Classification of Sir2–HerA systems reveals a multilayered regulatory cascade gating the type III antiphage activity

Abstract

Sir2–HerA systems are abortive infection defenses that integrate multiple enzymatic activities to induce growth arrest, thereby limiting phage propagation. However, the molecular logic that couples phage sensing to a precisely gated antiviral response, thereby ensuring infection-specific activation, remains unclear. Through genomic mining of Escherichia coli, we show that E. coli Sir2–HerA immunity diversifies into functional subtypes and identify three types (I–III) with distinct protection profiles. Focusing on the most potent type III system, we identified its phage activators, Gp2.5 and Gp5.9, through an unbiased T7 proteome screen. Mechanistically, type III Sir2–HerA follows a multilayered gating logic. Phage proteins trigger both the HerA nickase and Sir2 NADase; however, the activated NADase remains dormant due to an ATP-mediated checkpoint. The HerA nickase introduces DNA nicks, leading to the accumulation of end-exposed DNA intermediates that engage the complex-associated ATPase to drive ATP consumption. This process relieves the checkpoint, thereby unleashing the full trigger-dependent NADase activity and enabling robust NAD⁺ depletion. Together, our findings reveal a sophisticated molecular logic that integrates diverse enzymatic activities into a tiered gating architecture, ensuring high-fidelity phage defense while preventing inadvertent activation.

Graphical Abstract

Graphical Abstract.

Introduction

Phages are a major cause of bacterial mortality across diverse environments, imposing strong selection for a wide range of antiviral defenses [1]. Bacterial antiphage immunity is commonly organized into layered strategies that act at successive stages of infection, including mechanisms that prevent phage adsorption [2–4], impede genome delivery [5], or recognize and destroy invading phage genomes [6, 7]. Beyond these interception and clearance strategies, recent work has revealed a broad and rapidly expanding repertoire of abortive infection systems that suppress phage propagation by terminating infection before progeny assembly is complete, often by triggering death of the infected cell [8–14]. Although abortive infection can provide powerful population-level protection, it poses a substantial risk to host viability, as unintended activation would be catastrophic for uninfected cells.

Accordingly, to safeguard the host from unintended cell death, many abortive infection systems employ multilayered regulatory architectures that strictly couple infection sensing to downstream toxic effectors, thereby ensuring lethal activities remain quiescent during normal growth and are activated only upon phage infection [15–19]. For example, in CBASS and Thoeris systems, infection activates dedicated sensor modules that generate sufficient levels of small-molecule second messengers, which subsequently engage downstream effectors to execute abortive infection [8, 20, 21]. A common terminal route to abortive infection is the rapid collapse of nicotinamide adenine dinucleotide (NAD⁺) metabolism: many defense systems encode NAD-consuming enzymes whose activation disables essential cellular processes and thereby halts phage reproduction [8, 9, 22, 23]. However, how NAD-consuming activity is precisely gated during infection remains poorly defined for multiple NAD-consuming systems, including the Sir2–HerA [22–24].

Recent studies have elegantly established the Sir2–HerA system, also referred to as Nezha, as a sophisticated multi-enzyme abortive infection defense that integrates NADase, ATPase, helicase, and nuclease activities [22, 24–26]. This mechanistic framework makes Nezha a tractable model for dissecting how complex enzymatic defense system is gated during infection. During phage infection, Escherichia coli cells carrying Nezha exhibit a rapid decline in intracellular ATP, whereas ATP levels remain comparatively stable in cells lacking the system [22]. This infection-associated ATP depletion is proposed to relieve ATP-mediated inhibition of NADase, helicase, and nuclease activities, thereby restricting phage replication through abortive infection [22]. Despite this progress, several central aspects of Sir2–HerA system remain unresolved. First, the phage-derived triggers that initiate Sir2–HerA activation during phage infection are largely undefined, precluding a molecular understanding of how the system distinguishes viral intrusion from host metabolic noise. Second, the hierarchical coordination between these integrated enzymatic activities, as well as the regulatory circuitry that couples them to ensure a precise, infection-linked NAD⁺ collapse, remains elusive [22, 24–26]. Third, it remains unknown whether Sir2–HerA similarly forms discrete functional subtypes within a single bacterial species, a diversification pattern frequently observed in other defense systems such as Thoeris, CBASS, and TIR [10, 27, 28].

In this work, we systematically mine E. coli genomes and reveal that Sir2–HerA immunity is not a singular entity but a subtype-diversified family beyond the previously described Sir2–HerA/Nezha system. These types exhibit distinct antiphage spectra, indicating hitherto unrecognized functional diversification, exemplified by the type III system as the most potent subtype. While sharing the conserved ATPase and NADase activities of previously described Sir2–HerA/Nezha, type III Sir2–HerA uniquely incorporates a specialized nickase, rather than the canonical helicase and nuclease, to drive its gating logic. Focusing on this subtype, we identify phage single-stranded DNA-binding proteins (SSB) as universal triggers that can be sensed across phages. We further show that Sir2–HerA follows a multilayered gating logic in which a phage SSB trigger concurrently activates both the HerA nickase and Sir2 NADase. However, the triggered Sir2 NADase activity is initially masked by an ATP-mediated checkpoint. The HerA nickase generates DNA nicks, leading to the accumulation of end-exposed DNA intermediates that engage the complex-associated ATPase to drive ATP consumption. This in turn relieves ATP-mediated inhibition of pre-activated Sir2 and potentiates trigger-dependent Sir2 NADase activity to drive NAD⁺ depletion.

Our findings establish Sir2–HerA as a diversified E. coli defense family that senses phage replication proteins and executes abortive infection through a sequential, ATP-checkpointed gating program, thereby providing a general framework for how multi-enzyme NAD-consuming systems achieve potent antiviral immunity while preventing inadvertent activation.

Materials and methods

Bacteria and phages

Escherichia coli strain DH5α (hsdR17(rK- mK+) sup2) was used to construct plasmids. Escherichia coli BL21 (DE3) [F⁻ ompT hsdSB(rB⁻ mB⁻) gal dcm (DE3)] was used for protein expression. Escherichia coli MG1655 (F⁻ λ⁻ rph-1) was used to identify the antiphage activities of Sir2–HerA defense systems. A collection of 816 E. coli strains from our laboratory stock was used to clone candidate defense systems containing Sir2–HerA. Phages T1, T4, and T7, together with 107 additional E. coli phages previously isolated on E. coli MG1655, were used to identify the antiphage activities of each candidate system [5].

Plasmid construction

To clone candidate Sir2–HerA defense systems, we used a collection of 816 E. coli isolates as templates to polymerase chain reaction (PCR) amplify Sir2–HerA target fragment. The PCR products were cloned into the empty vector and transformed into E. coli MG1655 strains. The primers used to amplify the Sir2–HerA systems are provided in Supplementary Table S1. To co-express the Sir2–HerA complex, the HerA and Sir2 were cloned as a polycistronic cassette into the pET28a vector under control of the T7 promoter, with HerA carrying an N-terminal 6 × His tag. For single-protein expression, Sir2 and Gp2.5 were each cloned into the pET28a vector with an N-terminal 6 × His tag followed by a tobacco etch virus (TEV) protease recognition site. HerA was expressed alone from the pET32a vector. For pulldown assays, plasmids encoding untagged Gp2.5 and C-terminally Strep-tagged Gp2.5 were constructed using the pET28a vector. Gp5.9 was also constructed with a C-terminal Strep tag in the same vector. All mutant variants were generated by site-directed mutagenesis.

