The Panoptes system uses decoy cyclic nucleotides to defend against phage

Abstract

Bacteria combat phage infection using antiphage systems and many systems generate nucleotide-derived second messengers upon infection that activate effector proteins to mediate immunity¹. Phages respond with counter-defences that deplete these second messengers, leading to an escalating arms race with the host. Here we outline an antiphage system we call Panoptes that indirectly detects phage infection when phage proteins antagonize the nucleotide-derived second-messenger pool. Panoptes is a two-gene operon, optSE, wherein OptS is predicted to synthesize a nucleotide-derived second messenger and OptE is predicted to bind that signal and drive effector-mediated defence. Crystal structures show that OptS is a minimal CRISPR polymerase (mCpol) domain, a version of the polymerase domain found in type III CRISPR systems (Cas10). OptS orthologues from two distinct Panoptes systems generated cyclic dinucleotide products, including 2',3'-cyclic diadenosine monophosphate (2',3'-c-di-AMP), which we showed were able to bind the soluble domain of the OptE transmembrane effector. Panoptes potently restricted phage replication, but phages that had loss-of-function mutations in anti-cyclic oligonucleotide-based antiphage signalling system (CBASS) protein 2 (Acb2) escaped defence. These findings were unexpected because Acb2 is a nucleotide 'sponge' that antagonizes second-messenger signalling. Our data support the idea that cyclic nucleotide sequestration by Acb2 releases OptE toxicity, thereby initiating inner membrane disruption, leading to phage defence. These data demonstrate a sophisticated immune strategy that bacteria use to guard their second-messenger pool and turn immune evasion against the virus.

Fig. 1. The Panoptes system restricts phage.

a, Accession number and genome coordinates for V. navarrensis Panoptes operon genes, optS and optE. Domain architectures and accession numbers for encoded proteins are shown. b, Heat map of fold defence provided by VnOptSE for a panel of diverse phages from the BASEL collection. Escherichia coli expressing VnOptSE were challenged with phages and fold defence was calculated for each phage by dividing the efficiency of plating (in PFU per millilitre) on empty vector by the efficiency of plating on VnOptSE-expressing bacteria. Family names are above each indicated set of phages; numbers indicate BASEL phage numbers. c, Efficiency of plating of indicated phages infecting E. coli expressing a plasmid with either VnOptSE or an empty vector (EV). Data represent the mean ± s.e.m. of n = 3 biological replicates, shown as individual points. TM, transmembrane. Scale bar, 25 amino acids. Source Data

Fig. 2. OptS is a minimal CRISPR polymerase.

a, The crystal structure of the KpOptS apoprotein shows a putative tetrameric architecture. Two rotated views are depicted with each protomer shown in a different colour. b, Structure of the isolated protomer (monomer) highlighting overall domain fold and putative active site (orange sticks for Asp6 and Asp57). c, Superposition of type III Cas10 CRISPR polymerase palm domain (Protein Data Bank (PDB) 6O75, light orange) with KpOptS (purple). The remainder of the Cas10 structure depicted in grey highlights the minimal nature of the mCpol. d, Superposition of GGDEF-containing enzyme PleD (PDB 1W25, light blue) with KpOptS, with unaligned regions shown in grey. e, HPLC analysis of an ATP chemical standard compared with the product of KpOptS when incubated with ATP alone. f, HPLC analysis of 2′,2′-c-di-AMP, 2′,3′-c-di-AMP and 3′,3′-c-di-AMP chemical standards, compared with the product of KpOptS. g, Chemical structure of 2′,3′-c-di-AMP. h, Bottom, mass spectrometry spectra of a 2′,3′-c-di-AMP standard. Top, the KpOptS ATP-dependent product. A value of 330.05 m/z indicates a substantial proportion of the doubly charged species. i, Magnified view of the KpOptS putative active site highlighting catalytic Asp6 and Asp57 (orange sticks). j, HPLC analysis of the reaction products when wild-type (WT) or mutant KpOptS is incubated with ATP, compared with an ATP chemical standard. k, Thermal shift data showing the melt profile for a soluble version of VnOptE incubated with several c-di-AMP analogues. Buffer indicates protein alone with no extra cyclic dinucleotide added. For e,f,j and k, data are representative of n = 3 biological replicates. l, Quantification of intracellular 2′,3′-c-di-AMP from VnOptSE or empty vector-expressing E. coli. Data represent the mean ± s.e.m. of n = 3 biological replicates, shown as individual points. A two-sided Student’s t-test was used. *P < 0.05. LOD, limit of detection; mAU, milli-absorbance unit. Source Data

