A miniature CRISPR–Cas10 enzyme confers immunity by inhibitory signalling

Bacterial strains and growth conditions

Standard bacterial cultures were grown in lysogeny broth (LB Lennox) at 37 °C and 250 rpm except where stated otherwise. When necessary, 34 µg ml⁻¹ chloramphenicol (+Ch), +50 µg ml⁻¹ kanamycin sulfate (+K) or +50 µg ml⁻¹ carbenicillin (+C) was supplemented to media and LB agar plates. For cellular assays, all bacterial strains were stored at −80 °C in 25% glycerol (Sigma) when not in use. Cloning and cellular assays were generally performed in E. coli DH10b genotype cells (F- mcrA Δ(mrr-hsdRMS-mcrBC) endA1 recA1 ϕ80dlacZΔM15 ΔlacX74 araD139 Δ(ara, leu)7697 galU galK rpsL (StrR) nupG λ-) (Intact Genomics, NEB). For phage experiments involving G17 or Goslar, E. coli strain ECOR47 was used⁶².

Phage propagation and scaling

Phages were propagated using standard protocols. Typically, phage production was performed at 37 °C in LB Lennox media infecting E. coli BW25113 (F- DE(araD-araB)567 lacZ4787(del)::rrnB-3 LAM- rph-1 DE(rhaD-rhaB)568 hsdR514)⁶³ at an initial MOI of 0.1. Phages G17 and Goslar were propagated on E. coli ECOR47. Phage titres were determined on their propagation hosts. Bacteriophages used in this study and their sources^35,36,51 have been described at length in Supplementary File 2.

Plasmid construction

Plasmids were assembled using PCR, gel extraction (Zymo) and Golden Gate assembly⁶⁴ or Gibson Assembly using 25–30 bp of overlapping homology⁶⁵. For some assemblies, Golden Gate ready DNA was ordered from TWIST BioSciences in lieu of generation by PCR. Native promoter Panoptes includes 300 bp upstream of the gene encoding mCpol and 150 bp downstream of the gene encoding 2TMβ. In general, plasmids were propagated in DH10b genotype E. coli (Intact Genomics). For G17 and Goslar assays, select plasmids were transferred to E. coli ECOR47 through electroporation (Intact Genomics) using standard electroporation parameters (2,000 V and 200 Ω). For protein expression and purification, plasmids were transferred into BL21 AI genotype E. coli (F- ompT hsdSB (rB-mB-) gal dcm araB::T7RNAPtetA) (Thermo Fisher). For assays involving two plasmids, plasmids were first cloned individually and co-transformed into DH10b genotype E. coli (Intact Genomics) through electroporation. All plasmids and co-transformations used in this study were sequenced confirmed by full-plasmid sequencing (Plasmidsaurus). To verify co-transformed plasmids, raw reads were mapped against reference plasmid sequences using Geneious. Plasmids used in this study are listed in Supplementary File 1.

Phage infection assays

Phage infection plaque assays were performed via double agar overlay. Overlays were formed by mixing 100 µl of saturated overnight cultures grown at 37 °C at 250 rpm with 5 ml molten LB Lennox agar (0.7% w/v, 60 °C). For G17 and Goslar assays, a lower percentage top agar concentration was used (0.35% w/v). The agar–bacterial mixture was supplemented with kanamycin to a final overlay concentration of 50 µg ml⁻¹. For plaque assays involving crystal violet (Sigma), crystal violet was supplemented to a final overlay concentration as specified. The top agar and bacterial mixture was poured onto a 5 ml LB agar and kanamycin plate and left to dry under microbiological flame for 15 min. Phages were diluted tenfold in SM buffer (Teknova) and 2 μl of each dilution were spotted onto the top agar overlay and dried under microbiological flame. Once dry, plates were incubated at 30 °C for 12–16 h. Plates were scanned in a standard photo scanner (Epson) and PFU were enumerated, keeping note of changes in plaque size relative to a negative control. ‘Lysis from without’⁶⁶ phenotypes were interpreted as a lack of productive phage infection and approximated as 1 PFU. Efficiency of plaquing calculations were calculated as mean(PFU in condition)/mean(PFU negative control). The negative control is defined as a strain harbouring a plasmid encoding red fluorescent protein (RFP) instead of Panoptes (pBA1326). All plaque assays were performed in biological triplicate from independent overnight cultures. Visualizations were performed using Seaborn (v0.13.2) in Python.

