Phage defence by deaminase-mediated depletion of deoxynucleotides in bacteria

Abstract

Vibrio cholerae biotype El Tor is perpetuating the longest cholera pandemic in recorded history. The genomic islands VSP-1 and VSP-2 distinguish El Tor from previous pandemic V. cholerae strains. Using a co-occurrence analysis of VSP genes in >200,000 bacterial genomes we built gene networks to infer biological functions encoded in these islands. This revealed that dncV, a component of the cyclic-oligonucleotide-based anti-phage signalling system (CBASS) anti-phage defence system, co-occurs with an uncharacterized gene vc0175 that we rename avcD for anti-viral cytodine deaminase. We show that AvcD is a deoxycytidylate deaminase and that its activity is post-translationally inhibited by a non-coding RNA named AvcI. AvcID and bacterial homologues protect bacterial populations against phage invasion by depleting free deoxycytidine nucleotides during infection, thereby decreasing phage replication. Homologues of avcD exist in all three domains of life, and bacterial AvcID defends against phage infection by combining traits of two eukaryotic innate viral immunity proteins, APOBEC and SAMHD1.

Extended Data Fig. 1 |. VSP-1 and VSP-2 schematic and predicted gene networks (MRS).

Cartoon of VSP-1 (A) and VSP-2 (B) from El Tor V. cholerae N16961 and gene network predictions from GeneCoOccurrence. Arrows indicate the highest partial correlation Wij each gene has to another (ovals). Two arrows are presented pointing in opposing directions where the highest correlation Wij is reciprocal between two genes. MRS = maximum relatedness subnetwork.

Extended Data Fig. 2 |. Complementation of various ig²²² constructs to prevent AvcD induced cell filamentation.

Cell length distributions of Δig²²² V. cholerae expressing pAvcD. All cell length distributions represent ~750–1000 cells measured per strain (n = 3 biological samples), with summary statistics: mean (diamonds), median (horizontal black line), interquartile range (box), and data below and above the interquartile range (vertical lines). Different letters indicate significant differences at p < 0.05, according to Two-way ANOVA with Tukey’s multiple comparison post-hoc test.

Extended Data Fig. 3 |. AvcD C-terminal 6x Histidine fusion maintains the same activity as the WT AvcD enzyme and the presence of avcI does not reduce the abundance of AvcD.

(A) Representative images of WT V. cholerae and Δig²²² cultures maintaining an empty vector plasmid (pVector1) or P_tac-inducible avcD-6xHIS plasmid (pAvcD^6xHis) grown in the presence of 100 μM IPTG for 2 h. Cells were stained with FM4–64 prior to imaging and performed in biological triplicate. (B) Representative coomassie stained PAGE gel (top) and matched anti-6x His antibody Western blot (bottom) of whole cell lysates normalized to total protein from V. cholerae WT and Δig²²² cultures maintaining pVector1 or pAvcD^6xHis. Black triangles correspond to AvcD^6xHis (60.6 kDa). Analysis was performed in biological triplicate and the relative signal intensity (C) was the determined by comparing the intensities of AvcD^6xHis from paired WT and Δig²²² lysates probed on the same blots. Data represent the mean ± SEM of three biological replicate. Statistical significance was determined using two-sided Student’s t-test. P values between WT and Δig²²² is 0.185. ns indicate not significant.

Extended Data Fig. 4 |. AvcD-Avci complex formation in solution and Denaturing urea PAGE analysis of Avci and Avci-RC.

(A) AvcD forms a complex with AvcI in an AvcD concentration-dependent manner as determined by EMSA. Trace quantities of AvcI reverse complement (AvcI-RC) binding to AvcD is observed. (B) AvcI and AvcI-RC run at essentially equivalent molecular weights on a 7 M urea denaturing PAGE. Low range ssRNA ladder (NEB). This was performed at least three times, yielding similar results.

Extended Data Fig. 5 |. multiple sequence alignment of AvcD homologs and Avci homologs explored in this study.

(A) Amino acid alignment of the V. cholerae AvcD and three homologs using EMBL-EBI ClustalW. ‘*’ indicates 100% identity, ‘:’ indicates >75%, and ‘.’ Indicates >50% similarity. Black triangles indicate conserved residues in V. cholerae AvcD targeted for site-directed mutagenesis. (B) Nucleotide alignment of V. cholerae AvcI and three homologs using LocARNA. The average secondary structure is indicated in dot-bracket notation (top). Consensus identities are correlated with the height of the bars below the corresponding nucleotide. Compatible base pairs are colored according to the number of different types C-G (1), G-C (2), A-U (3), U-A (4), G-U (5) or U-G (6) of compatible base pairs in the corresponding columns. The color saturation decreases with the number of incompatible base pairs.

