The CRISPR ring nuclease Csx15 oligomerises on cyclic nucleotide binding to regulate antiviral defence

Abstract

Prokaryotic type III CRISPR systems signal infection by generating cyclic oligoadenylate (cOA) second messengers, which activate defence proteins allosterically, providing immunity. cOA molecules are typically degraded by extrinsic, stand-alone ring nuclease (RN) enzymes with phosphodiesterase activity or by the intrinsic RN activity of the effectors themselves. Viruses and plasmids also encode RNs, which can function as anti-CRISPRs. Eight different families of extrinsic RNs are currently known. Here, we report the structural and biochemical analysis of one of these families: Csx15. We show that Csx15 is a dimeric protein of the CRISPR-associated Rossmann fold (CARF) superfamily with the ability to bind cyclic tetra-adenylate (cA₄) molecules in a shared binding site formed by the head-to-tail stacking of dimers in a filament conformation. Some family members are non-enzymatic, relying on the sequestration (sponging) of cA₄ to regulate the host immune response, while others act as canonical RNs, slowly degrading cA₄.

Introduction

CRISPR systems provide adaptive prokaryotic immunity by capturing and storing short segments of DNA from invading mobile genetic elements in a CRISPR array in the host genome. This is transcribed and processed into CRISPR RNAs (crRNAs), guiding CRISPR defence systems to target cognate invading nucleic acids during subsequent infection. Seven major types of CRISPR system have been described [1]. Type III CRISPR systems consist of a multi-subunit interference complex with a Cas10 catalytic subunit. Upon binding to invading RNA, the cyclase domain of Cas10 is activated, producing cyclic oligoadenylates (cOA) or SAM-AMP as secondary messengers [2–4]. These messengers activate various downstream effectors, initiating diverse immune responses, including nonspecific RNA or DNA cleavage [2,5–8], proteolysis [9–11], transcriptional and translational regulation [12,13], degradation of nucleotides and cofactors [14,15], and membrane disruption [16–18]. The activation of these effectors can cause cell growth arrest or death, thus providing protection at the population level [19].

In reality, many type III CRISPR systems include a specialised class of enzymes, collectively known as ‘ring nucleases’ (RN), that degrade the cyclic nucleotide, providing a mechanism to deactivate the immune response and avoid unnecessary cell death. The first CRISPR-associated ring nuclease (Crn1) was identified biochemically [20], followed quickly by Crn2 and Crn3 [21,22]. These enzymes are all specific for the degradation of cA₄, the most commonly used signalling molecule in type III CRISPR systems [23]. Recently, a further enzyme, Crn4, was described that has a much broader substrate specificity, degrading all cOA species tested [24]. In addition, many of the effector proteins activated by cOA species have intrinsic RN activity and are therefore self-limiting [12,25–31]. A recent bioinformatic analysis revealed that the ring nucleases are diverse and abundant in type III CRISPR systems, adding two more cA₄-specific enzymes (Csx16 and Csx20) to the growing list [32]. Viruses also encode ring nucleases in their genomes, where they function as anti-CRISPR (Acr) proteins by intercepting and destroying cOA second messengers [21,33].

Csx15 was originally proposed as a candidate RN based on its amino acid sequence and genomic neighbourhood analysis [34]. Analysis of ∼1000 type III CRISPR loci in 40,000 complete prokaryotic genomes revealed that Csx15 was present in 18 loci, often co-occurring with Crn1 [32]. About 25% of csx15 genes in the dataset were orphans, not associated with a CRISPR locus. AlphaFold3 (AF3) [35] modelling predicted a CARF (CRISPR-associated Rossmann fold)-like fold for Csx15. A representative example from Pseudomonas fluorescens (PfCsx15) was tested for RN activity but found to have little or no activity in vitro [32]. However, PfCsx15 is an orphan protein and could thus have diverged in function.

Here, we investigate two Csx15 orthologues, the orphan PfCsx15 and the CRISPR-associated Chlorobaculum limnaeum Csx15 (CliCsx15). PfCsx15 specifically sequesters cA₄ to inhibit the cA₄-activated ribonuclease activity of a CRISPR effector protein. CliCsx15 exhibits cA₄-specific RN activity in vitro and can neutralise CRISPR defence in vivo. Both proteins adopt higher oligomerisation states induced by cA₄ binding. Crystal structures of PfCsx15 and CliCsx15 reveal a conserved dimeric structure and the molecular basis for multimerisation in a head-to-tail conformation that sandwiches cA₄ between dimers. The likely functions of these unusual cA₄ binding proteins are discussed.

