2024 05 08 593200v1 Full

bioRxiv preprint doi: https://doi.org/10.1101/2024.05.08.593200; this version posted May 8, 2024.
The copyright holder for this preprint (which

was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
available under aCC-BY-NC-ND 4.0 International license.
De novo gene synthesis by an antiviral reverse transcriptase

Stephen Tang1, Valentin Conte1†, Dennis J. Zhang2†, Rimantė Žedaveinytė1, George D. Lampe1,
Tanner Wiegand1, Lauren C. Tang2, Megan Wang1, Matt W.G. Walker2, Jerrin Thomas George1‡,
Luke E. Berchowitz3,4, Marko Jovanovic2, Samuel H. Sternberg1*
1
Department of Biochemistry and Molecular Biophysics, Columbia University, New York, NY, USA.
2
Department of Biological Sciences, Columbia University, New York, NY, USA.
3
Department of Genetics and Development, Columbia University, New York, NY, USA.
4
Taub Institute for Research on Alzheimer’s and the Aging Brain, New York, NY, USA.
†
These authors contributed equally to this work.
‡
Present address: Department of Microbiology and Cell Biology, Montana State University, Bozeman, MT, USA.
*Corresponding author. Email: shsternberg@gmail.com
Abstract: Bacteria defend themselves from viral infection using diverse immune systems, many of which
sense and target foreign nucleic acids. Defense-associated reverse transcriptase (DRT) systems provide
an intriguing counterpoint to this immune strategy by instead leveraging DNA synthesis, but the iden-
tities and functions of their DNA products remain largely unknown. Here we show that DRT2 systems
execute an unprecedented immunity mechanism that involves de novo gene synthesis via rolling-circle
reverse transcription of a non-coding RNA (ncRNA). Unbiased profiling of RT-associated RNA and DNA
ligands in DRT2-expressing cells revealed that reverse transcription generates concatenated cDNA repeats
through programmed template jumping on the ncRNA. The presence of phage then triggers second-strand
cDNA synthesis, leading to the production of long double-stranded DNA. Remarkably, this DNA product
is efficiently transcribed, generating messenger RNAs that encode a stop codon-less, never-ending ORF
(neo) whose translation causes potent growth arrest. Phylogenetic analyses and screening of diverse DRT2
homologs further revealed broad conservation of rolling-circle reverse transcription and Neo protein func-
tion. Our work highlights an elegant expansion of genome coding potential through RNA-templated gene
creation, and challenges conventional paradigms of genetic information encoded along the one-dimen-
sional axis of genomic DNA.
One-Sentence Summary: Bacterial reverse transcriptases synthesize extrachromosomal genes via roll-
ing-circle amplification to confer potent antiviral immunity.
bioRxiv preprint doi: https://doi.org/10.1101/2024.05.08.593200; this version posted May 8, 2024. The copyright holder for this preprint (which
Tang et al. (2024), Main text 2
INTRODUCTION
Mobile genetic elements (MGEs), including viruses, plasmids, and transposons, act as a major
driving force to shape genome evolution by expressing protein enzymes that catalyze diverse DNA rear-
rangements (1). Many of these enzymes are encoded by the most abundant gene family in nature (2), and
over sufficiently long timescales, MGEs can become dominant constituents of their host genomes due to
their proliferative properties (3). In response, cells harbor arsenals of defense mechanisms to counter the
pervasive spread of MGEs, which range in complexity from single-gene modules to the multi-organ ver-
tebrate immune system (4). Strikingly, MGEs themselves often provide the source material from which
anti-MGE mechanisms are derived (5). In vertebrates, for example, domestication of an ancestral trans-
poson led to the evolution of V(D)J recombination, laying the groundwork for protein-based adaptive
antiviral immunity by facilitating antigen receptor diversification (6). And in bacteria, recurrent exaptation
of transposon-encoded genes enabled the emergence of CRISPR–Cas systems, which exploit guide RNA
molecules to recognize and cleave foreign targets during nucleic acid-based adaptive antiviral immunity
(7–9). MGEs have thus contributed to the perennial host–parasite arms race as both aggressor and defend-
er, across all domains of life.
Inspired by the innovative molecular mechanisms of immunity born out of MGE co-option, we
set out to investigate additional host pathways used by bacteria to defend against genetic invaders, and
in particular, bacteriophages (10, 11). We were especially intrigued by recent discoveries of diverse an-
tiphage defense systems that encode reverse transcriptase (RT) enzymes (12, 13). In stark contrast to
well-studied phage defense pathways that target and destroy foreign nucleic acids using nucleases, such
as restriction-modification (RM) and CRISPR–Cas, these RT-based systems presumably confer immunity
via nucleic acid synthesis.
Prokaryotic RTs are thought to all descend from a common retroelement ancestor, the Group II in-
tron, which catalyzes self-splicing and site-specific DNA insertion, and is also thought to be the precursor
to the eukaryotic spliceosome (14). The antiviral roles of multiple classes of domesticated retroelements
have emerged in recent years (14), reinforcing the notion that genetic conflicts position MGEs in both of-
fensive and defensive roles. One class, for example, comprises fusions or operonic associations between
an RT and the Cas1 integrase, itself a domesticated transposase, to record molecular memories from past
infections by writing fragments of viral RNA into the CRISPR DNA (15–17). Retrons constitute another
class that mediate innate antiphage immunity using complex operons that typically consist of an RT do-
main, a non-coding RNA (ncRNA), and a toxin protein (12, 13, 18). In uninfected cells, the RT reverse
transcribes the ncRNA to form a complementary DNA (cDNA) product, which is thought to maintain
the toxin in an inactive state; phage infection then drives cDNA modification, leading to activation of the
toxin and initiation of cell suicide (12, 18). This general strategy, known as abortive infection, prevents
the invading phage from completing its lytic cycle and provides population-level immunity at the expense
of the infected cell.
Defense-associated RT (DRT) systems comprise a third class of retroelements with antiphage
function, which originate from a monophyletic clade of RTs termed the Unknown Group (UG) (13, 19).
In contrast to RT-Cas1 and retron systems, many DRTs feature single-gene operons (19), implying that
reverse transcription activity alone could be responsible for providing phage defense. Initial experimental
efforts have confirmed the phage defense functions of 9 UG subgroups (DRT1-9), in addition to demon-
strating the functional requirement for an intact RT domain (13, 19). However, the identities of the DRT
cDNA products, and their mechanisms of immunity, have not been studied.
Here we develop a systematic approach to profile the cDNA products of any RT enzyme of interest,
and apply it to DRT2-family phage defense systems. This strategy revealed an unprecedented rolling-cir-
cle reverse transcription activity, leading to the production of repeated, concatenated cDNA products.
Remarkably, we discover that these cDNAs contain an open reading frame (ORF) that remains in-frame
through each repeat, and that promoter sequences formed across the repeat junction result in abundant
mRNA transcription. Expression of the never-ending ORF (neo) gene, which is stringently regulated by
the presence or absence of phage, leads to rapid growth arrest and programmed dormancy. Beyond reveal-
ing an elegant example of retroelement exaptation for host–MGE defense, we present evidence supporting
the broad conservation of this mechanism for RNA-templated creation of extrachromosomal genes.
RESULTS
cDNA synthesis by a defense-associated reverse transcriptase
We focused our attention on DRT2 systems because of their intriguing minimal architecture, con-
sisting of a single ORF and upstream ncRNA described previously (19). Unlike most other DRT and
retron systems, which typically encode a reverse transcriptase enzyme alongside one or more additional
protein domains predicted to function as effectors of the immune response, DRT2 systems lack additional
protein-coding genes, and the RT protein lacks domains beyond the predicted RNA-directed DNA poly-
merase (11, 19) (fig. S1A). We therefore hypothesized that the cDNA product of the RT enzyme likely
plays a critical and central role in the DRT2 immune mechanism. To identify this cDNA, we developed
a sequencing approach to systematically identify RT-associated cDNA synthesis products based on im-
munoprecipitation of FLAG-RT fusions, which we termed cDNA immunoprecipitation and sequencing
(cDIP-seq) (Materials and Methods), and performed these experiments alongside traditional RNA im-
munoprecipitation (RIP)-seq to also capture RT-associated RNA substrates (Fig. 1A and fig. S1B). We
validated this approach on the well-studied Retron-Eco1 (formerly Ec86) (20) after first verifying that
the FLAG-tagged retron RT retained phage defense activity comparable to the WT RT (fig. S1C). RIP-
seq and cDIP-seq of Retron-Eco1 recapitulated all known features of both the msRNA and msDNA,
including RNase H processing of msRNA, and precise 5′ and 3′ ends of the msDNA (21) (fig. S1D,E).
These results increased our confidence that a similar approach could provide new insights into the DRT2
molecular mechanism, and so we turned our attention to a candidate system from Klebsiella pneumoniae
(KpnDRT2).
After confirming that fusing the KpnDRT2 RT with a FLAG epitope tag did not affect defense activ-
ity against T5 phage (fig. S1C), we performed RIP-seq and cDIP-seq from cells constitutively expressing
plasmid-encoded ncRNA and RT from their native promoter, and then performed genome-wide analyses
to identify RNA and cDNA molecules enriched by IP. The resulting datasets revealed that the highest
enriched RNA and cDNA transcripts mapped to the KpnDRT2 ncRNA locus (Fig. 1B), suggesting that
the primary substrate for reverse transcription by DRT2 is encoded in cis, similar to retron systems. We
ascribed the apparent RIP-seq enrichment of RT mRNA to the likely presence of read-through transcripts
extending from the ncRNA into the coding sequence. Other enriched hits from cDIP-seq data were also
found in control experiments using a catalytically inactive RT mutant (13) (hereafter YCAA), suggesting
a spurious origin (fig. S2A). Interestingly, RIP-seq and cDIP-seq experiments in the presence of T5 phage
also revealed a strong and specific enrichment of transcripts derived from the DRT2 ncRNA locus (fig.
S2B), indicating that RT substrate choice is largely unchanged during phage infection.
Mapping of RIP-seq and cDIP-seq data onto the KpnDRT2 locus revealed the presence of a large
ncRNA and a seemingly well-defined cDNA with the opposite strandedness relative to the ncRNA, as
expected for reverse transcription (Fig. 1C). Control experiments with the inactive YCAA RT mutant
demonstrated that ncRNA enrichment occurred independently of reverse transcriptase activity, whereas
cDNA enrichment from this locus required an intact RT active site (Fig. 1C and fig. S2A). We next lever-
aged custom RNA-seq library preparation protocols based on dRNA-seq (22) and Term-seq (23) to demar-
cate the precise 5′ and 3′ ends of the 281-nucleotide (nt) ncRNA (Fig. 1C), while analyzing start and end
Fig. 1. Systematic discovery of DRT2 reverse transcription substrates and products in vivo. (A) Schematic of RNA im-
munoprecipitation (RIP) and cDNA immunoprecipitation (cDIP) sequencing approaches to identify nucleic acid substrates of
FLAG-tagged reverse transcriptase (RT) from KpnDRT2. The plasmid-encoded immune system is schematized top left. (B)
MA plots showing the RT-mediated enrichment of RNA (top) and DNA (bottom) loci from RIP-seq and cDIP-seq experiments,
relative to input controls. Each dot represents a transcript, and red dots denote transcripts with > 20-fold enrichment and false
discovery rate (FDR) < 0.05. (C) dRNA-seq, Term-seq, RIP-seq, and cDIP-seq coverage tracks, from top to bottom, for either
WT RT or a catalytically inactive RT mutant (YCAA). dRNA-seq and Term-seq enrich RNA 5′ and 3′ ends, respectively, where-
as RIP-seq and cDIP-seq identify RT-associated RNA and DNA ligands. Red and pink denote top and bottom strands, respec-
tively, and the DRT2 locus is shown at bottom; coordinates are numbered from the beginning of the K. pneumoniae-derived
sequence on the expression plasmid. Data are normalized for sequencing depth and plotted as counts per million reads (CPM).
(D) Predicted secondary structure of the KpnDRT2 ncRNA. The cDNA template region is colored in pink, and the gray dotted
line denotes the direction of reverse transcription. (E) Coverage over the DRT2 ncRNA locus from total DNA sequencing of
cells +/- T5 phage infection (left), and bar graph of cDNA counts for the same samples alongside the YCAA mutant (right). Red
and pink denote top and bottom strands, respectively; data are mean ± s.d. (n = 3).
coordinates from cDIP-seq alignments to define the 5′ and 3′ ends of the 119-nt cDNA (fig. S2C). Interest-
ingly, dRNA-seq data revealed a single transcription start site (TSS) upstream of the ncRNA, but not the
RT gene (Fig. 1C), suggesting that the ncRNA and RT share an upstream promoter, and are separated into
mature transcripts via an unknown processing step. We next used a multiple sequence alignment (MSA)
of DRT2 homologs to generate a covariance model of the ncRNA (fig. S2D), which was in excellent
agreement with the KpnDRT2 ncRNA secondary structure predicted by in silico RNA folding (24) (Fig.
1D). The ncRNA features numerous conserved stem-loop (SL) elements, a template region corresponding
to the cDNA product abutted by a short basal stem, and a large 3′ region that we hypothesize serves as a
scaffold for sequence and/or structure-guided recruitment of the RT.
We next investigated how cDNA synthesis is altered as cells are actively infected by, and defend-
ing against, T5 phage. cDIP-seq data largely recapitulated the same observations made in the absence of
phage (fig. S2B,E), but we were wary of drawing conclusions about reverse transcription output based
on an approach that would only quantify cDNAs still bound to the RT after immunoprecipitation. We
therefore turned to total DNA sequencing using the input controls from cDIP-seq experiments, and our
analyses revealed a strong induction in KpnDRT2 cDNA levels upon phage infection (Fig. 1E). Surpris-
ingly, while cDNA synthesis products in the absence of phage were predominantly single-stranded, with
opposite strandedness to the ncRNA, the presence of phage induced higher levels of both the initial cDNA
product and its reverse complement (Fig. 1E). Although these experiments do not reveal the identity of
the polymerase necessary for second-strand synthesis, we hypothesize that the RT likely possesses both
RNA-templated and DNA-templated DNA polymerase activity, similar to other well-studied bacterial re-
verse transcriptases (25), and that conversion of ssDNA to dsDNA may be a key step within the antiviral
defense pathway.
Rolling-circle reverse transcription generates concatenated cDNA products