Extended Data Fig. 6 |. Cross-species inhibition of avcD and avcI homologs.

Cell length distributions of E. coli co-expressing various combinations of P_tac-inducible plasmids encoding homologs of avcD and avcI. All cell length distributions represent ~1000–3000 cells measured per strain (n = 3 biological samples), with summary statistics: mean (diamonds), median (horizontal black line), interquartile range (box), and data below and above the interquartile range (vertical lines). Different letters indicate significant differences at p < 0.05, according to Two-way ANOVA with Tukey’s post-hoc test. VC = Vibrio cholerae, VP = Vibrio parahaemolyticus, PM = Proteus mirabilis, ETEC = E. coli ETEC.

Extended Data Fig. 7 |. Phylogenetic analysis and domain architectures of the six AvcD query proteins.

(A) Phylogenetic tree of AvcD homologs from representative phyla across the tree of life. Stars indicate the six proteobacterial starting points for the homology search, as well as the eukaryotic Saccharomyces cerevisiae dcd1 (triangle). (B) Domain architecture and secondary structure predictions for the six proteobacterial starting points (query proteins) were predicted using InterProScan (Methods). Results from six main analyses are shown here for the query proteins: Gene3D (including CATH structure database), Pfam, ProSiteProfiles, PANTHER, and SUPERFAMILY protein domain profile databases, and MobiDBLite for disorder prediction. No transmembrane regions (using TMHMM) or membrane/extracellular localization were predicted for any of the proteins (using Phobius); hence not shown. Numbers (bottom) indicate the amino acid position of predicted domains and features. (C) Cell length distributions of E. coli expressing pAvcD, a P_tac-inducible plasmid encoding dcd1 from S. cerevisiae (pDcd1^Sc), or pVector1. All cell length distributions represent ~1000–3000 cells measured per strain (n = 3 biological samples), with summary statistics: mean (diamonds), median (horizontal black line), interquartile range (box), and data below and above the interquartile range (vertical lines). Different letters indicate significant differences at p < 0.05, according to Two-way ANOVA with Tukey’s post-hoc test.

Extended Data Fig. 8 |. Mutations in conserved residues of AvcD do not affect the stability or function of the protein.

(A) Phyre2 (ref. ) predicted structure of AvcD from V. cholerae El Tor. Insets highlight conserved residues of the PLN (top) and DCD (bottom) domains selected for mutagenesis. (B) Representative Coomassie stained gel (top) and anti-6x His antibody Western blot (bottom) of whole cell lysates from E. coli BL21(DE3) cells maintaining an empty vector (pVector^6xHis), inducible C-terminal 6x histidine tagged avcD (WT) or avcD variants (S52K, D162A + Q163A, E384A, and C411A + C414A) grown in the presence of 1 mM IPTG for 3 h. Sample inputs were normalized by culture OD₆₀₀ and resolved by SDS-PAGE. Three biological replicates of each strain were analyzed with similar results. Black triangles correspond to the predicted molecular weight of the AvcD tagged fusions (60.6 kDa). M = molecular weight marker. (C) V. cholerae mutant expressing the indicated AvcD variants. ori/ter ratios of Chromosome 1 in Δig²²² V. cholerae strains expressing the indicated pAvcD construct and quantified using qRT-PCR. Each bar represents the mean ± SEM, n=3. Different letters indicate significant differences (n=3) at p < 0.05, according to Two-way ANOVA with Tukey’s post-hoc test. (D) Representative images of Δig²²² cultures maintaining an empty vector plasmid pVector 1 or pAvcD grown in the presence of 100 μM IPTG for 8 h. Cells were stained with FM4–64 prior to imaging and performed in biological triplicate. (E) Relative difference in avcD expression between Δig²²² and WT V. cholerae at three different growth phases using qRT-PCR and an endogenous gyrA control. Data represent the mean ± SEM of three biological replicates.

Extended Data Fig. 9 |. Cessation of global translation, by treatment with spectinomycin, does not liberate AvcD enzymatic activity.

Intracellular abundance of dCTP (A), dCMP (B), dUTP (C), and dUMP (D) of WT and ΔavcD V. cholerae during spectinomycin treatment (200 μg/mL) measured by UPLC-MS/MS. Data represent the mean ± SEM of three biological replicate cultures. No statistically significant differences in nucleotide concentrations were observed between strains at any time point as determined by Two-way ANOVA with Two-way ANOVA with Šídák’s multiple-comparison test.

Extended Data Fig. 10 |. Ectopic expression of DncV and AvcD does not lead to filamentation in the ΔcapV mutant of V. cholerae.