Fig. 3. Panoptes orthologues have altered cyclic nucleotide products.

a, Crystal structure of KpOptS co-crystallized with non-hydrolysable ATP analogue ApCpp. Only two protomers are shown to highlight the orientation of analogues between two binding pockets (insufficient density to model the ribose and nucleobase for the second molecule of ApCpp). Inset, dashed line used to visualize the angle and close distance between reactive groups in the two substrate molecules. Divalent cation (probably Mg²⁺) shown as orange spheres. b, HPLC analysis of nucleoside triphosphate (NTP) (ATP, GTP, UTP, CTP) standards compared with the products of VnOptS incubated with NTP. Arrows indicate known products verified through further experiments (Extended Data Figs. 2, 5 and 6). c, Thermal shift data showing the melt profile for a soluble version of VnOptE incubated with the main cyclic dinucleotides produced by VnOptS in vitro. Buffer indicates protein alone with no extra cyclic dinucleotide added. d, Nucleobase recognition site of KpOptS with ApCpp highlighting potential interactions with Ser30 and Asn34. The metal/phosphate coordinating residues are in one protomer and the base recognition residues are in the other. e, HPLC analysis of reactions of KpOptS (wild type or S30N mutant) with ATP and GTP showing altered product ratio. Inset, alignment of KpOptS beginning at Ile29 and VnOptS beginning at Ile30. Yellow asterisks indicate relevant positions of KpOptS Ser30 and Asn34 from d. f, HPLC analysis of reactions of VnOptS (wild type or N31S mutant) with ATP and GTP showing altered product ratio. For b,c,e and f, data are representative of n = 3 biological replicates. Source Data

Fig. 4. Acb2 is necessary and sufficient to activate the Panoptes system.

a, Schematic of phage T4 escape mutations from this paper (purple) and a previously investigated cyclic dinucleotide binding-site mutation (black^,). b, Efficiency of plating of T4 phages with the indicated genotype on E. coli expressing an empty vector or the VnOptSE operon from a plasmid. Images are representative of n = 3 biological replicates. c, Colony formation of E. coli expressing empty vector or VnOptSE from a plasmid and indicated empty vector or acb2 allele on an IPTG-inducible plasmid. Data are mean ± s.e.m. of n = 3 biological replicates, shown as individual points. A two-sided Student’s t-test was used. *P < 0.05. d, ITC of 2′,3′-c-di-AMP interacting with Acb2. Data are n = 3 technical replicates. K_d, ∆H (change in enthalpy) and ∆S (change in entropy) were determined by generating a global fit for the three indicated replicates. Raw data are shown in Extended Data Fig. 8. e, Colony formation of E. coli co-expressing indicated VnOptSE genes from IPTG-inducible (plasmid 1) and arabinose-inducible (plasmid 2) plasmids. CD, catalytically dead, OptS^D58A. Data represent the mean ± s.e.m. of n = 3 biological replicates, shown as individual points. f, Fluorescence microscopy of E. coli co-expressing the indicated VnOptSE gene(s) from one plasmid and an empty vector or acb2 allele on a second IPTG-inducible plasmid 0 min, 30 min, 60 min and 90 min after induction with IPTG. Bacteria were stained with membrane stain FM1-43 (green) and propidium iodide (PI; red). Images are representative of n = 2 biological replicates. Scale bar, 5 μm. Source Data

Fig. 5. Phage anti-defence proteins activate Panoptes.

a, Colony formation of E. coli expressing an empty vector or the VnOptSE operon from one plasmid and empty vector or indicated phage protein on a second IPTG-inducible plasmid. Data represent the mean ± s.e.m. of n = 3 biological replicates, shown as individual points. b, Venn diagram of pairwise co-occurrence for mCpol and SMODS protein domains across prokaryotes. The P value (purple) represents the significance of pairwise co-occurrence from Fisher’s exact test. c, Normalized phage defence provided by VnOptSE or CBASS for wild-type T4 or T4 ∆acb2 phage. Escherichia coli expressing VnOptSE or CBASS were challenged with phages and fold defence was calculated for each phage by dividing the efficiency of plating (in PFU per millilitre) on empty vector by the efficiency of plating on VnOptSE or CBASS-expressing bacteria. Data represent the mean ± s.e.m. of n = 3 biological replicates, shown as individual points. Efficiency of plating raw data are shown in Extended Data Fig. 7c (VnOptSE) and Extended Data Fig. 7f (CBASS). A two-sided Student’s t-test was used. *P < 0.05. d, Model depicting the Panoptes system in steady state (top) and during infection by a phage that expresses Acb2 (bottom). OptS constitutively synthesizes cyclic nucleotides, which bind OptE and restrain activity. Upon infection, Acb2 sequesters the OptS-derived cyclic nucleotides from OptE. Activation of OptE ultimately leads to membrane disruption and restriction of phage replication. Source Data