We next sought to investigate the sequence requirements and potential antiviral function of cDNA
synthesis. We began by mutating the sequence of individual SLs throughout the ncRNA in order to elim-
inate base-pairing, focusing on SL1 at the 5′ end, SL2 at the base of the template region, SL5 within
the template region, and SL6 within the scaffold region (Fig. 2A). Mutations to all four regions led to a
complete loss of phage defense activity, indicating possible defects in ncRNA binding, cDNA synthesis,
or both (Fig. 2B). When we directly interrogated ncRNA binding and cDNA synthesis by the RT using
RIP-seq and cDIP-seq, respectively, we found that SL1 and SL6 mutants led to either a partial or complete
loss of cDNA synthesis, likely due to disruptions in the positioning of the RT on the ncRNA (Fig. 2C).
The SL5 mutant exhibited strong ncRNA and cDNA enrichment, as did an additional mutant in which
the region surrounding the cDNA synthesis start site was scrambled (fig. S3A-C), suggesting that defense
activity depends on not only cDNA synthesis, but also on the sequence of the cDNA product itself. The
phenotype of the SL2 mutant, however, was puzzling: the sequence of the template region was complete-
ly unchanged and cDNA production resembled the WT system, and yet phage defense was completely
abolished (Fig. 2B,C). This apparent discrepancy indicated that, beyond production of cDNA with the
appropriate ncRNA-specified sequence, additional features of the cDNA product underlie phage defense
activity.
Given that the stem of SL2 borders the template region, we hypothesized that disruption of this
structural element might lead to imprecise initiation or termination of cDNA synthesis, which in turn
might explain the altered immune function. We inspected the 3′ termini of cDIP-seq reads more closely,
and to our surprise, we found that the large majority of reads extended well beyond the boundary defined
by the coverage signal; these extensions had been soft-clipped from the reads by conventional alignment
Fig. 2. Rolling-circle reverse transcription generates a concatenated cDNA product. (A) Schematic of DRT2 ncRNA
secondary structure, with stem-loops (SL) numbered 1–8 and selected perturbations highlighted in red. SL1MUT, SL5MUT, and
SL6MUT correspond to ncRNA mutants in which the SL bases were scrambled, resulting in the elimination of sequence motifs
and secondary structure. SL2MUT abolishes base pairing within the SL2 stem. Sequences of all mutants are presented in Sup-
plementary Table S3. (B) Plaque assay showing loss of phage defense activity for all SL mutants from A (left), and bar graph
quantifying the reduction in efficiency of plating (EOP, right); data are mean ± s.d. (n = 3). (C) RIP-seq and cDIP-seq coverage
tracks for the indicated SL mutants alongside input controls, revealing a range of defects in either RNA binding, cDNA syn-
thesis/binding, or both. (D) Top: Schematic of terminal portions of cDIP-seq reads (light gray) failing to align to the cDNA
reference, resulting in ‘soft clipping’ and exclusion from coverage plots. A donut plot reporting the proportion of cDNA-map-
ping reads with the indicated lengths of 3′-clipped sequences is shown at left for DRT2 WT cDIP-seq. Bottom: Mapping of
3′-soft-clipped sequences from cDIP-seq experiments back to the DRT2 locus, demonstrating that they derive from the cDNA
5′ end. SL2MUT exhibits an aberrant pattern relative to WT. (E) Schematic of sequencing reads that map across two concatenated
cDNA repeats (top), and bar graph quantifying the abundance of junction-spanning reads from sequencing of total DNA in the
indicated conditions (bottom). Red and pink denote top and bottom strands, respectively; data are mean ± s.d. (n = 3). (F) Sche-
matic of long-read Nanopore sequencing workflow with DNA from phage-infected cells (top), and histogram of cDNA repeat
length distribution for WT KpnDRT2 from Nanopore sequencing (bottom). (G) Inferred mechanism of rolling-circle reverse
transcription (RCRT) mediated by sequence and structural features of SL2. After synthesis of 5′-TGT-3′ templated by ACA-1 at
the end of one cDNA repeat (top), the nascent DNA strand dissociates from its template and reanneals with the complementary
ACA-2 following SL2 melting (middle). Template jumping initiates a subsequent round of reverse transcription, with concate-
nation of one cDNA repeat to the next and incorporation of one additional base at the repeat junction, ultimately leading to long
rolling-circle cDNA products (bottom).
algorithms (Fig. 2D). To determine the identity of these soft-clipped extensions, we extracted their se-
quences and mapped them back to the plasmid and E. coli genome. Remarkably, these sequences in fact
derived from the 5′ end of the cDNA (Fig. 2D), suggesting a template jumping mechanism whereby the
RT proceeds from the end of the template region back to the start, resulting in concatenated cDNA repeat
products. Whereas the concatenated cDNAs generated by the WT system had a precise and uniform head-
to-tail junction, including one additional nucleotide immediately adjacent to SL2, junction sequences for
the SL2 mutant were more heterogeneous (Fig. 2D). Indeed, when we quantified the frequency of the
expected junction sequence across all tested ncRNA mutants, we found that only the SL5 and cDNA start
mutants retained WT levels of the repeat junction, whereas all other SL mutants nearly eliminated the
expected template jumping products (fig. S3D).
We next quantified concatenated cDNA products in total DNA samples from cells +/- T5 phage
infection. T5 phage infection triggered a large increase in the abundance of bottom-strand junction-span-
ning reads, corresponding to the initial products of RNA-templated DNA synthesis (Fig. 2E). This was
matched with a concomitant increase in top-strand junction-spanning reads (Fig. 2E), suggesting that
concatenated cDNA synthesis products are efficiently converted into dsDNA in a phage and RT-dependent
manner. Similar analyses from cDIP-seq datasets showed a lesser decrease in top-strand junction-span-
ning reads during phage infection (fig. S3E), which we attribute to the RT likely having lower affinity for
the dsDNA generated by second-strand cDNA synthesis, such that it releases these products in cells and/
or during immunoprecipitation.
Although short-read sequencing enabled accurate determination and quantification of cDNA repeat
junctions, we next leveraged long-read Nanopore sequencing to assess the length of concatenated cDNA
products, and to obtain orthogonal evidence of template jumping with a PCR-free approach. Remarkably,
KpnDRT2 cDNA products from phage-infected cells spanned a dramatic range of repeat lengths from
1–40 (Fig. 2F), revealing that reverse transcription by KpnDRT2 is highly processive and involves many
consecutive rounds of template jumping to generate long concatenated cDNA (from 120 to ~5000 bp). Fi-
nally, we carefully inspected the sequence and secondary structure of the ncRNA in order to better under-
stand the mechanism of template jumping. Concatenation of cDNA repeats occurs between the sequences
directly abutting SL2, and we noticed that the terminal 3-nt of each repeat are templated by a conserved 3′-
ACA-5′ (ACA-1) whose sequence perfectly matches the right half of SL2 (ACA-2; Fig. 2G). We therefore
hypothesized that the RT may dynamically melt SL2 during each round of reverse transcription, allowing
the terminal 5′-TGT-3′ of nascent cDNA transcripts to equilibrate between hybridization to ACA-1 and
ACA-2, and thus prime a subsequent round of cDNA synthesis (Fig. 2G). This model was supported by
a complete loss of defense activity in ncRNA mutants disrupting homology between the ACA motifs (fig.
S3F). We note that the proposed cDNA concatenation mechanism resembles rolling-circle DNA replica-
tion (26), and henceforth refer to the generation of concatenated cDNA products as rolling-circle reverse
transcription (RCRT; Fig. 2G).
Concatenated cDNAs encode a translated open reading frame (ORF)