For the antiphage activity assays, the Sir2–HerA locus together with its native promoter was cloned into the expression plasmid. For toxicity assays of phage genes, the phage trigger genes were cloned into the pBAD vector under the control of the arabinose-inducible promoter. For inter-system communication assays, chimeric Sir2–HerA systems were constructed by swapping partners between different systems. Specifically, plasmids carrying the mutated herA gene from Nezha system were co-transformed with plasmids carrying the mutated sir2 gene from Sir2–HerA Ⅲ system into E. coli MG1655, thereby reconstituting recombinant Sir2–HerA systems.

Identification of the Sir2–HerA candidate defense systems in E. coli

The laboratory previously analyzed all 2289 E. coli genomes available in the NCBI database as of October 2020 [29]. Protein-coding sequences were predicted using Prodigal (release 2.6.3) [30], and protein domains were annotated with HMMER version 3.3 [31]. Proteins annotated as containing a Sir2 domain were retrieved from the NCBI database. In total, 2892 Sir2-domain-containing proteins were collected, and their amino acid sequences were extracted for subsequent clustering. Sequence clustering was performed using CD-HIT with a sequence identity threshold of 0.7 (-c 0.7) and a word size parameter of 5 (-n 5), resulting in 58 clusters. To explore potential operonic structures, the genomic context of each Sir2-domain-containing gene was examined by analyzing the three upstream and three downstream genes. Protein domains of these neighboring genes were predicted using the GenomeNet Motif Search server (https://www.genome.jp/tools/motif/) to infer their potential functions. This analysis identified HerA genes adjacent to Sir2-domain-containing genes in eight of the 58 clusters, suggesting that Sir2 and HerA may constitute candidate defense systems for further evaluation. The amino acid sequences of HerA proteins from these candidate systems were then extracted and clustered using CD-HIT with the same parameters to classify HerA proteins into distinct clusters. Structural models of the Sir2 and HerA domain proteins were generated using AlphaFold [32] and analyzed in PyMOL. Structural alignments were performed using the align command.

Validation of the candidate Sir2–HerA defense systems

Phages previously propagated on E. coli MG1655 were tenfold serially diluted in Pi-Mg buffer. Test strains were inoculated 1:100 into 10 ml fresh lysogeny broth (LB) containing the appropriate antibiotic and grown overnight at 37°C with shaking at 180 revolutions per minute (rpm). Tryptic soy broth (TSB) soft agar was melted, cooled to 60°C, and mixed with 500 μl of bacterial culture and 8 ml soft agar supplemented with antibiotic, then poured onto solid agar plates. After solidification, 2.5 μl of each serially diluted phage suspension was spotted onto the surface and plates were incubated at 37°C overnight for plaque enumeration. Using MG1655 as the control, the efficiency of plating was determined for strains carrying the Sir2–HerA system.

Sequence identity analysis of Sir2 and HerA proteins

Amino acid sequences of Sir2 and HerA from Nezha and the types I–III Sir2–HerA systems were obtained from the corresponding gene annotations and saved in FASTA format. Sir2 and HerA were analyzed separately. Sequences were imported into DNAMAN (v6) and aligned using the built-in multiple sequence alignment function with default settings for protein sequences. Pairwise amino acid sequence identities were calculated in DNAMAN for all sequence pairs within each protein family. The resulting pairwise identity matrices were exported and assembled in GraphPad Prism 10, with percent identity values displayed in each cell. For the composite heatmap, Sir2 identity values were plotted on one axis and HerA identity values on the other, and the color scale indicated amino acid sequence identity (%).

Phylogenetic analysis

From the above screening, we selected 41 Sir2 proteins that exhibited phage resistance, together with previously reported Sir2 proteins from E. coli (NCBI Reference Sequence: WP_021577683.1) [22, 24], Paenibacillus sp. 453MF (NCBI Reference Sequence: WP_091160507) [25], and Staphylococcus aureus (NCBI Reference Sequence: WP_000513494.1) [26]. The amino acid sequences were aligned using MUSCLE (v3.8.1551) with default parameters [33]. The phylogenetic tree was then inferred with FastTree (v2.1.10) [34] and visualized using the iTOL online tool (v5) [35].

Bacterial growth curve assay

Strains carrying the Sir2–HerA defense system or the empty vector control were streaked onto LB agar plates supplemented with the appropriate antibiotics and incubated at 37°C for 12–16 h. Single colonies were then inoculated into 10 ml LB broth and grown overnight at 37°C with shaking at 180 rpm. Subsequently, 100 μl of the overnight culture was inoculated into 10 ml fresh LB medium containing kanamycin and incubated with shaking at 37°C, 180 rpm, until the cells reached the logarithmic growth phase. Phages were diluted to the desired concentrations corresponding to the indicated multiplicities of infection (MOIs). For the growth curve assay, 50 μl of the diluted phage suspension was added to each well of a 100-well plate, while 50 μl Pi-Mg buffer was added to control wells. Then, 150 μl of the log-phase E. coli culture, with the Sir2–HerA system or the empty vector control, was added to each well and gently mixed. The 100-well plate was placed in an automated growth-curve analyzer (Bioscreen C) set to 37°C, and OD₆₀₀ was measured at 10 min intervals. Bacterial growth curves were finally plotted as OD₆₀₀ versus time post infection using GraphPad Prism 10.

Toxicity assays of phage genes in the presence of the Sir2–HerA systems

Plasmids encoding the corresponding Sir2–HerA systems or the empty vector control were co-transformed together with the wild-type or mutant phage triggers into E. coli MG1655 and incubated on plates at 37°C in an inverted position for 12–16 h. Single colonies were inoculated into 10 ml LB broth and grown at 37°C with shaking at 180 rpm for 8–10 h. To assess the toxicity of the phage proteins on solid medium, 400 μl of the fresh E. coli MG1655 culture was harvested by centrifugation at 6000 rpm for 5 min, washed once with 400 μl phosphate buffered saline (PBS), and resuspended in 400 μl PBS. Ten-fold serial dilutions were prepared, and 5 μl of each dilution was spotted onto tryptic soy agar (TSA) plates with or without 0.2% arabinose. Plates were incubated at 37°C overnight to examine bacterial colony formation.

Protein expression and purification

Recombinant Sir2–HerA, Sir2, HerA, Gp2.5, and Gp5.9 proteins were expressed in E. coli BL21 (DE3). Overnight cultures grown in LB medium containing the appropriate antibiotic were inoculated 1:100 into fresh LB with antibiotic and incubated at 37°C with shaking at 180 rpm until the OD₆₀₀ reached 0.6–0.8. Protein expression was induced by the addition of 0.3 mM isopropyl β-D-1-thiogalactopyranoside (IPTG), after which cultures were shifted to 16°C and incubated for a further 12 h with shaking at 140 rpm. Cells expressing His-tagged proteins were harvested by centrifugation at 8000 rpm for 10 min at 4°C, resuspended in binding buffer (25 mM Tris-HCl, 150 mM NaCl, pH 8.0) and disrupted at 4°C using a high-pressure homogenizer. Cell debris was removed by centrifugation at 30 000 × g for 20 min at 4°C, and the clarified supernatant was loaded onto the HisSep Ni-NTA 6FF chromatography column pre-equilibrated in binding buffer. The column was washed with washing buffer (25 mM Tris-HCl, 150 mM NaCl, 20 mM imidazole, pH 8.0) and target proteins were eluted with elution buffer containing 250 mM imidazole. Proteins expressed from the pET28a vector were incubated with TEV protease to remove His tag, whereas proteins expressed from the pET32a vector were incubated with enterokinase to remove the TrxA fusion tag. After digestion, the reaction mixtures were passed through a HisSep Ni-NTA 6FF chromatography column to deplete residual tags and proteases, and the flow-through containing the cleaved target proteins was collected. The proteins were further purified by gel-filtration column on HiLoad^TM 16/600 Superdex^TM 200 pg (Cytiva) and Superose^TM 6 Increase 10/300 GL column (Cytiva) in a buffer consisting of 25 mM Tris-HCl, 150 mM NaCl, pH 8.0.