For plaque assays involving CRISPRi-ART³⁹, experiments were performed with the following modifications. Kanamycin, chloramphenicol and aTc (Sigma) were additionally added to final overlay concentrations of 50 µg ml⁻¹, 34 μg ml⁻¹ and 20 nM, respectively, and plated onto a 5-ml LB agar, kanamycin and chloramphenicol plate. The negative control is defined as Panoptes (pBA1751) alongside a CRISPRi-ART vector with an RFP-targeting guide RNA (pBA635)³⁹. Guide sequences are listed in Supplementary File 1. The positive control is defined as a pJEx vector without Panoptes (pBA1326) alongside a CRISPRi-ART guide RNA targeting a phage not present in the experiment. Titres were calculated as mean (PFU condition). All plaque assays were performed in biological triplicate from independent overnight cultures. Visualizations were performed using matplotlib (v3.7.2) and Seaborn (v0.13.2) in Python. CRISPRi-ART guide RNAs were chosen using gRNA1 and gRNA4 designs as previously described³⁹.

Bacteriophage liquid growth and phage production

Liquid phage experiments were performed in a Biotek plate reader using LB Lennox + kanamycin media. Strains containing native context Panoptes (pBA1751) or a negative control (pBA1326) plasmid were grown overnight at 37 °C and 250 rpm. Of overnight culture, 8 × 10⁶ CFU were seeded into each well of a 96-well plate (3903, Corning) in 200 µl media. For phage experiments, T4 was diluted in LB + kanamycin media to achieve defined MOIs during infection except for MOI = 0 in which no phage was added. Growth was monitored in a Biotek Cytation 5 plate reader for 12 h at 800 rpm shaking at 30 °C with OD₆₀₀ readings every 5 min. At the end of the experiment, cultures were pelleted and the supernatant from investigated wells was collected. Phage production was estimated via plaque assay (above) on E. coli harbouring a negative control (pBA1326) plasmid and dividing by the effective titre at time 0. All liquid phage experiments were performed in biological triplicate and replicate conditions from independent overnights. Visualizations were performed using matplotlib (v3.7.2) and Seaborn (v0.13.2) in Python.

Estimation of free phage particle production from a single round of infection was performed in 5 ml cultures. Cultures were inoculated with 2 × 10⁸ CFU E. coli harbouring native context Panoptes (pBA1751) or a negative control (pBA1326) plasmid in LB + kanamycin media and incubated at 30 °C at 250 rpm for 15 min. Following incubation, approximately 2 × 10⁶ CFU of phage T4 was added for a low MOI infection of approximately 0.01. Infections proceeded at 30 °C at 250 rpm and 200 µl sampled every 30 min for 6.5 h. For each 200 µl sample, remaining cells were pelleted, 100 µl supernatant was extracted and stored on ice until every sample was collected. Phage titres were enumerated via plaque assay on E. coli harbouring a negative control plasmid (pBA1326). Free phage production was calculated by dividing the sample titre by the number of added phages. Replicates were performed in biological triplicate, sourcing samples from parallel cultures seeded from independent overnights. Visualizations were performed using matplotlib (v3.7.2) and Seaborn (v0.13.2) in Python.

Panoptes toxin–antitoxin assays

To test for a potential toxin–antitoxin relationship in Panoptes, we cloned candidate antitoxin mCpol or mCpol D57A under control of the pTet promoter in a p15a-CmR plasmid and candidate toxin 2TMβ under control of the pJex promoter⁴¹ in a low-copy SC101-KanR plasmid. dCas13d under pTet control on a p15a-CmR plasmid (pBA635)³⁹ was used as an antitoxin negative control. RFP expressed under pJex control on a SC101-KanR plasmid was used as a toxin negative control. Both sets of plasmids were co-transformed into DH10b E. coli (Intact Genomics) and sequence verified using whole-plasmid sequencing (Plasmidsaurus), followed by read alignment (Geneious). Sequence-confirmed co-transformants were stored at −80 °C in 20% glycerol (v/v) until further use.