Conventional rolling-circle amplification utilizes a circular template, and thus we sought to rule out
the alternative explanation that RCRT occurs as a result of ncRNA circularization at the repeat junction.
To investigate this hypothesis, we reanalyzed our RIP-seq input controls, which represent total RNA-seq
datasets, for the presence of reads spanning the repeat junction. Strikingly, we detected such reads abun-
dantly in KpnDRT2 samples, but they strictly depended on the presence of phage and an active RT, and
even more unexpectedly, their strandedness was opposite to that of the ncRNA (Fig. 3A). This observation
raised the intriguing possibility that cDNA second-strand synthesis might generate a template strand for
another round of transcription by RNA polymerase (Fig. 3B). The resulting transcript would have opposite
strandedness to the initial ncRNA and would contain multiple repeats of the cDNA sequence.
In agreement with this hypothesis, inspection of the cDNA sequence produced by RCRT revealed
consensus promoter elements spanning the repeat junction (Fig. 3B), highly reminiscent of transposon
promoters that are selectively formed upon DNA circularization during the transposon excision step (27,
28) These observations began to suggest that phage-induced second-strand cDNA synthesis might serve to
trigger the production of a high-copy concatenated RNA molecule with downstream antiphage function.
Consistent with this idea, we found that concatenated RNAs were strongly induced shortly after phage
infection by ~10,000-fold, in a time-course infection experiment (Fig. 3C). Northern blot analysis from
phage-infected cells using a probe selective for the chimeric junction revealed a broad size distribution
spanning hundreds to thousands of nucleotides (Fig. 3C), in excellent agreement with the large size of
cDNA products observed via Nanopore sequencing (Fig. 2F).
What could be the function of transcribing a repetitive cDNA sequence into RNA during an anti-
phage immune response? We closely examined the sequence of the cDNA and noticed that if we translat-
ed this sequence in silico, one out of three open reading frames (ORF) lacked any stop codons (Fig. 3D
and fig. S4A). This observation led us to hypothesize that the concatenated RNA produced during phage
Fig. 3. The concatenated cDNA product contains a never-ending ORF (neo). (A) Bar graph quantifying RNA-seq reads
that map across two concatenated cDNA repeats, for the indicated conditions. Red and pink denote top and bottom strands, re-
spectively; data are mean ± s.d. (n = 3). (B) Model showing the consecutive production of ncRNA (transcription), concatenated
double-stranded cDNA (reverse transcription and second-strand synthesis), and concatenated RNA (transcription), all encoded
by the DRT2 locus. Dashed lines indicate repeat–repeat junctions resulting from rolling-circle reverse transcription, and the
inset (top left) shows the predicted promoter formed across the junction. (C) Bar graph quantifying relative concatenated RNA
abundance in a phage infection time course experiment using RT-qPCR with repeat junction primers (top), and Northern blot
of concatenated RNA using a junction-spanning probe (bottom). RT-qPCR data are normalized to WT uninfected cells (t =
0); data are mean ± s.d. (n = 3). (D) Putative open reading frame (ORF) encoded by the concatenated RNA. The start of cDNA
synthesis and putative start of translation are indicated (pink and blue arrows, respectively), and the repeat–repeat junction is
denoted with a dashed line. (E) Schematic of the cDNA template region (pink), with the putative start codon and experimentally
tested mutations indicated. (F) Plaque assay showing that phage defense activity is eliminated with a single-bp substitution that
introduces an in-frame stop codon, but is only modestly affected by synonymous or missense mutations. EV, empty vector. (G)
Bar graph quantifying phage defense activity for insertions within SL3, SL4, or SL5, of the indicated length. Reduction in EOP
is calculated relative to an EV control; data are mean ± s.d. (n = 3). The only mutants that retain phage defense activity have
insertion lengths of a multiple of 3 bp.
infection might be translated to generate an antiviral polypeptide. This hypothesis was supported by the
presence of a predicted ribosome binding site upstream of the predicted start codon (Fig. 3D and fig.
S4B), and by the observation that programmed template jumping adds one additional nucleotide during
each round of cDNA synthesis (Fig. 2G). This activity generates a 120-bp cDNA repeat unit comprising
exactly 40 sense codons, such that the reading frame would be preserved through each repeat to yield a
continuous ORF (Fig. 3D).
We set out to comprehensively test the hypothesis that translation of the continuous ORF within
the concatenated RNA is necessary for phage defense. First, we identified a region within SL3 that was not
strongly conserved in sequence, and introduced single-bp mutations that would generate a synonymous,
missense, or nonsense codon (Fig. 3E). While the synonymous and missense mutations had mild effects
on defense activity that we attributed to perturbation of the ncRNA secondary structure, the nonsense
mutation completely abolished phage defense (Fig. 3F). We also mutated the predicted start codon and
found that mutation to the non-canonical GUG start codon partially preserved defense activity, but all
other mutations were inactive (fig. S4C,D). To assess whether translation of multiple contiguous repeats
of the ORF was necessary for phage defense, we tested mutations that would introduce stop codons near
the end of one full ORF repeat and found that these, too, led to a loss of defense (fig. S4C,E). Finally,
inspired by classic experiments performed by Crick and Brenner to deduce the triplet nature of the genetic
code (29), we selected three ncRNA loop regions and designed insertions ranging from 1–9 bp in length.
We hypothesized that if translation of the ORF was required for defense, then out-of-frame mutations
would lead to a loss of defense, where in-frame mutations (i.e., insertions of a multiple of 3 bp) would be
partially or completely tolerated. Remarkably, we found that all out-of-frame perturbations across three
non-conserved loop regions, including minimal 1-bp insertions, caused a >103-fold decrease in phage de-
fense, while insertions of 3, 6, or 9 bp maintained near-WT activity levels (Fig. 3G).
Collectively, these experiments provided compelling genetic evidence for the existence and ex-
pression of a cryptic gene produced by RNA-templated concatenation of DNA repeats. Intriguingly, this
de novo gene exhibits a heterogeneous length distribution and lacks any in-frame stop codons, and thus
we refer to it as neo (never-ending ORF).
Neo-encoded polypeptides induce cell dormancy