Extended Data Fig. 1. Panoptes protects against phages from the Straboviridae family.

(a) Above: growth curve of E. coli expressing the indicated plasmid. Arrows indicate the time each culture was infected with phage T4 at the indicated multiplicity of infection (MOI). Below: efficiency of plating of the phage present in each sample at the indicated time points. Data represent the mean ± SEM of n = 3 biological replicates. (b) Efficiency of plating of indicated phages infecting E. coli expressing a plasmid with either VnOptSE or an empty vector (EV). Data represent the mean ± SEM of n = 3 biological replicates, shown as individual points. (c) Domain architectures of Klebsiella pneumoniae Panoptes operon genes, optS and optE. (d) Heatmap of fold defense provided by KpOptSE for a panel of diverse phages from the BASEL collection. E. coli expressing KpOptSE were challenged with phages and fold defense was calculated for each phage by dividing the efficiency of plating (in PFU/mL) on EV by the efficiency of plating on KpOptSE-expressing bacteria. Family names are above each indicated set of phages. (e) Efficiency of plating of indicated phages infecting E. coli expressing a plasmid with either KpOptSE or an EV. Data represent the mean ± SEM of n = 4 biological replicates, shown as individual points. A two-sided Student’s t-test was used. ns = not significant. (f) Efficiency of plating of indicated phages infecting E. coli expressing a plasmid with either KpOptSE or an EV. Data represent the mean ± SEM of n = 3 biological replicates, shown as individual points. Source Data

Extended Data Fig. 2. Oligomeric state analysis and biochemical profiling of KpOptS.

(a) Multiple sequence alignment of representative Panoptes OptS, type-III CRISPR polymerase palm domains, and GGDEF palm domains. Residues important for catalysis are indicated with black asterisks. Orange asterisk indicated nucleobase recognition residues for Kp and VnOptS. (b–e) The KpOptS tetramer contains three possible dimer (pair of protomers) interfaces shown in order of decreasing buried surface area (determined with PISA analysis, see Supplementary Table 1). (f) Size exclusion chromatography reveals a monodisperse signal for KpOptS with an elution volume consistent with an oligomer >29 kDa (monomer = 14 kDa). Grey dots indicate molecular weight standards: Blue dextran, 46.2 mL, 2000 kDa; Ferritin, 59.8 mL, 440 kDa; Aldolase, 71.0 mL, 158 kDa; Conalbumin, 78.9 mL, 75 kDa; Ovalbumin, 85.3 mL, 43 kDa; Carbonic anhydrase, 91.5 mL, 29 kDa; Ribonuclease A, 98.6 mL, 13.7 kDa; Aprotinin, 104.4 mL, 6.5 kDa. The trace is representative of at least two biological replicates. (g, h) Mass photometry results showing distribution of sizes detected for KpOptS in solution at 50 nM. (i) HPLC analysis of mixture of 4×NTPs (ATP, GTP, UTP, CTP) compared with the product of KpOptS when incubated with NTPs. Traces representative of at least three biological replicates. (j) HPLC analysis of GTP compared with the product of KpOptS when incubated with GTP. Traces representative of at least three biological replicates. (k) HPLC analysis of ATP and GTP compared with the product of KpOptS when incubated with ATP and GTP. Traces representative of at least three biological replicates. (l) HPLC analysis of 2′,2′-cGAMP, 2′,3′-cGAMP, 3′,2′-cGAMP, and 3′,3′-cGAMP chemical standards compared with the product of KpOptS incubated with ATP and GTP. Traces representative of at least three biological replicates. (m) P1 nuclease and CIP treatment of KpOptS ATP-only product. Traces representative of at least three biological replicates. Source Data

Extended Data Fig. 3. Thermal stability assays for VnOptE.