To perform solid agar toxin–antitoxin assays, LB agar plates were freshly poured and dried under flame with the following supplements: 35 µg ml⁻¹ chloramphenicol, 50 µg ml⁻¹ kanamycin, variable amounts of crystal violet to induce toxin expression and variable amounts of aTc to induce antitoxin expression. For toxin expression conditions, +50 nM crystal violet was added. For antitoxin expression conditions, +2 nM aTc was added. Once the agar was dried, three independent overnight cultures containing candidate toxin and antitoxin plasmids were plated in 10× serial dilutions with 5-µl spots and let dry under flame. Once dried, plates were incubated at 30 °C overnight. To let colonies mature for imaging, plates were transferred to a 37 °C incubator for an additional 24 h. Colonies were imaged and counted.

To perform liquid culture toxin–antitoxin assays, three independent overnight cultures containing candidate toxin and antitoxin plasmids were inoculated in LB media at 8 × 10⁶ CFU in a Corning 3903 plate with the following supplements: 35 µg ml⁻¹ chloramphenicol, 50 µg ml⁻¹ kanamycin, variable amounts of crystal violet to induce toxin expression and variable amounts of aTc to induce antitoxin expression. For toxin expression conditions, +125 nM crystal violet was used. For antitoxin expression conditions, +20 nM aTc was used. The plate was monitored in a Cytation5 plate reader (Biotek) at 30 °C at 807 rpm and OD₆₀₀ measured every 5 min for 12 h. Data were plotted using the Seaborn (v0.13.2) package in Python.

Panoptes activator assays

To test for an activator relationship with Panoptes, we cloned Panoptes in its native context into a p15a-CmR plasmid and candidate activators T4Acb1, T4Acb1(H44A,H113A), T4Acb2, T4Acb2(Y8A) and N4Acb4 under control of the pJex promoter⁴¹ in a low-copy SC101-KanR plasmid. dCas13d under pTet on a p15a-CmR plasmid (pBA635)³⁹ was used as the Panoptes negative control. RFP under pJex on a SC101-KanR plasmid was used as an activator negative control.

To perform liquid culture activator assays, three independent overnight cultures containing candidate Panoptes and candidate activator plasmids were inoculated in LB media at 8 × 10⁶ CFU in a Corning 3903 plate with the following supplements: 35 µg ml⁻¹ chloramphenicol, 50 µg ml⁻¹ kanamycin, variable amounts of crystal violet to induce candidate activator expression and no aTc for Panoptes expression. For candidate activator expression conditions, +125 nM crystal violet was used. The plate was monitored in a Cytation5 plate reader (Biotek) at 30 °C, 807 rpm and OD₆₀₀ measured every 5 min for 12 h. Data were plotted using the Seaborn (v0.13.2) package in Python.

To measure viability from liquid culture activator assays, we used a modified protocol. Three independent overnight cultures containing candidate pTet-native promoter Panoptes or pTet-dCas13d (negative control) and pJEX-T4Acb2 or pJEX-RFP (negative control) were inoculated in LB media at 8 × 10⁶ CFU in a Corning 3903 plate with the following supplements: 35 µg ml⁻¹ chloramphenicol, 50 µg ml⁻¹ kanamycin and no inducers. Cells were grown for 3 h in a Cytation5 plate reader (Biotek) at 30 °C and 807 rpm (OD₆₀₀ ~ 0.2). At this point (t = 0), either +125 nM crystal vioelt or +0 nM was supplemented to the cultures to activate or not activate Panoptes and grown for an additional 2 h (t = 120). At both t = 0 and t = 120, 5 µl of cells were sampled, tenfold serially diluted in inducer-free media and spotted on inducer-free LB agar with antibiotics. Plates were incubated overnight at 37 °C and colonies counted to estimate viability from t = 0 and t = 120. Data were plotted using the Seaborn (v0.13.2) package in Python.

Protein expression and purification

Expression sequences for ECOR31 mCpol were cloned into a custom pET-based vector by Gibson assembly to yield an N-terminal His₁₀-MBP-TEV or C-terminal TEV-MBP-His₁₀ construct. Expression plasmids for ECOR31 TM2β residues 72–200 (2TMβ(∆2TM)) were cloned into a pET Duet-1 vector for co-expression of His₆-SUMO2-TM2β and MBP-TEV-mCpol constructs.

Proteins were expressed in E. coli Rosetta 2 (DE3) pLysS by growing cells to an OD₆₀₀ of 0.4–0.6 in 2× yeast extract tryptone (2×YT) medium at 37 °C and induced with 0.5 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) following a cold shock at 4 °C. After induction, cells expressing each protein were grown overnight at 16 °C. Cells were collected by centrifugation for 20 min at 12,300g and 4 °C and resuspended in 20 mM Tris-HCl, pH 8.0, 10 mM imidazole, 2 mM MgCl₂, 500 mM KCl, 10% (v/v) glycerol, 0.5 mM Tris (2-carboxyethyl) phosphine and Roche protease inhibitor.

