Architecture and mechanism of a dual-enzyme retron system in prokaryotic immunity

Abstract

Retrons are bacterial genetic retroelements encoding a reverse transcriptase (RT) and a non-coding RNA (ncRNA)-multi-copy single-stranded DNA (msDNA) hybrid. Diverse effector proteins or domains are found to associate with retrons, typically forming tripartite toxin-antitoxin systems involved in anti-phage defense. Although retrons have attracted growing interest in genome editing technologies, the mechanisms underlying most retron-mediated immune systems remain poorly understood. Here, we characterize a distinct quaternary retron system, Ec78, harboring a dual-component effector complex in which the PtuA ATPase and PtuB nuclease act in concert to mediate phage clearance. The cryo-EM structure of the Ec78 complex adopts a flower-basket-like architecture, with two Ec78 retrons engaging the PtuAB effector complexes through a msDNA-insertion assembly mechanism. Shortening of msDNA in length releases the PtuAB from Ec78 retron and triggers its activation. The cryo-EM structure of the retron-unbound effector complex further reveals an arginine-lysine finger loop on the PtuB nuclease that undergoes an ordered-to-disordered transition during enzymatic activation. These findings delineate the molecular basis underlying the Ec78 system in antiviral defense and highlight the mechanistic diversity of retron systems in prokaryotic immunity.

Retrons are bacterial genetic retroelements implicated in anti-phage defense. Here, the authors characterize the quaternary retron system Ec78 and elucidate the biochemical mechanisms underlying Ec78-mediated prokaryotic immunity.

Introduction

Bacterial retrons are specialized genetic elements comprising a reverse transcriptase (RT) and a non-coding RNA (ncRNA)¹. The ncRNA consists of two parts, msr and msd (referred to msrRNA and msdRNA thereafter). During retron function, the msrRNA forms a scaffold that supports the RT protein, while the msdRNA segment serves as a template for the RT to produce multi-copy single-stranded DNA (msDNA)². Reverse transcription initiates from the 2’-hydroxyl group of a conserved guanosine in the msrRNA, typically resulting in a msrRNA-msDNA hybrid that contains a unique 2’,5’-phosphodiester bond between the branching guanosine and the first nucleotide of the msDNA^3–6. Since their discovery decades ago, studies of retrons have primarily focused on the mechanism of msDNA biogenesis. Taking advantage of the ability to produce msDNA in situ, retrons have been coupled with CRISPR-Cas9 systems to enable targeted insertion of DNA fragments, offering promising tools for precision genome editing across diverse cell types^7–10.

However, the biological function of retrons in bacteria remained obscure. Recent studies revealed that retrons act as components of multigene anti-phage defense systems, typically comprising a RT, a non-coding RNA, and an accessory defense-associated protein or domain^1,4,11,12. These systems are widely distributed, classified into nearly 13 types according to the accessory proteins or domains⁴. Despite their widespread distribution and emerging applications, only a few retron systems have been mechanistically characterized. Among them, the Ec86 and Sen2 systems function as tripartite toxin-antitoxin (TA) systems, in which the retron suppresses the toxic activity of the associated effectors^13–15. Structural analysis of the Ec86 interference filament revealed that its msDNA physically cages the N-glycosidase effector in an inactive state. Moreover, effector activation can be triggered by msDNA degradation in the Sen2 system via phage-encoded exonucleases, or msDNA methylation in the Ec86 system, suggesting that modification of msDNA may represent a general strategy for retron activation.

Surprisingly, in contrast to these well-characterized systems which utilize one effector protein for function, the type I-A retron system harbors two enzymatic effector components derived from the Septu system^1,4. Septu systems are prokaryotic defense systems¹⁶, featuring an ATPase named PtuA and an HNH nuclease named PtuB, and are classified into two types¹⁷. Type I Septu systems utilize a catalytically inactive PtuA to form a hexamer and recruit two PtuB nucleases for phage defense through degradation of phage DNA¹⁸. Type II Septu systems, on the other hand, are typically associated with retron genetic elements. Actually, the type II Septu system and the type I-A retron system refer to the same system, but are named differently based on structural or genetic perspectives. However, the mechanism by which the retron cooperates with PtuAB to mediate anti-phage defense remains unclear.

Here, we biochemically and structurally characterize a type I-A retron system, Ec78, and define it as a quaternary TA system, in which the retron functions as the antitoxin, and the two enzymatic effectors, PtuA and PtuB, form a complex and act in concert to mediate toxicity. We determine high-resolution cryo-EM structures of the two-component effector complex in both retron-bound and -free states. Together with biochemical and mutagenesis analysis, we delineate the molecular basis of retron-mediated inhibition, dual-enzymatic effector activation, and their synergistic toxicity, revealing diverse and conserved features among retron-associated defense systems. Our work provides mechanistic insights into a dual-enzymatic effector retron system and also offers a foundation for future development of retron-based tools in biotechnological applications.

Results

Ec78 system is a TA system

The Ec78 system belongs to the type I-A retron family and harbors a two-component effector complex¹⁷, consisting of an ATPase termed PtuA and an HNH endonuclease termed PtuB (Fig. 1a). Retron genetic elements consist of an RT protein and an ncRNA. The RT protein utilizes the msdRNA as a template for reverse transcription (Fig. 1a). A specialized msDNA molecule is generated and covalently linked to the msrRNA, forming an RNA-DNA hybrid¹. The ncRNA components in retron systems exhibit considerable variation in both sequence and length, suggesting distinct molecular mechanisms of their nucleic acid elements^4,19. Therefore, we treated the purified Ec78 complex with proteinase K and RNase A to isolate msDNA for sequencing. DNA sequencing revealed a 78-nucleotide (nt) msDNA synthesized by the RT. We then mapped the sequencing reads along the ncRNA locus and generated msDNA coverage plots for the Ec78 retron (Fig. 1b).