Results

PfCsx15 specifically inhibits cA₄-activated CRISPR effectors

Although PfCsx15 was previously identified as a candidate RN, very limited nuclease activity was observed using HPLC analysis [32]. To further investigate the function of PfCsx15, we tested whether the protein could inhibit a range of type III CRISPR effectors in vitro. We examined the cA₃-activated DNase Vibrio metoecus (Vme) NucC [8], the cA₄-activated ribonuclease TTHB144 from Thermus thermophilus HB8 [25], and the cA₆-activated ribonuclease Mycobacterium tuberculosis (Mtb) Csm6 [36] using an established fluorescent assay. We observed a progressive inhibition of TTHB144 in the presence of high concentrations of PfCsx15. At three concentrations of cA₄ (1, 2, or 4 μM), preincubation with PfCsx15 for 15 min led to a reduction in TTHB144 ribonuclease activity when the molar ratio of PfCsx15 dimer to cA₄ concentration was higher than approximately 0.6 (Figure 1). Weaker inhibition was observed for VmeNucC (Supplementary Figure S1) and MtbCsm6 (Supplementary Figure S2). These data suggest that PfCsx15 is specific for cA₄ and does not primarily function as an RN but rather exhibits properties more characteristic of a cyclic nucleotide-binding, or sponge, protein, analogous to the anti-CBASS protein Acb2 [37].

Figure 1. PfCsx15 specifically inactivates cA₄-induced RNase TTHB144.

(A) RNase alert assay with detection of fluorescence signal emitted by RNA cleavage when RNases were present. RNaseAlert substrates (100 nM) were incubated with the Csx15 (0–5 μM dimer) and cA₄ (1, 2, and 4 μM, respectively) at 35°C for 15 min before adding RNase TTHB144 (500 nM dimer). The fluorescence signal was plotted against time. (B) Plot of baseline-corrected fluorescence signal against the ratio of Csx15 (dimer) to cA₄. The average fluorescence signal from 59–61 min was corrected using the average baseline signal from 9–10 min. Data of two replicates were presented as mean ± s.d. The RNase activity of TTHB144 was inhibited when the ratio of Csx15 (dimer) to cA₄ exceeded 0.6. (C) Schematic representation of the fluorogenic biochemical assay

PfCsx15 oligomerises on cA₄ binding

Oligomeric states of PfCsx15 were examined using dynamic light scattering (DLS) in the presence or absence of different cOA species. Addition of cA₄, but not cA₃ or cA₆, to PfCsx15 resulted in a significant increase in global particle size (Figure 2A). Additionally, size-exclusion chromatography (SEC) showed that PfCsx15 eluted with a retention time consistent with a dimer in the absence of cA₄, while a significant shift to a shorter retention time (and thus larger size) was observed in the presence of cA₄ (Figure 2B). These observations suggested that cA₄ induces PfCsx15 to form high molecular weight complexes.

Figure 2. cA₄ induces the oligomerisation of PfCsx15.

(A) DLS of PfCsx15 in the presence or absence of a range of cOAs. PfCsx15 assembles into significantly larger molecular species only in the presence of cA₄. Statistical analysis was performed using two-way ANOVA multiple comparisons, followed by Šídák’s multiple comparisons test (****P <0.0001). (B) SEC of PfCsx15 with and without cA₄. A molar ratio of PfCsx15 (dimer) to cA₄ is 2:1. Gel filtration standard (Bio-Rad) includes thyroglobulin (670,000), γ-globulin (158,000), ovalbumin (44,000), myoglobin (17,000), and vitamin B12 (1350).

Crystal structure of PfCsx15

To better understand the structural organisation of PfCsx15, we crystallised the protein and solved the structure at a resolution of 1.88 Å (Supplementary Table S1). The dimeric structure corresponds closely to that predicted by AF3 previously [32], comprising a fold with a central 5-stranded β-sheet flanked by α-helices (Figure 3A). Structural comparison with the ring nuclease Crn1, which adopts a canonical CARF fold, shows that core topological features are conserved, with the exception of β3, which is absent (Supplementary Figure S3). It is important to note that the CARF protein family deviates from the classical Rossmann fold in an important respect: the right-handed helical crossover between strands β5 and β6 is replaced with a reverse turn in CARF proteins [38]. This results in an antiparallel arrangement of β5 and β6, which are parallel in the canonical Rossmann fold, potentiating the dimeric organisation of CARF proteins. Although the CARF acronym is well established, it has been suggested that a useful alternative definition might be ‘cyclic oligoadenylate recognition fold’ [38].

Figure 3. Crystal structure of PfCsx15.