Encouraged by our genetic assays supporting the translation of neo, we sought unambiguous bio-
chemical evidence of Neo protein products. After analyzing the predicted amino acid composition of Neo,
we designed a custom protease cocktail that would yield unique peptide fragments suitable for mass spec-
trometry (MS)-based proteomics (fig. S5A). We then extracted proteins from KpnDRT2-expressing cells
and performed liquid chromatography with tandem mass spectrometry (LC-MS/MS) analysis (Fig. 4A).
Neo-derived peptides were exclusively detected in phage-infected cells that expressed the WT RT enzyme
(Fig. 4B), and their abundances were substantial when compared to the rest of the E. coli proteome (Fig.
4C). These results provide concrete proof that neo mRNAs transcribed from concatenated cDNA genes
are translated into protein.
To gain further insights into the physiological consequences of Neo expression, we performed
additional MS-based proteomics experiments using a more standard trypsin-based digestion procedure,
and analyzed the differential protein abundance between T5 phage-infected cells expressing WT or
YCAA-mutant KpnDRT2. Phage proteins were widely depleted in WT cells, as expected for a protective
immune response (Fig. 4D). On the host side, two significantly enriched cellular factors immediately
captured our attention — ArfA and RMF — due to their associations with ribosome stress and ribosome
Fig. 4. Neo proteins induce programmed cellular dormancy. (A) Schematic of experimental approach to detect Neo in
phage-infected cells by liquid chromatography with tandem mass spectrometry (LC-MS/MS). (B) Bar graph quantifying Neo
protein quantity from cells tested in the indicated conditions. Data are mean ± s.d. (n = 3). (C) Abundance of RT and Neo pro-
teins relative to the E. coli proteome in phage-infected cells expressing WT DRT2. (D) Differential protein abundance in T5-in-
fected cells expressing DRT2 WT or YCAA. Phage proteins are colored in brown, and ArfA and RMF are colored in red and
labeled. All other differentially abundant proteins (fold change > 2 and FDR < 0.05) are colored in dark blue. (E) Schematic of
alternative ribosome rescue pathway mediated by ArfA, which would release Neo proteins from ribosomes stalled on non-stop
neo mRNAs without targeting them for degradation (right), unlike the tmRNA pathway (left). (F) Growth curves of strains
transformed with empty vector (EV) or the WT DRT2 system, +/- T5 phage at the indicated multiplicity of infection (MOI).
Shaded regions indicate the standard deviation across independent biological replicates (n = 3). (G) Schematic of cloning and
inducible expression strategy to monitor the physiological effects of Neo polypeptides of variable repeat length. (H) Growth
curves of strains transformed with WT or scrambled Neo sequences of the indicated repeat lengths, alongside an empty vector
(EV) control. The dashed line indicates the point of induction with arabinose and theophylline. Shaded regions indicate the
standard deviation across independent biological replicates (n = 3).
hibernation, respectively (Fig. 4D). ArfA (Alternative ribosome-rescue factor A) is a translation factor that
specifically rescues ribosomes stalled on aberrant mRNAs lacking a stop codon, acting as an alternative
to the tmRNA pathway that tags nascent polypeptide chains for degradation (30, 31). ArfA is known to
be specifically upregulated under conditions of tmRNA depletion (32), and its ribosome rescue activity
in neo-expressing cells would elegantly resolve the conundrum of how stop codon-less neo mRNAs are
nonetheless translated into functional proteins (Fig. 4E). Meanwhile, RMF (Ribosome modulation factor)
is a ribosome-associated protein that directs the assembly of 70S ribosomes into inactive 100S dimers
during stationary phase (33, 34), and is activated by the alarmone ppGpp, a known trigger of growth arrest
and cellular dormancy (35, 36). Thus, the upregulation of ArfA supports the likely mechanism by which
Neo polypeptides are translated, and the induction of RMF suggests that Neo production is associated with
cellular dormancy.
Abortive infection and programmed dormancy have emerged in recent years as common mecha-
nisms by which bacterial immune systems provide population-level immunity against phage infection, as
host shutdown of metabolic processes prevents phage replication, and consequently, viral spread (11, 37).
To investigate whether DRT2 uses a similar immune mechanism, we performed phage infection assays in
liquid culture at varying multiplicities of infection (MOI). DRT2-expressing cultures survived T5 phage
infection at low MOI, but infection at high MOI led to stalling of growth (Fig. 4F). Further analysis of
cultures infected at high MOI revealed that DRT2 effectively blocked T5 replication (fig. S5B), and that
the growth-arrested cells remained viable (fig. S5C), altogether supporting a mechanism of phage defense
via programmed dormancy.
We next sought to test the physiological effects of recombinant Neo expression, reasoning that the
many intricate steps involved in Neo production from the ncRNA locus might add confounding factors
to protein function analyses. We initially attempted to clone neo onto a standard inducible expression
vector and test the hypothesis that neo expression would be sufficient to trigger cellular dormancy. Yet
repeated attempts to clone expression vectors with more than 2 repeats of WT neo proved unsuccessful,
compared to scrambled control sequences that could be cloned with high efficiency, and the few colonies
that emerged consistently exhibited frameshift mutations or lacked the neo insert altogether (fig. S5D,E).
These qualitative results suggested that Neo may potently arrest cell growth, and that its leaky expression
had prevented the isolation of positive clones. Intriguingly, this effect was only observed with Neo repeat
lengths of 3 or more.
To circumvent this challenge, we adopted an alternative strategy (38), in which neo genes were
placed on a low-copy vector under the control of a tightly regulated pBAD promoter and theophylline
riboswitch (Fig. 4G). This multilayered strategy for control of neo expression — which evokes the elab-
orate regulation of neo expression by native DRT2 loci — enabled the isolation of the desired clones. We
then transformed cells with expression vectors encoding WT or scrambled neo with 1-3 repeats, and mon-
itored cell growth in liquid culture before and after inducing neo expression with arabinose and theophyl-
line. Strikingly, only the 3-repeat WT Neo construct exhibited any growth defect compared to an empty
vector control (Fig. 4H). To assess whether the growth-arrested cells could recover from dormancy, we
plated cells from the final time point of the liquid culture experiment on solid media supplemented with
either repressor (glucose) or inducer (arabinose and theophylline). We found that cells expressing 3-repeat
WT Neo exhibited a ~102-fold increase in colony-forming units when plated on repressor versus inducer
(fig. S5F), indicating strong recovery from Neo-induced dormancy.
Considered together, these results suggest that the intricate gene synthesis mechanism encoded by
KpnDRT2 may have evolved in order to strictly control the production of an effector protein whose toxic-
ity is too potent to be safely controlled by conventional regulatory strategies.
Fig. 5. Concatenated neo genes and programmed dormancy are a broadly conserved phage defense mechanism. (A) Sche-
matic for the automated detection of putative Neo proteins in homologous DRT2 operons. (B) Phylogenetic tree of KpnDRT2
homologs, with outer rings showing the widespread presence of RT-associated ncRNAs and putative Neo proteins. Homologs
selected for experimental testing are indicated with pink circles. (C) Multiple sequence alignment (MSA) and secondary struc-
ture prediction of Neo proteins identified in B. A single Neo repeat is shown for all homologs; shading indicates amino acid
conservation. (D) AlphaFold prediction of a 3-repeat Neo polypeptide, showing the sites of proline mutagenesis tested in E.
Prolines were inserted C-terminal to the indicated residues within each of 3 concatenated repeats. (E) Growth curves of strains
transformed with 3-repeat Neo constructs containing the indicated proline insertions, alongside an empty vector (EV) control.
The dashed line indicates the point of induction with arabinose and theophylline. Shaded regions indicate the standard deviation
across independent biological replicates (n = 3). (F) Heat map showing the distribution of neo cDNA repeat lengths in cells ex-
pressing the indicated DRT2 homologs. Data are plotted as log10(CPM) from Nanopore sequencing of total DNA. (G) Heat map
showing the growth rates of cells expressing Neo homologs with the indicated repeat lengths. Growth rates are normalized to
an EV control and represent the mean of independent biological replicates (n = 3). Empty cells with X indicate Neo expression
constructs that could not be successfully cloned, presumably due to toxicity. (H) Model for the antiphage defense mechanism
of DRT2 systems. RT enzymes bind the scaffold portion of associated ncRNAs and produce concatenated cDNA products via
rolling-circle reverse transcription (RCRT). Phage infection triggers second-strand synthesis, yielding a dsDNA molecule that
is transcribed into never-ending ORF (neo) mRNAs. Neo translation exploits a ribosome rescue pathway to produce Neo pro-
teins that potently arrest cell growth, protecting the larger bacterial population from the spread of phage.
Neo gene synthesis and Neo protein toxicity is a broadly conserved phage defense strategy
Equipped with a wealth of mechanistic information on the production of Neo protein by KpnDRT2,
we set out to explore the evolutionary conservation of this gene synthesis strategy for antiviral defense.
Starting with a large phylogenetic tree of DRT2 homologs (fig. S6A and Supplementary Table S1), we
used covariance models to annotate RT-associated ncRNAs and then extracted the putative neo gene and
Neo protein sequence based on the expected mechanism of template jumping and absence of in-frame stop
codons (Fig. 5A and Materials and Methods). Our pipeline identified candidate ncRNAs and Neo pro-
teins for the vast majority of DRT2 systems that were related to KpnDRT2 (Fig. 5B), revealing broad con-
servation of this unique mechanism for concatenated gene synthesis. Notably, sequence motifs expected to
be critical for neo gene synthesis and expression, including ACA-1, ACA-2, and repeat junction-flanking
promoter elements, were also strongly conserved across diverse homologs (fig. S6B). Iterative generation
of additional covariance models also enabled ncRNA prediction for more divergent DRT2 clades, but Neo