Fig. 1. Biochemical characterizations of the Ec78 system.

a Gene architectures of the Ec78 system from E. coli ECONIH5. b The msDNA sequencing reads extracted from Ec78 system were mapped to the corresponding ncRNA locus. c Schematic diagram of bacterial growth spot assay. Both Ec78 system and PtuAB effector complex were constructed using the native promoter. d, e Bacterial growth spot assay and CFU quantification of the Ec78 system or PtuAB effector complex. EV, empty vector. For both panels, corresponding CFU per mL was determined from serial dilutions and is shown as mean ± SEM from three independent experiments. Source data are provided as a Source Data file. f RT-qPCR quantification of tRNA^Tyr in cells expressing the PtuAB effector. The expression of PtuAB was regulated by glucose (repression) or arabinose (induction), and the arabinose induction resulted in degradation of tRNA^Tyr. g RT-qPCR quantification of tRNA^Tyr in cells expressing Ec78 system under T5 phage infection. For (f) and (g), data are presented as mean ± SEM (n = 3), and error bars represent the SEM. Source data are provided as a Source Data file.

Bacterial retron systems have been reported to function as TA systems, in which both the RT and the msDNA elements act as antitoxins¹⁵. To investigate whether the Ec78 system is a functional TA system, we cloned the Ec78 system (including retron, PtuA and PtuB) into one plasmid and transferred the vector into E. coli cells for expression. We also expressed the Ec78 PtuAB complex or individual Ec78 elements under native promoter (Fig. 1c). Expression of individual components within Ec78 system did not elicit cytotoxicity (Supplementary Fig. 1a, b). By contrast, expression of the PtuAB effector complex caused a ~ 10⁴-fold reduction in colony-forming unit (CFU) in E. coli, whereas co-expression of the Ec78 retron restored cell growth to near-control levels on solid media and in liquid culture (Fig. 1c–e and Supplementary Fig. 1a–c). Collectively, these results show that the Ec78 system behaves as a TA system, with PtuA and PtuB together functioning as the toxin. Catalytic inactivation of either the PtuA ATPase or the PtuB nuclease abolished effector toxicity and restored bacterial growth (Fig. 1d, e), demonstrating that both activities are indispensable for function.

Notably, while most well-characterized HNH nucleases, such as the CRISPR-associated HNH domains, are known to cleave DNA substrates for anti-phage immunity, PtuAB expression did not produce DNA damage–associated foci (puncta) in our confocal assays, suggesting that the Ec78 PtuAB is unlikely to induce chromosomal DNA degradation (Supplementary Fig. 1d). In contrast, the Ec78 effector was previously reported to preferentially target tRNA^Tyr in cells²⁰. Consistent with this, our RT-qPCR analysis showed a clear decrease in tRNA^Tyr upon PtuAB expression in E. coli (Fig. 1f). Moreover, a marked decrease in tRNA^Tyr was also observed when the Ec78 system was expressed in cells infected with T5 phage (Fig. 1g). Importantly, we observed nucleoid condensation in E. coli expressing PtuAB under phage infection in confocal imaging, which is consistent with translation inhibition caused by tRNA depletion (Supplementary Fig. 1d). In line with our observations, Ec78 PtuAB-mediated tRNA depletion and translation inhibition have also been reported in recent studies^21,22. Taken together, these results demonstrate that Ec78 functions as a TA system, with PtuAB acting as the effector mediating tRNA^Tyr degradation and translation inhibition in cells.

Cryo-EM structure of Ec78 system

To elucidate the structural basis of the retron Ec78 system, we co-expressed the Ec78 complex from E. coli ECONIH5 and determined its cryo-EM structure at 3.03 Å resolution (Fig. 2a, b and Supplementary Fig. 2). Cryo-EM structure of retron Ec78 system adopts a “flower-basket-like” architecture, comprising four PtuA proteins, one PtuB nuclease, and two Ec78 retrons, with a total molecular weight of 419 kDa (Fig. 2c–f). Two elongated α-helices (termed coiled-coil region) extending from PtuA protomers (A1 and B1) stagger to form the handle of the basket. The remaining portions of the PtuA dimers, together with the RT-msrRNA complex positioned alongside, form the base of the basket, which is split into two units by an extended msDNA stem-loop. Unit A contains a “hook-like” Ec78 retron, comprising the RT protein and a hybrid msrRNA-msDNA molecule. A PtuA dimer of the Ec78 effector faces the Ec78 retron in unit A. In unit B, the msDNA stem-loop of the retron is not involved in complex assembly and remains unresolved in the cryo-EM density, likely due to its intrinsic flexibility. Nevertheless, a complete Ec78 effector complex is clearly resolved, with the PtuB nuclease flanked by PtuA dimer.

Fig. 2. Overall structure of Ec78 system complex.

a Size exclusion chromatograms of Ec78 system. The peak fractions containing components of Ec78 system were analyzed by SDS-PAGE and urea PAGE. The gels were representative of three repeat experiments. Source data are provided as a Source Data file. b Representative 2D class average with particles of Ec78 complex in different orientations. c, d Cryo-EM density map and atomic model of Ec78 system (top view). The ncRNA and msDNA are colored in orange and black, respectively. Protein components are colored in the same scheme as in Fig. 1a. e, f Cryo-EM density map and atomic model of Ec78 complex (front view). The same color scheme as Fig. 2c is applied and the corresponding schematic is represented in the lower right corner.

Structure of Ec78 Retron

High-quality cryo-EM density of unit A enabled unambiguous tracing of most of the Ec78 retron (Fig. 3a–c and Supplementary Fig. 3). In particular, the 311-amino-acid RT protein adopts a canonical right-handed architecture, comprising palm, thumb, and finger domains, and harboring a conserved YADD catalytic core (Fig. 3b, c). A DALI search²³ of the Ec78 RT protein reveals the highest structural similarity to that in retron Ec86, with an RMSD of 3.0 Å, and a Z-score of 21.6. Several structural features are unique to Ec78 RT, including two distinctive loop regions (Fig. 3c). Loop I (α6-α7 loop, termed the sensing loop, discussed below), emerges near the finger domain and contacts the distal end of the msDNA stem-loop, and loop II (β5-α8 loop, discussed below) is responsible for interacting with PtuB, suggesting their specialized roles in Ec78 function (Supplementary Fig. 4a).