(A) Subunit of PfCsx15 with secondary structure elements labelled. Coloured from blue (N-terminus) to red (C-terminus). (B) Crystal structure of PfCsx15, showing dimer:dimer stacking in the crystal lattice with a head-to-tail organisation. Subunits are coloured green and teal in the lower dimer and orange and olive in the upper dimer. (C) Close-up view of the dimer:dimer interface, showing conserved residues H82, R116, R120, and R123 in the top dimer and S9 and H11 in the bottom dimer. The positions of two phosphate ions present in the crystal structure are shown in sphere representation.

The crystal packing positions dimers in a repeating head-to-tail conformation, generating a potential shared binding site between dimers (Figure 3B). Two β-hairpins formed by β-strands 4 and 5 form interactions with the dimer positioned above in the crystal lattice. Sequence comparisons (Supplementary Figure S4) indicate conservation of residues S9, H11, H82, R116, R120, and R123 in a range of Csx15 orthologues. In the dimer:dimer interface, we observed two phosphate ions, which might indicate the binding site of cA₄. The top dimer (orange/olive) contributes the conserved arginine residues (R116, R120, and R123), which interact with the phosphate ions, along with conserved H82 (Figure 3C). The bottom dimer (green/teal) contributes the conserved residues S9 and H11. The 6 arginine residues appear suitably positioned to interact with the phosphate groups in bound nucleotide ligands. Unfortunately, we were unable to co-crystallise the protein with cA₄, as the protein came out of solution when the nucleotide was added.

Mutation of conserved interface residues reduces filament formation

To test for an interaction with cA₄, residues H11 (bottom dimer) and R123 (top dimer) of PfCsx15 were independently mutated to alanine. DLS assays were first performed to examine the oligomeric states of PfCsx15 wild type and variants (Figure 4A). The formation of high molecular weight species upon addition of cA₄ was largely abolished for the H11A variant (Figure 4A). Although a higher oligomeric state was still observed for the R123A variant, the global particle size was significantly less than that observed for the wild type. These findings suggest that both residues play a role in the formation of higher-order oligomeric states and reinforce the conclusion that the dimer:dimer interface is the site of cA₄ binding.

Figure 4. Characterisation of PfCsx15 variants H11A and R123A.

(A) DLS of PfCsx15 wild type and variants. Mutation of conserved residues H11 and R123 abolished the cA₄-induced multimerisation of PfCsx15. Statistical analysis was performed using two-way ANOVA multiple comparisons, followed by Turkey’s multiple comparisons test (****P <0.0001, ^nsP = 0.0726). (B) Fluorescent assay of PfCsx15 wild type and variants. RNaseAlert substrates (100 nM) were incubated with the PfCsx15 (0-5 uM dimer) and cA₄ (1.1 μM) at 35°C for 15 min before adding RNase TTHB144 (500 nM dimer). The fluorescence signal was plotted against time. (C) Plot of baseline-corrected fluorescence signal in panel (B). Data of two replicates were presented as mean ± s.d. There was no observed inhibition of cA₄-induced TTHB144 ribonuclease activity when H11A was present

Ribonuclease inhibition assays were subsequently carried out to investigate the ability of the variant proteins to sequester cA₄ (Figure 4B,C). No inhibition of cA₄-activated TTHB144 ribonuclease activity was observed, even at very high concentrations of the H11A variant. For the R123A variant, full inhibition occurred at higher protein-to-cA₄ ratios than that required for the wild-type protein. Collectively, these observations indicate that H11 plays a major role in cA₄ binding and multimerisation, while R123 plays a secondary role.

CliCsx15, a type III CRISPR-associated Csx15, is a cA₄-specific ring nuclease

As PfCsx15 is encoded by an orphan gene that is not linked to a type III CRISPR locus, it may have diverged in function. We therefore selected and cloned a gene encoding Csx15 from the type III CRISPR locus of Chlorobaculum limnaeum strain DSM 1677, hereafter referred to as CliCsx15 (Figure 5A). The protein was expressed and purified to homogeneity. The RN activity was tested by incubating cOAs (fixed at 80 μM) with increasing concentrations of CliCsx15 for 60 min at 30°C, followed by HPLC analysis. Linear products A₂-p and A-p were observed at the higher protein concentrations (Figure 5B). The observation of 3′-phosphate termini (rather than 2′,3′-cyclic phosphate termini) on the products suggests that this is a hydrolytic reaction, involving a water nucleophile. This contrasts with ring nucleases such as AcrIII-1, which degrade cA₄ using an in-line attack of the adjacent 2′-hydroxyl group on the substrate [21]. The CliCsx15 RN activity was unaffected by the presence or absence of EDTA/Mg²⁺/Mn²⁺, suggesting that activation of the water nucleophile is not mediated by a divalent metal ion (Supplementary Figure S5). The turnover rate was very slow, with only ∼50% of cA₄ degraded over 60 min at an enzyme:substrate ratio of 1:8. No RN activity was detected for cA₃ or cA₆ substrates (Supplementary Figure S6).

