Cyclic nucleotide second messengers are used in all domains of life to amplify viral infection signals and activate cellular defences. In prokaryotes, CBASS (cyclic nucleotide-based antiphage signalling system) and type III CRISPR-Cas systems generate a range of cyclic nucleotides which bind and allosterically activate effector proteins to mount an anti-viral response. Viruses have evolved counter measures to antagonise these signalling pathways in the form of cyclic nucleotide sponges and phosphodiesterases that sequester or degrade these molecules to subvert immunity. Recently, the Panoptes system was shown to function as a guard against these viral tactics. The type I Panoptes polymerase, mCpol, generates cyclic dinucleotides as decoy molecules that, when sequestered by phage proteins, results in the activation of the membrane-permeabilising effector 2TMβ to halt the phage infection cycle. Here, we investigate the type II Panoptes system, demonstrating that it generates cyclic tri-adenylate (cA₃) to maintain a CRISPR-associated Rossmann fold-transmembrane (CARF-TM) effector in an inactive, dimeric state. When cA₃ is sequestered or degraded, the CARF protein undergoes conformational changes. In vivo, the absence of cA₃ results in membrane disruption and growth arrest. Type II Panoptes provides defence against phages that express the cA₃-degrading enzyme Acb1; phage escapers introduce mutations into the acb1 gene to avoid triggering the Panoptes system. These findings expand our understanding of the guard systems that constitute a fascinating component of the bacterial immune system.

Fig 1. Organisation and mechanism of the Panoptes System.

A. Operon organisation and major cyclic nucleotide signal for representative type I and II Panoptes systems. B. In type I Panoptes systems, mCpol is constitutively active, generating signalling molecules that inactivate the cognate effector, 2TMβ. When an infecting phage expresses a sponge protein, the cyclic oligonucleotide pool is depleted. The Panoptes effector oligomerises, leading to membrane disruption and growth arrest or cell death.

https://doi.org/10.1371/journal.pbio.3003934.g001

Type II Panoptes signals via cyclic tri-adenylate (cA₃)

While the characterised type I Panoptes systems synthesise cyclic dinucleotides and utilise a 2TMβ-family effector, type II systems with a CARF-TM effector remain unstudied. We designed synthetic genes encoding mCpol and CARF-TM from the cyanobacterium Cylindrospermum stagnale PCC7417 (Cst) for expression in E. coli and purified the recombinant proteins. The Cst mCpol protein was tested for nucleotide cyclase activity and shown to synthesise 3′3′3′-cA₃ in the presence of ATP and Mg²⁺ (Fig 2). The presence of other NTPs did not alter the reaction product. Two further mCpol enzymes associated with a CARF-TM effector, from Sphingomonas (Sph) and Ahniella affigens (Aaf), also synthesised 3′3′3′-cA₃ as shown by HPLC analysis and co-injection with synthetic 3′3′3′-cA₃ (S1 Fig). The turnover number for all three mCpol enzymes was approximately 1 h⁻¹ in vitro (S2 Fig), making mCpol significantly slower than Cas10 from Type III CRISPR systems [25]. Although in vitro reaction rates should be treated with caution, these data suggest mCpol may generate relatively low concentrations of cA₃ in cells. This could be consistent with a function as a decoy molecule for phage sponges and phosphodiesterases, providing sensitivity to changes in cA₃ concentration in vivo. Residues D11 and D63 of Cst mCpol correspond to the canonical metal-binding residues of Cas10 and Palm family polymerases more generally. We mutated D63 and the neighbouring D64 to generate Cst mCpol[D63N/D64N]. This variant was completely inactive, as expected (Fig 2).

Type II Panoptes is active in vivo and responds to cA₃ levels in cells

To test the activity of Panoptes type II in vivo, we performed transformation assays, where the effector was introduced to a strain containing the mCpol cyclase. A serial dilution of the transformation mixture was then plated onto selective medium and both cyclase and effector expression were induced. The plates were incubated overnight and inspected for colony growth. If the cyclase and effector combination give rise to a growth defect or kill the host cell, fewer colonies are observed compared to controls. We used Cst mCpol protein as the cyclase with either its cognate CARF-TM effector or NucC, a previously characterised dsDNA-endonuclease associated with CBASS and type III CRISPR defence [26,27]. NucC is activated by cA₃ and was included as a control to detect cA₃ production. In this experiment, mCpol was expressed under control of the T7 promoter (IPTG inducible), while NucC or CARF-TM were expressed under control of the arabinose-inducible pBAD promoter. As expected, expression of NucC with the catalytically inactive mCpol resulted in efficient transformation, as NucC is inactive in the absence of cA₃ (Fig 3A). In contrast, expression of the CARF-TM effector alone resulted in toxicity and therefore no viable colonies in the absence of mCpol. When the NucC plasmid was transformed into cells expressing Cst mCpol, no transformants were observed, consistent with activation of the nuclease by cA₃. In contrast, the toxicity of the CARF-TM effector was moderated with co-expressed mCpol. Truncation of the CARF-TM effector to remove the predicted transmembrane domain completely abolished activity, demonstrating that the transmembrane (TM) domain is essential. Expression of the inactive D63N, D64N variant of Cst mCpol did not alleviate the toxicity of CARF-TM, confirming the requirement for cA₃ synthesis.