Fig. 3. Structural basis of each component in Ec78 system complex.

a Schematic representation of the msrRNA-msDNA duplex in Ec78 system. Well-resolved nucleotides of msDNA in the duplex (black), msrRNA (orange) and conserved branching G (red) are highlighted, whereas the unresolved nucleotides are indicated by gray circles. b Domains distribution of Ec78 RT (upper panel). The atomic model of RT (lower panel) is shown with the same color scheme as the upper panel. c Structure alignment of Ec78 RT and Ec86 RT (PDB: 7V9X). Close-up view of the catalytic YADD motif (inserted panel). d Domain organization of Ec78 PtuA (upper panel). The superposition of Ec78 PtuA and Type I PtuA (PDB: 8EEA). Detailed insight into the catalytic pocket within the ATPase domain is shown in the inserted panel and key residues are shown in stick representation. e ATP hydrolysis of PtuA, WT Ec78 system, and Ec78 systems harboring catalytically inactive variants of PtuA (D395A/E396A and T100G/S101G). Three replicates were carried out for each measurement. The data were shown as mean ± SEM. Source data are provided as a Source Data file. f The atomic model of PtuB. Detailed insight into the catalytic pocket is shown in the inserted panel and catalytic residues are shown as sticks.

The msrRNA contains two stem-loop regions, RSLa (14–27 nt) and RSLb (28–48 nt) (Fig. 3a). Only RSLb is resolved in the cryo-EM density, wrapping around the RT thumb domain (Fig. 3b). The inverted repeat sequences (IRa1 and IRa2), located at the 5’-end of the msrRNA and the 3’-end of the msdRNA, along with the branching guanine required for priming msDNA synthesis, remain unresolved in the EM map (Fig. 3a). Additionally, six nucleotides at the 3’-end of the msDNA, positioned near the RT active site, are base-paired with the msrRNA. The 5’-region of the Ec78 msDNA extends outward to form a nail-like 28-base-pair stem-loop inserting into the entire complex, in stark contrast to the five-pointed star-shaped msDNA structure observed in retron Ec86²⁴. Moreover, the two retrons are almost identical to each other except for the msDNA stem loop (Supplementary Fig. 4b).

Structure of PtuAB effectors in Ec78 complex

Ec78 PtuA consists of a central ATPase domain, with an additional domain at its N-terminus and an α-helical bundle domain at its C-terminus (Fig. 3d and Supplementary Fig. 5). The ATPase domain adopts an ellipsoidal bilobed architecture, resembling typical ABC-type ATPases, such as Rad50²⁵. Lobe I of PtuA ATPase domain is composed of six β-sheets and three α-helices, and folds into an α/β roll structure (Supplementary Fig. 5a, b). The Lobe II contains five α-helices and six β-sheets and adopts a βα-sandwich fold. The two lobes are connected by a coiled-coil region (residues 257–305). Additionally, a distinctive encircling loop, spanning approximately 35 residues (residues 204–238), extends from the β8 sheet to the α6 helix in Lobe I and wraps around the surface of the ATPase domain. This structural feature is unique to PtuA and is not observed in other ABC-type ATPases (Fig. 3d and Supplementary Fig. 5c). DALI search indicates that Ec78 PtuA most resembles the PtuA in type I Septu system¹⁸, with the RMSD of 3.3 Å and DALI score of 30.7. However, the NTD and coiled-coil region are unique in the Ec78 PtuA (Fig. 3d), indicating their potential role in adapting Ec78 function (discussed below). Moreover, the β-hairpin and α-hairpin in the Lobe I and Lobe II of type I PtuA protein involving in its higher oligomerization, are absent in the Ec78 PtuA, suggesting the distinct anti-phage strategies of type I and II Septu systems¹⁸. The Ec78 complex harbors four copies of PtuA, which are almost identical (Fig. 2c–f and Supplementary Fig. 5d). Two ATPase domains of Ec78 PtuA dimerize in a face-to-face manner, forming two ATP-binding pockets at the dimer interface. Both ATP-binding pockets contain an ATP molecule in unit A, while only one pocket is occupied in unit B and the other site remains open, resulting in a closed conformation of the PtuA dimer in unit A, and a relatively open conformation in unit B (Supplementary Fig. 5e). Within PtuA dimer, each ATP-binding pocket is cooperatively formed by the two PtuA protomers (Fig. 3d). Residues from the Walker A motif (also termed the P-loop) of one PtuA protomer and the signature motif (also known as the C-loop) of the opposing protomer are involved in substrate binding through extensive polar and hydrophobic interactions. Two catalytic acidic residues, Asp395 and Glu396, are located in the conserved Walker B motif. Moreover, a glutamine residue on the Q-loop typically accommodates the Mg²⁺ ion, which is required for ATPase activity in ABC-type ATPases²⁵ (Supplementary Fig. 5f). The corresponding site in the PtuA of the type I Septu system is occupied by a glycine residue (Supplementary Fig. 5g), resulting in its catalytic inactivity¹⁸. In contrast, Ec78 PtuA bears an asparagine at the equivalent site (Fig. 3d), and in vitro ATP hydrolysis assays demonstrated that Ec78 PtuA is catalytically active. Importantly, the ATPase rates of isolated PtuA and the assembled Ec78 complex are comparable (Fig. 3e), suggesting that ATP hydrolysis may not be the direct driver of effector toxicity. Moreover, one PtuB molecule was observed in the Ec78 complex, comprising an N-terminal HNH nuclease domain and a C-terminal domain. The N-terminal HNH domain adopts a relatively loose conformation, featuring a two-stranded antiparallel β-sheet (β1-β2) flanked by two α-helices (α2-α3) (Fig. 3f and Supplementary Fig. 6), and harbors a catalytic dyad composed of two conserved residues (His57 and Asn73). The C-terminal domain consists of a two-stranded β-hairpin and a four-helix bundle, which are responsible for interaction with PtuA.