Strep-tagged Gp2.5 and Gp5.9 were purified using Strep(II)Sep 4FF affinity chromatography. Cells expressing Strep-tagged Gp2.5 or Gp5.9 were harvested, resuspended in binding buffer (100 mM Tris-HCl, 150 mM NaCl, pH 8.0), and disrupted at 4°C using a high-pressure homogenizer. Cell debris was removed by centrifugation at 30 000 × g for 20 min at 4°C, and the clarified supernatant was loaded onto a Strep(II)Sep 4FF chromatography column pre-equilibrated. The column was washed with the same buffer, and target proteins were eluted with elution buffer containing 2.5 mM desthiobiotin (100 mM Tris-HCl, 150 mM NaCl, 2.5 mM desthiobiotin, pH 8.0). Mutant proteins were expressed and purified using the same procedure as for the corresponding wild-type proteins.

Pulldown assays

The E. coli strains carrying plasmids encoding Sir2–HerA (6 × His-tagged) or T7 Gp2.5 (untagged) were grown in separate cultures to mid-log phase and induced with 0.3 mM IPTG at 16°C for 12 h with shaking at 140 rpm. Cells were harvested by centrifugation and resuspended in binding buffer (25 mM Tris-HCl, 150 mM NaCl, pH 8.0). Each cell suspension was disrupted separately at 4°C using a high-pressure homogenizer. The lysates were then combined, and the mixture was clarified by high-speed centrifugation to remove insoluble material. The resulting supernatant was further clarified by filtration through a 0.22 μm membrane filter prior to chromatography. The filtered lysate was loaded onto a Ni-NTA column to capture His-tagged Sir2–HerA and associated proteins. The resin was washed extensively with washing buffer (25 mM Tris-HCl, 150 mM NaCl, 20 mM imidazole, pH 8.0), and bound proteins were eluted with elution buffer containing 250 mM imidazole. Input and pulldown fractions were analyzed by 12% sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE).

For in vitro pulldown assays, His-tagged Sir2–HerA system proteins were mixed with Strep-tagged phage proteins, including Gp2.5-Strep or Gp5.9-Strep, in the presence or absence of DNA substrates. Where indicated, single-stranded DNA (ssDNA) or blunt double-stranded DNA (dsDNA) was added to the reaction mixture. The protein mixtures were incubated at 4°C for 1 h, followed by addition of BeyoGold™ His-tag Purification Resin (P2210) and further incubation at 4°C for 1 h. The resin-containing mixtures were loaded onto gravity columns, and the flow-through fractions were collected. The resin was washed with washing buffer (25 mM Tris-HCl, 150 mM NaCl, 20 mM imidazole, pH 8.0), and bound proteins were eluted with elution buffer containing 250 mM imidazole. Input, flow-through, and elution fractions were analyzed by 12% SDS-PAGE.

NADase activity assay

NADase activity was measured using the fluorescent NAD analogue nicotinamide 1, N⁶-ethenoadenine dinucleotide (ε-NAD⁺; Sigma). Standard reactions (50 μl total volume) were carried out in reaction buffer containing 20 mM HEPES (pH 8.0), 25 mM NaCl and 5 mM MnCl₂. The Sir2 alone, the Sir2–HerA complex, or the mutant Sir2–HerA complex was included at a final concentration of 0.5 μM, and the wild-type or mutant phage trigger proteins were added to 2.5 μM, in the presence or absence of 0.5 mM ATP/ATPγS or varying concentrations of ssDNA and 5′ overhang dsDNA. The reaction system was incubated at 37°C for 45 min. After incubation, 50 μM ε-NAD⁺ was added to the reaction system. Then the reaction mixtures were added to 96-well plates (Corning® 3690), and fluorescence was monitored using Tecan (Spark 10M) plate reader with measurements taken every 30 s for 30 min (excitation filter 310 nm, emission filter 410 nm). All assays were performed in triplicate. The means and standard deviations were calculated using GraphPad Prism 10.

ATPase activity assay

The ATPase activity was measured using Malachite Green Phosphate Detection Kit (Beyotime, S0196S). The assays were performed in reaction buffer containing 20 mM HEPES (pH 8.0), 25 mM NaCl, and 5 mM MnCl₂. Purified protein was added to a final concentration of 1 μM and ATP to a final concentration of 200 μM, with or without 1 to 50 nM DNA substrate. The mixtures were incubated at 37°C for 45 min. The release of inorganic phosphate (Pi) was monitored by Tecan (Spark 10M) plate reader at 630 nm. All assays were performed in triplicate. The means and standard deviations were calculated using GraphPad Prism 10.

DNA substrate preparation

The sequence of DNA substrates used in the NADase, ATPase and helicase assays are listed in Supplementary Table S2. For preparation of fluorescent DNA substrates, the ssDNA was 5′ Cy3-labeled. Equimolar amounts of Cy3-labeled ssDNA and the corresponding complementary strands were then mixed to generate blunt dsDNA, forked dsDNA, 5′ overhang dsDNA, and 3′ overhang dsDNA. The mixtures were heated to 95°C for 5 min and then slowly cooled to room temperature. The resulting DNA substrates were aliquoted and stored at −20°C until use.

Nuclease activity and nicking activity assay

Purified proteins were incubated with 200 ng pUC19 plasmid in reaction buffer (20 mM HEPES, pH 8.0, 25 mM NaCl, 1 mM ATP, and 5 mM MnCl₂) in a total volume of 10 μl. Incubations were performed at 37°C for 1 h, then stopped by adding ethylenediaminetetraacetic acid to a final concentration of 10 mM, and the samples were analyzed by electrophoresis on 1% agarose gels.

Helicase activity assay

Helicase assays were carried out in 10 μl reaction buffer containing 20 mM HEPES (pH 8.0), 25 mM NaCl, and 5 mM MnCl₂, with or without 1 mM ATP. Sir2–HerA proteins or HerA only proteins were added to a final concentration of 1 μM in the presence or absence of 2 μM Gp2.5. 5′ Cy3-labeled ssDNA was used as a reference substrate, whereas blunt dsDNA, forked dsDNA, 5′ overhang dsDNA and 3′ overhang dsDNA were used as substrates. Reaction mixtures were incubated at 37°C for 1 h. 12% native polyacrylamide gels were pre-run in 0.5 × tris-borate-EDTA (TBE) buffer for 30 min, after which samples were loaded and electrophoresed in 0.5 × TBE for a further 2 h.

Electrophoretic mobility shift assay

Electrophoretic mobility shift assays were performed to assess the binding of Gp2.5 to different DNA substrates. A 5′ Cy3-labeled ssDNA substrate was used as a reference, and blunt dsDNA, forked dsDNA, 5′ overhang dsDNA, and 3′ overhang dsDNA were used as dsDNA substrates. Binding reactions were assembled in a final volume of 10 μl containing DNA substrate, purified Gp2.5 protein at the indicated concentration, 20 mM Tris-HCl, pH 8.0, and 5 mM MgCl₂. Reactions were incubated at 37°C for 30 min. Following incubation, samples were mixed with native loading buffer and resolved on a 12% native polyacrylamide gel in 0.5 × TBE buffer. The gel was pre-electrophoresed in 0.5 × TBE buffer on ice for 30 min prior to sample loading, and electrophoresis was subsequently performed on ice for 2 h. Cy3-labeled DNA substrates and protein-DNA complexes were detected using a ChemiDoc MP Imaging System (Bio-Rad).