Figure 5. CliCsx15 is a cA₄-specific ring nuclease.

(A) Genomic context of the Csx15-associated type III-A CRISPR system in C. limnaeum strain DSM 1677. (B) Ring nuclease activity of CliCsx15. cA₄ (80 μM) incubated with CliCsx15 (0, 1, 10, and 40 μM) for 60 min at 30°C was analysed using HPLC. (C) DLS analysis. cA₄ promoted an increase in global particle size. Statistical analysis was performed using two-way ANOVA multiple comparisons, followed by Šídák’s multiple comparisons test (****P <0.0001).

We also tested whether cA₄ promoted the oligomerisation of CliCsx15. DLS was used to examine oligomeric states in the presence or absence of different cOAs. A significant increase in particle size was observed only upon cA₄ addition to CliCsx15, consistent with the observations for PfCsx15 (Figure 5C).

Crystal Structure of CliCsx15

We crystallised the apo form of CliCsx15, solving the structure at the extremely high resolution of 0.9 Å (Figure 6A). A sample of the electron density for residues at the interface is shown in Supplementary Figure S7. Overall, the dimeric structures of PfCsx15 and CliCsx15 are similar, with an RMSD of 3.0 Å over 208 residues. Notably, the same head-to-tail packing was observed in both structures, strengthening the hypothesis that this might represent the biologically relevant interface. To investigate this further, we generated a model of four monomers of CliCsx15 along with 4 AMP molecules (analogous to cA₄) using AF3. The AF3 model (Figure 6B), predicted with high confidence (ipTM 0.81; pTM = 0.84), matched very well to the organisation of CliCsx15 dimers observed in the crystal lattice. The four AMP molecules adopt orientations sandwiched in the dimer:dimer interface (Figure 6B,C), in suitable positions to interact with the conserved arginine residues contributed by the top dimer. Conserved residues H13 and Y113, equivalent to H11 and F106 in PfCsx15 (Supplementary Figure S4), flank the adenosine base of AMP at two positions in the dimer:dimer interface (Figure 6C). Conserved residue H90 sits in a central position in the top dimer, suitably located to participate in catalysis. The structural data suggest that Csx15 may multimerise in a concentration-dependent manner in the absence of cA₄, while the DLS data emphasise the importance of cA₄ in stabilising these multimers.

Figure 6. Crystal structure and AF3 modelling of CliCsx15.

(A) Crystal structure of CliCsx15, showing dimer:dimer stacking in the crystal lattice with a head-to-tail organisation, similar to PfCsx15. (B) AF3 model of CliCsx15 together with 4 AMP ligands. The AMP molecules cluster at the dimer:dimer interface, suggesting the expected binding site for cA₄. (C) Close-up view of the dimer:dimer interface in the AF3 model, showing the positions of conserved residues (labelled). (D) Structural alignment of CliCsx15 dimer (green) with AcrIII-2 SAVED monomer (light grey). cA₄ from CalpL SAVED domain [10] is shown with carbon atoms coloured yellow. Csx15 H13 and AcrIII-2 H128 are indicated. (E) Top-down view of panel (D).

Comparison with other ring nucleases

As noted, Csx15 adopts the structure found in the CARF and SAVED protein superfamily [34]. The top hit in a DALI [39] search (Z-score 8.4; RMSD 2.8 Å over 238 residues) is the AcrIII-2/Crn5 protein, a ring nuclease that uses a SAVED domain to bind cA₄. Similarly to Csx15, AcrIII-2 SAVED domains stack in a head-to-tail conformation to bind and cleave cA₄ [33]. To extend this comparison, we added the cA₄ molecule from the SAVED domain of the CalpL effector, which is closely related to AcrIII-2 [10,31]. This modelling exercise provides a good prediction for the cA₄ binding site of Csx15. Notably, the catalytic residue H13 of CliCsx15 is suitably positioned to participate in phosphodiester bond cleavage and coincides with the essential catalytic residue H128 of AcrIII-2 (Figure 6E,F). The loop containing H13, found between strand β1 and helix α1, is a conserved site for nucleotide interaction in the Rossmann fold superfamily, designated as Motif-I [34]. The ring nuclease Crn1 has a catalytically important serine residue in the equivalent position [20].