Retron-mediated inhibition of PtuAB complex

Next, we investigated the assembly of the PtuAB effector complex and how it is neutralized by the Ec78 retron (Fig. 4). In contrast to the previously reported Ec86 system, which contains two complete copies of the effector in each filament segment¹³, the Ec78 complex harbors a single PtuAB effector complex in one of its units, and the other contains only a PtuA dimer without the associated PtuB endonuclease (Fig. 2c–f). Similar features were also observed in type I Septu system, where three PtuA dimers form a horse-shaped hexamer, but the PtuB is recruited to the two side PtuA dimers¹⁸. In the Ec78 complex, two α-helical bundle CTDs from the PtuA dimers flank the α5-α7 helices of the PtuB C-terminus, establishing extensive polar and hydrophobic interactions (Fig. 4a). Deletion of the PtuA CTD likely disrupts PtuB recruitment, impairs effector complex assembly, and consequently abolishes cytotoxicity (Fig. 4e).

Fig. 4. Key interfaces of Ec78 system are essential for functions.

a Close-up view of the PtuAB interface. Key interaction residues are shown in stick representation. b Detailed insight into the RT-msDNA-PtuA interface. Corresponding residues are shown as sticks. c Close-up view of the PtuA-PtuA interface. Residues involved in interactions are shown in stick representation. d The interfaces between the PtuA dimer and msDNA. Key residues that participated in the interaction are shown as sticks. e Bacterial growth spot assay of PtuAB effector complex with truncation mutations in PtuA. The RT-binding NTD (residues 1–62), coiled-coil region (residues 261–304) and PtuB-binding CTD (residues 478–550) were deleted. Point mutations were generated in the encircling loop and coiled-coil region. CFU per mL was calculated from colony counts of serial dilutions. Data are presented as mean ± SEM from three biological replicates. Source data are provided as a Source Data file.

To investigate how the effector complex is anchored by the retron, we analyzed the interactions between the retron and the PtuAB effector complex. Unlike the Ec86 retron system, which utilizes two msDNA molecules to cage the effector for inhibition, the Ec78 system features a long stem-loop structure of the msDNA embraced by two PtuA dimers. Additionally, the RT-msrRNA complex caps the side of the ATPase dimer. Together, the retron elements form two major interfaces with PtuA dimers to inhibit the Ec78 effector complex, including the msDNA-PtuA interface and the RT-msrRNA-PtuA interface (Fig. 4b, d).

In particular, the PtuA dimers in units A and B adopt a head-to-head configuration, with the A2 and B2 protomers contacting each other through the α6 helix and β9-β10 loop on the back side (Supplementary Fig. 7a, b). The front side remains open, forming a positively charged cleft approximately 95 Å in length and 30 Å in width that accommodates the msDNA (Supplementary Fig. 7c, d). Stretches of positively charged residues on Lobe I of the ATPase domain, especially within the β8-α6 encircling loop, directly engage the stem-loop structure of the msDNA (Fig. 4d). In addition, the N-terminal domain (NTD) of PtuA, comprising α1-α3 helices and a β1-β2 hairpin, faces the RT thumb domain along with the msrRNA (Fig. 4b and Supplementary Fig. 5a). Stretches of basic residues on the RT-binding NTD form extensive hydrogen-bonding interactions with the thumb domain and the msrRNA backbone (Fig. 4b). These structural observations indicate that both the RT-msrRNA and msDNA elements of the Ec78 retron contribute to its antitoxin function. To validate these findings, we tried to generate PtuAB and Ec78 mutants with a deletion in the RT-binding NTD and the msDNA-interacting encircling loop (Supplementary Fig. 7e). We first purified RT-binding NTD deletion of PtuA and confirmed its correct folding with SEC. However, we failed to purify the PtuA truncation mutant lacking the msDNA-interacting encircling loop, suggesting that the deletion caused misfolding of PtuA. We therefore introduced alanine substitutions at msDNA-interacting residues within the PtuA ATPase domain and expressed the corresponding PtuAB complexes in the Ec78 system in E. coli. Both the deletion of the NTD region and charge-swap mutations in the encircling loop abolished effector function, indicating that these structural features are essential for PtuAB toxicity (Fig. 4e). The retron neutralizes the effector complex by engaging these features, thereby suppressing its toxic activity. Taken together, both the RT–msrRNA and msDNA elements contribute to effector inhibition by anchoring functional regions of PtuA.

The coiled-coil regions in ABC-type ATPases have been reported to play a crucial role in complex assembly. For instance, the long coiled-coil regions in Rad50/SMC-family ATPases mediate intra-dimer interactions along their distal regions²⁶. In contrast, the coiled-coil region in the Aria protein of the PARIS defense system folds into a helical bundle and mediates hexamer assembly through inter-dimer contacts^27,28. Distinct from these, the two-helix coiled-coil region in PtuA-A1 protomer bridges with that in PtuA-B1 protomer, mediating inter-unit contacts through both polar and hydrophobic contacts (Fig. 4c). The hinge-like arrangement of the coiled-coil regions anchors the msDNA, further stabilizing the PtuA-msDNA assembly (Supplementary Fig. 7d). We therefore hypothesized that the assembly of this coiled-coil region in the Ec78 complex may mediate a unique auto-inhibition mechanism of the effector complex. Deletion of the coiled-coil region in PtuA had no impact on the Ec78 system, but remarkably reduced the toxicity of PtuAB complex (Fig. 4e and Supplementary Fig. 7f), indicating that the PtuA coiled-coil does not contribute to the Ec78 assembly, but is functionally important for PtuAB activity. Moreover, the back sides of the PtuA-A2 and PtuA-B2 protomers are in close contact, which prevents their coiled-coil regions from interacting with each other. As a result, these regions are only visible in the cryo-EM map at a relatively low threshold (Supplementary Fig. 7g). Together, both the msDNA and RT–msrRNA elements of the Ec78 retron are responsible for the inhibition of PtuAB, primarily through engagement of functionally important structural features on the PtuA subunit.