protein annotation was more challenging, suggesting the possibility of alternative mechanisms of RCRT
and neo gene expression (fig. S6A,C).
Next, we investigated the amino acid sequence of diverse Neo proteins in more detail. Bioinfor-
matics analyses of multiple sequence alignments failed to identify any functional domains or similarities
to known proteins, but they did reveal high-confidence predictions of α-helical secondary structural ele-
ments (Fig. 5C). Using a 3-repeat Neo sequence, we predicted the 3D protein fold using multiple indepen-
dent methods, which yielded a structure reminiscent of HEAT repeats and other alpha solenoids consisting
of repeating antiparallel α-helices (39, 40) (Fig. 5D and fig. S6D). To test this model, we introduced
helix-breaking proline residues into either the loop connecting two α-helices, or into the helices directly
(Fig. 5D), and assessed the effects of these perturbations on cell growth. Consistent with our structural
model, insertions into either helix eliminated Neo-induced growth arrest, whereas the loop insertion mu-
tant exhibited a dormancy phenotype similar to the WT Neo sequence (Fig. 5E).
Having found that Neo proteins exhibit conserved α-helical folds, we sought to experimentally test
the conservation of additional critical features of Neo production and cellular function. We selected and
cloned 5 diverse DRT2 homologs (Fig. 5B and fig. S7A), and performed Nanopore sequencing of total
DNA from cells transformed with DRT2 expression vectors to assess the distribution of neo cDNA repeat
lengths. Remarkably, nearly all of the tested systems exhibited RCRT upon heterologous expression in
E. coli, and the cDNAs spanned a wide range of abundance levels and repeat lengths (Fig. 5F). We also
tested the effect of recombinant expression of the Neo proteins predicted to be encoded by these concate-
nated cDNAs. In all cases, Neo homolog expression led to repeat length-dependent growth arrest, further
confirming the requirement for cDNA repeat concatenation in the programmed dormancy mechanism to
defend against phage infection (Fig. 5G and fig. S7B).
Collectively, these experiments establish the generalizability of the KpnDRT2 gene synthesis and
antiphage defense mechanisms across a large swath of related immune systems.
DISCUSSION
Our work reveals an unprecedented mechanism of antiviral immunity mediated by DRT2 defense
systems (Fig. 5H). In uninfected cells, the ncRNA and RT enzyme are constitutively expressed from a sin-
gle promoter, leading to synthesis of a repetitive single-stranded cDNA via precise, programmed template
jumping that mediates RCRT. Upon phage infection, second-strand synthesis is triggered, leading to the
accumulation of double-stranded, concatenated cDNA molecules. A promoter created across the junction
between adjacent cDNA repeats then leads to abundant expression of heterogeneously sized mRNA en-
coding a stop codon-less, never-ending ORF (neo). It is the production of Neo protein, we propose, that
acts as the effector arm of the immune system by rapidly arresting cell growth and inducing programmed
dormancy, thus protecting the larger bacterial population from the spread of phage.
This pathway complicates textbook descriptions of the central dogma of molecular biology by
highlighting complex and repeated transitions back and forth between DNA- and RNA-based carriers
of genetic information, before translation finally yields a protein product. Furthermore, it challenges the
universal paradigm that genes are encoded linearly along the chromosomal axis. Genes across all three
domains of life are arranged in a polarized and singular orientation from head to tail, even considering
the existence of intron splicing and discontinuous exon joining in eukaryotes. Although neo proto-genes
are similarly arranged, synthesis of the mature gene form requires RNA-templated concatenation of the
tail of one proto-gene to the head of another. When considered alongside other examples of strategies
used by mobile elements to compactly encode genetic information, including ribosomal frameshifting,
overlapping ORFs, and nested genes, our work adds another layer of complexity to the ways in which
protein-coding sequences can be stored in the genome.
Why would such an elaborate immune pathway and gene synthesis mechanism have evolved?
One potential explanation is the need for stringent control of Neo expression. Our initial attempts to
clone recombinant Neo for functional testing were fraught with technical challenges, including the rapid
selection of loss-of-function mutants due to cellular toxicity, suggesting an extreme fitness cost associ-
ated with even low levels of Neo expression. It is likely that genomic encoding of pre-assembled neo
genes, under standard transcriptional control mechanisms, would pose intolerable autoimmune risk to the
host. We therefore hypothesize that gating Neo expression behind multiple layers of regulation, including
RNA-templated cDNA synthesis, phage-triggered dsDNA synthesis, and ArfA-mediated ribosome rescue,
likely allowed a potent dormancy factor to be stably maintained as part of the immune response.
Perhaps the most mysterious aspect of DRT2 immunity that remains to be elucidated is the structure
and molecular function of Neo. Although we detected unambiguous Neo-derived peptides in phage-in-
fected cells via mass spectrometry, the necessary protease digestion steps precluded determination of its
size. Our results indicate that a Neo polypeptide of at least 3 repeats in length is necessary and sufficient to
induce cell dormancy, but it is worth noting that neo mRNAs range in size from 200 to >5,000 nt, and that
translation could in theory initiate from within any neo repeat, each of which contains its own RBS. Thus,
we anticipate that native Neo proteins are similarly heterogeneous in size, potentially spanning hundreds
to thousands of amino acids. Mass spectrometry-based proteomics also highlighted ribosome hibernation
as a downstream effect of DRT2 immune system function, though additional experiments will be neces-
sary to determine whether RMF activation is a direct of Neo, or, more likely, an indirect consequence of
programmed dormancy. The fact that RMF activation is driven by the alarmone ppGpp (35), which causes
dormancy and growth arrest and is itself synthesized on ribosomes (41), raises the possibility that Neo
directly induces this cellular pathway. Finally, the conservation of α-helical repeat (αRep) domains fused
to reverse transcriptase domains in Class 1 DRT systems (19), which strongly resemble the predicted
α-helical fold of Neo, tantalizingly suggest a potential unifying theme for the effector functions across all
DRT systems.
Our discovery of highly efficient RCRT activity represents a unique biochemical behavior that
produces concatenated, repetitive cDNA molecules with precise junction sequences. Although many other
characterized reverse transcriptases exhibit intermolecular template switching activity in vitro (42, 43),
the intramolecular template jumping mechanism we describe here is the first report of such an activity
creating de novo protein-coding genes in vivo, with direct implications for biological function. Additional
work will be needed to determine the specific adaptations in the RT enzyme that facilitate this activity, but
our bioinformatics analyses and experimental results point to the critical importance of conserved ncRNA
structure and sequence motifs — in particular, the ACA motifs found abutting and within SL2. These mo-
tifs likely guide reannealing of the nascent cDNA transcript after one round of cDNA synthesis to a second
template immediately upstream of the cDNA start site, thereby initiating a new round of cDNA synthesis.
When considered alongside other well-studied examples of programmed template switching — such as
the synthesis of subgenomic RNAs by coronaviral RNA-dependent RNA polymerases (44, 45), and the
synthesis of full-length genomic cDNAs by retroviral reverse transcriptases (46) — our work expands
the diversity of products that can be generated by a single polymerase enzyme from its substrate. Fur-
thermore, DRT2 represents the first biological system in which a reverse transcriptase natively performs
RCRT. Intriguingly, it does so using a template that is not a closed circle, and thus differs from classic
examples of rolling circle amplification associated with plasmid, phage, and viroid replication (26). These
unique properties of DRT2-encoded enzymes offer considerable potential for biotechnology applications
that leverage templated DNA production in vivo (47–49), but with the added advantage of programmed
amplification. Notably, our data demonstrate that RCRT is maintained with mutation of SL5 or the reverse
transcription start site, suggesting that DRT2 could be harnessed to produce high-copy concatenated cD-
NAs with user-defined sequences.
The identification of Neo proteins disrupts conventionally held notions of features that define
protein coding genes, as well as our broader understanding of genome composition. Current genome
annotations algorithms generally rely on the definition of an ORF as a translated sequence of at least 30
amino acids bounded by a start and stop codon (50). However, only 26 codons of neo are identifiable from
a linear view of the K. pneumoniae genome, and the proto-gene lacks a stop codon. Thus, neo genes are
hidden in regions of genomes previously thought to be exclusively non-coding, suggesting that alternative
bioinformatics approaches will be needed in order to discover similar genes that elude standard methods
of ORF prediction. These findings seem especially important when considering the large proportion of
non-coding DNA in higher eukaryotes. For example, only ~1.5% of the human genome is thought to en-
code proteins (51). While much of the remaining ~98.5% of genome content encodes RNAs with known
or predicted gene regulatory functions, we posit that additional examples of Neo-like, non-canonical pro-
tein coding genes likely await future discovery in our own genomes.
REFERENCES
1. L. S. Frost, R. Leplae, A. O. Summers, A. Toussaint, Mobile genetic elements: the agents of open source
evolution. Nat Rev Microbiol 3, 722–732 (2005).
2. R. K. Aziz, M. Breitbart, R. A. Edwards, Transposases are the most abundant, most ubiquitous genes in
nature. Nucleic Acids Research 38, 4207–4217 (2010).
3. A. Canapa, M. Barucca, M. A. Biscotti, M. Forconi, E. Olmo, Transposons, Genome Size, and Evolutionary
Insights in Animals. Cytogenetic and Genome Research 147, 217–239 (2016).
4. E. V. Koonin, M. Krupovic, Evolution of adaptive immunity from transposable elements combined with
innate immune systems. Nat Rev Genet 16, 184–192 (2015).
5. E. V. Koonin, K. S. Makarova, Y. I. Wolf, M. Krupovic, Evolutionary entanglement of mobile genetic ele-