We also performed the plasmid challenge assay using the previously characterised Vibrio metoecus (Vme) type III CRISPR-Cas Cmr complex instead of the type II Panoptes mCpol as the cyclase component (S3 Fig). When activated, the Vme Cmr complex produces 3′3′3-cA₃ with a trace of cA₄, and its cognate effector is the NucC nuclease [26]. The Vme Cmr complex was able to fulfil the same role as Cst mCpol, reducing the growth defect imparted by CARF-TM. This confirmed that 3′3′3′-cA₃ was the relevant signalling molecule for type II Panoptes.

Overall, these data establish Panoptes type II as a defence system that signals via cA₃, using a signalling logic similar to Panoptes type I. To explore this further, we challenged the defence system with two different proteins that are known to bind or degrade cA₃: the phage sponge protein Acb2 [14] and the cA₃-degrading ring nuclease Crn4 [28]. Expression of either of these proteins from a plasmid resulted in the activation of Panoptes type II, reflected in cell toxicity when the CARF-TM effector was induced by arabinose (Fig 3B). This mimics the situation where a phage utilising a sponge or ring nuclease to subvert CBASS or type III CRISPR defence would activate Panoptes type II defence.

To extend this analysis, we turned to the BASEL collection of E. coli phages [29]. Some phages in this collection encode the anti-CBASS phosphodiesterase Acb1 [30], which degrades a range of cyclic di- and tri-nucleotides including cA₃. Doherty and colleagues demonstrated that the acb1 gene is disrupted in phage Bas43, but intact in closely related phages such as Bas35-37 of the Tevenvirinae family [23]. They showed that disruption of acb1 correlated with resistance to the type I Panoptes system. We recapitulated this analysis using the type II Panoptes system, observing a clear resistance phenotype for Bas43, where the cA₃-degrading Acb1 protein is non-functional (Fig 3C). Bas35-37, with functional Acb1, were all sensitive to type II Panoptes defence, resulting in fewer and smaller plaques (Fig 3D and 3E), consistent with the data obtained from plasmid challenge experiments. In the course of these experiments, we isolated a number of escaper mutants for phage Bas35 which generated larger plaques in the presence of active type II Panoptes defence. Sequencing of the PCR-amplified acb1 region in these escapers revealed single-nucleotide mutations predicted to deactivate the acb1 gene (Fig 3F). Although the complete phage genomes were not sequenced, these mutations in acb1 are likely to be responsible for the observed phenotypes.

In sum, the in vivo data all point to a key role for cA₃ depletion in the activation of type II Panoptes. Phage challenged by Panoptes defence escape by deactivating the genes encoding cA₃-depletion proteins. This, of course, could render them sensitive to CBASS and type III CRISPR defence.

The CARF-TM effector binds cA₃

Modelling with Alphafold3 [31] strongly supported a dimeric structure for the CARF-TM effector (Fig 4A), consistent with the dimeric organisation of many other CRISPR-associated CARF effectors [22]. cA₃ was modelled bound at the interface between the two subunits. To express and purify recombinant CARF-TM, we had to co-express the protein with mCpol, generating cA₃ to keep the effector in an inactive state. Under these conditions, CARF-TM with a non-cleavable N-terminal polyhistidine tag could be purified to homogeneity via immobilised metal affinity and size exclusion chromatography in the presence of the detergent DDM (S4A and S4B Fig). We also purified a truncated form of the effector (CARF), where the TM domain was replaced with a non-cleavable C-terminal hexahistidine tag. The truncated protein also required the presence of the cyclase for efficient purification. We observed two bands in the final purified sample for full-length CARF-TM and for the truncated CARF protein (S4C and S4D Fig); both were confirmed by mass spectrometry to be the effector. An elevated absorbance at 260 nm for both the full-length and truncated proteins was suggestive of the presence of nucleic acid. We extracted nucleotides from both proteins and analysed them by HPLC, confirming the presence of cA₃ in roughly equimolar amounts to the protein dimer (Fig 4B).