Cells were lysed by sonication, and cell lysate was clarified by centrifugation at 17,000g and 4 °C for 0.5 h. The supernatant was bound to Nickel-NTA affinity resin pre-equilibrated with wash buffer (20 mM Tris-HCl, pH 8.0, 500 mM KCl, 30 mM imidazole, 10% (v/v) glycerol and 0.5 mM Tris(2-carboxyethyl) phosphine (TCEP)) for 1 h at 4 °C. Supernatant was discarded and resin was washed 5 × 30 ml wash buffer (20 mM Tris-HCl, pH 8.0, 500 mM KCl, 30 mM imidazole, 10% (v/v) glycerol and 0.5 mM TCEP). All buffers for the purification of ECOR31 TM2β were additionally supplemented with 1 µM 2′3′-c-di-AMP (BioLog Life Science Institute) in the washing and subsequent steps. Protein was eluted in 5 ml elution buffer (20 mM Tris-HCl, pH 8.0, 500 mM KCl, 300 mM imidazole, 10% (v/v) glycerol and 0.5 mM TCEP). Recombinant TEV protease or SUMO protease 2 (SENP2) with an N-terminal His-tag (BPS Bioscience) was added to the elution for cleavage and dialysed overnight at 4 °C in a 3.5 kDa MWCO dialysis cassette (Thermo Fisher Scientific) with dialysis buffer (20 mM Tris-HCl, pH 8.0, 500 mM KCl, 30 mM imidazole, 10% (v/v) glycerol and 0.5 mM TCEP). The resultant solution was passed over a 5-ml Ni-NTA Superflow cartridge (Cytiva). Flow-through was collected, concentrated to less than 2 ml using a 3 kDa MWCO centrifugal filter (Amicon), and loaded onto a HiLoad 16/600 Superdex 200-pg column (Cytiva). Elution was isocratic (20 mM Tris-HCl, pH 8.0, 500 mM KCl, 10% glycerol and 1 mM TCEP) and monitored by absorbance at 280 nm. Peaks were pooled (approximately 86 ml for ECOR31 TM2β; approximately 83 ml for ECOR31 mCpol), concentrated to 5 mg ml⁻¹ (ECOR31 TM2β) or 3 mg ml⁻¹ (ECOR31 mCpol) as determined by absorbance at 280 nm, snap frozen and stored at −80 °C.

Analysis of recombinant protein

Purified protein was analysed by SDS–PAGE. Samples were prepared in 1X protein loading dye (50 mM Tris-HCl, pH 6.8, 15 mM EDTA, 6% (v/v) glycerol, 10% SDS and bromophenol blue), heated at 95 °C for 3 min, loaded onto a 12% Mini-Protean TGX precast protein gel (Bio-Rad) and run at 125 V until the dye front reached the bottom of the gel. Gels were stained in 30% ethanol, 10% glacial acetic acid in water and 0.1% (w/v) R-250 Coomassie, and destained in 40% ethanol and 10% glacial acetic acid in water.

Crystallization and structure determination

Crystals of mCpol bound to a hydrolysis-resistant amine-modified analog of ATP (ApNHpp; NU-449-1, Sapphire North America) were grown at 20 °C using the hanging drop vapour diffusion method. A 1-µl solution of 3 mg ml⁻¹ mCpol in 20 mM Tris-HCl, pH 8.0, 500 mM NaCl and 5% glycerol was mixed with 1 µl 0.1 M bicine, pH 8.0, and 15% PEG 1500 and supplemented with 0.2 µl 10 mM MgCl₂ with 10 mM of ApNHpp. Single crystals appeared within 3 days and were cryoprotected in a solution of mother liquor with 30% glycerol before being flash cooled in liquid nitrogen.

Data for mCpol bound to ApNHpp were collected via fine-phi slicing using 0.2° oscillations at beamline 12-2 at Stanford Synchrotron Radiation Lightsource at SLAC National Accelerator Laboratory. X-ray diffraction data were measured to 2.28 Å resolution.

