CRISPR antiphage defence mediated by the cyclic nucleotide-binding membrane protein Csx23

Abstract

CRISPR-Cas provides adaptive immunity in prokaryotes. Type III CRISPR systems detect invading RNA and activate the catalytic Cas10 subunit, which generates a range of nucleotide second messengers to signal infection. These molecules bind and activate a diverse range of effector proteins that provide immunity by degrading viral components and/or by disturbing key aspects of cellular metabolism to slow down viral replication. Here, we focus on the uncharacterised effector Csx23, which is widespread in Vibrio cholerae. Csx23 provides immunity against plasmids and phage when expressed in Escherichia coli along with its cognate type III CRISPR system. The Csx23 protein localises in the membrane using an N-terminal transmembrane α-helical domain and has a cytoplasmic C-terminal domain that binds cyclic tetra-adenylate (cA4), activating its defence function. Structural studies reveal a tetrameric structure with a novel fold that binds cA4 specifically. Using pulse EPR, we demonstrate that cA4 binding to the cytoplasmic domain of Csx23 results in a major perturbation of the transmembrane domain, consistent with the opening of a pore and/or disruption of membrane integrity. This work reveals a new class of cyclic nucleotide binding protein and provides key mechanistic detail on a membrane-associated CRISPR effector.

Graphical Abstract

Figure 1.

(A) Organisation of Vibrio type III-B CRISPR (Cmr) loci. (B) Predicted structure of V. cholerae HE-45 Csx23 (AF2) coloured by local distance difference test (LDDT) scaled from blue (high prediction confidence) to red (low).

Figure 2.

Plasmid challenge assay. (A) Schematic representation of the basis for the assay. (B) A serial dilution of the transformation mixture of E. coli cells expressing the VmeCmr complex with plasmids carrying a target sequence and varying effector genes was spotted onto agar plates containing antibiotics to select for all plasmids. VmeCmr complexes loaded with crRNA targeting the tetracycline resistance gene of the incoming plasmid (targeting crRNA, VmeCmr[TetR]) or loaded with crRNA targeting a sequence that is not present in the host genome or plasmids (non-targeting crRNA, VmeCmr[pUC]) were used.

Figure 3.

Csx23 is a tetrameric protein and binds cA₄. (A) Cross-linking of FL or CTD Csx23 in the presence and absence of cA₄ or cA₃ and increasing amounts of BS3 cross-linking reagent. The FL protein formed tetramers in detergent regardless of cOA presence; the CTD could form tetramers in solution only in the presence of cA₄. Blue dots show the predicted quaternary structure corresponding to each band on the SDS-PAGE. (B) EMSA showing binding of radioactive cA₄ by the FL and CTD Csx23 proteins. Protein concentrations were 25, 50, 100, 200, 400, 800 and 1600 nM for FL Csx3 and 0.4, 1, 4, 10, 40 and 100 μM for CTD Csx23; the cA₄ concentration is indicated beneath the gel.

Figure 4.

Structure of the tetrameric Csx23 CTD bound to cA₄. Structure of tetrameric Csx23 CTD (green cartoon, with exception of flexible loop (residues 107–113) shown in pink) in complex with cA₄ (sticks coloured by element, with carbon in yellow) from (A) ‘side’, (B) ‘top’ and (C) ‘bottom’ views. The sodium ion is shown as a blue sphere. R95 (sticks coloured by element, with carbon in green) and R110 and F111 (sticks coloured by element, with carbon in pink) are also shown. (D) Interactions formed between cA₄ (yellow) and residues R95 (green), R110 and F111 (both pink) in Csx23 CTD. Colouring as in panels A–C. Residues from just one subunit are shown for clarity. Black dotted lines represent electrostatic interactions; grey dotted lines represent hydrogen bonds; purple dotted lines represent water-mediated hydrogen bonds. (E) AF2 model of the FL Csx23 tetramer coloured by LDDT scaled from high (blue) to low (red) prediction confidence (top), and the same model with one subunit shown in light blue and the other three in light green (bottom).

Figure 5.

Investigation of roles of conserved residues. (A) Plasmid challenge assay for wild-type and variant Csx23 proteins. The R95A and R110A variants did not provide immunity, suggesting Csx23 function was disrupted. (B) EMSA showing that the R95A and R110A variants are unable to bind the cA₄ activator in vitro. Weak binding of cA₅ was observed for the R95A variant. Each binding reaction contained a two-fold molar excess of Csx23 tetramer over cOA. The amount of [³²P]-cOA was kept constant, unlabelled cA₄ was added to give final cA₄ concentrations of 0.25 μM (lanes 2, 5, 8), 2.5 μM (lanes, 3, 6, 9) and 25 μM (lanes, 1, 4, 7, 10).

Figure 6.

Pulse EPR demonstrates cA₄ mediated perturbation of the transmembrane domain. Pulse electron-electron double resonance (PELDOR) data of the Csx23 AALA V52R1 variant reconstituted in nanodiscs (ND) in the presence (red) or absence (black) of cA₄. (A) Raw PELDOR data. (B) Background-corrected traces with fits. (C) Overlay of corresponding distance distributions shown as 95% confidence bands with predicted distributions from MMM and mtsslWizard based on the AF2 predicted tetrameric structure; colour bars indicate reliability ranges (green: shape reliable; yellow: mean and width reliable; orange: mean reliable; red: no quantification possible). (D) Cartoon representation of AF2 predicted tetrameric structure of the spin-labelled Csx23 V52R1 variant; each subunit is shown in a different colour.

Figure 7.

Phage P1 challenge of E. coli expressing VmeCmr and varying effector genes. (A) Growth curves of cells with no effector, Csx23 or NucC with increasing amounts of phage P1. MOI: multiplicity of infection. The growth curve in the absence of phage infection is shown for comparison. (B) Growth curves from cells carrying Csx23 or no effector after phage P1 infection at MOI 15 and expanded view to show early time points. (C) The number of viable phage particles was determined by applying a dilution series of cleared supernatant from infected cultures to agar plates overlaid with BL21(DE3) Star cells, the same E. coli strain as used in all in vivo assays. The plaques from ≥3 independent experiments consisting of two biological replicates each were counted and plotted using Graphpad Prism. Mann–Whitney test (Prism 10.0) was used to determine statistical difference (not significant (ns): P-value ≥0.05, ****P-value < 0.0001). Cultures of strains containing either Csx23 or NucC contained 100 times fewer viable phage compared to those that did not carry any effector.