A DNA-gated molecular guard controls bacterial Hailong anti-phage defence

Abstract

Animal and bacterial cells use nucleotidyltransferase (NTase) enzymes to respond to viral infection and control major forms of immune signalling including cGAS-STING innate immunity and CBASS anti-phage defence^1-4. Here we discover a family of bacterial defence systems, which we name Hailong, that use NTase enzymes to constitutively synthesize DNA signals and guard against phage infection. Hailong protein B (HalB) is an NTase that converts deoxy-ATP into single-stranded DNA oligomers. A series of X-ray crystal structures define a stepwise mechanism of HalB DNA synthesis initiated by a C-terminal tyrosine residue that enables de novo enzymatic priming. We show that HalB DNA signals bind to and repress activation of a partnering Hailong protein A (HalA) effector complex. A 2.0-Å cryo-electron microscopy structure of the HalA-DNA complex reveals a membrane protein with a conserved ion channel domain and a unique crown domain that binds the DNA signal and gates activation. Analysing Hailong defence in vivo, we demonstrate that viral DNA exonucleases required for phage replication trigger release of the primed HalA complex and induce protective host cell growth arrest. Our results explain how inhibitory nucleotide immune signals can serve as molecular guards against phage infection and expand the mechanisms NTase enzymes use to control antiviral immunity.

Extended Data Figure 1 |. Structural and biochemical analysis of HalB.

a, Structure guided multiple sequence alignment of HalB protein homologs from indicated bacterial species. Shading indicates degree of residue conservation. b,c, Protein structural homology of HalB WT N-terminal lobe (b) and HalB catalytically inactive mutant (DDAA) C-terminal lobe (c) against all entries in PDB showing the DALI Z score of the top 30 and 150 hits respectively. HalB DDAA was used in (c) as the structure of the mutant protein resolves additional residues in the C-terminal tail that are not visible in the wildtype structure. d, E. coli STEC1178 HalB was expressed as an N-terminal 6×His-SUMO2 fusion and purified by Ni-NTA and separated from His-SUMO2 by size exclusion chromatography. e, Coomassie-stained SDS-PAGE analysis of fully purified HalB from R. bacterium QY30. f, Overview of tetrameric HalB crystal structure with one protomer in blue and three protomers in different shades of grey. Numbered boxes highlight dimerization interface described in detail in (h). g, Size exclusion chromatography with multi-angle static light scattering (SEC-MALS) analysis of purified HalB confirms HalB tetramerization. h, Detailed view of interacting residues as in (f) showing dimerization interface between HalB protomers. i, Urea-PAGE analysis of HalB WT and LEFE mutant ODA synthesis. j, Size exclusion chromatography overlay of HalB WT and LEFE mutant showing a rightward shift indicating a loss of tetrameric complex formation. k, Bacterial growth assay of E. coli expressing HalA with HalB active site and dimerization mutants. LEFE, L24E and F67E double mutant. l, Urea-PAGE analysis of HalB substrate and metal specificity using NTPs and α³²P-labeled NTPs as indicated. m, Coomassie-stained SDS-PAGE analysis of ODA synthesis with purified HalB WT and catalytically inactive mutant (DDAA) given indicated μM concentrations of dNTP. n, Urea-PAGE analysis of ODA synthesis over a 24 h period with purified HalB given α³²P-dNTP, the indicated μM concentrations of dNTP, and treated with proteinase K. Expression of HalB used in this figure was from R. bacterium QY30. Data shown are representative of at least three independent experiments.

Extended Data Figure 2 |. Mechanism of ODA synthesis.