These observations demonstrated that cA₃ remains bound to both the full-length and truncated effector during purification, suggesting a low dissociation constant for cA₃ binding. To explore ligand specificity, we investigated binding of ³²P-radiolabelled cA₃ by electrophoretic mobility shift assay (EMSA). At high concentrations of CARF-TM or CARF domain (approximately 1 µM dimer), a single retarded species consistent with cA₃ binding became visible (Fig 4C). This high apparent dissociation constant likely reflects the fact that both proteins were already bound to unlabelled cA₃, requiring displacement by competition with the radioactive species. The retarded radioactive cA₃ moiety was abolished when a 100-fold molar excess of cold 3′3′3′-cA₃ was added, but binding was unaffected by a range of other cyclic nucleotide species (Fig 4D). Overall, these data demonstrate that both the CARF-TM effector and its truncated, soluble CARF domain bind cA₃ specifically and with a high enough affinity to persist during protein purification. However, binding is not irreversible and is subject to competition by free cA₃ in vitro.

The soluble CARF variant multimerises when cA₃ is not bound

We next investigated whether the CARF effector underwent any conformational changes upon loss of cA₃, mimicking the effect of a cA₃-scavenging phage protein. Although we could purify small quantities of the full-length CARF-TM effector bound to cA₃, plasmid challenge experiments indicated that the presence of N- or C-terminal tags abolished function in vivo (S5 Fig). We therefore focussed on the truncated CARF version of the effector for further in vitro studies, cognisant of the limitations inherent in this approach. To deplete cA₃ from the CARF protein, we incubated the purified protein with increasing concentrations of the ring nuclease Crn4 [32]. HPLC analysis of the nucleotide species present following Crn4 incubation confirmed that a significant proportion of the cA₃ present was converted to smaller products, confirming that the ring nuclease can compete with the CARF-TM effector for cA₃ ligands in vitro (S6 Fig). This is consistent with the competition observed by free cA₃ in the EMSA experiments and with the in vivo data that demonstrate Panoptes activation by Crn4 (Fig 3).

To examine the effect of cA₃ removal on the effector, we probed structural changes using the fluorogenic dyes Sypro Orange and BODIPY FL L-cystine (BFC). Both dyes are typically used with thermal shift assays; however, we conducted isothermal assays following changes in fluorescence after addition of Crn4 to the cA₃:CARF complex. The fluorescence intensity of Sypro Orange increases upon binding to hydrophobic regions, which are normally buried in a folded protein but become exposed when the protein denatures and in turn become hidden when an unfolded protein aggregates (Fig 5A). The fluorescence of BFC in its disulfide dimer form is quenched. Cleavage by free thiols, such as exposed cysteine residues, releases the monomeric, fluorescent dye (Fig 5B). Unfolding and/or increased protein chain mobility can expose buried cysteine residues leading to an increase in fluorescence. We observed a decrease in Sypro Orange fluorescence intensity upon Crn4 treatment, consistent with burial of hydrophobic patches upon ligand removal (Fig 5C). For the BFC dye, isothermal incubation of the CARF protein with Crn4 resulted in a significantly larger increase in fluorescence than in the absence of Crn4 or with the catalytically inactive Crn4 [H15A] variant [28] (Fig 5D).

The effect of cA₃ removal on the quaternary structure of the CARF protein was also tested using chemical cross-linking and SDS-PAGE (Fig 5E). In the absence of Crn4, the CARF protein was observed primarily as monomeric and dimeric after crosslinking with very minor amounts of larger species. With increasing amounts of Crn4, progressively higher amounts of large (>dimeric) cross-linked species were observed (Fig 5F). This assay benefited from the unusual amino acid composition of Crn4, which lacks lysine residues and thus cannot cross-link [32]. We also investigated the removal of cA₃ from the CARF domain by dynamic light scattering (DLS). Notably, a precipitate formed when Crn4 was incubated with the CARF domain. As the samples were centrifuged, the precipitate was removed prior to DLS analysis. An approximately 100-fold increase in size of the remaining soluble CARF protein was detected upon cA₃-depletion (Fig 5G).

Together, these observations suggested that cA₃ removal results in structural perturbations of the CARF dimer, leading to unfolding or aggregation in vitro. The absence of the TM domain in these studies is a significant caveat, however. Given the difficulty in studying the full-length CARF-TM protein in vitro, we turned to in vivo analyses.

The CARF-TM effector causes growth arrest in the absence of cA₃

We set out to monitor changes in cell viability upon expression of CARF-TM using microscopy. The full-length, untagged CARF-TM was expressed from the T7 promoter in pEHisV5TEV [25], allowing induction of expression using IPTG. We omitted the mcpol gene from these strains, so that the effector would be in the active apo form when expressed. E. coli harbouring CARF-TM, truncated CARF or the empty vector were grown to mid-log phase before induction with 0.1 mM IPTG. Cells collected 1–2 h after induction were stained with Syto9 and propidium iodide (PI) [33]. Both are nucleic acid staining fluorescent dyes; Syto9 is a membrane-permeable dye that will stain live and dead cells, whereas PI can only enter cells with significantly damaged membranes. PI has a higher affinity for nucleic acids than Syto9 and tends to displace Syto9 from the nucleic acid given a high enough concentration. We determined the percentage of PI-dyed cells using fluorescent confocal microscopy (Figs 6A and S7). We observed minor differences between CARF-TM-expressing (7.2% PI-stained) or non-expressing cells (3.5% PI-stained for empty vector, 2.3% for CARF). This suggested that CARF-TM is not a fast-acting bactericidal agent when expressed in E. coli, but rather induces a growth arrest phenotype.