ments and host defence systems: guns for hire. Nat Rev Genet 21, 119–131 (2020).
6. C. Liu, Y. Zhang, C. C. Liu, D. G. Schatz, Structural insights into the evolution of the RAG recombinase.
Nat Rev Immunol 22, 353–370 (2022).
7. T. Karvelis, G. Druteika, G. Bigelyte, K. Budre, R. Zedaveinyte, A. Silanskas, D. Kazlauskas, Č. Venclovas,

V. Siksnys, Transposon-associated TnpB is a programmable RNA-guided DNA endonuclease. Nature 599,
692–696 (2021).
8. H. Altae-Tran, S. Kannan, F. E. Demircioglu, R. Oshiro, S. P. Nety, L. J. McKay, M. Dlakić, W. P. Inskeep,

K. S. Makarova, R. K. Macrae, E. V. Koonin, F. Zhang, The widespread IS200/IS605 transposon family
encodes diverse programmable RNA-guided endonucleases. Science 374, 57–65 (2021).
9. C. Meers, H. C. Le, S. R. Pesari, F. T. Hoffmann, M. W. G. Walker, J. Gezelle, S. Tang, S. H. Sternberg,

Transposon-encoded nucleases use guide RNAs to promote their selfish spread. Nature 622, 863–871
(2023).
10. D. Mayo-Muñoz, R. Pinilla-Redondo, N. Birkholz, P. C. Fineran, A host of armor: Prokaryotic immune

strategies against mobile genetic elements. Cell Reports 42, 112672 (2023).
11. H. Georjon, A. Bernheim, The highly diverse antiphage defence systems of bacteria. Nat Rev Microbiol,
1–15 (2023).
12. A. Millman, A. Bernheim, A. Stokar-Avihail, T. Fedorenko, M. Voichek, A. Leavitt, Y. Oppen-

heimer-Shaanan, R. Sorek, Bacterial Retrons Function In Anti-Phage Defense. Cell 183, 1551-1561.e12
(2020).
13. L. Gao, H. Altae-Tran, F. Böhning, K. S. Makarova, M. Segel, J. L. Schmid-Burgk, J. Koob, Y. I. Wolf, E.