a, Urea-PAGE analysis of ODA synthesis with purified HalB and α³²P-dATP supplemented with different ddNTP substrates to induce chain termination and determine specificity to adenine. b, Schematic of LC-MS workflow to determine the chemical composition of ODA. c, Left, LC-MS analysis of purified HalB WT or DDAA mutant incubated with dNTP. Right, Characterization of the extracted signal molecule by LC-MS in negative mode. Formate and chloride ions formed the major adducts [dA+formate]⁻ and [dA+Cl]⁻ respectively observed with deoxyadenosine. d, Uncropped Urea-PAGE analysis of ODA cleavage from the HalB–ODA complex using α³²P-labeled dATP treated with proteinase K and E. coli soluble lysate fraction as in Figure 2. α³²P-labeled dAMP made with apyrase-treated dATP and nuclease P1 were used as controls to visualize single nucleotide product. e, Schematic depicting process of biochemical fraction of E. coli BL21 cell lysate to enrich for HalB–ODA cleavage activity using heparin IEX or ammonium sulfate precipitation followed by phenyl hydrophobic interaction. f, Urea-PAGE analysis of ODA cleavage from the HalB–ODA complex using fractions obtained after S200 size-exclusion chromatography. Active fractions used for mass spectrometry analysis are highlighted in bold. g, Summary of mass spectrometry results from both fractionation methods. HIC, hydrophobic-interaction chromatography; IEX, ion exchange. h, List of enriched candidates shared between both purification schemes. Data shown as mean for two independent experiments. i, Urea-PAGE analysis of ODA cleavage from the HalB–ODA complex using α³²P-labeled dATP treated with purified reconstitution of candidate proteins. Two common contaminating ribosomal proteins (RplP and RplE) and one catalytically inactive protein due to a frameshift mutation (Rph) were excluded from further analysis. j, Urea-PAGE analysis of ODA cleavage from the HalB–ODA complex using α³²P-labeled dATP treated with purified H-NS and StpA. k, Overview of tetrameric HalB catalytically inactive mutant crystal structure with one protomer in light blue and three protomers in different shades of grey. l, R. bacterium QY30 HalB catalytically inactive mutant was expressed as an N-terminal 6×His-SUMO2 fusion and purified by Ni-NTA and separated from 6×His-SUMO2 by size exclusion chromatography. m, Coomassie-stained SDS-PAGE analysis of fully purified HalB catalytically inactive mutant. n, Size exclusion chromatography with multi-angle static light scattering analysis of purified HalB catalytically inactive mutant. o, Overview of tetrameric HalB R164A crystal structure with one monomer in dark blue and three monomers in different shades of grey. p, Coomassie-stained SDS-PAGE analysis of fully purified HalB R164A. q, Overview of HalB and residue R164 in an open conformation prior to ODA synthesis and an active confirmation during ODA synthesis. r, Detailed view of 1) HalB residues stabilizing incoming dATP substrate and 2) polder omit map contoured at 5.5 σ of Y227 and non-hydrolyzable dATP. HalB adenine discrimination occurs through residues T129 that coordinate sequence-specific contacts with the adenine nucleobase Watson-Crick edge, and L60 and R128 that restrict guanine and pyrimidine base recognition. Detailed view of 3) newly synthesized ODA bound to HalB C-terminal tyrosine residue and 4) polder omit map contoured at 3.5 σ of dAMP bound to Y227. Expression of HalB used in this figure was from R. bacterium QY30. Data shown are representative of at least three independent experiments.

Extended Data Figure 3 |. HalA substrate binding specificity.

a, R. bacterium QY30 HalA was expressed as an N-terminal 6×His fusion with untagged HalB and purified by Ni-NTA in the presence of DDM detergent and separated by size exclusion chromatography in the presence of GDN detergent. b, Coomassie-stained SDS-PAGE analysis of fully purified HalA. c, Electrophoretic mobility shift assay of HalA–ODA complex formation with indicated FAM-labeled (abbreviated as F) DNA substrates. Expression of HalA used in this figure was from R. bacterium QY30. Data shown are representative of at least three independent experiments.

Extended Data Figure 4 |. Cryo-EM data processing for the HalA–ODA complex.

a, Cryo-EM data processing scheme. To facilitate visual comparison, each volume’s hand was flipped wherever necessary to match the hand of the final reconstruction. b, Left, example motion-corrected micrograph, subjected to a 5-Å long-pass filter. Right, the same image subjected to CryoSPARC’s Micrograph Denoiser. Particles present in the final stack are circled in green. c, Example 2D class averages from the particle curation stage. d, Gold-standard Fourier shell correlation (GSFSC) curves after FSC-mask auto-tightening, as produced by CryoSPARC. The horizontal black line represents the FSC = 0.143 threshold. e, Local resolution of the final map (unsharpened and without density modification).