Cell length distributions measured from three biological replicates of ΔcapV V. cholerae cultures co-expressing either two empty vectors, pDncV and an empty vector, pAvcD and an empty vector, or pDncV and pAvcD grown in the presence of 100 μM IPTG for 8 h. Distributions represent ~1200–1700 cells measured per strain (n=3 biological samples). Different letters indicate significant differences at p < 0.05, according to Two-way ANOVA with Tukey’s post-hoc test.

Fig. 1 |. AvcD-induced filamentation is inhibited by sRNA Avci.

Growth curves (a) and representative images (b) of WT El Tor V. cholerae and ΔVSP-1/2 strains expressing avcD from a P_tac-inducible plasmid (pAvcD) or an empty vector control (pVector1). Cells were stained with FM4–64 before imaging. Scale bar, 2 μm. This experiment was repeated at least three times. c, Cell length distributions of WT V. cholerae and VSP island mutants expressing pAvcD or pVector1. d, Cell length distributions of VSP-1 gene locus mutants expressing pAvcD in combination with either pIg²²² or a vector control (pVector2). e, Table reporting the capacity of various ig²²² P_tac-inducible constructs to prevent AvcD-induced cell filamentation when expressed in combination with pAvcD in Δig²²² V. cholerae. Dotted line denotes a non-native RBS, ‘*’ indicates a putative start codon mutated to a stop. f, An AvcD–AvcI complex formed in an AvcD concentration-dependent manner as determined by EMSA. Trace quantities of non-specific binding of AvcD to the AvcI reverse complement (AvcI-RC) were observed. This experiment was repeated at least three times, yielding similar results. g, Cell length distributions of E. coli co-expressing P_tac-inducible plasmids encoding avcD homologues and their cognate avcI or vector controls. VC, Vibrio cholerae; VP, Vibrio parahaemolyticus; PM, Proteus mirabilis; ETEC, E. coli ETEC. h, Representative images of E. coli co-expressing various combinations of P_tac-inducible plasmids encoding homologues of avcD and avcI. Scale represents 2 μm. Error bars represent s.e.m. from three biological replicates. All violin plots represent ~1,000–3,000 cells measured per strain (n = 3 biological samples) with summary statistics: mean (diamonds), median (horizontal black line), interquartile range (box) and data below and above the interquartile range (vertical lines). Different letters indicate significant differences using two-way analysis of variance (ANOVA) with Tukey’s post-hoc test (c and d) or to two-sided Dunnett’s post-hoc test (g) against the control strain (pVector1 + pVector2) at P < 0.05.

Fig. 2 |. AvcD is a DCD.

a, Cell length distributions of E. coli expressing pAvcD, P_tac-inducible plasmids encoding a variety of AvcD active-site variants, or pVector1. The violin plot represents ~1,700–3,000 cells measured per strain (n = 3 biological samples), and different letters indicate significant differences at P < 0.05, according to two-way ANOVA with Tukey’s multiple comparison post-hoc test. b, Colorimetric assay detecting the evolution of ammonium from lysates of E. coli, previously expressing pAvcD or pAvcD^E384A, incubated with 12 amine-containing nucleotide substrates (37.7 mM cytidine and 7.5 mM for all other substrates) for 30 min. Data represent the mean ± s.e.m., n = 3, two-way ANOVA Šídák’s multiple-comparison test. NS, not significant. P values for dCTP and dCMP: WT versus E384A, <0.0001. c,d, Quantification of dUTP (c) and dUMP (d) using UPLC–MS/MS, in the indicated cell lysates before and after addition of 1 μM dCTP. Each bar represents mean ± s.e.m., n = 3. e, In vivo nucleotide concentrations of E. coli expressing pAvcD, AvcD active-site variants (pAvcD^S52K pAvcD^E384A), or an AvcD homologue (pAvcD^ETEC) for 1 h measured by UPLC–MS/MS and normalized to a vector control strain. Data are graphed as mean ± s.e.m., n = 3, two-way ANOVA with Tukey’s multiple-comparison test. P values for dCTP: WT versus S52K, 0.0083; WT versus E384A, 0.0193; S52K versus ETEC, 0.0019; E384A versus ETEC, 0.0236. P values for dCMP: WT versus S52K, 0.044; WT versus E384A, 0.0045; S52K versus ETEC, 0.0039; E384A versus ETEC, 0.049. ND, none detected; NS, not significant.