Benzonase treatment of protein samples

To remove contaminating nucleic acids from purified protein preparations, Sir2–HerA complex protein and T7 Gp2.5 trigger protein were treated with BeyoZonase^TM Super Nuclease, ≥99% (Beyotime). The reaction buffer contained 50 mM Tris-HCl (pH 8.0), 1 mM MgCl₂ and 0.1 mg/ml BSA. BeyoZonase^TM Super Nuclease was added to a final concentration of 200 U/ml, and the mixture was incubated at 4°C for 36 h. The resulting protein preparation was stored at −80°C for subsequent experiments.

Sample preparation for transmission electron microscopy

Freshly purified protein samples were prepared for negative-stain transmission electron microscopy (TEM) using carbon-supported copper grids (XP-CF200). Briefly, 20 μl of sample was applied to the grid for 1 min and blotted with filter paper. This application step was repeated once. The grid was then stained with 20 μl uranyl acetate for 1 min, blotted, and stained a second time for 1 min before blotting again. Grids were air-dried at room temperature for 5–10 min. Micrographs were collected on Thermo Fisher Scientific Talos L120C G2 transmission electron microscope operated at 120 kV.

Statistics and reproducibility

Statistical analyses were carried out using GraphPad Prism 10 (GraphPad Software Inc., USA). For comparisons among multiple individual reaction groups, such as those in Fig. 2H, data were analyzed by ordinary one-way analysis of variance (ANOVA) followed by Tukey’s multiple comparisons test, and significance annotations were based on adjusted P-values. For two-factor comparisons involving different reaction conditions and peak fractions, such as those in Fig. 5E, data were analyzed by two-way ANOVA followed by Tukey’s multiple comparisons test, and significance annotations were based on adjusted P-values. For comparisons between two groups across multiple experimental conditions, such as those in Fig. 4D and E, multiple t-tests were performed with correction for multiple comparisons using the Benjamini, Krieger, and Yekutieli, and significance annotations were based on Q values. Data are presented as mean ± standard deviation (SD) from three replicates, with individual data points shown on the bar graphs, unless otherwise indicated. Statistical significance was defined as adjusted P <.05 for ANOVA followed by Tukey’s multiple comparisons test and Q < 0.05 for multiple t-tests. Details for specific experiments can be found in the corresponding figure legends. All plaque assays were performed twice and the representative images were shown.

Figure 2.

Type III Sir2–HerA is triggered by multiple phage-encoded proteins. (A) Schematic of the inducible trigger screen. Individual T7 genes were cloned under an arabinose-inducible promoter and co-transformed with type III Sir2–HerA plasmid into E. coli MG1655. Trigger-dependent toxicity was assessed by spot-dilution assays and by time-course growth measurements. (B) Spot-dilution assays showing induction-dependent toxicity upon co-expression of type III Sir2–HerA with T7 Gp2.5 or Gp5.9. Cells carrying the corresponding plasmids were spotted on plates without (−Ara) or with (+Ara) arabinose. (C–F) Growth dynamics following arabinose induction of the indicated constructs. Co-expression of type III Sir2–HerA with T7 Gp2.5 (C) or T7 Gp5.9 (D) caused strong growth inhibition relative to controls. T4 SSB Gp32 (E) and phage #34 SSB (F) similarly elicited type III Sir2–HerA-dependent growth inhibition. Data are representative of three replicates. (G) Size-exclusion chromatography (SEC) profile of purified type III Sir2–HerA showing two oligomeric states (peaks A and B). Inset, SDS-PAGE analysis of elution (E) and peak fractions (A and B) (M, marker). (H) In vitro NADase assays testing purified Sir2 and Sir2–HerA peak fractions in the absence or presence of the indicated triggers. Gp2.5 or Gp5.9 stimulated NADase activity predominantly in the Sir2–HerA complex peak B fraction, with Gp2.5 yielding stronger activation than Gp5.9. Bars show mean ± SD. Statistical significance was determined by one-way ANOVA followed by Tukey’s multiple-comparison test. ns, not significant; *P <.05; **P <.01; ***P <.001; ****P <.0001.

Figure 5.

Phage Gp2.5 remodels Sir2–HerA oligomerization to partition nickase, ATPase, and NADase activities into distinct assemblies. (A) SEC profiles of purified type III Sir2–HerA complex alone, Gp2.5 alone, and Sir2–HerA after pre-incubation with Gp2.5. Gp2.5 induces redistribution of Sir2–HerA into four major peaks: peaks 2 and 3 overlap with the basal Sir2–HerA peak A and peak B positions, whereas a higher-molecular weight peak 1 and a lower-molecular weight peak 4 (co-eluting near Gp2.5) appear upon Gp2.5 incubation. (B) SDS-PAGE analysis of SEC fractions. (C) NADase assays of isolated SEC peaks. NADase activity is detected only in peak 4. Data are shown as mean ± SD. ****P <.0001. (D) DNA nicking assays of isolated SEC peaks. Nickase activity is detected only in peak 1, whereas peaks 2–4 show little to no nicking. (E) ATPase assays of isolated SEC peaks. ATPase activity is detected in peaks 2 and 3, but not in peak 1 or peak 4. Data are shown as mean ± SD. Statistical significance was determined by two-way ANOVA followed by Tukey’s multiple-comparisons test. Significance annotations are based on adjusted P-values. ns, not significant; *P <.05; **P <.01; ***P <.001; ****P <.0001. (F) Model: phage Gp2.5 initiates a multilayered activation of type III Sir2–HerA. Gp2.5 activates both the HerA nickase and the Sir2 NADase of the Sir2–HerA complex. Nickase activity can lead to the accumulation of end-exposed DNA intermediates that enhance NADase activity through two routes: potentiating Gp2.5-triggered NADase activity and engaging the HerA ATPase to drive ATP hydrolysis, thereby relieving ATP-dependent inhibition of the Sir2 NADase. Together, these two end-exposed DNA-dependent routes enabling robust NAD⁺ depletion.

Figure 4.

DNA ends activate Sir2–HerA ATPase to relieve ATP-mediated inhibition of Sir2–HerA NADase and potentiate trigger-dependent NADase activity. (A) ATPase assays show that neither HerA alone nor the assembled type III Sir2–HerA complex (peaks A and B) exhibits detectable ATP hydrolysis, and Gp2.5 co-incubation does not elicit ATPase activity. ns, not significant. (B) Linear DNA substrates with exposed ends stimulate ATPase activity of the Sir2–HerA complex, whereas supercoiled plasmid DNA does not. ****P <.0001. (C) Mutation of the ATPase site (HerA K188A) abolishes DNA end-stimulated ATPase activity of the Sir2–HerA complex. (D) ATP gates Sir2–HerA NADase activity through HerA. ATP inhibits NADase activity of wild-type Sir2–HerA, and ATPγS similarly inhibits, whereas the K188A mutant retains Gp2.5-stimulated NADase activity that is no longer inhibited by ATP. Significance annotations are based on q values. ns, not significant; *q < 0.05; **q < 0.01; ***q < 0.001; ****q < 0.0001. (E) End-exposed DNA potentiates trigger-dependent NADase activity. 5′ overhang dsDNA and ssDNA alone do not activate NADase activity of the Sir2–HerA peak B complex, but strongly enhances Gp2.5-triggered NADase activity; a similar DNA-dependent potentiation is observed for the K188A complex. **q < 0.01. (F) Benzonase treatment of purified Gp2.5 markedly reduces its ability to stimulate Sir2–HerA NADase activity, indicating that Gp2.5 alone is a weak activator and that trace nucleic-acid contamination can inflate apparent activation. ****P <.0001. (G) Dose dependence of DNA-mediated licensing. Titration of ssDNA in reactions containing Gp2.5 and the Sir2–HerA peak B complex shows that ATP inhibits NADase at low ssDNA concentrations, but this inhibition is progressively relieved as ssDNA increases and is abrogated at high ssDNA. In the absence of ATP, ssDNA potentiation of Gp2.5-triggered NADase activation is dose dependent. For the K188A mutant complex, ssDNA potentiation remains dose dependent, but ATP no longer inhibits. ns, not significant; *q < 0.05; **q < 0.01; ***q < 0.001; ****q < 0.0001. (H) Model: Phage SSB (Gp2.5) activates the Sir2–HerA complex to unlock HerA nickase and, in parallel, stimulating Sir2 NADase activity. Nickase activity leads to the accumulation of end-exposed DNA intermediates that (i) engage the complex-associated ATPase to drive ATP hydrolysis and relieve ATP-mediated inhibition of Sir2 NADase activity, and (ii) potentiate trigger-dependent NADase activation, together enabling robust NAD⁺ depletion.

