Tu. In this model, EF-Tu engages the aminoacyl acceptor end of the tRNA, consistent with experimentally determined structures of EF-Tu–aminoacyl–tRNA complexes [42, 43], whereas TreThr binds the opposite region around the anticodon loop (Supplementary Fig. S10). The predicted binding surfaces therefore do not substantially overlap, suggesting that TreThr recognition of aminoacyl–tRNAThr may be compatible with EF-Tu binding. Discussion In this work, we characterized the T6SS-associated RhsF polymorphic toxin from C. violaceum, which is encoded in the vgrG3 island. C. violaceum is an environmental β-proteobacterium and opportunistic pathogen that uses a quorum sensing-regulated T6SS for interbacterial competition. Previous work showed that this bacterium encodes a single active T6SS containing multiple nonredundant VgrG proteins, among which VgrG3 plays a major role in T6SS function [32]. The VgrG spike proteins usually constitute one of the loading sites of effectors on the T6SS, and effectors are usually encoded at close proximity to their carrier [13, 44]. Indeed, the vgrG3 gene is followed by four genes encoding the full-length RhsF effector, its immunity protein, a truncated Rhs element which likely can re-organize with RhsF to diversify the Rhs repertoire, as shown previously in Salmonella Typhimurium [40], and its potential immunity protein. The organization of the VgrG3 island therefore suggests that RhsF might be loaded on VgrG3 for delivery into target cells. In most Rhs polymorphic toxins, the N-terminal trafficking domain specifies the association with its cargo [20]. Indeed, the N-terminal region of RhsF displays similarity to the N-terminal α-helical clade domain of the Tse15 effector from A. baumannii (Supplementary Fig. S2), which corresponds to a FIX domain [22, 41]. The Tse15 FIX domain was recently reported to interact with the C-terminal extension of its cognate VgrG15 spike protein, which correspond to an Ig-like transthyretin domain followed by α-helices [22]. Interestingly, VgrG3 also bears a similar extension, suggesting that RhsF follows a comparable mode of loading in which its FIX module mediates contacts with the VgrG3 C-terminus, allowing positioning the toxin domain for injection into neighboring cells. This hypothesis is supported by recent co-immunoprecipitation assays demonstrating interaction between VgrG3 and RhsF [45]. More broadly, these data show that T6SS-associated Rhs toxins use at least two major architectural solutions for loading onto the secretion apparatus: (i) PAAR domains that dock directly at the tip of the spike, and (ii) FIX clade domain, used by Tse15 and RhsF, which interacts with a dedicated VgrG extension. This broadens the current view of how Rhs polymorphic toxins are recruited by the T6SS by diversifying the N-terminal trafficking modules. Further work on T6SS-associated Rhs proteins will likely identify new modules and new modes of interaction with VgrG. Besides its N-terminal trafficking domain, RhsF has a typical Rhs organization, with the central domain composed of YD-repeats and folded as a shell or cocoon delimitated by autoproteolysis signatures, which likely encapsulates the C-terminal extension. We showed that this C-terminal extension, which we designated TreThr, impairs translation in vitro and inhibits E. coli growth when produced in the cytoplasm. We demonstrated that TreThr is an ART with a catalytic pocket comprised of Arg1403 and Glu1497 residues. These results confirm the data from a recent study reporting the structural and functional characterization of the RhsF C-terminal extension showing that this domain acts as an ART toxin that inhibits protein synthesis. This study also provided further information on the catalytic mechanism as mutagenesis assays showed that the third residue of the active site is not a serine or threonine, as found in R-[ST]-E-family ARTs, but rather the Tyr1456 residue, thereby classifying TreThr as an R-Y-E ART [45]. However, the molecular target and the mechanism by which this toxin inhibits translation remained unknown. Here, we further show that TreThr impairs translation, through specific targeting of threonine tRNAs. Our further data support a model in which this C-terminal TreThr domain ADP-ribosylates the second and/or third base of the GGU anticodon of the thrT and thrV tRNAThr isoacceptors, hence preventing base pairing and ACC codon decoding during translation. This activity expands the functional repertoire of T6SS-associated ART toxins, which illustrates the remarkable versatility of this enzyme family. While S. proteomaculans Tre1 and X. bovienii TreX block cell division by modifying the FtsZ ring component, P. laumondii Tre23, S. enterica Typhimurium TreTu and P. aeruginosa RhsP2 inhibits protein synthesis through ADP-ribosylation of 23S ribosomal RNA, of the elongation factor EF-Tu, or of structured noncoding RNAs, including the tRNA pool and RNase P [24–28]. Rather than targeting the ribosome itself, a translation factor, or multiple RNA species, TreThr acts with high selectivity on two threonine tRNAs. By introducing a focused lesion within the decoding apparatus, it shuts down translation, revealing a previously unrecognized strategy for disrupting tRNA function. Indeed, many bacterial toxins, such as colicins E5 and D, PrrC, or VapC, have been shown to inhibit translation by cleaving tRNAs, most commonly within the anticodon loop [46–55]. Other systems, including CDI toxins, extend this diversity by targeting distinct regions of the tRNA molecule, such as the acceptor stem or T-loop, further emphasizing the versatility of tRNAs as substrates [55–57]. In contrast to this tRNA hydrolysis mechanisms, TreThr does not cleave the tRNA backbone, but instead modifies it. This represents a distinct mode of action in which the tRNA is not degraded but inactivated. However, tRNA modification is also a relatively common strategy: GCN5-related N-acetyltransferase (GNAT)-family toxins such as AtaT, TacT, or ItaT inhibit translation by acetylating the aminoacyl moiety of charged tRNAs, whereas MenT3 adds poly-cytidines to the aminoacyl acceptor ends of most tRNA but with a marked preference for tRNASer, thereby arresting elongation [58–62]. However, while TreThr targets charged tRNA, it does not target the aminoacyl moiety, but the anticodon loop itself. ADP-ribosylation of tRNA is not unprecedented. Streptomyces coelicolor ScARP protein was previously shown to ADP-ribosylate yeast tRNA [63]. ScARP catalyzes ADP-ribosylation of the N2 position of guanosine, establishing guanine as a possible acceptor for RNA ADP-ribosylation. Interestingly, our sequencing analysis restricts the TreThr-modified region to the GU dinucleotide at positions 35–36 of the GGU anticodon of the thrT and thrV isoacceptors. Although identification of the precise ADP-ribose acceptor will require direct chemical characterization, G35 represents an attractive candidate for the TreThr-modified nucleotide. Given the central role of the anticodon in codon recognition, one may hypothesize that such a modification directly interferes with decoding by preventing proper codon-anticodon pairing. This mechanism therefore combines the specificity typically associated with tRNases with a nondestructive mode of inhibition, potentially allowing rapid and efficient shutdown of translation by reducing the pool of translation-competent tRNAThr(GGU). As our toxicity assays demonstrated that TreThr is a highly potent toxin and given that the T6SS is likely to inject only a limited number of Rhs effector molecules per firing event, it will be important to determine the catalytic efficiency of TreThr-mediated ADP-ribosylation. The T6SS effector Tse6, which depletes cellular NAD+, was shown to exhibit a high catalytic rate, hydrolyzing ∼105 NAD+ molecules per minute [64]. Such high turnover rates may be critical for enabling potent intoxication despite the limited number of effector molecules delivered into target cells. A striking feature of TreThr is its specificity for particular threonine tRNA isoacceptors. Despite the overall structural conservation of tRNAs, several toxins are able to discriminate between closely related isoacceptors, as illustrated by colicin D targeting only tRNAArg species, VapC-mt4 toxins acting on tRNAAla and tRNASer, the PARIS defence system AriB effector hydrolyzing tRNALys(UUU) or AtaT acetylating the methionine amine group of the charged initiator tRNAfMet [47, 54, 59, 65, 66]. Similarly, TreThr does not act broadly on the tRNA pool, but rather displays exquisite substrate specificity toward the two tRNAThr(GGU) species thrT and thrV. The targeting of thrT and thrV by TreThr suggests that specific determinants govern substrate recognition. Our results indicate that TreThr recognizes its substrate in a very specific biochemical state, i.e. aminoacylated. Indeed, purified free tRNAs were not detectably modified, whereas modification occurred in the presence of amino-acids and aminoacyl–tRNA synthetases and was retained with charged tRNA preparations. One may hypothesize that aminoacylation alters the conformational ensemble or dynamics of tRNAThr(GGU), thereby increasing the accessibility or productive presentation of the anticodon loop to TreThr. In the cell, however, aminoacylated elongator tRNAs are rapidly captured by EF-Tu–GTP, which recognizes the aminoacylated acceptor end of the tRNA. Structural studies of EF-Tu–aminoacyl–tRNA complexes have established that this interaction primarily involves the acceptor arm and aminoacyl moiety, leaving the anticodon region exposed [42, 43]. Consistent with these structures, our AlphaFold3 model of the TreThr–NAD–tRNAThr(GGU)–EF-Tu complex positions EF-Tu and TreThr at opposite ends of the tRNA, without substantial overlap between their predicted binding interfaces (Supplementary Fig. S10). Although this model