To test the functional roles of H13 and H90 in CliCsx15, we mutated each residue to alanine and examined the ability of the variants to multimerise on cA₄ binding and to degrade the cyclic nucleotide. The RN activity of H13A was abolished, while only very low RN activity was observed for the H90A variant (Figure 7A), implicating both residues in the catalytic mechanism. The cA₄-induced oligomerisation was abolished when the residue H90 was mutated, whereas variant H13A had subtle effects on its oligomeric state (Figure 7B).

Figure 7. Ring nuclease activity and oligomeric states of CliCsx15 variants.

(A) RN activity of CliCsx15 wild-type and variants. cA₄ (50 μM) incubated with CliCsx15 wt and variants (25 or 50 μM dimer) for 60 min at 30°C was analysed using HPLC. No ring nuclease activity was detected in the presence of variant H13A, whereas limited cleavage activity was observed in the high concentration of variant H90A. (B) DLS analysis. cA₄-induced multimerisation was completely abolished when a conserved residue H90 was mutated, while the mutation of H13 had less effects on its oligomerisation state. Statistical analysis was performed using two-way ANOVA multiple comparisons, followed by Šídák’s multiple comparisons test (****P <0.0001)

Taken together, our biochemical, X-ray structural, and modelling studies provide a strong prediction that Csx15 adopts a linear head-to-tail multimeric structure when bound to cA₄, reminiscent of the ring nucleases Csx3 [22] and AcrIII-2 [33].

CliCsx15 can neutralise CRISPR defence in vivo

To assess the in vivo function of Csx15, we tested CliCsx15 in a well-established recombinant type III-A CRISPR system from Mycobacterium tuberculosis (MtbCsm), expressed in E. coli. This system generates a range of cOA species and activates the cA₄-dependent effector Thioalkalivibrio sulfidiphilus Csx1 to confer plasmid immunity. Escherichia coli C43 cells expressing MtbCsm were transformed with a pRATDuet plasmid encoding Csx1 alone or in combination with CliCsx15, together with a tetracycline resistance gene (tetR). When Csx1 was activated by cA₄ produced by the MtbCsm complex programmed with a crRNA targeting tetR transcripts, a reduction in transformation efficiency was observed, as anticipated. Co-expression with CliCsx15 restored cfu to near-control levels (Figure 8). As a negative control, MtbCsm did not produce cA₄ when programmed with a crRNA targeting the pUC plasmid, and no reduction in transformants was observed (Figure 8). The neutralisation of cA₄-mediated plasmid immunity suggested that CliCsx15 effectively reduced the intracellular cA₄ concentration.

Figure 8. CliCsx15 relieves cA₄-mediated plasmid immunity.

(A) Plasmid challenge assay. Schematic representation of the experimental system. (B) Serial dilutions of E. coli transformants expressing the MtbCsm system together with effector Csx1, ring nuclease CliCsx15, or both were spotted onto antibiotic-selective agar plates with inducer to activate expression of all plasmids. Co-expression of CliCsx15 restored transformation efficiency in cells with cA₄-activated Csx1. Two biological replicates, each with two technical replicates, were performed. Representative plates are shown.

Discussion

Cyclic nucleotide second messengers are a key feature of both eukaryotic and prokaryotic antiviral immunity, providing a convenient means to amplify the primary signal of infection. Typically, these cyclic nucleotides bind and allosterically activate defensive effector proteins to elicit an immune response. An unusual feature of type III CRISPR systems is the near-ubiquity of a mechanism to degrade these cyclic nucleotides via intrinsic or extrinsic ring nucleases, providing a means to switch off this defence pathway when it is no longer needed. Six families of extrinsic cA₄-specific RNs have been described [32,33], in addition to an unusual RN (Crn4), which is capable of degrading all tested cOA species [24]. Here, we tested the candidate RN Csx15, demonstrating that it binds specifically to cA₄, forming a head-to-tail, extended filament conformation. Some family members are capable of slowly degrading cA₄, while others appear to sequester (sponge) the cyclic nucleotide.

Sponging of cyclic nucleotides is commonly encountered as an anti-defence strategy to combat the CBASS, Pycsar, and Thoeris immune systems in bacteria [37,40–42]. This appears effective when the levels of second messenger produced are quite low. In response, some bacteria encode the Panoptes system, which constitutively generates ‘decoy’ cyclic nucleotides to keep a membrane-bound effector deactivated [43,44]. When a phage with a sponge invades cells with Panoptes, sequestration of the decoy molecules results in the activation of Panoptes defence. In contrast to these pathways, type III CRISPR systems typically make much larger quantities of second messenger molecules [8,45,46], which could make sponging a fruitless approach. In response, viruses encode RNs and phosphodiesterases to degrade these molecules [47].