To follow on from the live/dead staining, we determined the cell density and culturability of the three strains for the first 3 h after IPTG induction. The optical density initially increased for all three strains, levelling off for the CARF-TM-expressing strain after 2 h, consistent with growth arrest upon effector expression (Fig 6B). The growth of cells with the empty vector control slowed significantly on IPTG addition—a phenomenon that was consistently observed and which may relate to the burden IPTG places on cells in a context-dependent manner [34]. We therefore investigated the equivalent growth curves using the Aaf Panoptes system under the control of an arabinose promoter, confirming that Aaf CARF-TM, but not truncated CARF, induced growth arrest in the absence of mCpol, with no growth defect in the presence of mCpol (S8 Fig).

Microdilution and colony counting assays for the Cst effector demonstrated that whilst the strains with empty vector or expressing the CARF domain yielded a similar number of colony-forming units (cfu), those for the CARF-TM-expressing strain dropped by 2–3 orders of magnitude (Fig 6C and 6D). Combined with the data from PI staining, this suggests cells experience growth arrest when the CARF-TM effector is activated. While the cells don’t appear to lyse, as observed for type I Panoptes [23], they may still be functionally “dead” to all intents and purposes.

Expression of apo CARF-TM may alter cell permeability in E. coli

We next assessed the effect of Cst CARF-TM on metabolic changes in response to altered cell permeability by monitoring the cellular reduction of resazurin to resorufin [35] (Fig 7). Before induction, the fluorescent signal was indistinguishable between the three strains (CARF-TM, CARF and eV). By 1 h after induction, CARF-TM-containing cells showed a faster rate of resorufin production compared to control cells (Fig 7A). This could indicate an increase in NAD(P)H levels. However, we observed that the addition of the antibiotic polymyxin B, which permeabilises the outer membrane (OM) of gram-negative cells [36,37], yielded very similar results, though with a faster timescale (Fig 7B). Indeed, there was no additive effect for CARF-TM expression and polymyxin B exposure. These data may indicate that the observed increase in resazurin reduction on both CARF-TM expression and polymyxin B treatment may be a consequence of increases in cell permeability, rather than a simple increase in intracellular NAD(P)H levels.

To explore this further, we investigated cell staining using the lipophilic dye Nile Red (NR) [38]. Cells expressing the apo CARF-TM effector were readily stained by NR (Fig 7C and 7D). Under the same conditions, only a small fraction of cells expressing the CARF domain or containing the empty vector were stained with NR (Fig 7C and 7D). When growing in liquid culture in the presence of NR, cells expressing CARF-TM gave rise to significantly enhanced NR fluorescence over time (Fig 7E). These data support a model where expression of apo CARF-TM in the inner membrane of E. coli results in permeabilization of the OM and growth arrest, but not rapid cell death. It should be noted; however, that effects observed on membrane permeabilization are indirect, with the significant caveat that the system is being studied in a heterologous host.

Cloning

Enzymes were purchased from Thermo Scientific or New England Biolabs and used according to manufacturer’s instructions. Oligonucleotides and codon-optimised synthetic genes were obtained from Integrated DNA Technologies (Coralville, Iowa, USA). All constructs were verified by sequencing (Eurofins Genomics, DE).

The mCpol-encoding genes from Cylindrospermum stagnale PCC 7417 (Cst; WP_015209627.1), Sphingomonas sp. (Sph; WP_230771556.1), and Ahniella affigens D13 (Aaf; WP_106889607.1) were cloned into pEHisV5TEV [25] for expression with a cleavable N-terminal His₈-tag. Active site residue variants were obtained by primer-directed mutagenesis [51]. Cst CARF-TM (WP_015209626.1) and Cst mCpol were cloned into a modified pRSFDuet-based plasmid by restriction assembly to yield an N-terminal His6-TEV-CARF-TM and an untagged Cst mCpol. The expression construct for soluble Cst CARF was obtained by PCR amplification of the CARF domain, followed by restriction digest and ligation into 5′-NdeI and 3′-XhoI sites of pEV5hisTEV [25]. This allows expression of Cst CARF with a native N-terminus and the LEHHHHHH affinity tag following residue T207 of Cst CARF. The Cst mCpol gene was cloned into the 5′-NdeI, 3′-XhoI sites of pRAT [52] to allow co-expression of the native Cst mCpol from an arabinose-inducible promoter with IPTG-inducible Cst CARF-CHis.