Extended Data Figure 5 |. Structural analysis of HalA.

a, Isolated HalA protomer showing regions involved in membrane interaction and formation of the ion channel (transmembrane domain α6-α10) and regions involved in ODA binding (pentapeptide repeat domain). b, Isolated HalA tetrameric ion channel region used for structural homolog comparison. c,d, Protein structural homology of HalA transmembrane domain against all entries in PDB showing the DALI Z score and FoldSeek E value of the top 130 and 1200 hits respectively. Entries with annotated ion channel activity are highlighted. e, HalA comparison to structurally related and well-characterized ion channel proteins, and corresponding PDB entry IDs are in parentheses. 2TM domain of HalA and ion channel proteins were used to highlight their structural similarities within the ion selectivity filter and transmembrane regions. Highlighted 2TM domains are shown as individual proteins next to the full structure, or overlayed altogether with HalA. f, Detailed view of HalA interacting residues required for oligomerization. g, Detailed view of the ion channel comparing the closed state (cryo-EM structure) and AlphaFold model of HalA. The AlphaFold model shows a wider conformation of the ion conduction pathway, suggestive of conformational rearrangements that would lead to an open state in the absence of ODA. h, Detailed view of HalA residues involved in recognition of ODA binding specificity. Sequence-specific interactions occur between HalA residues T83, T104, and T124 with base dA5; S141 and N143 with base dA3; and side-chains W26, K109, and F139 facilitate additional interactions that control selective ODA recognition and restrict guanine and pyrimidine base recognition.

Extended Data Figure 6 |. HalA sequence analysis and growth assay.

a, Structure guided multiple sequence alignment of HalA protein homologs from indicated bacterial species. Shading indicates degree of residue conservation. b, Bacterial growth assay of E. coli expressing HalA mutants required for channel assembly with and without HalB. Expression of HalA used in this figure was from R. bacterium QY30. Data shown are representative of at least three independent experiments.

Extended Data Figure 7 |. Flow cytometry and live cell imaging analysis of HalA activation.

a,b, Flow cytometry quantification of DiBAC4-positive cells (a) and PI-positive cells (b) from E. coli expressing HalA or HalAB treated with glucose, polymyxin B, or induced with arabinose for the indicated times. c, Left, gating strategy: bacterial cells were selected using side scatter height versus forward scatter height (SSC-H vs. FSC-H) and side scatter height versus width (SSC-H vs. SSC-W). Right, representative plots from cells treated with polymyxin B, or cells expressing arabinose-inducible plasmids containing HalA or HalAB. Flow cytometry data shown are collected from 100,000 events and are representative of at least two independent experiments. d,e, Live cell imaging analysis of membrane depolarization using DiBAC4 in E. coli containing plasmids expressing Hailong defence system and infected with (d) WT or escape mutant phage SECφ4 at a calculated MOI of 10 or 50, or (e) no phage infection. Scale bars, 5 μm. f, Cellular localization of HalA visualized in E. coli expressing EGFP-HalA and periplasimic-mCherry. Scale bars, 1 μm. Expression of HalA and HalAB used in this figure was from R. bacterium QY30. Live cell imaging data shown are representative of at least three independent experiments.

Extended Data Figure 8 |. Identification and structural characterization of Hailong phage escape mutants.

a, Full analysis of isolated phage escape mutants. Left, sequenced genes containing the indicated point mutations are highlighted in light green within the SECφ4 genome. Right, representative plaque assays and heatmap illustrating fold defence of E. coli expressing Hailong from E. coli STEC1178 and challenged with WT SECφ4 phage and SECφ4 escape mutant phages. b, Multiple sequence alignment of SECφ4 escape mutant gp43 proteins from indicated phage and bacterial homologs. Shading indicates degree of residue conservation. c, AlphaFold modeled structure of SECφ4 Exo (left) and comparison with human EXO5 (right) with (PDB ID: 7LW9) and without DNA (PDB ID: 7LW7). Boxes highlight nuclease active site. d, Detailed view of nuclease active residues in SECφ4 Exo (left) and human EXO5 (right).

