Protein-primed DNA homopolymer synthesis by an antiviral reverse transcriptase

Abstract

Bacteria defend themselves from viral predation using diverse immune systems, many of which sense and target foreign DNA for degradation¹. Defense-associated reverse transcriptase (DRT) systems provide an intriguing counterpoint to this strategy by leveraging DNA synthesis instead^2,3. We and others recently showed that DRT2 systems use an RNA template to assemble a de novo gene, leading to expression of an antiviral effector protein, Neo^4,5. It remains unknown whether similar mechanisms of defense are employed by other DRT families. Focusing on DRT9, here we uncover an unprecedented mechanism of DNA homopolymer synthesis, in which viral infection triggers polydeoxyadenylate (poly-dA) accumulation in the cell to drive abortive infection and population-level immunity. Cryo-EM structures reveal how a conserved noncoding RNA serves as both a structural scaffold and reverse transcription template to direct hexameric complex assembly and RNA-templated poly-dA synthesis. Remarkably, biochemical and functional experiments identify conserved tyrosine residues within the reverse transcriptase itself that prime DNA synthesis, leading to the formation of high-molecular weight protein-DNA covalent adducts. Synthesis of poly-dA in vivo is regulated by the competing activities of phage-encoded triggers and host-encoded silencers of DRT9. Collectively, our work unveils a novel nucleic acid-driven defense system that expands the paradigm of bacterial immunity and broadens the known functions of reverse transcriptases.

Figure 1 |. Systematic discovery of DRT9 reverse transcription substrates and products in vivo.

a, Phylogenetic tree of bacterial reverse transcriptase (RT) homologs within the so-called Unknown Group (UG), which are broadly implicated in antiphage defense. DRT9 (UG28) and DRT2 (UG2) are highlighted, as are other systems that have been the subject of experimental studies; genetic architectures of archetypal DRT9 and DRT2 systems are shown below. The tree was adapted from Mestre et al. by collapsing each UG/DRT clade into a single representative. b, Schematic of RNA immunoprecipitation (RIP) and cDNA immunoprecipitation (cDIP) sequencing approaches to identify nucleic acid templates and products of FLAG-tagged reverse transcriptase (RT) from SenDRT9. The plasmid-encoded immune system is schematized at the top. c, MA plots showing the RT-mediated enrichment of RNA (top) and DNA (bottom) loci from RIP-seq and cDIP-seq experiments, respectively, relative to input controls. Each dot indicates a transcript, and red dots indicate transcripts with log₂(enrichment) > 5 and false discovery rate (FDR) < 0.05. d, RIP-seq and cDIP-seq coverage tracks (top to bottom) for either WT RT or a catalytically inactive RT mutant (MUT), in the presence of T5 phage infection, for SenDRT9 (left) and KpnDRT2 (right). Genomic locus schematics are shown below each graph, and data are normalized for sequencing depth and plotted as counts per million reads (CPM). Coordinates for DRT9 are numbered from the beginning of the S. enterica-derived sequence on the expression plasmid. e, Bar graph analyzing the percentage of unmapped cDIP-seq reads from WT or MUT SenDRT9 cDIP-seq datasets in the absence or presence of T5 phage infection. Reads were mapped to a concatenated reference genome comprising the E. coli chromosome, phage genome, and expression plasmid. f, Schematic of unmapped read analytical pipeline (top), and MEME results that revealed poly-dT and poly-dA motifs enriched in unmapped reads from the WT + T5 cDIP-seq dataset in e (bottom). E, E-value significance of the motif; n, number of sites contributing to the motif. g, Bar graph of normalized dA₂₅, dT₂₅, dG₂₅, and dC₂₅ counts from WT and MUT cDIP-seq datasets in the absence or presence of T5 phage infection. h, Southern blot analysis of total DNA isolated from cells expressing WT or MUT SenDRT9 in the absence or presence of T5 phage infection. Duplicate blots were probed with either oligo-dT₄₀ (left) or oligo-dA₄₀ (right) to detect poly-dA and poly-dT species, respectively. Sizes from a double-stranded DNA (dsDNA) ladder are marked, from the methylene blue-stained membrane after transfer (Extended Data Fig. 2d). i, Bar graph of chimeric dA₁₀dT₁₀ and dT₁₀dA₁₀ counts from WT and MUT cDIP-seq datasets in the presence of T5 phage infection. Data in e,g,i are shown as mean ± s.d. for n = 3 independent biological replicates.