For plasmid challenge assays, Cst mCpol was cloned into the 5′-NdeI, 3′-XhoI sites of pACE (Geneva Biotech, Genève, CH) and Cst CARF-TM was inserted into MCS1 of pRATDuet [13] to allow induction with IPTG and l-arabinose, respectively. Cst CARF was obtained by primer-directed mutagenesis to introduce a stop codon after residue T207 of the native sequence. The ring nuclease Crn4 from Actinomyces procaprae [28] and the cyclic nucleotide sponge Acb2 from Pseudomonas phage PaMx33 [14] were each cloned into MCS2 of pRATDuet alongside Cst CARF-TM in MCS1. pRATDuet-NucC has been previously described [53].

The Aaf type II Panoptes system (WP_106889607.1 and WP_106889608.1) was used for phage assays. Synthetic versions of the wild-type Aaf-mCpol-CARFTM locus and Aaf-CARFTM alone were ligated into pRFSara. This plasmid was derived from pRSFDuet-1 (Novagen, Merck-Millipore) by replacing the lacI and T7 promoter region with the araC and pBAD promoter region from pBADHisTEV [54] whilst removing multiple-cloning-site 1 [52]. The transmembrane domain of CARF-TM was removed by mutating the L212 codon to a stop codon.

Untagged Cst CARF-TM and CARF constructs used in microscopy and in vivo fluorescence studies were obtained through restriction digest and ligation. Full-length Cst CARF-TM was ligated into the 5′-NdeI, 3′-XhoI sites of pEHisV5TEV, which removed the N-terminal affinity tag, to give pE-Cst CARF-TM. The truncated Cst CARF was obtained by primer-directed mutagenesis to introduce a stop codon after residue T207 of the native sequence to give pE-Cst CARF.

Protein purification

mCpol activity

Nucleotide cyclase activity was assayed in 12.5 mM Tris-HCl, pH 8.0, 125 mM NaCl, 10 mM MgCl₂ containing 500 µM ATP or ATP analogue and 5–25 µM (dimer) Cst mCpol wild type or mutant. The reaction was incubated for 3 h at 37 °C, unless stated otherwise. Enzymes were removed by ultracentrifugation (3 kDa MWCO) or by heat deactivation (95 °C for 5 min) followed by centrifugation (16,000g, 10 min, 4 °C). Reaction products were analysed by HPLC (Dionex UltiMate 3000 equipped with single-wavelength UV detector, Thermo Scientific) as follows. Compounds from Cst mCpol reactions were separated on a Synergi Fusion RP column (2.5 µm, 2 × 50 mm, Phenomenex) at 0.3 ml min⁻¹ and 40 °C column temperature with a linear gradient of 2%–50% methanol against 20 mM ammonium bicarbonate over 4.8 and 0.2 min gradient delay. Elution was monitored by UV at 260 nm. Data were analysed using Chromeleon and visualised in Prism 10 (GraphPad). Cyclic nucleotides were identified by comparison to and co-injection with synthetic standards (Biolog, DE) or by mass spectrometry.

Selected samples were analysed by liquid chromatography-mass spectrometry on an LCQ Fleet mass spectrometer coupled to an Ultimate 3000 UHPLC module (Thermo Scientific) using the same liquid chromatography conditions as above. Electrospray ionisation was monitored in positive ion mode without fragmentation. Data were analysed with Excalibur Qual Browser software and visualised in Prism 10.

cA₃ production rate

cA₃ production by mCpol enzymes was determined using the EnzChek Pyrophosphate Assay kit (Invitrogen, Fisher Scientific). Reactions contained 12.5 mM Tris-HCl, pH 8.0, 100 mM NaCl, 10 mM MgCl₂, 0.2 mM 2-amino-6-mercapto-7-methylpurine ribonucleoside (MESG), 1 U ml⁻¹ purine nucleoside phosphorylase, 0.1 U ml⁻¹ pyrophosphatase, 0.5 mM ATP (Fermentas, Fisher Scientific) and 2.5–5 µM mCpol dimer. Background hydrolysis was determined by omitting mCpol. Pyrophosphate (1–10 µM) in the reaction mixture without ATP and mCpol were used to obtain the standard curve. The reaction was started by the addition of ATP or pyrophosphate standards and the absorption at 360 nm was monitored at 37 °C using a FluoStar Omega (BMG Labtech) plate reader. Data analysis was performed in Prism 10 (Graphpad). Rates were determined by linear regression, corrected for enzyme concentration with the assumption that for each cA₃ molecule, three molecules of pyrophosphate are generated, and that non-production (pyro)phosphate release is negligible. The reported rate was the mean from three independent experiments.

Dynamic light scattering

DLS measurements were performed with the Zetasizer Nano S90 (Malvern) instrument. Samples contained 60–100 µM protein dimer in 20 mM Tris-HCl, 75 mM NaCl, 10 mM MgCl₂ and 0.1% DDM as required. After centrifugation at 12,000g for 10 min at room temperature, the sample was centrifuged and loaded into a quartz cuvette (ZMV1012). Measurements were performed at 25 °C with three measurements of 13 runs.