Extended Data Figure 9 |. Phylogenetic and biochemical analysis of SECφ4 Exo.

a, Phylogenetic analysis of ~1,700 SECφ4 Exo sequence homologs obtained using NCBI BLAST. b, Genera of phages encoding SECφ4 Exo. c, SECφ4 gp43 DNA exonuclease (Exo) was expressed as an N-terminal 6xHis-SUMO2 fusion and purified by Ni-NTA and separated by size exclusion chromatography. d, Coomassie-stained SDS-PAGE analysis of fully purified SECφ4 Exo. e, Analysis of 5 nt fluorescein-labeled deoxynucleotide substrates (F-dA, dA-F, dC-F), 5 nt single-stranded RNA (rA-F), or 20 bp thymidine-labeled double-stranded DNA (dA-dT-F). Direction of fluorescein (labeled as green F) tagged to the oligonucleotide indicates either a 5′-tagged DNA (indicating a free 3′ end) or a 3′-tagged DNA (indicating a free 5′ end). Data shown are representative of at least three independent experiments. f, AlphaFold modeled structure of SECφ4 Exo with escape mutant residues within scaffolding regions.

Figure 1 |. Diverse Hailong systems protect bacteria from phage infection.

a, Representative Hailong operons and gene neighborhoods from S. pruni NBRC 15498, E. coli 300073, P. manganicus JH-7, and E. yinggardensis WSM1721. Defence systems were annotated using Defence Finder. b, Schematic of HalA and HalB proteins. c, Frequency of HalA and HalB found in all Hailong systems. d, Genera of bacteria encoding Hailong. e, Representative plaque assays of E. coli expressing empty vector (control) or Hailong from E. coli STEC1178 and challenged with phage SECφ6. Serial dilution factors represent phage dilution to visualize plaque forming units. f, Heatmap illustrating fold defence of E. coli expressing Hailong from gammaproteobacteria (E. coli, Klebsiella, Yersinia) or alphaproteobacteria (Rhodobacteraceae) species and challenged with indicated phages. Data shown are representative of three independent experiments.

Figure 2 |. HalB synthesizes oligodeoxyadenylate.

a, Bacterial growth assay of a 10-fold dilution series of E. coli containing arabinose-inducible plasmids expressing HalA alone, HalA with HalB, or HalA with a HalB catalytically inactive mutant (DDAA, D21A/D23A). b, Crystal structure of HalB in complex with the nucleotide product oligodeoxyadenylate (ODA). Inset depicts cartoon representation of tetrameric structure. c, Polder omit map contoured at 3.4 σ of the first two deoxyadenosine molecules of ODA. d, Coomassie-stained SDS-PAGE analysis of purified HalB with and without nuclease P1 treatment. e, Urea-PAGE analysis of HalB ODA synthesis reactions labeled with α³²P-dATP and treated with and without proteinase K. f, Urea-PAGE analysis of HalB substrate specificity using dNTPs and α³²P-labeled dNTPs as indicated. g, Urea-PAGE analysis of ODA cleavage from the HalB–ODA complex using α³²P-labeled dATP treated with and without proteinase K or E. coli soluble lysate fraction. For additional controls, see Extended Data Figure 2d. Expression of HalB used in this figure was from R. bacterium QY30. Data shown are representative of at least three independent experiments. For gel source data, see Supplementary Figure 1.

Figure 3 |. Mechanism of HalB ODA synthesis.