(a) Thermal shift data showing the melt profile for a soluble version of VnOptE incubated with several cGAMP linkage isomers. Data representative of nine independent reactions per tested condition (3 replicate plates, 3 technical replicates per plate). (b) Thermal shift data showing the melt profile for a soluble version of VnOptE incubated with several c-di-GMP linkage isomers. Data representative of nine independent reactions per tested condition (3 replicate plates, 3 technical replicates per plate). (c) Thermal shift data showing the melt profile for a soluble version of VnOptE incubated with several cUAMP linkage isomers. Data representative of nine independent reactions per tested condition (3 replicate plates, 3 technical replicates per plate). (d) Thermal shift data showing the melt profile for a soluble version of VnOptE incubated with several cyclic nucleotides which have been shown to be important in type III CRISPR signaling. (e) Table of melting temperatures and thermal shift values for all tested cyclic nucleotides with VnOptE. Data representative of nine independent reactions per tested condition (3 replicate plates, 3 technical replicates per plate). Error in values is reported as standard deviation. Source Data

Extended Data Fig. 4. AbCap5 is a sensor of 2′,3′-c-di-AMP.

(a) Visualization of DNA degradation by AbCap5 when incubated with either 2′,3′-c-di-AMP, 3′,3′-c-di-AMP, 2′,2′-cGAMP, 2′,3′-cGAMP, 3′,2′-cGAMP, or 3′,3′-cGAMP. Data are representative images of n = 2 technical replicates. (b–d) Visualization of DNA degradation by AbCap5 when incubated with a dilution series of 2′,3′-c-di-AMP (b) or extracted nucleotides from E. coli expressing an empty vector (EV; c) or the VnOptSE operon (d). Data are representative images of n = 3 biological replicates. (e) Fluorescence-based DNase activity assay of AbCap5 incubated with the indicated concentration of 2′,3′-c-di-AMP. Cleavage of the DNA substrate results in fluorescence and was measured as relative fluorescence units (RFU), which was monitored over time. The gray shaded area represents the linear range of the assay. Data are the mean ± SEM of n = 3 technical replicates and are representative of n = 2 biological replicates. (f) Fluorescence-based DNase activity assay of AbCap5 incubated with nucleotide extracts from E. coli expressing either an EV or VnOptSE. Cleavage of the DNA substrate results in fluorescence and was measured as relative fluorescence units (RFU), which was monitored over time. The gray shaded area represents the linear range of the assay. Data are the mean ± SEM of n = 3 technical replicates and are representative of n = 2 biological replicates. (g) Standard curve of velocity (RFU/min) versus concentration of 2′,3′-c-di-AMP in the reaction well. Velocities were determined for each 2′,3′-c-di-AMP concentration from the points within the linear range shown in (e). (h) Quantification of 2′,3′-c-di-AMP from VnOptSE or EV-expressing E. coli in the reaction well. Data represents the mean ± SEM of n = 3 biological replicates, shown as individual points. A two-sided Student’s t-test was used. *P  <  0.05. Source Data

Extended Data Fig. 5. HPLC analysis of VnOptS catalytic activity.

(a) Electron density map for the ligands in the KpOptS structure at two sigma cutoff values (1 above, 2 below) showing the weak density quality for the ribose and adenine base in ApCpp but strong density supporting analog binding through phosphate-metal ion interactions at all four possible binding sites within the tetramer. (b) ApCpp interactions with divalent cation (likely Mg²⁺) through catalytic residues Asp6 and Asp57. (c) HPLC analysis of the product of VnOptS when incubated with ATP. Traces representative of at least three biological replicates. (d) HPLC analysis of the product of VnOptS when incubated with GTP. Traces representative of at least three biological replicates. (e) HPLC analysis of the product of VnOptS when incubated with UTP. Traces representative of at least three biological replicates. (f) HPLC analysis of the product of VnOptS when incubated with CTP. Traces representative of at least three biological replicates. (g) HPLC analysis of the product of VnOptS when incubated with ATP and GTP. Traces representative of at least three biological replicates. (h) HPLC analysis of the product of VnOptS when incubated with ATP and UTP. Traces representative of at least three biological replicates. (i) HPLC analysis of 3′,3′-cUAMP, 2′,3′-cUAMP, and 3′,2′-cUAMP chemical standards compared with the product of VnOptS incubated with ATP and UTP. Traces representative of at least three biological replicates. (j) HPLC analysis of the reaction products when wild-type or mutant VnOptS is incubated with ATP. Traces representative of at least three biological replicates. (k) HPLC analysis of P1 nuclease and CIP treatment of VnOptS product of incubation with ATP and GTP (3′,2′-cGAMP). Traces representative of at least three biological replicates. (l) HPLC analysis of P1 nuclease and CIP treatment of VnOptS product of incubation with ATP and UTP (3′,2′-cUAMP). Traces representative of at least three biological replicates. Source Data

Extended Data Fig. 6. Mass spectrometry data for VnOptS cyclic dinucleotides.