Electrophoretic mobility shift assay

[α-³²P]-Radiolabelled cA₃ was prepared using 45 µM Cst mCpol dimer in 12.5 mM Tris-HCl, pH 8.0, 100 mM NaCl, 10 mM MgCl₂ and 300 µM [³²P]-ATP (0.15 MBq). The reaction was incubated at 37 °C for 3 h, protein was removed by phenol-chloroform extraction and excess phenol was removed by an additional chloroform extraction step. A typical EMSA assay contained 100 nM [α-³²P]-cA₃ and 0–4 µM His₆-CARF-TM monomer in 30 mM HEPES, pH 7.5, 90 mM NaCl, 10 mM MgCl₂, 10% glycerol, 0.4 mM DDM or 100 nM [α-³²P]-cA₃ and 0–14 µM CARF-CHis₆ monomer in 12.5 mM Tris-HCl, pH 8.0, 200 mM NaCl, 10 mM MgCl₂, 10% glycerol, unless stated otherwise. The mixture was incubated at room temperature for 30 min before addition of G-250 Native Gel Sample Loading Buffer (Invitrogen) to 0.005% final concentration. Samples were immediately loaded onto a pre-run 6% acrylamide gel (29:1 acrylamide:bis-acrylamide). Radiolabelled material was separated for 2 h at 200 V in 1× TBE buffer and visualised by exposure to a phosphor storage screen and phosphorimaging (Cytiva).

Protein crosslinking

The cross-linking reaction was carried out in 20 mM HEPES, pH 7.5, 250 mM NaCl, 10% glycerol with 20 µM Cst CARF dimer in the absence or presence of 2.5–40 µM Crn4. The reaction was incubated at 37 °C for 20 min. The BS-3 (bis(sulfosuccinimidyl)suberate-d₀, Thermo Scientific) crosslinker was then added to 0.2 mM final concentration and the mixture was incubated with gentle agitation for 30 min at 20 °C. SDS-PAGE sample loading buffer was added to each reaction and the samples were heated to 95 °C for 2 min before analysis by SDS-PAGE.

Fluorescence assays

Sypro Orange was purchased as a 5,000× stock in DMSO (Invitrogen) and used at 5X final concentration in 20 mM HEPES, 250 mM NaCl, 10% glycerol, pH 7.5 with 9 µM CARF dimer and 3 µM Crn4 dimer wild type or [H15A] variant. The reaction was preincubated with fluorescence monitoring (ex/em 485/580 nm) at 37 °C in a FluoStar Omega plate reader (BMG Labtech) for 15 min before addition of Crn4. BODIPY FL L-Cystine (BFC, Invitrogen, Fisher Scientific) was dissolved in DMSO and used at 1 µM final concentration in the same buffer as for Sypro Orange with 0.5 µM CARF and 1 µM Crn4. The reaction was preincubated for 3 min at 37 °C before addition of Crn4 with fluorescence monitoring at ex/em 485/520 nm. Data were analysed using Graphpad Prism 10.

Plasmid challenge assay

The plasmid challenge assay was performed as described in [53] with the following modifications. Competent cells were prepared from individual colonies of E. coli C43(DE3) transformed with pACE-Cst mCpol, pACE-Cst mCpol[D63N/D64N] or pACE empty vector. pRATDuet, pRATDuet-Cst CARF-TM, pRATDuet-Cst CARF, pRATDuet-Cst CARF-TM-Crn4 or pRATDuet-Cst CARF-TM-Acb2 were used for transformation. Selective conditions were LB agar containing 100 µg ml⁻¹ ampicillin for the acceptor cells only and containing ampicillin and 12 µg ml⁻¹ tetracycline for uninduced transformants. Cst mCpol and Cst CARF-TM ± Crn4 or Acb2 expression was induced with 100 µM IPTG and 0.0125% or 0.05% l-arabinose, respectively. Plates were incubated overnight at 37 °C.

Phage assays and Aaf-Panoptes growth curves

PRSFara constructs containing Aaf-CARFTM, truncated Aaf-CARF[L212*], wild-type Aaf-Panoptes (mCpol-CARFTM), or Aaf-Panoptes-ΔTM (mCpol-CARF[L212*]) were transformed into E. coli MG1655 alongside an empty vector. Individual colonies were used as biological replicates.

For growth curves, the strains were cultured in LB medium containing 50 µg ml⁻¹ kanamycin at 37 °C. Overnight cultures were diluted 100-fold into fresh medium and grown to an OD₆₀₀ of 1.0–1.5.

The cultures were then diluted in triplicate to a final OD₆₀₀ of 0.1 in clear, F-bottom 96-well plates (Greiner) with l-arabinose at a final concentration of 50 µM (CARFTM, CARF) or 0.5 mM (Panoptes) and a total volume of 200 µl. Growth was monitored by reading the absorbance at 600 nm (2× 2 matrix scan, 2 mm width, no pathlength correction) every 20 min and double-orbital shaking at 200 rpm for 15 s before each cycle using a FluoStar Omega (BMG Labtech) plate reader.