Figure 2 |. ncRNA sequence determinants of SenDRT9-mediated phage defense and poly-dA synthesis.

a, Predicted secondary structure of the SenDRT9 ncRNA. Stem-loop (SL) regions and uridine nucleotides implicated in RNA-templated DNA synthesis are labeled; coordinates are numbered based on the mature ncRNA species identified by RIP-seq. b, Bar graph quantifying SenDRT9 defense activity against T5 phage for scrambled ncRNA SL mutants (MUT), quantified as the fold reduction in efficiency of plating (EOP) relative to an empty vector (EV) control. Data are from n = 2 technical replicates. c, Heat map quantifying SenDRT9 defense activity for the indicated ncRNA point mutations, quantified as the fold reduction in EOP relative to an empty vector (EV) control. Uridine nucleotides implicated in RNA-templated DNA synthesis are highlighted in bold; data are shown as the mean of n = 2 technical replicates. d, Heat map quantifying SenDRT9 defense activity for the indicated ncRNA mutations, shown as in c. e, Plaque assay showing loss of SenDRT9 defense activity against T5 phage for substitutions within the putative ncRNA template region. Residues U123–U126 (U₄) were mutated to each of the indicated nucleotides; EV, empty vector. f, Southern blot analysis of total DNA isolated from cells expressing the WT or U₄>A₄ MUT ncRNA from e, in the absence or presence of T5 phage infection. Duplicate blots were probed with either oligo-dT₄₀ (left) or oligo-dA₄₀ (right) to detect poly-dA and poly-dT species, respectively. Sizes from a double-stranded DNA (dsDNA) ladder are marked, from the methylene blue-stained membrane after transfer (Extended Data Fig. 3g).

Figure 3 |. The SenDRT9 RT-ncRNA complex performs protein-primed, RNA-templated DNA homopolymer synthesis.

a, Denaturing 5% urea-PAGE analysis of DNA polymerization assays that investigated the effect of time (left), dATP concentration (middle), or nucleotide competition (right) on poly-dA synthesis using radiolabeled [α−³²P]-dATP. All reactions contained 20 nM RT-ncRNA complex and were incubated at 37 °C. Left, reactions contained 100 μM dATP; Middle, reactions were incubated for 10 min; Right, reactions were incubated for 10 min with 100 μM radiolabeled dATP and the indicated, additional unlabeled nucleotides at 100 μM. M1 and M2 denote DNA ladder markers. b, Denaturing 5% urea-PAGE analysis of DNA polymerization assays, after reactions were treated with the indicated proteinase or nuclease reagents prior to electrophoretic separation, with or without boiling samples first. All reactions contained 150 nM RT-ncRNA and 100 μM [α−³²P]-dATP, and were incubated at 37 °C for 10 min prior to proteinase/nuclease addition, with the exception of lane 15, which was pre-treated with RNase A/T1 prior to dATP addition. The prominent mobility shift of the poly-dA product upon proteinase K treatment, irrespective of boiling, suggests the existence of a covalent protein-DNA conjugate. Note that single-strand-specific endonuclease P1 readily degrades poly-dA to small fragments, whereas TURBO DNase, which prefers dsDNA, has little activity on the poly-A product under these conditions. c, SDS-PAGE analysis of DNA polymerization assays, in which RT-ncRNA complexes (0.15 μM) were incubated with the indicated dNTPs or rNTPs (0.9 mM) for 60 min at 37 °C, before reactions were quenched and resolved electrophoretically. Red arrows denote smears corresponding to high-molecular weight (MW) protein-DNA conjugates in reactions that contained dATP. Incomplete tag removal during purification accounts for the residual His₆-GST-RT band; * refers to a purification contaminant; M, protein marker. d, AlphaFold 3 structure prediction of a SenRT monomer (left), with magnified view (middle) showing the close proximity of C-terminal residues Y496 and Y498 (yellow) to the YADD active site (red). The multiple sequence alignment (right) highlights the strong conservation of both tyrosine residues. e, Plaque assay showing loss of SenDRT9 defense activity against T5 phage for single Y496F and Y498F mutants, as well as a double Y496F,Y498F mutant; EV, empty vector. f, SDS-PAGE analysis of DNA polymerization assays as in c, but with a Y496F,Y498F RT mutant. g, SDS-PAGE analysis of DNA polymerization assays as in c, with either WT or a U₄>A₄ ncRNA mutant. Mutation of the U₄ template region to A₄ completely switches substrate specificity from dATP to dTTP.