(a–c) MS spectra of chemical standard (bottom) and the VnOptS nucleotide-dependent product (top) for 2′,3′-c-di-AMP, 3′,2′-cGAMP, and 3′,2′-cUAMP. (d–e) Chemical structure of 3′,2′-cyclic guanosine monophosphate-adenosine monophosphate (3′,2′-cGAMP) and 3′,2′-cyclic uridine monophosphate-adenosine monophosphate (3′,2′-cUAMP). Source Data

Extended Data Fig. 7. Phage T4 escapers evade Panoptes-mediated defense.

(a) Efficiency of plating of T4 parent and escaper phages on E. coli expressing an empty vector (EV) or the VnOptSE operon from a plasmid. Images are representative of n = 3 biological replicates. (b) Efficiency of plating of T4 parent phages and corresponding escaper mutant phages (Escaper, T4.X.Y, where X indicates the parent phage number and Y indicates the escaper phage number) on E. coli expressing an EV or the VnOptSE operon from a plasmid. Data are the mean ± SEM of n = 3 biological replicates, shown as individual points. (c) Efficiency of plating of T4 phages with the indicated genotype on E. coli expressing an EV or the VnOptSE operon from a plasmid. Data are the mean ± SEM of n = 3 biological replicates, shown as individual points. (d) Efficiency of plating of T4 phages with the indicated genotype on E. coli expressing an EV or the VnOptSE operon from a plasmid. Data are the mean ± SEM of n = 3 biological replicates, shown as individual points. (e) Efficiency of plating of T4 phages with the indicated genotype on E. coli expressing an EV or the KpOptSE operon from a plasmid. Data are the mean ± SEM of n = 3 biological replicates, shown as individual points. (f) Efficiency of plating of T4 phages with the indicated genotype on E. coli expressing an EV or the CBASS operon from a plasmid. Data are the mean ± SEM of n = 3 biological replicates, shown as individual points. For c-f, a two-sided Student’s t-test was used. *P  <  0.05, **P  <  0.001. Source Data

Extended Data Fig. 8. Acb2 binds the product of KpOptS.

(a) HPLC analysis of the ability of Acb2 to bind the main products of KpOptS (ATP and GTP reaction). Traces representative of at least three biological replicates. (b) HPLC analysis of the ability of Acb2 to bind and release the product of KpOptS when treated with proteinase K (ATP only reaction). Traces representative of at least three biological replicates. (c) HPLC analysis of the ability of mutant Acb2 to bind the ATP-derived product of KpOptS. Traces representative of at least three biological replicates. (d–f) Replicates of ITC assays to test binding of 2′,3′-c-di-AMP to Acb2. Source Data

Extended Data Fig. 9. mCpol distribution and gene co-occurrences.

(a) Heat map depicting the phyletic distribution (degree of presence) of selected conflict systems in prokaryotic lineages. Rainbow coloring corresponds to the percentage of presence in a lineage as indicated by the key to the left of the figure: dark gray indicates the absence of a system in each lineage. The phyletically restricted CATRA (Caspase and TPR repeat-associated)- and GSDM (Gasdermin)-containing systems are included as comparisons to the similarly restricted mCpol system. (b) Tables depict: (top) percentages of degree of presence in all sampled prokaryotic lineages (self-cell/diagonal) and pairwise co-occurrence (off-diagonal) for selected systems across all prokaryotes and (bottom) p-values for the significance of pairwise co-occurrences. (c) Venn diagram representing selected system pairs from the above tables. For b-c, a two-sided exact Fisher’s test was performed. Co-occurrences with p-values of less than 1e-06 (passing the stringent co-occurrence criterion) are shaded (red) or provided below the diagram (red), respectively. (d) Conserved mCpol gene neighborhood associations, with individual genes depicted as box arrows, colored by their discrete domains. GenBank accessions of mCpol genes are provided below the neighborhoods. “Core” mCpol-encoding operons are separated from discrete co-associating systems by blue circles.