For phage assays, Panoptes strains were cultured in LB medium containing 50 µg ml⁻¹ kanamycin, 0.5 mM l-arabinose, 10 mM MgCl₂ and 5 mM CaCl₂. Strains, grown at 37 °C to early exponential phase, were diluted into soft LB agar, kept at 42 °C, to an OD₆₀₀ of 0.1; the inoculated soft agar (7 ml) was immediately poured onto 20 ml LB agar in 10 × 10 cm square Petri dishes. Phages Bas31, Bas35-37 and Bas43 from the BASEL collection [29] were serially diluted 10-fold into SM buffer (50 mM Tris-HCl, 100 mM NaCl, 10 mM MgSO₄, pH 7.5); 2.5 µl of each dilution was spotted onto inoculated LB agar plates in duplicate and the plates incubated overnight at 28 °C. The efficiency of plating (EOP) was determined by counting the number of plaques at the lowest dilution factor that provided discernible individual plaques. Where plaque size was too small, the last spot that gave a clear zone in the bacterial lawn was taken to be 100 plaques. The concentration of plaque-forming units was calculated taking dilution factor and applied volume into account.

Escaper phages were isolated by three passages of microdilution assays against MG1655/ Aaf type II Panoptes. Culture conditions were as described above. Three plaques were selected for isolation; plaques 1 and 3 were isolated and sequenced as biological duplicates, giving rise to 1a,b and 3a,b. As expected, the duplicates carried identical mutations of the acb1 sequence. The acb1 region was PCR-amplified using MyTaq Red (Bioline) and primers 5′-ggaagcagaaattgcaatagca and 5′-cacatattacagaagctggattgc. PCR products were purified using the Monarch PCR and DNA Clean-up kit (New England Biolabs) according to the manufacturer’s recommendation. PCR primers were used for sequencing by Eurofins Genomics. The sequences of the acb1 gene product from escapers 1a, 1b, 2, 3a and 3b are provided in file S2 Data.

Microscopy

E. coli C43(DE3) was transformed with empty vector, pE-Cst CARF-TM or pE-Cst CARF. Single colonies were used to inoculate LB containing 50 µg ml⁻¹ kanamycin. After culturing overnight at 37 °C with shaking, the cultures were diluted 100-fold into fresh selective medium. Cells were grown at 37 °C to an OD₆₀₀ of 0.5–0.8 and induced with 0.1 mM IPTG. Incubation was continued at 37 °C. Samples (0.5 ml) were taken before induction or 1 h after induction. Cells were collected by centrifugation at room temperature (2 min at 1,500g) and the pellet was washed once with PBS buffer. The pellet was resuspended in PBS buffer to an approximate OD₆₀₀ of 5–10. Nile Red (Merck) was added from a 2 mg ml⁻¹ stock solution in DMSO to a final concentration of 20 µg ml⁻¹ and incubated in a warm environment for 30–45 min. Syto 9 (Invitrogen) and propidium iodide (Merck) were added to 2.5 µM and 20 µg ml⁻¹ final concentrations, respectively, and incubated at room temperature for 15–30 min.

Samples (4 µl) were applied to a thin agarose slab (1% agarose in water) on a standard glass microscope slide and allowed to dry for a few minutes before covering with a coverslip. Images were acquired with a fluorescence confocal microscope (DeltaVision Imaging System), connected to a digital camera. The system was configured for both fluorescence and transmitted light DIC imaging. Images were recorded with an Olympus 100×/1.40 oil immersion objective lens, deconvolved using softWoRx Explorer 1.3 and further processed with Fiji image analysis software [55].

Culturability assay

The same strains as for microscopy were used: E. coli C43(DE3) transformed with empty vector, pE-Cst CARF-TM or pE-Cst CARF. Starter cultures were diluted into fresh LB medium containing 50 µg ml⁻¹ kanamycin and incubated at 37 °C with shaking. When the OD₆₀₀ reached 0.8–1.0, cultures were diluted with selective medium to an OD₆₀₀ of 0.5 and induced with 0.1 mM IPTG. Incubation was continued at 37 °C. Samples were taken before induction or 1–3 h after induction. For each sample, the OD₆₀₀ was measured, and a 10-fold serial dilution was prepared that was applied to LB agar plates containing kanamycin but no IPTG. Plates were incubated overnight at 37 °C. Spots with isolated colonies were used to count colony-forming units (cfu), from which the cfu ml⁻¹ of the original sample was calculated based on dilution and applied volume (2.5 µl). The results are derived from four biological replicates. Mean and standard deviation were plotted using Prism 10 (Graphpad).

