Phage-triggered reverse transcription assembles a toxic repetitive gene from a noncoding RNA

Abstract

Reverse transcription has frequently been co-opted for cellular functions and in prokaryotes is associated with protection against viral infection, but the underlying mechanisms of defense are generally unknown. Here, we show that in the DRT2 defense system, the reverse transcriptase binds a neighboring pseudoknotted noncoding RNA. Upon bacteriophage infection, a template region of this RNA is reverse transcribed into an array of tandem repeats that reconstitute a promoter and open reading frame, allowing expression of a toxic repetitive protein and an abortive infection response. Biochemical reconstitution of this activity and cryo-electron microscopy provide a molecular basis for repeat synthesis. Gene synthesis from a noncoding RNA is a previously unknown mode of genetic regulation in prokaryotes.

Fig. 1.. DRT2 reverse transcriptase produces concatemeric cDNA upon phage infection.

(A) Diagram of the DRT2 defense system and schematic for assay for cDNA production by DRT2 reverse transcriptases. (B) Tagmentation reads were annotated by sliding an 11 bp window along their lengths and coloring according to the mapping position of the center of the window to the DRT2 ncRNA. Disjunct read coloring represents chimeric reads, an example of which is shown below the plot for the DRT2 system. Coverage is calculated using soft-clipped reads. (C) Small RNA sequencing reads mapped to the DRT2 locus after expression in E. coli. The diagram shows the proposed formation of concatemeric cDNA (ccDNA) by reverse transcription. (D) PCR and qPCR assay for ccDNA formation. Top: schematic for specific detection of ccDNA using outward-facing primers. Bottom left: PCR of DRT2-expressing E. coli with and without infection with T5 phage. YCDD>YCAA, an active site mutant (D269A and D270A) of the DRT2 reverse transcriptase. Plasmid primers are specific for a 120 bp region of the DRT2 expression plasmid backbone. Bottom middle: quantitative PCR (qPCR) of the same conditions, with two biological and two technical replicates. Bottom right: qPCR of ccDNA production during a time course of T5 infection. Shading represents the standard error of three independent replicates. (E) Schematic of strand-specific DNA sequencing library preparation. V indicates the DNA bases A, C, or G. (F) Strand-specific reads mapped to the DRT2 ncRNA, colored as in B, with the position above or below the x-axis corresponding to plus- or minus-strand mapping respectively. Read height is weighted by the number of reads in the same sample mapping to a single-strand DNA control. (G) Normalized strand-specific read counts for the reads shown in F. “Template” counts any read mapping within the 120 bp template region of the DRT2 ncRNA. “Repeat junction” counts any read containing juxtaposed template regions.

Fig. 2.. ccDNA reverse-transcribed by DRT2 can be transcribed into RNA.

(A) Conservation at each position for 44 homologs of the K. pneumoniae DRT2 ncRNA. Conservation was calculated as the difference in entropy from maximum entropy at each position. The tan shaded background represents the template region. The moving average was calculated with a window size of 10 bp. (B) Sequence logo derived from an alignment of 42 homologs of the DRT2 ncRNA template region (shown in reverse complement to the ncRNA sequence). (C) The same sequence logo rearranged to correspond to a ccDNA repeat junction. Putative conserved promoter elements are shaded. (D) ccDNA assay by PCR, and RT-PCR assay to test for transcription of the ccDNA. WT, wild-type DRT2 system; YCAA, reverse-transcriptase active site mutant of DRT2 system; –10, mutation of the putative –10 sequence from TATAAT to CGCGGC; –10*, the same mutation combined with mutation of ncRNA bases ₂₄AUU₂₆ to ₂₄GGC₂₆ to restore pairing with the mutant –10 sequence; –35, mutation of the putative –35 sequence from TTGACA to AAAGGC; EV, empty vector control. (E) T5 defense (fold reduction in PFUs) and T5-induced ccDNA production (qPCR normalized to WT) for each of the promoter mutants. (F) GFP reporter assay. GFP production was assessed as the background-corrected GFP fluorescence when cells reached OD₆₀₀=0.5. Reporter 1 contains the reverse-complement of the 120-bp ncRNA template region upstream of the GFP start codon. Reporter 2 has the same repeat permuted as found in ccDNA leading up to the start codon. Reporters 3 and 4 are identical to 2 but with the –10 or –35 mutants described above. Fluorescence was determined for three independent replicates.

Fig. 3.. DRT2 ccDNA contains a toxic open reading frame.