Processing and refinement of crystallographic data

Crystallographic data were processed with the SSRL autoxds script with an I/sigI cut-off ≥ 1.50. Crystals displayed moderate anisotropic X-ray diffraction with some diffraction extending beyond 2.0 Å resolution. However, the resolution was isotropically truncated to 2.28 Å resolution to generate a robust complete dataset. The structure was solved by molecular replacement⁶⁷ using a ColabFold-generated⁶⁸ model (pLDDT = 95.12) of mCpol with residual residues from the C-terminal cleavage site (mCpol-ENLYFQ) in PHENIX⁶⁹. Molecular replacement successfully identified the placement of two mCpol monomers as indicated by the log-likelihood gain of 556.03 and the translation-function Z score of 23.6. The structure was refined using PHENIX⁷⁰ including simulated annealing, non-crystallographic symmetry and TLS (translation, libration, screw-rotation) parameters. In each protein monomer (chains A and B) residues 121–125 were disordered and thus not included in the model. The model was built and adjusted using COOT⁷¹. The structure was refined to a final R_free and R_work of 24.88% and 22.43%, respectively (Extended Data Table 1). Extended Data Table 1 shows data processing and model refinement statistics. Atomic coordinates and structure factors have been deposited to the Protein Data Bank.

Sequence alignment of mCpol domains

Proteins in the Pfam entry PF18182 ‘minimal CRISPR polymerase domain’ were downloaded and used to create an alignment of 163 mCpol sequences. Sequences were trimmed and aligned using MAFFT alignment with default parameters in Geneious Prime (v2023.2.1).

TLC of cyclase products

Recombinant enzymes were assessed for cyclase activity by in vitro reactions with nucleoside triphosphates and analysis by TLC. Cyclase activity assays were initiated by the addition of recombinant enzyme (40 μM final) in reaction buffer (50 mM Tris, pH 8.0, 10 mM MgCl₂ and 100 mM NaCl) to 5 mM NTP (Thermo Scientific). The reaction mixture was incubated at 37 °C for 18 h and stopped by vortexing for 20 s.

Recombinant enzymes were assessed for cyclic dinucleotide (c-di-AMP) degradation in biochemical reactions containing buffered c-di-AMP and products were analysed by TLC. Reactions were initiated by the addition of recombinant enzyme (40 μM) in reaction buffer (50 mM Tris, pH 8.0, 10 mM MgCl₂ and 100 mM NaCl) to 1.25 mM c-di-AMP (Biolog or MedChemExpress). The reaction mixture was incubated at 37 °C from 10 min to 18 h and stopped by vortexing for 20 s.

To silica gel 60 matrix TLC plates (5 cm × 10 cm, glass support with fluorescent indicator 254 (Supelco)), was added 2 μl in vitro enzymatic reaction mix. Separation was performed in an eluent of n-propanol:ammonium hydroxide:water (11:7:2 v/v/v) for 45 min. The plate was allowed to dry fully and visualized with a short-wave ultraviolet light source at 254 nm. Uncropped TLC images are available in Supplementary Fig. 1.

S1 nuclease digestion of cyclase products

To determine the linkage identity of products formed in vitro, cyclase reactions described previously were subjected to digestion by S1 nuclease. To a reaction containing 1X S1 nuclease buffer (40 mM sodium acetate (pH 4.5 at 25 °C) and 300 mM NaCl₂), 20 μl of the vortex-inactivated cyclase reaction and 200 U S1 nuclease (Thermo Scientific) were added to a total volume of 40 μl. The reaction was incubated at 37 °C for 4 h and inactivated by the addition of 2 μl of 0.5 M EDTA with heating (70 °C for 10 min).

Preparation of extracts for mass spectrometry

mCpol–ATP reactions were diluted 1:2 in deionized H₂O, centrifuged at 13,000g for 15 min and used directly in subsequent analysis. Species identified by TLC were directly analysed by TLC–MS. Silica containing the product was scraped away from the TLC plate and added to 40 μl water. The resultant slurry was vortexed and heated at 30 °C for 10 min and then centrifuged at 13,000g for 15 min. The supernatant was removed for subsequent analysis.