V. Koonin, F. Zhang, Diverse enzymatic activities mediate antiviral immunity in prokaryotes. Science 369,
1077–1084 (2020).
14. A. González-Delgado, M. R. Mestre, F. Martínez-Abarca, N. Toro, Prokaryotic reverse transcriptases: from

retroelements to specialized defense systems. FEMS Microbiology Reviews 45, fuab025 (2021).
15. S. Silas, G. Mohr, D. J. Sidote, L. M. Markham, A. Sanchez-Amat, D. Bhaya, A. M. Lambowitz, A. Z. Fire,

Direct CRISPR spacer acquisition from RNA by a natural reverse transcriptase–Cas1 fusion protein. Science
351, aad4234 (2016).
16. F. Schmidt, M. Y. Cherepkova, R. J. Platt, Transcriptional recording by CRISPR spacer acquisition from
RNA. Nature 562, 380–385 (2018).
17. A. González-Delgado, M. R. Mestre, F. Martínez-Abarca, N. Toro, Spacer acquisition from RNA mediated
by a natural reverse transcriptase-Cas1 fusion protein associated with a type III-D CRISPR–Cas system in
Vibrio vulnificus. Nucleic Acids Research 47, 10202–10211 (2019).
18. J. Bobonis, K. Mitosch, A. Mateus, N. Karcher, G. Kritikos, J. Selkrig, M. Zietek, V. Monzon, B. Pfalz, S.
Garcia-Santamarina, M. Galardini, A. Sueki, C. Kobayashi, F. Stein, A. Bateman, G. Zeller, M. M. Savitski,
J. R. Elfenbein, H. L. Andrews-Polymenis, A. Typas, Bacterial retrons encode phage-defending tripartite
toxin–antitoxin systems. Nature 609, 144–150 (2022).
19. M. R. Mestre, L. A. Gao, S. A. Shah, A. López-Beltrán, A. González-Delgado, F. Martínez-Abarca, J.

Iranzo, M. Redrejo-Rodríguez, F. Zhang, N. Toro, UG/Abi: a highly diverse family of prokaryotic reverse
transcriptases associated with defense functions. Nucleic Acids Research 50, 6084–6101 (2022).
20. A. J. Simon, A. D. Ellington, I. J. Finkelstein, Retrons and their applications in genome engineering. Nucleic
Acids Research 47, 11007–11019 (2019).
21. Y. Wang, Z. Guan, C. Wang, Y. Nie, Y. Chen, Z. Qian, Y. Cui, H. Xu, Q. Wang, F. Zhao, D. Zhang, P. Tao,
M. Sun, P. Yin, S. Jin, S. Wu, T. Zou, Cryo-EM structures of Escherichia coli Ec86 retron complexes reveal
architecture and defence mechanism. Nat Microbiol 7, 1480–1489 (2022).
22. C. M. Sharma, J. Vogel, Differential RNA-seq: the approach behind and the biological insight gained. Cur-
rent Opinion in Microbiology 19, 97–105 (2014).
23. D. Dar, M. Shamir, J. R. Mellin, M. Koutero, N. Stern-Ginossar, P. Cossart, R. Sorek, Term-seq reveals
abundant ribo-regulation of antibiotics resistance in bacteria. Science 352, aad9822 (2016).
24. R. Lorenz, S. H. Bernhart, C. Höner zu Siederdissen, H. Tafer, C. Flamm, P. F. Stadler, I. L. Hofacker, Vien-
naRNA Package 2.0. Algorithms for Molecular Biology 6, 26 (2011).
25. S. K. Park, G. Mohr, J. Yao, R. Russell, A. M. Lambowitz, Group II intron-like reverse transcriptases func-
tion in double-strand break repair. Cell 185, 3671-3688.e23 (2022).
26. P. Wawrzyniak, G. Płucienniczak, D. Bartosik, The Different Faces of Rolling-Circle Replication and Its
Multifunctional Initiator Proteins. Front. Microbiol. 8 (2017).
27. B. Ton-Hoang, M. Bétermier, P. Polard, M. Chandler, Assembly of a strong promoter following IS911 circu-
larization and the role of circles in transposition. EMBO J 16, 3357–3371 (1997).
28. D. Perkins-Balding, G. Duval-Valentin, A. C. Glasgow, Excision of IS492 Requires Flanking Target Se-
quences and Results in Circle Formation in Pseudoalteromonas atlantica. Journal of Bacteriology 181,
4937–4948 (1999).
29. F. H. C. Crick, L. Barnett, S. Brenner, R. J. Watts-Tobin, General Nature of the Genetic Code for Proteins.
Nature 192, 1227–1232 (1961).
30. Y. Chadani, K. Ono, S. Ozawa, Y. Takahashi, K. Takai, H. Nanamiya, Y. Tozawa, K. Kutsukake, T. Abo,
Ribosome rescue by Escherichia coli ArfA (YhdL) in the absence of trans-translation system. Molecular
Microbiology 78, 796–808 (2010).
31. K. C. Keiler, Mechanisms of ribosome rescue in bacteria. Nat Rev Microbiol 13, 285–297 (2015).
32. F. Garza-Sánchez, R. E. Schaub, B. D. Janssen, C. S. Hayes, tmRNA regulates synthesis of the ArfA ribo-
some rescue factor. Molecular Microbiology 80, 1204–1219 (2011).
33. A. Wada, Y. Yamazaki, N. Fujita, A. Ishihama, Structure and probable genetic location of a “ribosome mod-
ulation factor” associated with 100S ribosomes in stationary-phase Escherichia coli cells. Proceedings of the
National Academy of Sciences 87, 2657–2661 (1990).
34. T. Prossliner, K. Skovbo Winther, M. A. Sørensen, K. Gerdes, Ribosome Hibernation. Annu Rev Genet 52,
321–348 (2018).
35. K. Izutsu, A. Wada, C. Wada, Expression of ribosome modulation factor (RMF) in Escherichia coli requires
ppGpp. Genes to Cells 6, 665–676 (2001).
36. T. K. Wood, S. Song, Forming and waking dormant cells: The ppGpp ribosome dimerization persister mod-
el. Biofilm 2, 100018 (2020).
37. A. J. Meeske, S. Nakandakari-Higa, L. A. Marraffini, Cas13-induced cellular dormancy prevents the rise of
CRISPR-resistant bacteriophage. Nature 570, 241–245 (2019).
38. J. Vandierendonck, Y. Girardin, P. De Bruyn, H. De Greve, R. Loris, A Multi-Layer-Controlled Strategy for