Figure 4 |. CryoEM structure of the hexameric SenDRT9 RT-ncRNA complex.

a, Domain architecture (left) of SenDRT9-encoded RT (PDB: 9NLX, this study) compared to AbiK (PDB: 7R06). The SenRT has an N-terminal extension (red) that forms a triangle-shaped protomer through contacts with the downstream thumb that are bridged by the palm domain (right). b, Surface representation of a single RT-ncRNA monomer, colored by relative electrostatic potential, reveals a positively charged tunnel that threads the U₄ template through the ‘triangle’ of each RT protomer. c, Cartoon representation of the ncRNA secondary structure observed in the cryo-EM density, revealing a complex RNA fold composed of several notable features. RNA structures implicated in inter-subunit interactions via kissing loop interactions are highlighted with asterisks. A cartoon representation of the RT domain architecture, colored as in a, highlights the path of the RNA across each protein domain. d, Structure of a single RT-ncRNA monomer, with RT colored as in a and ncRNA stem loops labelled as in c, highlighting how each RT protomer is enveloped by its associated ncRNA. e, 2.6 Å map and resulting model of the hexameric SenDRT9 RT-ncRNA complex. f, Kissing loop interactions between flipped out bases in SL5 (C93) and SL6 (G113) of adjacent ncRNA chains, which form an oligomerization interface above each trimer. g, In addition to stabilizing interactions within the trimer, the ncRNA also supports the back-to-back arrangement of trimers via kissing loop interactions between flipped out bases G68 and U69. h, A close-up view of the polymerase active site demonstrates how priming tyrosines Y496 and Y498 lie between the YADD(243–246) catalytic motif and templating nucleotides U123-U126, in an orientation poised for homopolymer synthesis.

Figure 5 |. Identification of a viral trigger that activates SenDRT9-mediated abortive infection.

a, Schematic of workflow to isolate phage variants that escape detection and/or elimination by a SenDRT9 immune response. Individual plaques escaping defense are isolated and sequenced to identify mutations affecting candidate trigger genes. b, Representative coverage tracks from whole-genome sequencing of T5 ‘escaper’ phage variants that bypass SenDRT9 immunity, shown above the corresponding annotated genomic locus. All mutation classes shown (point mutations and larger deletions) perturb the putative P11 promoter driving expression of four protein-coding genes. c, Colony formation assays to determine cell viability upon co-expression of WT or RT-inactive (MUT) SenDRT9 with candidate phage genes regulated by the P11 promoter, identified in b. The gene product of T5.058 (gp58) prevents cell growth exclusively in WT SenDRT9 cells, whereas gp55 causes generic toxicity regardless of SenDRT9 activity; the remaining candidates exhibit no phenotype. d, Southern blot analysis of total DNA isolated from cells expressing WT or MUT SenDRT9, with or without gp58 induction. DNA was probed with oligo-dT₄₀ to detect poly-dA species. Sizes from a double-stranded DNA (dsDNA) ladder are marked, from the methylene blue-stained membrane after transfer (Extended Data Fig. 9d). e, Growth curves of cells expressing WT or MUT SenDRT9, in the presence or absence of T5 phage at the indicated multiplicity of infection (MOI). Shaded regions indicate the SD across n = 3 independent biological replicates. f, Predicted AlphaFold 3 structure of T5 gp58 (teal) superimposed onto the structure of the E. coli replication restart helicase PriA (grey) bound to a replication fork DNA substrate (orange; PDB ID: 8FAK). The magnified inset highlights gp58 residues that overlap PriA residues implicated in binding to DNA 3′ ends. RMSD = 0.95 Å over 47 Cα atoms between gp58 and PriA. g, Bar graphs quantifying cell viability in colony forming units (CFU) upon co-expression of the indicated gp58 alanine substitution variants with SenDRT9 WT (left) or MUT (right). Asterisks (*) indicate amino acid residues predicted to interact with the DNA substrate 3′ end, as highlighted in f. Data are shown as mean ± s.d. for n = 3 technical replicates.