Fig. 3 |. Avci–AvcD is a toxin–antitoxin system.

a, To-scale genomic diagram of avcI and avcD and the primers (a, b, c and d) used for generating diagnostic PCR products. b, PCR products amplified from nucleic acid templates (above) using the indicated primer pairs (below) resolved in a 1% agarose gel. All reactions were performed in duplicate using biologically independent samples with similar results. No RT, non-reverse transcribed RNA template control. gDNA, genomic DNA control. This was repeated three times, yielding similar results. c, Relative difference in transcript abundance between avcI and avcD at different growth phases in WT V. cholerae normalized to an endogenous gyrA control using qRT–PCR. Data are graphed as mean ± s.e.m., n = 3. d–g, In vivo abundance of dCTP (d), dCMP (e), dUTP (f) and dUMP (g) of WT and ΔavcD V. cholerae before and after addition of rifampicin (250 μg ml⁻¹) measured using UPLC–MS/MS and normalized to total protein. Data represent the mean ± s.e.m. of three biological replicate cultures, two-way ANOVA with Šídák’s multiple-comparison test. NS, not significant. For dCTP, P values for WT versus ΔavcD mutant: 5 min, 0.0081; 15–40 min, <0.0001. For dCMP, P values at 27.5 min: 0.0274. For dUMP, P values at 27.5 min and 40 min: 0.0414 and 0.0001, respectively.

Fig. 4 |. avcID homologues provide phage defence.

a, Fold reduction in the number of plaques conferred by four homologous avcID systems to a naïve E. coli host challenged with a panel of coliphages. Fold reduction determined by serial dilution plaque assays comparing the efficiency of plaquing on an E. coli host maintaining a plasmid borne avcID system and its native promotor against a vector control strain. b, Efficiency of plaquing on strains encoding WT avcID^VP from V. parahaemolyticus (pAvcID^VP) or point mutations in the PLN (S49K), DCD (E376A) or a double domain point mutant (pAvcID^{VP-avcDS49K+E376A}) against a vector control strain. Data represent the mean ± s.e.m. of three biological replicate cultures, one-way ANOVA with Dunnett’s post-hoc test. P values for T3: WT versus S49K, < 0.001; WT versus E376A, <0.001; WT versus S49K + E376A, <0.001. P values for T5: WT versus S49K, <0.001; WT versus E376A, 0.0002; WT versus S49K + E376A, <0.0001. P values for T6: WT versus S49K, 0.0012; WT versus E376A, 0.0008; WT versus S49K + E376A, 0.0011. P values for SECФ18: WT versus S49K, <0.0001; WT versus E376A, <0.0001; WT versus S49K_E376A, <0.0001. c, The relative genome abundance of T5 infecting E. coli expressing pAvcID^VP or its double point mutation variants pAvcID^{VP-avcDS49K+E376A}. Data represent the mean ± s.e.m. of three biological replicate cultures, two-way ANOVA with two-sided Šídák’s multiple-comparison test. P values at 40 min: 0.0004. NS, not significant. d–g, In vivo abundance of dCTP (d), dCMP (e), dUTP (f) and dUMP (g) in an E. coli host carrying a vector control, pAvcID^VP or the avcID system with its native promoter from E. coli ETEC (pAvcID^ETEC) before and after addition of T3 phage (MOI 5). Nucleotides measured using UPLC–MS/MS and normalized to total protein. Data represent the mean ± s.e.m. of three biological replicate cultures, two-way ANOVA with Dunnett’s post-hoc test (d–g). NS, not significant. For dCTP, P values: EV versus ETEC at 5 min, <0.0001; EV versus VP at 12.5 min, <0.0001; EV versus VP at 20 min, 0.0009; EV versus ETEC at 20 min, 0.001. For dCMP, P values: EV versus VP at 12.5 min, <0.0001; EV versus ETEC at 12.5 min, 0.0358. For dUMP, P values: EV versus ETEC at 5 min, 0.0017; EV versus VP at 12.5 min, <0.0001; EV versus ETEC at 12.5 min, 0.0001.

Fig. 5 |. AvcD mediates nucleotide pool depletion.

a–d, In vivo abundance of dCTP (a), dCMP (b), dUTP (c) and dUMP (d) of an E. coli host carrying vector control, the avcID system from V. parahaemolyticus with its native promoter (pAvcID^VP), or an inactive avcD mutant (pAvcID^{VP-avcDS49K+E376A}) before and after treatment with rifampicin or infected with phage (T3, MOI 5; SECΦ18, MOI 10). Nucleotides were measured using UPLC–MS/MS, normalized to total protein. Data represent the mean ± s.e.m. of two biological replicate cultures.

Fig. 6 |. Model for AvciD-based anti-phage activity in bacteria.

Top: before infection, AvcD is maintained in an inactive state by the abundant sRNA, AvcI. Bottom: following infection, AvcD is liberated from AvcI, probably by the cessation of global transcription or the enhanced degradation of AvcI. Active AvcD rapidly depletes available dCMP and dCTP substrates promoting the accumulation of dUMP, via deamination, which probably impairs efficient phage DNA replication and new phage virion production.