requires experimental validation, it suggests that TreThr could recognize tRNAThr(GGU) thrT and thrV while bound to EF-Tu, or that TreThr binding may not preclude subsequent association of EF-Tu. AlphaFold3 predictions further suggest that TreThr interacts with the thrT and thrV tRNAs with relatively high confidence scores (ipTM > 0.7) in the presence of NAD. Interestingly, models generated in the absence of NAD yielded low confidence scores (ipTM scores 0.2 and 0.15 for thrT and thrV), suggesting that NAD positioning within the TreThr pocket may help organize or stabilize the productive TreThr-tRNA interface rather than simply serving as a passive donor molecule after substrate binding. Further biochemical and structural studies will likely provide information on the mode of recruitment of TreThr on its target tRNAs. Contrarily to colicin D, which cleaves the four tRNAArg isoacceptors [47], TreThr targets only two of the four threonine tRNA isoacceptors. These two tRNAs, thrT and thrV, share the same GGU anticodon and hence decode ACC threonine codons. By specifically impairing decoding of ACC threonine codons, TreThr is unlikely to affect the translation of proteins that lack ACC codons. However, given that ACC is the most frequently used threonine codon in E. coli (accounting for ∼40% of threonine codons), inactivation of the corresponding tRNA pool is expected to impair translation of the vast majority of proteins [67–69]. Through highly specific targeting of key isoacceptors, TreThr is likely to achieve a global effect, comparable to that of broad-spectrum tRNases that degrade multiple tRNAs, thereby combining molecular precision with widespread physiological impact. The broad variety of toxins targeting tRNAs is a recurrent strategy in bacterial competition and host-phage interactions. Studies have shown that bacteriophage effectors cleave host tRNAs [70] or encode tRNAs that are resistant to host tRNases, often through mutations in the anticodon loop [71], highlighting an evolutionary arms race centered on tRNA integrity. As many counter-defense mechanisms have been described to prevent damages caused by T6SS attacks or effectors [72], it will also be important to investigate whether target cells possess mechanisms to repair or tolerate TreThr-mediated modification, as has been described for other forms of tRNA damage [55]. Supplementary Material Acknowledgements We thank members of the Cascales laboratory, especially Laure Journet, Thierry Doan, and Julie Viala for discussions and support, and Melissa Guilbert and Audrey Gozzi for technical support. Author contributions Maïalène Chabalier (Formal analysis [supporting], Investigation [equal], Writing—review & editing [supporting]), Jonas B Desjardins (Investigation [supporting], Resources [supporting], Writing—review & editing [supporting]), Julius Martinkus (Investigation [supporting], Resources [supporting], Writing—review & editing [supporting]), Eric Cascales (Conceptualization [lead], Data curation [lead], Formal analysis [lead], Funding acquisition [lead], Investigation [equal], Project administration [lead], Resources [lead], Supervision [lead], Validation [lead], Writing—original draft [lead], Writing—review & editing [lead]). Contributor Information Maïalène Chabalier, Laboratoire d’Ingénierie des Systèmes Macromoléculaires (LISM, UMR7255), Aix-Marseille Université, CNRS, 31 Chemin Joseph Aiguier CS7071, Marseille Cedex 20 13402, France. Jonas B Desjardins, Laboratoire d’Ingénierie des Systèmes Macromoléculaires (LISM, UMR7255), Aix-Marseille Université, CNRS, 31 Chemin Joseph Aiguier CS7071, Marseille Cedex 20 13402, France. Julius Martinkus, Laboratoire d’Ingénierie des Systèmes Macromoléculaires (LISM, UMR7255), Aix-Marseille Université, CNRS, 31 Chemin Joseph Aiguier CS7071, Marseille Cedex 20 13402, France. Eric Cascales, Laboratoire d’Ingénierie des Systèmes Macromoléculaires (LISM, UMR7255), Aix-Marseille Université, CNRS, 31 Chemin Joseph Aiguier CS7071, Marseille Cedex 20 13402, France. Supplementary data Supplementary data is available at NAR online. Conflict of interest None declared. Funding This work was supported by the Centre National de la Recherche Scientifique, the Aix-Marseille Université, and by grants from the Agence Nationale de la Recherche (ANR-20-CE11-0017) and the Fondation Bettencourt Schueller. J.B.D. was supported by a doctoral fellowship and a 4th year extension from the Fondation pour la Recherche Médicale (contracts ECO202206015565 and FDT202504020886). J.M. was supported by the ANR grant. Funding to pay the Open Access publication charges for this article was provided by ANR. Data availability The data underlying this article are available in the article and in its online supplementary material. References Associated Data Supplementary Materials Data Availability Statement The data underlying this article are available in the article and in its online supplementary material.