a, Crystal structure of catalytically inactive HalB from R. bacterium QY30 revealing re-positioning of the flexible C-terminal tail within the NTase active site. b, Structure of HalB as in Fig. 2b with a dashed line indicating the projected position of the covalent attachment of ODA to the protein flexible C-terminal tail. c, Top, overview of HalB D21A/D23A active site and the C-terminal tail. Bottom, overview of WT HalB active site and a detailed view of residues interacting with ODA. d, Top, schematic of protein–protein contacts with HalB and its C-terminal tail. Bottom, schematic of protein–DNA contacts in the HalB–ODA complex. HalB residues are highlighted in blue. e, Schematic of the step-by-step mechanism of ODA synthesis: Step 1) A tyrosine in the C-terminal tail acts as a protein primer within the active site. Step 2) dATP binding is coordinated by metal ions and the tyrosine repositions to attack the α-phosphate. Step 3) Tyrosine forms a covalent bond with dAMP and the protein–DNA complex repositions for the next attack on dATP. Step 4) Fully synthesized ODA. f, Coomassie-stained SDS-PAGE analysis of purified HalB mutants demonstrating the NTase active site, C-terminal tail, and tyrosine priming residue are required for ODA synthesis and covalent attachment. g, Urea-PAGE analysis of mutant HalB ODA synthesis reactions labeled with α³²P-dATP and treated with and without proteinase K. h, Bacterial growth assay of E. coli expressing either HalA alone, or HalA with HalB mutants. Expression of HalB used in this figure was from R. bacterium QY30. Data shown are representative of at least three independent experiments. Manganese ions are depicted as purple spheres. For gel source data, see Supplementary Figure 1.

Figure 4 |. Structure and function of HalA.

a, Electrophoretic mobility shift assay of ODA binding to HalA–ODA complex. b, Quantification of binding affinity of HalA–ODA complex with different lengths of ODA. Data are presented as mean ± s.d. from n=3 independent experiments. c, EMSA analysis of HalA–ODA complex binding to different 5 nt single-stranded deoxy-nucleotide substrates (dA, dT, dC, dN), 5 nt single-stranded RNA (rA), or 20 bp double-stranded DNA (dA-dT). d, Cryo-EM structure of HalA from R. bacterium QY30 in complex with ODA. e, Zoom-in view of the ODA binding pocket at the interface between two HalA pentapeptide repeat domains and residues interacting with ODA. f, Schematic of protein–DNA contacts in the HalA–ODA complex. g, Structural model of HalA activation. Cryo-EM structure of ODA-bound HalA in a closed conformation and an AlphaFold modeled apo structure of HalA showing a wider diameter along the channel, suggesting a more open conformation. h, Bacterial growth assay of E. coli expressing HalA mutants with and without HalB. i, Heatmap illustrating fold defence of E. coli expressing E. coli STEC1178 Hailong operons with HalA channel assembly mutants and challenged with indicated phages. K232E and ΔC mutants used for the phage challenge assay are equivalent to R. bacterium QY30 K211E mutant and truncation of residues 352–359 respectively. j, Kernel density estimates of fluorescence distributions measured by flow cytometry (left) and quantification of membrane depolarization (right) of E. coli cells containing arabinose-inducible plasmids expressing HalA alone or HalAB, treated with glucose or arabinose for 8 h and incubated with the voltage-sensitive dye DiBAC4. Bar graph data are presented as means from 4 independent experiments. Polymyxin B was used as a positive control for membrane depolarization. k, Fluorescence microscopy analysis of ion flow using cell permeant fluorescent Na⁺ indicator Sodium Green Tetraacetate in E. coli containing arabinose-inducible plasmids expressing HalA alone or HalAB. Scale bars, 5 μm. l, Live cell imaging analysis of membrane depolarization using DiBAC4 in E. coli containing plasmids expressing Hailong defence system and infected with WT or escape mutant phage SECφ4. Scale bars, 1 μm. Hailong systems used in this figure were from R. bacterium QY30 (a–h,j,k) or E. coli STEC1178 (i,l). Data shown are representative of at least three independent experiments.

Figure 5 |. Phage DNA exonuclease activates Hailong.

a, Representative plaque assays of E. coli containing empty vector (control) or Hailong from E. coli STEC1178 plasmids and challenged with WT SECφ4 phage and SECφ4 escape mutant phage in tenfold serial dilution. b, Schematic depicting the genomic loci of the SECφ4 exonuclease gp43 (Exo) and annotation of gp43 escape mutations conferring resistance to Hailong anti-phage defence. SSB, single-stranded DNA binding protein. c, DNA cleavage assay of ODA with purified WT and mutant Exo from phage SECφ4. d, Bacterial growth assay of E. coli transformed with two separate plasmids expressing Hailong from R. bacterium QY30 and SECφ4 WT or mutant Exo. e, Model of Hailong anti-phage defence system signaling and phage restriction. All data shown are representative of at least three independent experiments.