What then is the rationale for Csx15—a protein that is a good cA₄ sponge but a very slow ring nuclease? Firstly, we note that Csx15 was not observed in any phage genomes [32], suggesting that it may not have utility as an anti-CRISPR. We have observed effective neutralisation of a cA₄-activated CRISPR effector by Csx15 in vivo (Figure 6), but it should be borne in mind that this involves over-expression of the protein from a multicopy plasmid, which could enhance its effectiveness. Perhaps, when part of a CRISPR system, Csx15 serves to mop up any small amounts of cA₄ that are generated aberrantly, for example, by the binding of a cellular transcript with a partial match to the crRNA of the type III system. Csx15 could thus act as a dampener to reduce autoimmunity while avoiding the unwanted neutralisation of the system when fully activated by phage infection.

Finally, Csx15 joins a growing list of CARF/SAVED superfamily proteins that multimerise by sandwiching cyclic nucleotide activators between consecutive structural units in a head-to-tail conformation. Notable examples are the SAVED proteins Cap4 [48], TIR-SAVED [49], AcrIII-2 [33], and CalpL [10,31]. For CARF proteins, the repeating unit is a dimer, and stacking interactions are less common but have been observed recently in the CRISPR effector Cat1 [14]. In many of these proteins, cyclic nucleotide-promoted filamentation provides a means to activate a fused effector domain, while in Csx15 and AcrIII-2 the relevant outcome is cA₄ degradation. In CalpL, both activities are evident, and cyclic nucleotide degradation could provide a means to disassemble active filaments and deactivate the immune response. Although clearly a CARF superfamily member, Csx15 has features that distinguish it from the other characterised members of the clade. In particular, one of the six canonical strands of the CARF fold (strand 3 in Crn1) has been lost, while strands 4 and 5 form an extended β-hairpin that mediates dimer:dimer contacts. Filamentation of CARF family proteins in prokaryotic immunity pathways could be an emerging paradigm.

Methods

Cloning

Synthetic genes encoding PfCsx15 (WP_191946534.1) and CliCsx15 (WP_069809204.1), purchased from Integrated DNA Technologies (IDT), were codon optimised for expression in Escherichia coli and subcloned into the pEHisV5TEV vector between the NcoI and BamHI restriction sites. Mutants were created by site-directed mutagenesis using primers with the desired mutations. The E. coli DH5α strain was used for cloning and mutagenesis. Sequence integrity and mutation were confirmed by sequencing (Eurofins Genomics). All synthetic genes and primers used in the present study are listed in Supplementary Table S2.

For the plasmid challenge assay, the CliCsx15 gene was cloned into multiple cloning site -1 (MCS-1) between the NdeI and XhoI restriction sites of pRATDuet, under the control of a T7 promoter. For co-expression, the Csx1 gene was cloned into MCS-2 of the pRATDuet-CliCsx15 plasmid between NcoI and SalI sites, under the control of a pBAD promoter.

Protein expression and purification

The expression and purification of PfCsx15 were previously described [32]; the same protocols were used for CliCsx15. In brief, the E. coli C43 (DE3) strain was used for protein expression. Two litres of cell culture were induced with 0.4 mM isopropyl-β-d-1-thiogalactoside (IPTG) when OD₆₀₀ reached 0.6–0.8 and grown overnight at 25°C. For the determination of molarity, Csx15 was considered to be dimeric. Purified proteins and variants are shown in Supplementary Figure S8.

Ring nuclease assay

RN activity was examined by incubating enzymes (1, 10, and 40 μM) with 80 μM cOA species (Biolog) in the reaction buffer 20 mM Tris–HCl, 250 mM NaCl, pH 7.5, at 30°C for 60 min. The reactions were terminated by adding two equivalent volumes of cold methanol and vortexing. Samples were vacuum-dried before resuspension in water for HPLC analysis.

HPLC analysis

HPLC was conducted on an Ultimate3000 UHPLC system (Thermo Fisher Scientific) with a C18 column (Kinetex EVO 2.1 × 50 mm, particle size 2.6 μm). The column temperature was set at 40°C, and the absorbance was monitored at 260 nm. Samples were analysed by gradient elution with solvent A (20 mM ammonium acetate, pH 8.5) and solvent B (methanol) at a flow rate of 0.3 ml/min as follows: 0–0.5 min, 1% B; 0.5–6 min, 1%–15% B; 6–7 min, 100% B.