Activation mechanism of Ec78 effector complex

To further elucidate the molecular basis underlying the activation mechanism of Ec78 effector complex, we attempted to determine the structure of the PtuAB complex (Supplementary Fig. 8). We generated catalytically inactive mutants of the PtuAB effector complex to prevent toxicity and purified the complex for cryo-electron microscopy. Intriguingly, we could only obtain a partial reconstruction of PtuAB without ATP incubation, with PtuB flanked by the two-helix bundles of the PtuA dimer but lacking the ATPase domains of PtuA (Supplementary Fig. 9). By contrast, addition of ATP enabled reconstruction of the full PtuAB complex at high resolution, indicating that the ATP molecules are essential for stabilizing the architecture of the PtuAB complex. We therefore determined the structure of the ATP-bound PtuAB effector complex at a resolution of 2.70 Å (Fig. 5a, b and Supplementary Fig. 8). The overall architecture of the effector complex resembles that in the Ec78 system, and the interaction modes between PtuA and PtuB are almost identical (Supplementary Fig. 10a, b). However, the helix bundle in the NTD and the encircling loop in the ATPase domain were not resolved in the PtuAB complex, likely due to increasing flexibility (Fig. 5c).

Fig. 5. The key finger loop plays an important role in Ec78 system when transitioning from Ec78 complex to Ec78 effector complex.

a Cryo-EM map of Ec78 effector complex. b Atomic model of Ec78 effector complex. c Structure comparison between Ec78 effector complex in inhibition and activation states. The arginine-lysine finger loop is marked with a rectangular box. d Close-up view of the interaction between the arginine-lysine finger loop of PtuB and RT protein. Key residues are shown as sticks. e Bacterial growth spot assay of WT and mutant effector complex. Truncation mutation in the finger loop resulted in the loss-of-function in the effector complex. Data are shown as CFU counts, with values representing the mean ± SEM of three independent biological replicates. Source data are provided as a Source Data file. f Bacterial growth spot assay of Ec78 system variants bearing msDNA truncation mutations. Corresponding CFU counts are shown in (g) as mean ± SEM from three biological replicates. Source data are provided as a Source Data file.

Notably, a unique arginine-lysine finger loop region (β2-α3, residues 78-90) on the PtuB nuclease exhibited increasing flexibility and could not be resolved in the PtuAB effector complex (Fig. 5c and Supplementary Fig. 10c). In contrast, this loop region could be clearly modeled according to high-quality EM density in the Ec78 complex (Fig. 5d). In the inhibition state, the loop II on RT protein lifts the arginine-lysine finger loop on the PtuB through direct contacts. Specifically, D197 on the RT loop hydrogen bonds with Arg82 and Lys83 on the β2-α3 loop of PtuB, maintaining its extended configuration. The order-to-disorder transition in this dynamic loop indicates its regulatory role in modulating effector activity (Fig. 5c, d). Truncation of the arginine-lysine finger loop or substitution of these two basic residues abolished effector-mediated growth arrest, confirming its essential role in PtuAB function (Fig. 5e). Taken together, the arginine-lysine finger loop in PtuB is functionally important and may act as a molecular “switch” governing the transition between effector inhibition and activation.

Previous studies have shown that msDNA serves as a sensor of phage infection. For instance, retron-Sen2 is activated by phage-encoded nucleases that degrade msDNA^15,20, suggesting that reduction in msDNA length can serve as a molecular trigger for retron-associated TA systems. To test this in Ec78, we generated a series of stem-loop truncation mutants (22, 16, 10, 6, 5, and 0 bp; Fig. 5f). The truncated constructs showed a progressive drop in CFU, with variants bearing stem-loops of 10 bp or less reducing CFU by over 10-fold compared to wild type (Fig. 5g), supporting the msDNA-length triggered activation mechanism. Conversely, extending the msDNA stem-loop in length also resulted in growth arrest (Supplementary Fig. 10e), reinforcing the importance of precise msDNA length. We further identified a loop region in Ec78 RT (α6–α7 loop, residues 135–155) that contacts the distal end of the msDNA and appears to sense its length through electrostatic interactions. We therefore refer to it as the “sensing loop” (Supplementary Fig. 10d). Despite repeated cloning attempts, deletion variants of this loop could not be obtained, likely owing to misfolding or toxicity, underscoring its importance. Moreover, the mutation K141D/R142D within the loop had little effect on toxicity, indicating that inhibition of PtuAB toxicity by the sensing loop likely depends on its overall structural integrity rather than single side-chain contacts (Supplementary Fig. 10f). Together, these findings support a model in which reduction of msDNA length serves as the activation trigger of Ec78 system.

To better understand the mechanism of PtuAB-mediated cytotoxicity, we next focused on the HNH nuclease activity of PtuB to biochemically characterize the PtuAB-mediated cytotoxicity. Expression of the PtuAB complex in cells led to a pronounced decrease in tRNA^Tyr levels (Fig. 1f), prompting us to purify tRNA^Tyr and perform in vitro cleavage assays. However, the PtuAB complex did not cleave tRNA^Tyr in vitro (Supplementary Fig. 11a). Most well-characterized HNH nucleases target DNA substrates, and some specific family members can act on RNA²⁹. We further assayed PtuAB cleavage activity on a panel of DNA and RNA substrates (Supplementary Fig. 11b–f). We also performed ion-substitution assays by replacing Mg²⁺ with other metal ions, but no cleavage activity was detected (Supplementary Fig. 12). Nonetheless, no cleavage activity was observed under any tested conditions, and the direct molecular target of the Ec78 PtuAB complex therefore remains elusive. We therefore hypothesize that PtuAB triggers tRNA^Tyr depletion indirectly, possibly via a more complex mechanism or by requiring additional cellular cofactors.