Results

Diverse Sir2–HerA systems in E. coli exhibit distinct antiphage spectra

To test whether Sir2–HerA systems diversify into discrete functional subtypes within a single bacterial species, we systematically mined publicly available E. coli genomes (n = 2289) for instances in which sir2 and herA genes are co-located in close genomic proximity (Fig. 1A), a hallmark of Sir2–HerA systems. This analysis revealed eight Sir2–HerA clusters defined by Sir2 sequence similarity, of which cluster 1 corresponds to the previously described Sir2–HerA/Nezha system, whereas the remaining clusters represent divergent Sir2–HerA lineages (Fig. 1B). Independent clustering of the corresponding HerA proteins showed a broadly similar grouping pattern, largely consistent with the Sir2-based classification (Supplementary Fig. S1). To determine their antiphage activity, representative loci from each cluster were cloned and transformed into E. coli MG1655. The resulting strains were challenged with a panel of 110 E. coli phages spanning at least nine genera [5] (Fig. 1C and Supplementary Table S3). Beyond Nezha, three additional systems, designated Sir2–HerA types I, II, and III, showed robust antiphage activity and displayed distinct antiphage spectra (Fig. 1C), establishing functional diversification of Sir2–HerA immunity within E. coli.

Figure 1.

Diverse Sir2–HerA systems in E. coli exhibit distinct antiphage spectra. (A) Bioinformatic pipeline used to mine publicly available E. coli genomes for loci in which sir2 and herA occur in close genomic proximity. (B) Eight Sir2–HerA clusters defined by Sir2 sequence identity. Shown are representative domain organizations (Sir2, green; HerA, orange) and the number of loci assigned to each cluster. (C) Antiphage profiling of representative Sir2–HerA systems from each cluster expressed in E. coli MG1655 against a panel of E. coli phages. Heat map indicates restriction efficiency across phages; bar plot summarizes the number of phages restricted by each system. (D) Pairwise amino-acid identity matrix for Sir2 and HerA across Nezha and types I–III, highlighting divergence among subtypes. (E) Phylogenetic analysis showing that Nezha and types I–III form distinct clades within E. coli and relate to homologs found in other bacterial taxa. (F) Structural comparison of the Sir2 and HerA proteins from Nezha (PDB ID 8J4U) and types I–III (AlphaFold3 prediction), illustrating subtype-specific differences in predicted architectures. (G) Growth curves following phage challenge at high MOI (5), low MOI (0.05), or uninfected control, showing rapid growth collapse at high MOI for strains expressing Nezha or types I–III relative to vector controls, consistent with an abortive infection-like outcome. Data are representative of three replicates.

Sequence comparisons based on Sir2 and HerA amino acid sequences showed that types I–III are highly diverged from the previously described Nezha system and from one another (∼13%–22% identity at the Sir2 and HerA level), and phylogenetic analysis of Sir2 sequences placed types I–III into distinct clades separate from Nezha within E. coli (Fig. 1D and E). Consistent with subtype-specific functional organization, swapping the HerA protein of type III into type I, type II, or Nezha abolished antiphage protection in each chimeric system (Supplementary Fig. S2). To compare their architectures, we performed AlphaFold-based structural superposition analyses (Supplementary Fig. S3 and Fig. 1F). Although the core Sir2 and HerA domains were relatively conserved in domain-level comparisons (Supplementary Fig. S3), the full-length proteins exhibited greater structural divergence (Fig. 1F). Together, the sequence divergence, distinct antiphage spectra, subtype-specific HerA compatibility, and full-length structural differences support the classification of these systems as distinct Sir2–HerA subtypes rather than Nezha variants.

Given that previously described Sir2–HerA/Nezha has been linked to abortive infection [9, 22, 24], we asked whether the newly identified Sir2–HerA subtypes similarly restrict phage propagation by sacrificing infected cells. Bacterial growth curves showed that a high dose of phage infection (MOI = 5) induced rapid growth collapse of E. coli cells expressing each Sir2–HerA subtype relative to uninfected controls (Fig. 1G), which is usually a feature of the abortive infection mechanism of many defense systems [8, 9, 14, 16]. To define the basis of Sir2–HerA immunity, we chose type III Sir2–HerA for subsequent analysis, as it provided the most robust protection and the broadest antiphage spectrum (Fig. 1C).

Type III Sir2–HerA is triggered by multiple phage-encoded proteins

Since type III Sir2–HerA restricts phage infection through abortive infection, we reasoned that specific phage-encoded proteins produced during infection might act as triggers to activate the defense module and induce host collapse [36]. To identify such triggers in an unbiased manner, we used phage T7 as a model and constructed an inducible expression library covering all 59 T7 genes [37, 38], each cloned into an arabinose-inducible plasmid (Fig. 2A). Each T7 gene plasmid was co-transformed with the type III Sir2–HerA system into E. coli MG1655, and transformants were screened for induction-dependent cytotoxicity upon arabinose addition (Fig. 2A). This screen revealed two T7 proteins that reproducibly elicited Sir2–HerA-dependent toxicity: Gp2.5, the SSB [39, 40], and Gp5.9, a small protein reported to inhibit RecBCD-mediated DNA end processing [41] (Fig. 2B and Supplementary Fig. S4). These results indicated that Gp2.5 and Gp5.9 are candidate triggers of type III Sir2–HerA activation.

Growth curves showed that arabinose induction of Gp2.5 caused little to no growth impairment in control cells, but triggered marked growth inhibition specifically in cells carrying type III Sir2–HerA expressed from its native promoter (Fig. 2C), consistent with trigger-dependent activation of an abortive infection-like outcome. A similar pattern was observed for Gp5.9, which also caused a strong growth defect specifically when co-expressed with type III Sir2–HerA upon induction (Fig. 2D). To test whether type III Sir2–HerA can sense functionally related proteins from other phages, we cloned SSB genes from the canonical T4 phage (Myoviridae) and the phage #34 (Dhillonvirus, Supplementary Table S3) for inducible expression, noting that neither T4 nor phage #34 encode a clear Gp5.9 functional equivalent [42]. Co-expression of type III Sir2–HerA with T4 Gp32 or phage #34 SSB resulted in marked growth inhibition, whereas expression of either component alone had minimal impact (Fig. 2E and F).

To further test whether these triggers directly activate type III Sir2–HerA, we purified Sir2 and the Sir2–HerA complex (Supplementary Fig. S5). SEC analysis showed that the type III Sir2–HerA complex assembled into two oligomeric states (peaks A and B) (Fig. 2G). In vitro NADase assays showed no detectable activity for Sir2 alone or for peak A or peak B of the Sir2–HerA complex under basal conditions. By contrast, both Gp2.5 and Gp5.9 stimulated NADase activity specifically in peak B, with Gp2.5 producing a modestly stronger activation than Gp5.9 (Fig. 2H). Co-incubation of Gp2.5 and Gp5.9 produced only a slight increase in NADase activity compared with Gp2.5 alone, suggesting a limited additive effect rather than strong synergistic activation (Fig. 2H). However, this trigger-induced activity was still relatively low compared with the robust NADase activation observed when Gp2.5 was combined with end-exposed DNA substrates, as described below. Together, these biochemical results support the conclusion that Gp2.5 and Gp5.9 are activators of type III Sir2–HerA.