Cyclase product analysis by HPLC–MS

Cyclase product extracts were analysed using a liquid chromatography system (1200 series, Agilent Technologies) that was connected in line with an LTQ-Orbitrap-XL mass spectrometer equipped with an electrospray ionization source (Thermo Fisher Scientific). The liquid chromatography system was equipped with a G1322A solvent degasser, G1311A quaternary pump, G1316A thermostatted column compartment and G1329A autosampler unit (Agilent). The column compartment was equipped with an Ultra C18 column (length of 150 mm, inner diameter of 2.1 mm and particle size of 3 µm; 9174362, Restek). Ammonium acetate (98% or more, Sigma-Aldrich), methanol (Optima LC–MS grade, 99.9% minimum; Fisher) and water purified to a resistivity of 18.2 MΩ cm⁻¹ (at 25 °C) using a Milli-Q Gradient ultrapure water purification system (Millipore) were used to prepare mobile-phase solvents. Mobile-phase solvent A was water and mobile-phase solvent B was methanol, both of which contained 10 mM ammonium acetate. The elution program consisted of isocratic flow at 0.5% (v/v) B for 2 min, a linear gradient to 99.5% B over 2 min, isocratic flow at 99.5% B for 4 min, a linear gradient to 0.5% B over 1 min, and isocratic flow at 0.5% B for 21 min, at a flow rate of 100 µl min⁻¹. The column compartment was maintained at 30 °C and the sample injection volume was 1 µl. External mass calibration was performed in the positive ion mode using the Pierce LTQ electrospray ionization positive ion calibration solution (88322, Thermo Fisher Scientific). Full-scan, high-resolution mass spectra were acquired in the positive ion mode over the range of m/z = 300–2,000, using the Orbitrap mass analyzer, in profile format, with a mass resolution setting of 60,000 (at m/z = 400, measured at full-width at half-maximum peak height). Tandem mass (MS/MS or MS²) spectra were acquired using collision-induced dissociation in the linear ion trap, in centroid format, with the following parameters: isolation width = 3 m/z units, normalized collision energy = 28%, activation Q = 0.25 and activation time = 30 ms. Data acquisition was controlled using Xcalibur software (v2.0.7, Thermo Fisher Scientific).

Mass spectrometry data processing

Raw data were converted to mzXML format using msconvert (v3.0.19052.1) from the Galaxy platform⁷². Data were then processed using the open source software MZmine (v3.9.0). Compound identification was performed by differential mass spectrometry of in vitro reactions with active and inactive enzymes, and by comparing retention times and m/z with those of chemical standards.

2′3′-c-di-AMP titration

Recombinant 2TMβ(∆2TM) was dialysed overnight at 4 °C into SEC buffer (20 mM Tris-HCl, pH 8.0, 500 mM KCl, 10% glycerol and 1 mM TCEP). Protein samples were diluted to 4.75 mg ml⁻¹ (323 µM) in SEC buffer with varying concentrations of 2′3′-c-di-AMP (0, 500 µM, 1 mM and 2 mM) and 100 µl was loaded into a Superdex 200 Increase 3.2/300 column (Cytiva). Elution was isocratic in SEC buffer and monitored by A280. Data were visualized in GraphPad Prism (v10.2.2).

Thermal shift assay of 2TMβ

Spectra were obtained using a Bio-Rad CFX Connect Real-Time PCR Detection System. Solutions contained 1 μM 2TMβ(∆2TM) in 20 mM Tris-HCl, pH 8.0, 500 mM KCl, 10% glycerol, 1 mM TCEP and 2X SYPRO orange dye with or without 20 μM 2′3′-c-di-AMP. To a 96-well half-skirted clear-bottom PCR plate (Axygen) was added 20 μl of each solution. Wells were sealed with Axyseal sealing film (Axygen). Fluorescence was measured as the solutions were heated from 18 °C to 90 °C at a rate of 2 °C per minute. The fluorescence signal as a function of temperature was plotted, and the background values of buffered solution (with or without 2′3′-c-di-AMP) without protein was subtracted from each sample. Measurements were performed in triplicate and the graph is the average of three replicates ± standard deviation.

Native PAGE

Recombinant 2TMβ(∆2TM) was dialysed overnight at 4 °C into 20 mM Tris-HCl, pH 8.0, 500 mM KCl, 10% glycerol and 1 mM TCEP. To 1X native loading buffer (50 mM Tris-Cl, pH 8.0, 0.1% bromophenol blue, 10% glycerol and 100 mM dithiothreitol) 4 μg 2TMβ(∆2TM) was added at 4 °C. The sample was analysed by electrophoresis on an 8% native polyacrylamide gel (5% 29:1 acrylamide:bis solution, 750 mM Tris-Cl, pH 8.8) in Tris-glycine running buffer (1.5% w/v Tris-Cl, pH 8.8, and 9.4% (w/v) electrophoresis grade glycine) at 4 °C and 150 V for 105 min. The gel was stained in 30% ethanol, 10% glacial acetic acid in water and 0.1% (w/v) R-250 Coomassie, and destained in 40% ethanol and 10% glacial acetic acid in water. Uncropped gel images are available in Supplementary Fig. 1.