Resazurin assay

We used the samples obtained from the culturability assay (above). For the resazurin assay, the cell density of each sample was adjusted to OD₆₀₀ 0.5 with LB medium; 45 µl OD₆₀₀-adjusted culture and 45 µl LB or LB containing 20 µg ml⁻¹ polymyxin B were mixed in a black, non-binding, PS 96-microtitre plate (F-bottom, Greiner). The plate was incubated at 37 °C in the plate reader (FluoStar Omega, BMG Labtech) with fluorescence monitoring before starting the experiment by addition of 10 µl 50 µg ml⁻¹ resazurin in LB. The production of resorufin was followed at 37 °C with excitation and emission wavelengths of 540 and 580 nm, respectively. The results are derived from four biological replicates. Initial rates were determined by linear regression of the first 10 min after resazurin addition. Prism 10 (Graphpad) was used for regression analysis and plotting results.

Alphafold predictions

Structural predictions of Cst CARF-TM were generated using Alphafold3 [31], with model weights provided by Google Deepmind. Structural predictions were made either in the presence or absence of ~50 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylglycerol (POPG) molecules and a cA₃ ligand. Initial predictions were made for both a dimeric and trimeric conformation; however, pTM and ipTM scores for a trimeric conformation fell below the cutoff for a reliable prediction (0.45 and 0.32, respectively, compared to 0.8 and 0.77 for a dimer).

S1 Fig. HPLC analysis of mCpol reaction product.

Aaf (A) or Sph (B) mCpol in vitro reactions in the presence of Mg²⁺ and ATP or NTP. Coinjection: HPLC samples of the 3′3′3′-cA₃ standard and mCpol reaction were mixed and reinjected to check for matrix effects. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003934.s001

(PDF)

S2 Fig. Colorimetric pyrophosphate release assays for mCpol activity.

Graphs for one representative experiment are shown. A. Standard curve. B. Progress curves for Sph mCpol. C. Progress curves for Cst mCpol. D. Standard curve for Aaf mCpol reaction. E. Progress curves for Aaf mCpol. Data for (D) and (E) only were blank (no mCpol, no pyrophosphate) corrected. The turnover was calculated based on the assumption that one cA₃ molecule is produced per three pyrophosphates and that non-productive pyrophosphate release was negligible. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003934.s002

(PDF)

S3 Fig. Plasmid challenge using the type III-B CRISPR system.

E. coli BL21(DE3) Star was transformed with Cst CARF-TM or the cA₃-activated CRISPR-associated nuclease NucC. The type IIIB (Cmr) CRISPR effector from Vibrio metoecus (Vme), charged with non-targeting crRNA (Cmr[pUC]) or targeting (Cmr[Tet]) crRNA, was used to allow regulated cA₃ production. A serial dilution of the transformation mixture was applied to selective plates. The cyclase component was under the control of the lactose-inducible T7 promoter, and the effector (CARF-TM or NucC) was under the control of the arabinose-inducible pBAD promoter. Upon induction of protein expression with 0.1% d-lactose and 0.2% l-arabinose, Vme Cmr deactivated Cst CARF-TM (marked by gold arrows).

https://doi.org/10.1371/journal.pbio.3003934.s003

(PDF)

S4 Fig. Purification of Cst CARF-TM and CARF proteins.

A, B: Chromatogram for size exclusion chromatography and SDS-PAGE analysis of selected fractions for Cst CARF-TM, respectively. C, D: Chromatogram for size exclusion chromatography and SDS-PAGE analysis of selected fractions for Cst CARF, respectively. Pooled fractions used for analysis are indicated by an orange box. The chromatography data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003934.s004

(PDF)

S6 Fig. HPLC analysis of supernatant from heat-denatured CARF: Crn4 mixture after 30 min incubation at 37 °C.

The ring nuclease Crn4 can degrade CARF-bound cA₃. The molar ratio of CARF to Crn4 is indicated. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003934.s006

(PDF)

S8 Fig. Growth curves of Aaf CARF-TM and the full Aaf Panoptes system in E. coli MG1655.

A. Upon induction of the Aaf CARF-TM effector under control of the arabinose promoter, a dose-dependent growth inhibition was observed (5 µM or 50 µM arabinose concentration) that was absent when the transmembrane domain was removed (CARF). B. When CARF-TM was expressed in the presence of Aaf mCpol, no growth defect was observed with 0.5 mM arabinose induction. Reflecting the situation in the Aaf genome, the carf-tm gene does not have its own ribosome binding site in this construct, which ensures translational coupling. ΔTM: mCpol-CARF. The mean and standard deviation from two biological replicates with three technical replicates each is shown. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003934.s008

(PDF)

S1 Data. Data underlying Figs 2, 3E, 4, 5C, 5D, 5F, 5G, 6A, 6B, 6C, 7A–7E, S1A, S1B, S2A–S2E, S4A, S4C, S6, S8A, and S8B.

https://doi.org/10.1371/journal.pbio.3003934.s010

(XLSX)