Phage Gp2.5 relieves Sir2-mediated repression to activate HerA nickase independently of Gp2.5-stimulated NADase activity

Having identified Gp2.5 as the primary trigger, we next sought to determine how it regulates the enzymatic outputs of the type III Sir2–HerA. To this end, we purified Sir2, HerA, and the assembled Sir2–HerA complex and evaluated their activities in vitro. Sir2 alone showed little or no detectable DNA cleavage at concentrations up to 1 μM, whereas apparent DNA degradation was observed only at a higher protein concentration of 2 μM and was not enhanced by Gp2.5 (Fig. 3A and B). Because such high protein concentrations can increase the contribution of trace nuclease contamination or other in vitro assay artifacts, we interpret this activity cautiously and do not consider it convincing evidence for trigger-dependent nuclease activity of Sir2 alone. This observation stands in stark contrast to the constitutive nuclease activity previously attributed to Sir2 in the Sir2–HerA/Nezha system [22, 24]. Furthermore, we found that purified HerA alone possesses intrinsic nickase activity, a functional feature absent in the previously characterized Sir2–HerA/Nezha system [22, 24]. Notably, the addition of Gp2.5 did not significantly alter HerA nicking (Fig. 3C and D). Together, these results indicate that Gp2.5 does not appreciably modulate the basal activities of Sir2 or HerA in isolation, suggesting a model in which Gp2.5 acts primarily by activating the assembled Sir2–HerA complex.

Figure 3.

Gp2.5 relieves Sir2-mediated repression to activate HerA nickase independently of Gp2.5-stimulated NADase activity. (A–D) Biochemical assays show that Sir2 displays detectable DNA cleavage only at high protein concentration (A) and is not enhanced by Gp2.5 (B), whereas HerA exhibits intrinsic nickase activity (C) that is largely insensitive to Gp2.5 (D). (E, F) The assembled type III Sir2–HerA complex [peak A in panel (E), peak B in panel (F)] is catalytically silent for nuclease/nickase activity. (G) Gp2.5 unlocks nickase activity in the assembled complex. Peak A and peak B Sir2–HerA were incubated with Gp2.5 as indicated and assayed for nicking (arrowhead indicates nicked product). (H) Nicking assay for HerA (D218A), showing loss of intrinsic nickase activity across the indicated protein concentrations. (I, J) The HerA (D218A) mutation abolishes type III Sir2–HerA antiphage protection (I) and eliminates Gp2.5-dependent growth collapse (J). Data are representative of three replicates. (K) NADase activity for Sir2–HerA peak B complexes (WT or D218A). Gp2.5 stimulates NADase activity of both WT and D218A complexes, with modest reduction for D218A, indicating that loss of nickase does not eliminate Gp2.5-stimulated NADase activation. Data are shown as mean ± SD. ****P <.0001. (L) Model: Sir2–HerA complex formation represses HerA nickase activity, and the phage SSB Gp2.5 relieves this repression to switch nickase “off” to “on”. In parallel, Gp2.5 also stimulates Sir2–HerA NADase activity, and separation-of-function (D218A) indicates that Gp2.5-triggered NADase activation can be retained even when nickase catalysis is disabled.

We next asked how Gp2.5 affects the enzymatic activities of the assembled Sir2–HerA complex. Under basal conditions, neither peak A nor peak B displayed detectable nuclease activity or nickase activity (Fig. 3E and F). Strikingly, incubation with Gp2.5 rendered both peak A and peak B competent for DNA nicking, indicating that complex formation locks the intrinsic HerA nickase in an inactive state and that Gp2.5 relieves Sir2-mediated repression to unlock nickase activity within the complex (Fig. 3G). In contrast to prior reports on Nezha, we detected no helicase activity for HerA alone or for the Sir2–HerA complex (peaks A and B), and Gp2.5 co-incubation did not reveal helicase activity in any preparation (Supplementary Fig. S6).

To test whether HerA nicking is required for immunity, we mapped a catalytic determinant of the nickase. Substitution of HerA Asp218 with Ala (D218A) abolished nickase activity (Fig. 3H) and eliminated type III Sir2–HerA defense (Fig. 3I). Consistently, co-induction of the D218A mutant system with Gp2.5 failed to elicit the marked growth inhibition observed for the wild-type system, indicating loss of Gp2.5-triggered toxicity (Fig. 3J). Notably, Gp2.5 still stimulated NADase activity of the D218A Sir2–HerA complex, although modestly reduced compared with wild type (Fig. 3K), demonstrating that Gp2.5-dependent NADase activation can be retained when nickase catalysis is disabled.

Together, these results define a dual-activation logic for type III Sir2–HerA (Fig. 3L). Assembly of Sir2 with HerA locks the intrinsic HerA nickase in an inactive state, establishing a quiescent baseline, and Gp2.5 relieves this Sir2-mediated repression to unlock nickase activity within the complex. In parallel, Gp2.5 stimulates Sir2–HerA NADase activity. The loss of defense upon disabling HerA nickase, despite preserved Gp2.5-dependent NADase activation, further indicates that both the nickase and NADase arms are required for robust antiphage immunity.

DNA ends activate Sir2–HerA ATPase to relieve ATP-mediated inhibition of Sir2–HerA NADase and potentiate trigger-dependent NADase activity

We next investigated ATPase regulation in the type III Sir2–HerA system and its relationship to NADase activation. Neither HerA expressed alone nor the assembled Sir2–HerA complex (peaks A and B) displayed detectable ATPase activity, and co-incubation with Gp2.5 did not elicit ATP hydrolysis (Fig. 4A). In striking contrast, ATPase activity of the Sir2–HerA complex, but not HerA alone, was robustly activated by various linear DNA forms, including ssDNA and dsDNA with blunt, forked, or overhanging ends (Fig. 4B and Supplementary Fig. S7A and B). In contrast, the complex failed to respond to supercoiled plasmid DNA (Fig. 4B and Supplementary Fig. S7A). Addition of Gp2.5 did not further increase ATPase activity in the presence of the tested DNA substrates (Supplementary Fig. S7C). These data indicate that ATPase activation is selectively triggered by DNA substrates with exposed ends rather than by Gp2.5 or closed circular DNA. Substitution of HerA Lys188 with Ala (K188A) abolished DNA end-stimulated ATPase activity of the Sir2–HerA complex (Fig. 4C), indicating that Lys188 is essential for ATPase activity.

High ATP concentrations inhibited NADase activity of the wild-type type III Sir2–HerA complex (Fig. 4D). Moreover, the non-hydrolysable ATP analog ATPγS similarly suppressed NADase activity of wild-type Sir2–HerA (Fig. 4D), indicating that ATP binding to HerA, without requiring ATP hydrolysis, is sufficient to gate NADase activation. The ATPase mutant complex (K188A) retained Gp2.5-stimulated NADase activity, but this Gp2.5-triggered NADase activity was no longer inhibited by high ATP concentrations (Fig. 4D), consistent with ATP binding to HerA acting as a checkpoint for NADase gating.