Discussion

Retrons were discovered more than 40 years ago, but their biological functions remained unclear for decades². Recent studies have revealed that retrons act as anti-phage defense systems, typically through abortive infection (Abi) strategies^1,30. Retrons are characterized by associated defense components encoded within their gene cassettes and can be grouped into approximately 13 types based on their associated proteins or domains, reflecting considerable mechanistic diversity⁴. However, only a few systems have been well studied^1,15,24. The type I-A retron systems are uniquely characterized by a dual-component effector complex¹⁷. In this study, we resolved high-resolution cryo-EM structures of both the Ec78 system and its isolated PtuAB effector complex. These structures, together with biochemical and mutational analyses, reveal the molecular basis of Ec78 system assembly, effector activation, and phage infection sensing. Our findings support the paradigm that retron systems function as TA systems, in which the retron serves as the antitoxin that neutralizes the toxic activity of associated effectors.

In the Ec78 effector complex, PtuA and PtuB assemble into a 2:1 stoichiometric unit (Fig. 5a, b), in contrast to the 6:2 PtuA:PtuB polymer observed in type I Septu systems¹⁸. Structural analysis shows that PtuA dimers flank PtuB on both sides, forming a compact and cooperative assembly. Both ATPase and nuclease activities are indispensable for cytotoxicity, indicating that PtuA and PtuB act in concert to execute effector toxicity in anti-phage defense. Notably, Ec78 PtuA has evolved an NTD for RT binding and an encircling loop for msDNA binding, playing critical roles in both effector function and retron-mediated inhibition.

Cryo-EM analysis revealed that the predominant particles of the Ec78 complex adopt a flower basket-like architecture, in which two Ec78 retrons neutralize one complete PtuAB complex and a PtuA dimer (Fig. 2c–f). We also observed particles containing solely retrons and two PtuA dimers (Supplementary Fig. 2c). This finding suggests a dynamic, stepwise assembly process of the Ec78 complex, in which PtuA dimers are first recruited by the retron and subsequently flank PtuB to complete the effector complex (Fig. 6). In addition, we observed Ec78 assemblies containing either a single PtuB on the other PtuA dimer or two PtuB subunits on both the PtuA dimers, indicating the presence of multiple, biologically relevant assembly states (Supplementary Fig. 2c).

Fig. 6. Ec78 retron associates with PtuAB to inhibit the activity.

Disruption of msDNA integrity triggers the release of the PtuAB effector complex, leading to cell death.

In line with earlier studies, the Ec78 complex employs both RT and msDNA components to inhibit effector activity. However, the effector composition, as well as the detailed structural basis for effector inhibition across retron systems, is notably distinct. In Ec86, the effectors mediating NAD⁺-hydrolyze are caged within msDNA molecules. The Ec86 retron, together with the effectors, forms extended filaments. Phage-triggered msDNA modification promotes filament disassembly and activation. By contrast, Sen2 relies on an accessory RcaT nuclease as the toxin, which is unleashed when phage nucleases degrade msDNA. Ec78 instead uses a dual-component ATPase–HNH effector complex, PtuAB, whose activity is restrained by msDNA/RT contacts. In the Ec78 complex, a nail-shaped msDNA stem-loop is embraced by two PtuA dimers and the shortening of msDNA in length triggers the release and activation of PtuAB. Moreover, although the catalytic dyad of the PtuB nuclease is solvent-exposed in the Ec78 complex, suggesting a potentially active conformation, a dynamic arginine–lysine loop essential for PtuB activity is fixed by the RT palm domain, thereby suppressing nuclease activity (Fig. 5d).

In the Ec78 complex, a nail-shaped stem-loop of msDNA is embraced by two PtuA dimers (Fig. 4a), in contrast to the Ec86 filament, where two msDNA molecules cage the effector dimer inside. The RT and msDNA elements tightly contact the PtuA dimers, capping their unique structural features essential for effector toxicity, thereby neutralizing the effector complex in an inhibited state (Fig. 4b, d). Moreover, the catalytic dyad of the PtuB nuclease is solvent-exposed in the Ec78 complex, probably indicating an active state. However, a dynamic arginine-lysine loop, essential for PtuB activity, is fixed by the RT palm domain, thereby suppressing the PtuB nuclease activity (Fig. 5d).

msDNA modification has been reported as a common strategy for toxin activation. Building on previous evidence that the phage-encoded nuclease may serve as an activator of the Ec78 system²⁰, we found that introducing msDNA truncation mutations into the Ec78 complex resulted in growth arrest in E. coli, highlighting the importance of msDNA in Ec78 function and supporting a potential activation mechanism through msDNA degradation (Fig. 5f, g). More interestingly, we identified a unique loop in the RT palm domain that senses msDNA length (Supplementary Fig. 10d). This sensing loop recognizes the distal end of the msDNA stem-loop and likely functions as a molecular checkpoint that monitors msDNA integrity, stimulating the release of the effector and facilitating toxicity activation upon phage infection. Upon release of the effector complex from the Ec78 retron, the positively charged structural features on PtuA and the arginine-lysine finger loop on PtuB undergo order-to-disorder transitions, which likely enables target engagement (probably tRNA as previously reported²⁰), thereby activating the toxin (Fig. 6). Altogether, our study not only provides in-depth insights into retron-mediated defense mechanisms but also paves the way for the future development of retron-based tools in biotechnology applications.