Fluorescence microscopy

Inducing agents or water were added to 400 µl or 200 µl of log phase cultures at OD₆₀₀ 0.12–0.2 (final concentrations: aTc = 200 nM and crystal violet = 50 nM), which were then incubated in a roller for 50 min at 30 °C. Stains were added to the cultures and incubated, rolling, for an additional 10 min at 30 °C. The final stain concentrations were 4 µg ml⁻¹ DAPI and either 2.5 µg ml⁻¹ MitoTracker Green FM (Fig. 3i–k) or 1 µg ml⁻¹ FM4-64 and 2.5 µg ml⁻¹ SYTOX Green (Extended Data Fig. 6). FM4-64 is a membrane stain and was used to verify the presence of membrane lesions in the SYTOX Green staining assay, but is not shown. Of culture, 20 µl was spotted on an agarose imaging pad (25% LB and 1% agarose) in single-well concavity slides for imaging. Cells were imaged at room temperature with eight 0.2-µm slices in the z axis. Images were collected on a DeltaVision Elite Deconvolution microscope (Applied Precision) using DeltaVision SoftWoRx (v6.5.2) and figure panels were prepared in Adobe Photoshop (v21.2.0) and Adobe Illustrator (v24.2).

Sequence mining and genomic neighbourhood analysis

A collection of mCpol encoding genes was curated using the following strategy. First, a DALI search was performed against representatives of the clustered AlphaFold Database (AFDB) using an AlphaFold2 model of the mCpol (A0A426EXS8). After filtering for a Z score of 12, a structurally informed multiple sequence alignment (MSA) was generated and used for phylogenetic analysis as previously described⁷³. The clade containing mCpol representatives (n = 5) was extracted from the resulting tree, and the selection was then expanded to include all cluster members (n = 29). A structurally informed MSA was generated from this set and used as input for an HMM search against the NCBI_NR_Aug_2018 database on the MPI Bioinformatics Toolkit website (MSA enrichment iterations using HHblits: 1, e-value cut-off for reporting: e−10, maximum target hits: 1,000). This set of mCpol candidates (n = 592) was aligned using MAFFT (v7.490) in Geneious Prime (v2023.0.1; algorithm: E-INS-i, scoring matrix: BLOSUM62 and gap open penalty: 1.3). Fragmented mCpol genes were removed from the alignment, and the remaining genes were used for neighbourhood analysis (n = 593).

A custom Python script was used to retrieve Genbank files of mCpol-containing loci. Only contigs longer than 20 kb were kept for further analysis (n = 505). The resultings contigs were converted to FASTA format and deduplicated using mmseqs⁷⁴ easy-cluster (-c 0.5 –min-seq-id 0.5 –dbtype 2 –cluster-mode 2). The final set of sequences (n = 252) were then subjected to defence association analysis via DefenseFinder Web Service (v2.0.0)^46,75 (https://defensefinder.mdmlab.fr/. The outputs of DefenseFinder were then further processed using a custom Python script to add defence system hit annotations to the Genbank files. This final output was used to investigate co-occurrence of mCpol with CBASS and other types of defence systems.

Taxonomic analysis of mCpol-encoding genes

To determine the taxonomic distribution of mCpol homologues, we retrieved the NCBI taxonomic identifier (taxID) associated with each protein sequence. Protein IDs were supplied to a custom script that queries the NCBI Entrez system via E-utilities (esearch, elink, efetch and xtract)⁷⁶ to extract both the taxID. For protein IDs that failed automated retrieval, taxonomic information was manually obtained using NCBI’s web interface. For each taxID, we used a separate custom Python script that calls the NCBI Datasets⁷⁷ command-line tool to obtain taxonomy summaries in JSON format and subsequently parses the taxonomic ranks saved as a tsv. A Jupyter notebook was used to further cross-referenced protein IDs with locus tag information from associated annotation files. Merging and reformatting of the dataset was performed in Python using pandas, yielding a finalized tab-delimited table containing accession IDs, taxonomic lineages and protein identifiers for downstream analysis.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.