Given that DNA substrates with exposed ends strongly stimulate ATP hydrolysis, we next examined how these DNA influences Gp2.5-dependent NADase activation. Using ssDNA and 5′ overhang dsDNA as representative end-exposed substrates, we found that these DNA alone did not activate NADase activity of the Sir2–HerA peak B complex (Fig. 4E). Interestingly, they markedly potentiated Gp2.5-triggered NADase activity when added together with Gp2.5, yielding substantially higher NADase activity than Gp2.5 alone (Fig. 4E). A similar DNA-dependent potentiation was observed for the K188A mutant complex, indicating that end-exposed DNA not only engages ATPase activity but also exerts a strong synergistic effect on Gp2.5-triggered NADase activation.

Since Gp2.5 has intrinsic ssDNA-binding activity and only weakly stimulated Sir2–HerA NADase activity on its own, we suspected that residual DNA carried through the Gp2.5 purification might contribute to the apparent activation of the Sir2–HerA NADase. To test this possibility, we treated purified Gp2.5 with benzonase, which degrades both ssDNA and dsDNA, before incubating it with Sir2–HerA. Results showed that benzonase treatment markedly reduced the ability of Gp2.5 to stimulate Sir2–HerA NADase activity (Fig. 4F), supporting the idea that Gp2.5 alone is a poor activator and that trace DNA contamination can inflate the observed signal. Together with our finding that end-exposed DNA substrates synergize with Gp2.5 to enhance NADase activation, these data indicate a synergistic requirement for both the phage SSB trigger and infection-relevant DNA substrates to achieve robust NADase activation.

Since Gp2.5 activates the Sir2–HerA nickase, ongoing infection would be expected to increase DNA nicking and thereby promote the accumulation of end-exposed DNA intermediates in the cell. We therefore asked whether the synergistic enhancement of Gp2.5-triggered NADase activation by end-exposed DNA is dose dependent. To this end, we titrated ssDNA in reactions containing Gp2.5 and the Sir2–HerA peak B complex and quantified NADase activity (Fig. 4G). At low ssDNA concentrations, ATP strongly suppressed NADase activity, but this inhibition weakened as ssDNA concentration increased and was fully relieved at ∼1 μM ssDNA. In the absence of ATP, ssDNA potentiation of Gp2.5-triggered NADase activation was also dose dependent, indicating that end-exposed DNA quantitatively amplifies the trigger response. For the ATPase-deficient K188A Sir2–HerA mutant complex, ssDNA similarly enhanced Gp2.5-triggered NADase activation in a dose-dependent manner. In contrast to the wild-type complex, ATP no longer inhibited NADase activity of the K188A mutant and instead caused a slight increase across the ssDNA titration. The basis of this modest increase is currently unclear and may reflect an indirect effect specific to the mutant reaction condition. Nevertheless, the key conclusion from the K188A mutant is that disruption of the HerA ATPase site bypasses ATP-dependent inhibition of NADase activation.

Together, these data support a gating model in which nickase activity leads to the accumulation of end-exposed DNA intermediates that drive robust Sir2 NADase activation through two coordinated routes: first, by engaging the ATPase to consume ATP and relieve ATP-mediated inhibition; and second, by directly potentiating Gp2.5-triggered NADase activity (Fig. 4H).

Phage Gp2.5 remodels Sir2–HerA oligomerization to partition nickase, ATPase, and NADase activities into distinct assemblies

We next asked how Gp2.5 engages the Sir2–HerA complex to activate NADase, nickase, and ATPase activities. A pulldown assay using His-tagged Sir2–HerA complex failed to capture untagged Gp2.5 (Supplementary Fig. S8A and B). Because DNA substrates strongly influence Sir2–HerA activation, we further tested Gp2.5-Sir2–HerA association in the presence of ssDNA or blunt dsDNA; however, Strep-tagged Gp2.5 still failed to efficiently pull down the Sir2–HerA complex under either condition (Supplementary Fig. S8C). These results suggest that Gp2.5 does not form a stable, high-affinity complex with Sir2–HerA, even in the presence of DNA, prompting us to test whether Gp2.5 instead alters the assembly state of Sir2–HerA. SEC revealed that incubation of Sir2–HerA with Gp2.5 redistributed the complex into four peaks (Fig. 5A). Two of these peaks (peaks 2 and 3) overlapped with the original Sir2–HerA peaks A and B, whereas two additional assemblies emerged: a higher-molecular weight assembly (peak 1) and a later-eluting assembly that co-migrated with Gp2.5 (peak 4). SDS-PAGE analysis indicated that peak 1 contained predominantly HerA, peak 4 was enriched for HerA together with Gp2.5, and peaks 2 and 3 contained both Sir2 and HerA (Fig. 5B).

We next mapped enzymatic activities across these Gp2.5-induced assemblies. NADase assays showed that only peak 4 displayed detectable NADase activity (Fig. 5C). Importantly, peak 4 generated using a catalytic Sir2 mutant (N238A) lacked NADase activity (Supplementary Fig. S9), indicating that the activity in peak 4 is Sir2 dependent even though Sir2 was not readily detected by Coomassie staining, because the overall concentration of this fraction is low and Sir2 is likely below the detection limit of Coomassie-based visualization. In contrast, nickase assays showed that only peak 1 exhibited DNA nicking activity, whereas peaks 2–4 were inactive in this assay (Fig. 5D). Finally, ATPase measurements revealed activity only in peaks 2 and 3, whereas peaks 1 and 4 lacked detectable ATPase activity (Fig. 5E).

Electron microscopy of the SEC fractions (Supplementary Fig. S10) further supported Gp2.5-induced assembly diversification. Peak 1 contained large but heterogeneous particles, and peak 2 showed ∼20–30 nm particles, a subset of which resembled previously reported Nezha-like assemblies [22]. In contrast, peak 3 was the most homogeneous, forming a flower-like ring (∼10–15 nm) that is notably smaller than Sir2–HerA particles described previously [22, 24–26]. Peak 4 showed low particle contrast and few Sir2–HerA-like particles, consistent with SDS-PAGE indicating dominance of Gp2.5 (Fig. 5B). Overall, the heterogeneity of peaks 1–2 and the low abundance of Sir2–HerA-like particles in peak 4 currently limit high-resolution structural analysis, motivating future optimization to stabilize and enrich discrete assemblies for structural characterization.

Together, these results indicate that Gp2.5 does not activate Sir2–HerA by forming a tight, stable complex, but instead remodels Sir2–HerA oligomerization to generate functionally specialized assemblies. In this model, Gp2.5 promotes formation of a higher-order HerA-rich assembly that is nickase active but ATPase inactive (peak 1), and a lower-order assembly that supports Sir2-dependent NADase activity (peak 4). Meanwhile, the assemblies whose elution overlaps the original Sir2–HerA peaks (peaks 2 and 3) retain DNA end-dependent ATPase activity, consistent with these assemblies representing Sir2–HerA complexes whose oligomeric states have not perturbed by Gp2.5.

Discussion

Sir2–HerA (Nezha) systems are multi-enzyme abortive infection defenses that abort phage propagation by depleting NAD⁺, yet how infection is sensed and how these activities are gated to prevent inadvertent activation in the absence of infection remain unclear [22, 24–26]. Here we show that Sir2–HerA achieves infection specificity through a subtype-diversified, trigger-linked and checkpointed control architecture: type III Sir2–HerA can be activated by phage proteins, and its lethal NAD⁺ depletion is licensed only after multilayered gating steps that couple triggering to infection-derived DNA intermediates and host ATP state (Fig. 5F). Building on this framework, our study advances three conceptual points.

First, we found that Sir2–HerA is not a single mechanistic entity. Within E. coli, Sir2–HerA immunity diversifies into discrete functional subtypes with distinct antiphage spectra, and subtype identity is not interchangeable even at the level of core components. This within-species diversification parallels the subtype structure described for defense families such as Thoeris and CBASS [27, 28]. Together, these observations suggest that Sir2–HerA systems can evolve distinct triggers and regulatory wiring, rather than operating through a single, Nezha-like mechanism.