Methods

Plasmid construction

The genomic locus of the Ec78 system (GenBank: CP026202.1) including the ncRNA, RT, PtuA, and PtuB coding genes, was synthesized (Sangon) and cloned into appropriate vectors. The co-expression plasmids for Ec78 system, excluding the ncRNA (RT, PtuA and PtuB) were cloned into 2HR-T vector (Addgene #29718) using the native genomic locus sequence, with a His-Strep tag attached to the N-terminus of RT, all under the control of a single T7 promoter. The ncRNA coding gene was cloned separately into the 13S-A vector (Addgene #48323). These two plasmids were co-transformed to express the full Ec78 system complex. For confocal fluorescence microscopy, tRNA^Tyr was cloned into the 13S-A vector.

The Ec78 system plasmid and the plasmid containing Ec78 effector complex were both inserted into pACYCDuet-1 (Addgene #128837) and pBAD LIC vectors (Addgene #37501), and were independently transformed into E. coli BL21 (DE3) cells for further experiments. Truncation mutants were generated by deleting the indicated residues, leaving the flanking residues directly joined. All mutations were generated by QuickChange mutagenesis (Takara) following the manufacturer’s instructions. The sequences of nucleic acids in this study were provided in Source Data.

The msDNA extraction and sequencing

The purified Ec78 complex was treated with proteinase K and then incubated with RNase A to digest RNA. The purified ssDNA was processed using the Vazyme ssDNA Library Prep Kit (ND620) following the protocol to prepare the sequencing library. The ssDNA sequencing was conducted at GENEWIZ.

Protein expression and purification

The BL21(DE3) competent cells were transformed with expression plasmids and plated on LB agar containing appropriate antibiotics. Single colonies were picked and inoculated into 5 mL of starter culture, grown overnight at 37 °C, and subsequently transferred to a 1 L culture. Protein expression was induced with 0.2 mM isopropyl-β-D-thiogalactoside (IPTG) when the optical density at 600 nm reached ~0.6. Cultures were grown at 16 °C with orbital shaking at 160 rpm for 16 h.

Cells were harvested by centrifugation at 4 °C for 10 min, and the resulting pellets were resuspended in binding buffer (25 mM HEPES pH 8.0, 500 mM NaCl, 2 mM β-mercaptoethanol). The cell suspension was lysed by sonication, and the lysate was centrifuged to remove debris. The supernatant was incubated with Ni-NTA or Strep-Tactin, depending on the tag of the target protein. Proteins were eluted with binding buffer supplemented with 300 mM imidazole or 5 mM biotin, respectively. Wild-type PtuAB was isolated from the Ec78 complex, in which the HNH protein carried a C-terminal Strep tag, using a HiTrap HP column.

The eluate was further purified by size-exclusion chromatography, with or without prior ion-exchange chromatography. Fractions were analyzed by SDS-PAGE to confirm protein components and by Urea-PAGE to detect nucleic acid components when present. Peak fractions with the correct size of proteins and/or nucleic acid complexes were concentrated, aliquoted, and stored at −80 °C prior to use.

ATPase assay

The ATPase assay was performed in a reaction buffer containing 25 mM Tris-HCl pH 8.0, 200 mM NaCl, 10 mM MgCl_2, and 2 mM DTT. Purified target proteins were incubated with 200 µM ATP at a final concentration ratio of 1:20 in the reaction buffer. The total volume of the reaction was 80 µL. The mixture was then incubated at 37 °C for 60 min. Malachite green reagent was added to enable colorimetric detection of released inorganic phosphate. Absorbance at 620 nm was detected using a microplate reader (BioTek Synergy H1 hybrid). The data are presented as mean ± standard error of the mean (SEM).

Spot assay

Both the Ec78 system plasmid and the plasmid containing Ec78 PtuAB complex were constructed using a native promoter, and the two recombinant plasmids were transformed into E. coli for amplification. Cells were cultured until the optical density at 600 nm reached ~0.3. After equal proportion dilution, these dilutions were spotted onto LB agar plates containing the appropriate antibiotic. The plates were incubated overnight at 37 °C and imaged using a gel imager (Azure Biosystems). Bacterial growth was quantified by counting colony-forming units (CFUs).

For msDNA mutations, the plasmids used the pBAD promoter and were transformed into E. coli. Individual colonies were grown in the liquid culture containing 0.4% glucose and antibiotics, and spotted onto LB agar plates in the presence of 0.2% arabinose and appropriate antibiotics.

Bacterial growth assay

E. coli cells with the Ec78 system, mutations, and Ec78 PtuAB complex were cultured at 37 °C. Then, 180 μL of culture was mixed with 20 μL of cells, and the mixture was transferred into a 96-well plate with the appropriate antibiotic and incubated at 37 °C with shaking. The OD₆₀₀ of the cells was measured using a microplate reader (Biotek) every 15 min for a total duration of 4 h.

Cryo-EM sample preparation and imaging

For both Ec78 system complex sample and Ec78 effector complex sample, an aliquot of 3.5 μL purified target protein complex at 1–2 mg/mL was applied to glow-discharged Au R1.2/1.3 holey carbon grids (300 mesh, Quantifoil). After a 5 s incubation, grids were blotted for 2 s at a blot force setting of 2. Then the grid was plunge-frozen into liquid ethane, which was pre-cooled by liquid nitrogen, using Vitrobot Mark IV (FEI Thermo Fisher) at 4 °C and 100% humidity.

Both the datasets of the Ec78 system complex and the Ec78 effector complex were collected on a FEI Titan Krios equipped with a Falcon4 detector at an acceleration voltage of 300 kV and at a nominal magnification of 165,000×, with calibrated physical pixel sizes of 0.75 and 0.81 Å/pixel. Cryo-EM images were collected by SerialEM³¹ or EPU³² with the defocus range of −1.2 to −1.6 μm. Each micrograph was dose-fractionated into 32 frames. The total dose was 50.0 e^-/Ų per micrograph. The dataset of the ATP-free Ec78 effector complex was collected using the same strategy as described above.