Figure 6 |. Host regulation of and response to poly-dA synthesis by DRT9 immune systems.

a, Bar graph quantifying cell viability in colony forming units (CFU) upon transformation of WT or ΔsbcB E. coli with either WT or RT-inactive (MUT) SenDRT9. The inability to propagate WT SenDRT9 in a ΔsbcB background suggests that the sbcB gene product, Exodeoxyribonuclease I (ExoI), neutralizes the otherwise toxic properties of DRT9 synthesis products. Data are shown as mean ± s.d. for n = 3 technical replicates. b, Denaturing 5% urea-PAGE analysis of DNA polymerization assays, after reactions were treated with ExoI (left) or DNase P1 (right) in the presence of increasing concentrations of T5 gp58. All reactions contained 20 nM RT-ncRNA and 100 μM [α−₃₂P]-dATP, and were incubated at 37 °C for 10 min prior to incubation with 0.015 Units/μL ExoI or 2 Units/μL nuclease P1 alongside gp58 ranging from 0.1–10 μM. Nuclease P1, which is an endonuclease, is not inhibited by gp58. c, Schematic of co-immunoprecipitation followed by mass spectrometry (co-IP MS) to identify protein interactors of SenDRT9; proteins that interact with RT-mediated DNA synthesis products will be enriched alongside direct interactors with the RT-ncRNA complex. d, UpSet plot showing the number of overlapping and unique protein interactors identified from co-IP MS experiments with either WT or MUT SenDRT9 in the absence or presence of T5 phage infection. The size of each significantly enriched protein set is displayed on the bar graph to the right, while the size of set overlaps is displayed on the bar graph above. Significantly enriched proteins were identified as those exhibiting >20-fold enrichment relative to the control IP, with a false discovery rate (FDR) < 0.05. Phage and host proteins enriched uniquely in the WT + T5 condition are shown at right, which notably includes the T5 gp58 trigger protein and host nucleoid-associated proteins (NAPs) such as H-NS and StpA, known to preferentially bind AT-rich DNA. e, Bar graph showing normalized protein intensity values for T5 SSB and E. coli H-NS and StpA across the indicated co-IP MS experiments from d, including an untagged (no FLAG) control. Data are shown as mean ± s.d. for n = 3 independent biological replicates. f, Model for the antiphage defense mechanism of DRT9 systems. The RT and ncRNA form a hexamer, leading to constitutive, protein-primed synthesis of the poly-dA cDNA by the U-rich template region of the ncRNA, and partial second-strand synthesis of poly-dT. Host ExoI normally degrades poly-dA products in the cell, but phage infection leads to trigger expression (gp58 in T5) that competes for 3′ end binding, thereby stabilizing poly-dA products and causing their accumulation. SenDRT9 poly-dA products are bound by additional viral and host factors in the cell (H-NS and StpA), leading to cell growth arrest through an unknown mechanism.