Dynamic light scattering

DLS measurements were conducted on a Zetasizer Nano S90 (Malvern) instrument to evaluate the hydrodynamic radii of PfCsx15 and CliCsx15 in the presence or the absence of cOAs. One milligram per millilitre of proteins was mixed with and without 70 μM cOAs in 20 mM Tris–HCl, pH 7.5, 150 mM NaCl, and 10% glycerol. Samples were centrifuged at 14,000 rpm for 10 min prior to measurement. The experiment was carried out at 4°C with 3 measurements of 13 runs in two independent experiments. Statistical analysis was performed using two-way ANOVA multiple comparisons, followed by Šídák’s multiple comparisons test (Prism, version 10.2.2).

Fluorogenic biochemical assay

Assays were carried out on a fluorescence plate reader (FLUOstar Omega, BMG Labtech) with a Greiner 96 half-area plate. PfCsx15 (0, 0.1, 0.2, 0.3, 0.6, 1.2, 2.5, and 5 μM dimer) was incubated with cA₄ (1, 2, and 4 μM) in the presence of RNAse-Alert (IDT) substrate (100 nM). Reactions were incubated at 35°C for 15 min in a buffer of 20 mM Tris–HCl and 100 mM NaCl, pH 7.5, before adding T. thermophilus HB8 cA₄-activated ribonuclease TTHB144 (500 nM dimer). Fluorescence intensity was monitored up to 90 min with 30 s intervals (λ_ex/em 485/520 nm). PfCsx15 with the same gradient was incubated with cA₃ (2 and 5 μM) in the presence of a FAM:Iowa Black^® double-stranded DNA substrate (100 nM) in a buffer of 20 mM Tris–HCl, pH 7.5, 250 mM NaCl, and 10 mM MgCl₂. Reactions were conducted at 35°C for 15 min before adding cA₃-activated DNase Vibrio metoecus NucC (250 nM trimer) and continued for 85 min with same fluorescence measurements. For cA₆-activated ribonuclease Mycobacterium tuberculosis Csm6, PfCsx15 was incubated with cA₆ in the same ratio of Csx15/cA₃ and RNase-Alert in a buffer of 20 mM Tris–HCl, pH 7.5, 250 mM NaCl, and 100 mM potassium l-glutamate with the same fluorescence measurements. The curves are the mean of replicates (Prism, version 10.2.2).

Protein crystallisation

Immediately prior to crystallisation both proteins, PfCsx15 at 15 mg ml⁻¹ and CliCsx15 at 16 mg ml⁻¹, were centrifuged at 14000×g. Sitting drop vapour diffusion experiments were set up at the nanoliter scale at 1:1 and 2:1 protein to reservoir solution ratios. Four commercially available crystallisation screens were set up, after which the plates were sealed and incubated at 20°C. Of the many conditions that yielded crystals, PfCsx15 crystals used in data collection were grown from 22.5% PEG smear medium, 10% glycerol, 0.2 M sodium potassium phosphate (pH 7.5), and 0.1 M HEPES (pH 7.5). CliCsx15 crystals used in data collection were grown from a 22% PEG smear broad, 0.1 M Tris (pH 8.8). All crystals were cryoprotected with 25% glycerol prior to harvesting and cryo-cooling in liquid nitrogen.

X-ray data collection, structure solution, and refinement

All X-ray data were collected at the Diamond Light Source. Data for PfCsx15 were collected at a wavelength 0.9537 Å, −173°C on beamline I04 to 1.88 Å, whereas data from CliCsx15 crystals were collected on IO3 at 0.7918 to 0.90 Å resolution. All data were automatically processed using Xia2 [50]. PfCsx15 data were processed in spacegroup P6₂ with unit cell dimensions a = b = 89.3 Å, c = 31.4 Å, ⍺ = β = 90°, and γ = 120° for one molecule in the asu. CliCsx15 also has one molecule in the asu and was processed in spacegroup C2, unit cell dimensions a = 83.3 Å, b = 32.5 Å, c = 57.8 Å, ⍺ = γ = 90°, and β = 132.8°. The structures were solved by phasing the data using PhaserMR [51] in the CCP4 suite [52], using models generated by AF3 [35]. Initial B-factors were modelled in Phenix [53]. Model refinement was achieved by iterative cycles of REFMAC5 [54] and Phenix, all hydrogens generated but not written to the output file, with manual model manipulation in COOT [55]. Anisotropic B-factor refinement was performed in the case of CliCsx15 and isotropic for PfCsx15. The second subunit of each dimer was generated from the crystallographic two-fold symmetry. Neighbouring symmetry-related dimers were generated from crystallographic translations, producing the dimer-on-dimer assembly.