Cryo-EM data processing

For the dataset of Ec78 system complex, 5499 micrographs were collected and subsequently processed in CryoSPARC³³, with the beam-induced motion by MotionCor2³⁴ and CTF estimation³⁵. Micrographs with an estimated CTF resolution higher than 5 Å were excluded, resulting in 4,892 micrographs retained for further analysis. Particles were picked independently using blob picking, template picking and Topaz picking³⁶, respectively. After removing duplicate particles from the three particle stacks, a final set of 1,557,724 particles was generated. Then multiple rounds of 2D classification were carried out, yielding 400,145 promising particles for ab-initio reconstruction. After obtaining volumes, multiple rounds of heterogeneous refinement and 3D classification were employed to refine these volumes. Particles re-extraction, local CTF refinement and non-uniform refinement were performed to improve the volume quality, and the final resolution of the best volume was 3.03 Å.

For the dataset of Ec78 effector complex, 7987 micrographs were collected and processed in CryoSPARC. Motion correction and CTF estimation were performed, followed by manual curation of exposures. 7937 micrographs were used for particle picking utilizing blob picking and template picking. Particles with protein features were selected to train a Topaz model. After Topaz extraction, 1,681,139 particles were picked. Multiple rounds of 2D classification were then performed to eliminate junk particles. Ab-initio reconstruction generated the initial volumes, and heterogeneous refinement was further conducted to screen good particles. Next, another round of Ab-initio reconstruction and non-uniform refinement was carried out, generating the final map at 2.70 Å. Data for the ATP-free Ec78 effector complex were processed following the same strategy as above.

Model building and refinement

For the Ec78 system complex, the initial atomic models of retrons, PtuA and PtuB subunits predicted by AlphaFold3³⁷ were manually fitted into the cryo-EM density map using ChimeraX³⁸. For modeling of the nucleotides, docked subunits were used to segment the density map, and the nucleotides were built into the remaining density manually in Coot³⁹. Then the nucleotides were manually refined to trace the density and mapped to the sequence. Further refinements were performed by phenix.real_space_refine⁴⁰. For the Ec78 effector complex, the protein subunits from the refined structure of Ec78 system complex were separately docked into the EM density in ChimeraX. Further iterative refinements were performed by phenix.real_space_refine. The quality of all the final models was validated by MolProbity in Phenix⁴¹. Data collection and model refinement statistics are concluded in Supplementary Table 1.

RT-qPCR

Cells expressing Ec78 or PtuAB were cultured in LB medium at 37 °C until the OD₆₀₀ reached 0.4. Arabinose and glucose were added to final concentrations of 0.2% and 0.4%, respectively. Cultures were incubated for 30 min. For the Ec78 defense system, T5 phage was added at an MOI of 2 and infection continued for 30 min, then cells were collected for total RNA extraction.

Total RNA was extracted using RNApure Bacteria Kit (CoWin Biosciences). cDNA was synthesized using StarScript III RT SuperMix (Genstar). Finally, 8 ng of cDNA was used for RT-qPCR reaction using RealStar Fast SYBR qPCR Mix (Genstar). Expression levels were normalized to 5S rRNA. Each condition included at least three replicates.

In vitro nucleic acid cleavage assay

For the in vitro tRNA transcription assay, a home-made T7 RNA Polymerase was used for transcription. The reaction was performed in a buffer containing 100 mM HEPES-K pH 7.9, 20 mM MgCl₂, 30 mM DTT, 2 mM spermidine, 2.5 mM NTPs. In addition, 250 µM Cy5-labeled UTPs were incorporated into the transcription reaction to enable visualization. For tRNA refolding, tRNAs were denatured at 95 °C for 5 min and then at 65 °C for 1 min, then allowed to slowly cool to room temperature in a buffer containing 10 mM MgCl₂.

An in vitro nucleic acid cleavage assay was performed in a buffer containing 25 mM Tris-HCl pH 8.0, 200 mM NaCl, 5 mM MgCl₂, 5% glycerol, 2 mM DTT. ssDNA, dsDNA, overhang dsDNA, ssRNA and dsRNA were synthesized by Genewiz and labeled with 5’-Cy3. For testing different metal ions, MgCl₂ in the buffer was replaced with 1 mM ZnCl₂ or 5 mM MnCl₂ in separate reactions. Proteins were incubated with 20 nM nucleic acid substrate at a ratio of 1:10 at 37 °C for 1 h. The reaction was stopped by adding 25 mM EDTA and 1 µL proteinase K (20 mg/mL). Samples were loaded onto 15% urea-PAGE gel for visualization.

Confocal fluorescence microscopy

Cells were cultured and induced at 37 °C until OD₆₀₀ reached 0.4. After collection, cells were washed three times with PBS. Subsequently, cells were stained with DAPI (50 µg/mL) and FM4-64 (5 µg/mL) to visualize the nucleoid and membrane, respectively. A total of 5 µL of cells was plated onto an agarose pad. Images were acquired on a Zeiss LSM 800 confocal microscope.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization: Z.Y., H.Zhu, and H.Z. Experimental studies: Q.H., Y.L., B.L., Z.L., H.C., S.Z., J.H., and J.Y. Data analysis: X.L., Q.H., Y.L., Z.L., S.Z., J.H., Y.L., J.Y., H.Y., Z.G., Y.W., H.Zhu, and H.Z. Supervision: H.Z. Manuscript writing: X.L., Z.L., and H.Z. with contributions from all authors.

Peer review

Peer review information

Nature Communications thanks Yue Feng and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

Sequencing data have been deposited in NCBI (BioProject: PRJNA1332408). The atomic coordinates have been deposited in the Protein Data Bank under accession codes 9L7P (Ec78 system complex) and 9U9Y (Ec78 effector complex). Cryo-EM maps have been deposited in the Electron Microscopy Data Bank under corresponding accession codes EMD-62876 and EMD-63972. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-67175-9.