(A) Example open reading frames for five different DRT2 orthologs, all of which span theoretical repeat junctions. (B) Distribution of putative template lengths for 42 DRT2 ncRNA orthologs. Major x-axis ticks are for multiples of three. (C) The indicated ccDNA codons were either mutated to stop codons (asterisks) or synonymous codons as indicated. The graphs show T5 defense (fold reduction in PFUs) and T5-induced ccDNA production (qPCR) for each of the codon mutants, compared to a wild-type control. (D) Protein structure prediction for 5 repeats of the K. pneumoniae DRT2 Neo ORF. For AlphaFold2 a multiple-sequence alignment (MSA) of 42 Neo orthologs was supplied. No MSA was used for AlphaFold3. Two representative models are shown for each prediction. (E) Inducible expression of 1–3 repeat units of the Neo protein leads to toxicity. 10-fold dilutions of exponentially growing E. coli were plated on repressive media (0.2% glucose) or inducing media (0.05% arabinose). P_BAD, arabinose-inducible promoter. (F) The same constructs as in E, grown in LB for two hours before induction with 0.05% arabinose. Solid lines represent uninduced controls. Shading represents the standard error of three replicates.

Fig. 4.. Structure of the DRT2 ribonucleoprotein complex.

(A) Cryo-EM structure of the DRT2 reverse transcriptase bound to the DRT2 ncRNA. Dashed lines represent nucleotides not resolvable in the cryo-EM density. (B) Structure of a group II intron-encoded RT in the same orientation as A (23). (C) The DRT2 RNP contains a 5 nt DNA primer in the active site. The inset shows how DRT2 G277 occupies the same position as the incoming thymidine triphosphate (dTTP) in a stalled elongating RT enzyme (57). (D) Nucleic acids extracted from DRT2 RNPs were sequenced using Ordered Two-Template Relay (24). Reads were aligned to the 3′ end of the DRT2 RNP and sequence logos calculated. PBS mutants derive from a library of PBS mutant sequences, see fig. S6 for all PBS mutations. YCDD>YCAA, RT active site mutation; DBR1, yeast debranching enzyme; PBS, primer-binding site (E) Histogram of distances between 3′ oxygens and 5′ phosphates for RNA bases separated by two intervening nucleotides, calculated from the DRT2 ncRNA structure itself. Dotted line indicates the distance between the last observed base of the DRT2 ncRNA and the first observed base of the primer (also shown in C).

Fig. 5.. Structure of DRT2 in the elongating state.

(A) The purified DRT2 RNP was incubated with deoxynucleotide triphosphates (dNTPs) at the indicated concentrations, and products were visualized by denaturing gel electrophoresis (left) or agarose gel electrophoresis (right). In the agarose gel each lane has 2-fold more dilute dNTPs compared to the neighbor on the right, except lane 1 has no dNTPs. The indicated concentrations are of each dNTP. (B) Each ccDNA repeat contains a natural HinfI restriction endonuclease site (staggered lines). Purified in vitro ccDNA products were incubated with HinfI and or RNase A. (C) Structure of the DRT2 RNP in the elongating state. Residues are colored by the root-mean-square deviation (RMSD) of their positions compared to the resting state. The template region is not well-resolved in the elongating state and is shown for reference as a transparent background. (D) Details of the DRT2 reverse transcriptase active site in the resting state or (E) in the elongating state (with dNTPs). (F) Secondary structure diagram of the DRT2 ncRNA in the resting and elongating states. Non-canonical base pairs are notated using the nomenclature of Leontis and Westhof (64).

Fig. 6.. Molecular mechanism of ccDNA synthesis.

(A) Purified in vitro ccDNA products were sequenced by Nanopore long-read sequencing, and reads were analyzed as sequential 11-mers (the minimum k-mer length for unique mapping to the DRT2 ncRNA). Jump maps were generated by plotting pairs of neighboring mapped 11-mers. The position on the y-axis corresponds to the mapped position of the 3′ base of the 5′ 11-mer, the position on the x-axis corresponds to the mapped position of the 5′ base of the 3′ 11-mer. Off-diagonal points represent jumps, or chimeric reads. (B) Distribution of mapping positions of the first 11-mer on the Nanopore read after the sequencing adapter. The most frequent mapping position corresponds to the “GATAT” primer. (C) Jump maps for in vitro ccDNA. Each point is sized and colored according to the observed frequency of that jump in all reads. The shaded tan background corresponds to the template region of the ncRNA. Windows mapping to the same strand as the ncRNA are called plus-strand windows; windows mapping to the antisense of the ncRNA (as expected from reverse transcription) are called minus-strand windows. Four categories of jump are labelled. “Category 1” jumps are the minus-to-minus strand jumps between the start and end of the template region, as observed in our other sequencing data, and account for 65% of all off-diagonal jumps. “Category 2” jumps are minus-to-plus strand reversals and likely represent the DRT2 RT switching direction. These collectively account for 11% of off-diagonal jumps. “Category 3” jumps are plus-to-plus across the repeat junctions, likely representing “second-strand synthesis” and account for 17% of off-diagonal jumps. “Category 4” jumps are plus-to-minus and likely represent read-through from the primer back into the ncRNA. These account for 7% of off-diagonal jumps. (D) Model for the mechanism of the DRT2 defense system. The jump categories from C are labelled.