PDBePISA analysis of both crystal forms supports a biologically relevant dimer. In the CliCsx15 structure, the interface buries 1730 Ų and is stabilised by 28 hydrogen bonds and 4 salt bridges (ΔG = –6.2 kcal mol^-1; CSS = 1.0). In the PfCsx15 structure, a similar interface buries 1786 Å² and exhibits an even more favourable interaction energy (ΔG = –22.5 kcal mol^-1; CSS = 0.63). These observations are consistent with SEC/DLS measurements indicating both proteins are primarily dimeric in the absence of cA₄. PDBePISA does not predict higher-order oligomers to be stable in the absence of cA₄. We also analysed the AF3 model of CliCsx15 together with 4 AMP ligands with PDBePISA, but this did not suggest any stable oligomer, possibly because AF3 models lack realistic side chain positioning.

As Pfcsx15 was crystallised in the presence of phosphate, phosphate ions were modelled in the structure. The geometric qualities of each structure were monitored throughout using MolProbity [56]. Data and refinement statistics are shown in Supplementary Table S1. Final coordinates have been validated and deposited in the Protein Data Bank with deposition codes 9TEU for CliCsx15 and 9TET for PfCsx15, along with the original data.

Analytical size-exclusion chromatography

To analyse the oligomeric state of Csx15, 100 μl of protein (1 mg ml⁻¹) with or without cA₄ was analysed using a Superose 6 Increase 10/300 GL size-exclusion column and equilibrated in 20 mM Tris–HCl, pH 8.0, 250 mM NaCl, 10% glycerol, and 1 mM DTT. The eluted fractions were then analysed by SDS–PAGE gel (Invitrogen). Standards (catalogue number 1511901, Bio-Rad) were analysed in the same running condition, comprising thyroglobulin (670,000 Da), γ-globulin (158,000 Da), ovalbumin (44,000 Da), myoglobin (17,000 Da), and vitamin B12 (1350 Da).

Plasmid challenge assays

Plasmids used in the present study for expression of the programmed type III MtbCsm system include pCsm1-5 (containing Csm interference genes cas10, csm3, csm4, and csm5 from M. tuberculosis and csm2 from M. canettii), pCRISPR-TetR, and pCRISPR-pUC (containing M. tuberculosis cas6 and a CRISPR array targeting a tetracycline-resistance gene or a pUC multiple cloning site, respectively). pRATDuet-derived plasmids carry a tetracycline-resistance gene, either a single gene or both genes encoding TsuCsx1 and CliCsx15. Escherichia coli C43 (DE3) carrying plasmids pCsm1-5 and pCRISPR-TetR or pCRISPR-pUC was transformed with pRATDuet-derived plasmids. The cells recovered for 2 h at 37°C and serially diluted before plating onto LB agar supplemented with 100 μg/ml ampicillin and 50 μg/ml spectinomycin for determination of the recipient cell count. Transformation efficiency was assessed using plates additionally containing 12.5 μg/ml tetracycline. The plasmid immunity was evaluated on plates further supplemented with 0.2% (w/v) d-lactose and 0.2% (w/v) l-arabinose, in addition to all three antibiotics. Plates were incubated overnight at 37°C, then imaged. Technical duplicates of two biological replicates were conducted.

Abbreviations

Acr

anti-CRISPRs

AF3

AlphaFold3

cA₄

cyclic tetra-adenylate

CARF

CRISPR-associated Rossmann fold

CliCsx15

Chlorobaculum limnaeum Csx15

cOA

cyclic oligoadenylate

crRNAs

CRISPR RNAs

DLS

dynamic light scattering

IDT

Integrated DNA Technologies

MCS-1

multiple cloning site -1

Mtb

Mycobacterium tuberculosis

SEC

size-exclusion chromatography

RN

ring nuclease

Vme

Vibrio metoecus

Data Availability

All experimental materials described in this paper are available from the corresponding author on request. The protein structure co-ordinates and data have been deposited in the Protein Data Bank with deposition codes 9TEU and 9TET [57]. Raw data underlying the fluorescence assays are provided as supplementary data.

Competing Interests

The authors declare that there are no competing interests associated with the manuscript.

Funding

This work was supported by a European Research Council Advanced Grant [grant REF 101018608 to M.F.W.].

Open Access

Open access for this article was enabled by the participation of the University of St Andrews in an all-inclusive Read & Publish agreement with Portland Press and the Biochemical Society under an agreement with JISC.