Second, we identify phage-encoded triggers that link Sir2–HerA activation to bona fide phage infection. An unbiased screen across the T7 proteome revealed two triggers, the ssDNA-binding protein Gp2.5 and Gp5.9, and the ability of other SSBs to elicit the same outcome indicates that phage replication-associated proteins can serve as infection-linked cues [10, 42–46]. This trigger logic is conceptually appealing because phage SSBs are produced during infection and can accumulate to high levels in infected cells; for example, T7 Gp2.5 is among the most abundant phage proteins and is detectable within minutes of infection, consistent with rapid accumulation during lytic growth [47, 48]. In the type III Sir2–HerA system, such phage protein cues can be integrated with end-exposed DNA intermediates that license downstream activation.

Third, our data define a sequential, multilayered gating architecture that addresses a central problem of abortive infection: how to deploy lethal effector activities while preventing inadvertent activation in the absence of infection (Fig. 5F). In the type III system, the phage trigger (Gp2.5) functions as a dual-activator that simultaneously unleashes the HerA nickase and primes the Sir2 NADase. However, while the nickase becomes immediately active, the NADase remains sequestered by an ATP-mediated checkpoint. The resulting DNA nicks lead to the accumulation of end-exposed intermediates that drive ATP depletion via the activation of HerA ATPase, thereby lifting the checkpoint. This sequential logic, requiring both Gp2.5-mediated priming and the relief of ATP inhibition, ensures that robust NADase activity is licensed only upon the convergence of these two signals. Crucially, neither Gp2.5 nor end-exposed DNA intermediates alone is sufficient for NADase activation, thereby preventing inadvertent immune execution. Sir2–HerA therefore operates not as a single on-off switch, but as an integrative logic gate. By coupling a specific phage protein cue with infection-state readouts including end-exposed DNA and ATP depletion, the system achieves the high-fidelity control required to restrict NAD⁺ collapse to bona fide infection contexts.

Our findings reveal an assembly-remodeling mechanism that regulates type III Sir2–HerA activation. Phage Gp2.5 drives a redistribution of Sir2–HerA into distinct assemblies that partition activities, including nickase, ATPase, and NADase, into separate states. Such partitioning offers a physical mechanism for ordered gating by preventing simultaneous engagement of incompatible outputs within a single homogeneous complex. Electron microscopy of basal and Gp2.5-induced assemblies further suggests substantial heterogeneity in the activated states (Supplementary Fig. S10), highlighting that Sir2–HerA may operate as a dynamic multistate machine whose functional transitions remain to be captured structurally.

A comparison with previously characterized Sir2–HerA/Nezha systems highlights a fundamental transition from passive metabolic sensing to multilayered gating. Nezha complexes exist in a primed state, where NADase, helicase, and nuclease activities are constitutively present but masked by high concentrations of ATP [23, 25]. Consequently, immune execution is triggered passively by the metabolic exhaustion (ATP drop) associated with phage replication, rather than by a specific phage protein. In sharp contrast, our type III Sir2–HerA is quiescent until a dedicated phage protein (Gp2.5) initiates a multilayered regulatory cascade (as described above): Gp2.5 activates the HerA nickase to generate end-exposed DNA intermediates, which in turn stimulate the ATPase to drive ATP depletion and ultimately unleash the Sir2 NADase. This rewiring into a multilayered signaling cascade transforms Sir2–HerA from a metabolic sensor into a sophisticated coincidence detector. Such a tiered architecture ensures that the lethal defense program is licensed only when a specific phage protein and a secondary infection readout (end-exposed DNA) are simultaneously present. Finally, activation proceeds via assembly remodeling, partitioning nickase, ATPase, and NADase into distinct assemblies. Together, these differences reveal subtype-specific rewiring of catalytic repertoire and gating logic, expanding Sir2–HerA from a uniform Nezha-like mechanism to an evolvable defense family tuned by triggers, checkpoints, and assembly-state control.

Although our findings support a model in which Gp2.5 remodels Sir2–HerA to activate its effector functions, several mechanistic questions remain unresolved. First, although our data establish assembly remodeling as a key mode of Sir2–HerA activation, we have not yet resolved the structural basis for nickase and NADase licensing. Co-incubation with Gp2.5 redistributes Sir2–HerA into remodeled assemblies that are highly heterogeneous by electron microscopy, precluding high-resolution structural analysis at present. Second, the mechanistic contribution of the DNA-binding activity of Gp2.5 remains unclear. Our data show that Gp2.5 preferentially recognizes ssDNA-containing substrates rather than blunt dsDNA ends (Supplementary Fig. S11A), and that disruption of its ssDNA-binding surface by the K152A mutation strongly impairs Sir2–HerA activation (Supplementary Fig. S11B–E). However, because Gp2.5 can activate Sir2–HerA in assays without added DNA, it remains unresolved whether the DNA-binding surface contributes by engaging DNA during activation, by mediating a DNA-independent interaction with Sir2–HerA, or by supporting both processes. Future structural and biochemical studies will be needed to capture more homogeneous activated states and to determine how Gp2.5 and DNA substrates control the transition from basal Sir2–HerA assemblies to nickase- and NADase-active complexes.

Additionally, our analysis of Gp5.9 suggests that type III Sir2–HerA can be activated by more than one phage-derived cue through a related remodeling-based mechanism. Although Gp2.5 was used as the main trigger for mechanistic dissection, Gp5.9 also activated the nickase (Supplementary Fig. S12) and NADase (Fig. 2H) activities of the Sir2–HerA complex. Like Gp2.5, Gp5.9 did not form a stable complex with Sir2–HerA in pull-down assays, either in the absence of added DNA or in the presence of ssDNA or blunt dsDNA (Supplementary Fig. S13). Instead, Gp5.9 remodeled Sir2–HerA oligomerization and partitioned nickase, ATPase, and NADase activities into distinct assemblies (Supplementary Fig. S12), suggesting that Gp5.9 and Gp2.5 converge on a related assembly-remodeling mechanism. Combined treatment with Gp2.5 and Gp5.9 also did not further enhance nickase and ATPase activation (Supplementary Fig. S12), further suggesting that the two triggers may engage overlapping rather than synergistic activation routes. Given the reported DNA-mimic properties of Gp5.9 [41], one possibility is that Gp5.9 functionally mimics certain features of a Gp2.5-DNA-associated activation state, thereby promoting Sir2–HerA remodeling. However, this model remains speculative at present; future structural studies will be required to define how Gp5.9 and Gp2.5 engage Sir2–HerA and whether they stabilize overlapping or distinct activated conformations.

In conclusion, we establish that Sir2–HerA immunity is defined by functional diversification rather than a single mechanistic archetype. By characterizing the type III system, we uncover a transition from passive sensing to a multilayered regulatory cascade that operates through coincidence-detection logic and assembly remodeling. This work reveals a fundamental principle of bacterial defense: the partitioning of disparate enzymatic activities into distinct assemblies to provide precise control over lethal effectors.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

The authors declare no competing financial interest.

Funding

This work was supported by the National Natural Science Foundation of China (U23A20241, 32570051, and 32503044), Fundamental Research Funds for the Central Universities (2662026DKPY005), Self-determined Project of State Key Laboratory of Agricultural Microbiology (AML2026D04), Fundamental and Interdisciplinary Disciplines Breakthrough Plan of the Ministry of Education of China (JYB2025XDXM704), China Postdoctoral Science Foundation (2025M783038), Postdoctoral Fellowship Program of CPSF (GZC20251990), and the earmarked fund for CARS-41.The Open Access publication charges were jointly supported by the funding sources listed above.

Data availability

The data underlying this article are available in the article and in its online supplementary material.