Cloning and Expression of Toxin Genes in Escherichia coli. Toxins 15, 508 (2023).
39. M. A. Andrade, C. Perez-Iratxeta, C. P. Ponting, Protein repeats: structures, functions, and evolution. J
Struct Biol 134, 117–131 (2001).
40. S. H. Yoshimura, T. Hirano, HEAT repeats – versatile arrays of amphiphilic helices working in crowded
environments? Journal of Cell Science 129, 3963–3970 (2016).
41. S. Ronneau, R. Hallez, Make and break the alarmone: regulation of (p)ppGpp synthetase/hydrolase enzymes
in bacteria. FEMS Microbiology Reviews 43, 389–400 (2019).
42. M. G. Wulf, S. Maguire, P. Humbert, N. Dai, Y. Bei, N. M. Nichols, I. R. Corrêa, S. Guan, Non-templated
addition and template switching by Moloney murine leukemia virus (MMLV)-based reverse transcriptases
co-occur and compete with each other. Journal of Biological Chemistry 294, 18220–18231 (2019).
43. A. M. Lentzsch, J. L. Stamos, J. Yao, R. Russell, A. M. Lambowitz, Structural basis for template switching
by a group II intron–encoded non-LTR-retroelement reverse transcriptase. Journal of Biological Chemistry
297, 100971 (2021).
44. S. Zúñiga, I. Sola, S. Alonso, L. Enjuanes, Sequence Motifs Involved in the Regulation of Discontinuous
Coronavirus Subgenomic RNA Synthesis. Journal of Virology 78, 980–994 (2004).
45. I. Sola, F. Almazán, S. Zúñiga, L. Enjuanes, Continuous and Discontinuous RNA Synthesis in Coronavirus-
es. Annu Rev Virol 2, 265–288 (2015).
46. K. Delviks-Frankenberry, A. Galli, O. Nikolaitchik, H. Mens, V. K. Pathak, W.-S. Hu, Mechanisms and Fac-
tors that Influence High Frequency Retroviral Recombination. Viruses 3, 1650–1680 (2011).
47. A. V. Anzalone, P. B. Randolph, J. R. Davis, A. A. Sousa, L. W. Koblan, J. M. Levy, P. J. Chen, C. Wilson,

G. A. Newby, A. Raguram, D. R. Liu, Search-and-replace genome editing without double-strand breaks or
donor DNA. Nature 576, 149–157 (2019).
48. S. C. Lopez, K. D. Crawford, S. K. Lear, S. Bhattarai-Kline, S. L. Shipman, Precise genome editing across
kingdoms of life using retron-derived DNA. Nat Chem Biol 18, 199–206 (2022).
49. S. Tang, S. H. Sternberg, Genome editing with retroelements. Science 382, 370–371 (2023).
50. D. Hyatt, G.-L. Chen, P. F. LoCascio, M. L. Land, F. W. Larimer, L. J. Hauser, Prodigal: prokaryotic gene
recognition and translation initiation site identification. BMC Bioinformatics 11, 119 (2010).
51. E. S. Lander, Initial impact of the sequencing of the human genome. Nature 470, 187–197 (2011).
52. M. Martin, Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet.journal
17, 10–12 (2011).
53. Md. Vasimuddin, S. Misra, H. Li, S. Aluru, “Efficient Architecture-Aware Acceleration of BWA-MEM for
Multicore Systems” in 2019 IEEE International Parallel and Distributed Processing Symposium (IPDPS)
(2019; https://ieeexplore.ieee.org/document/8820962), pp. 314–324.
54. H. Li, B. Handsaker, A. Wysoker, T. Fennell, J. Ruan, N. Homer, G. Marth, G. Abecasis, R. Durbin, 1000
Genome Project Data Processing Subgroup, The Sequence Alignment/Map format and SAMtools. Bioinfor-
matics 25, 2078–2079 (2009).
55. F. Ramírez, D. P. Ryan, B. Grüning, V. Bhardwaj, F. Kilpert, A. S. Richter, S. Heyne, F. Dündar, T. Manke,
deepTools2: a next generation web server for deep-sequencing data analysis. Nucleic Acids Res 44, W160-
165 (2016).
56. J. T. Robinson, H. Thorvaldsdóttir, W. Winckler, M. Guttman, E. S. Lander, G. Getz, J. P. Mesirov, Integra-

tive genomics viewer. Nat Biotechnol 29, 24–26 (2011).
57. Y. Liao, G. K. Smyth, W. Shi, featureCounts: an efficient general purpose program for assigning sequence
reads to genomic features. Bioinformatics 30, 923–930 (2014).
58. M. I. Love, W. Huber, S. Anders, Moderated estimation of fold change and dispersion for RNA-seq data
with DESeq2. Genome Biology 15, 550 (2014).
59. A. R. Quinlan, I. M. Hall, BEDTools: a flexible suite of utilities for comparing genomic features. Bioinfor-
matics 26, 841–842 (2010).
60. B. Langmead, S. L. Salzberg, Fast gapped-read alignment with Bowtie 2. Nat Methods 9, 357–359 (2012).
61. H. Li, Minimap2: pairwise alignment for nucleotide sequences. Bioinformatics 34, 3094–3100 (2018).
62. C. S. Hughes, S. Moggridge, T. Müller, P. H. Sorensen, G. B. Morin, J. Krijgsveld, Single-pot, sol-

id-phase-enhanced sample preparation for proteomics experiments. Nat Protoc 14, 68–85 (2019).
63. R. Bruderer, O. M. Bernhardt, T. Gandhi, S. M. Miladinović, L.-Y. Cheng, S. Messner, T. Ehrenberger,

V. Zanotelli, Y. Butscheid, C. Escher, O. Vitek, O. Rinner, L. Reiter, Extending the limits of quantitative
proteome profiling with data-independent acquisition and application to acetaminophen-treated three-dimen-
sional liver microtissues. Mol Cell Proteomics 14, 1400–1410 (2015).
64. K. M. McKenney, R. P. Connacher, E. B. Dunshee, A. C. Goldstrohm, Chemi-Northern: a versatile chemi-

luminescent northern blot method for analysis and quantitation of RNA molecules. RNA, rna.079880.123
(2024).
65. K. Katoh, K. Kuma, H. Toh, T. Miyata, MAFFT version 5: improvement in accuracy of multiple sequence
alignment. Nucleic Acids Research 33, 511–518 (2005).
66. A. M. Waterhouse, J. B. Procter, D. M. A. Martin, M. Clamp, G. J. Barton, Jalview Version 2—a multiple
sequence alignment editor and analysis workbench. Bioinformatics 25, 1189–1191 (2009).
67. F. Gabler, S.-Z. Nam, S. Till, M. Mirdita, M. Steinegger, J. Söding, A. N. Lupas, V. Alva, Protein Sequence
Analysis Using the MPI Bioinformatics Toolkit. Current Protocols in Bioinformatics 72, e108 (2020).
68. J. Jumper, R. Evans, A. Pritzel, T. Green, M. Figurnov, O. Ronneberger, K. Tunyasuvunakool, R. Bates, A.

Žídek, A. Potapenko, A. Bridgland, C. Meyer, S. A. A. Kohl, A. J. Ballard, A. Cowie, B. Romera-Paredes,
S. Nikolov, R. Jain, J. Adler, T. Back, S. Petersen, D. Reiman, E. Clancy, M. Zielinski, M. Steinegger, M.
Pacholska, T. Berghammer, S. Bodenstein, D. Silver, O. Vinyals, A. W. Senior, K. Kavukcuoglu, P. Kohli, D.
Hassabis, Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021).
69. M. Mirdita, K. Schütze, Y. Moriwaki, L. Heo, S. Ovchinnikov, M. Steinegger, ColabFold: making protein
folding accessible to all. Nat Methods 19, 679–682 (2022).
70. Z. Lin, H. Akin, R. Rao, B. Hie, Z. Zhu, W. Lu, N. Smetanin, R. Verkuil, O. Kabeli, Y. Shmueli, A. dos San-
tos Costa, M. Fazel-Zarandi, T. Sercu, S. Candido, A. Rives, Evolutionary-scale prediction of atomic-level
protein structure with a language model. Science 379, 1123–1130 (2023).
71. Z. Du, H. Su, W. Wang, L. Ye, H. Wei, Z. Peng, I. Anishchenko, D. Baker, J. Yang, The trRosetta server for
fast and accurate protein structure prediction. Nat Protoc 16, 5634–5651 (2021).
72. H. M. Salis, E. A. Mirsky, C. A. Voigt, Automated design of synthetic ribosome binding sites to control pro-
tein expression. Nat Biotechnol 27, 946–950 (2009).
73. P. Z. Johnson, A. E. Simon, RNAcanvas: interactive drawing and exploration of nucleic acid structures.
Nucleic Acids Research 51, W501–W508 (2023).
74. W. Li, A. Godzik, Cd-hit: a fast program for clustering and comparing large sets of protein or nucleotide
sequences. Bioinformatics 22, 1658–1659 (2006).
75. K. Katoh, D. M. Standley, MAFFT Multiple Sequence Alignment Software Version 7: Improvements in
Performance and Usability. Molecular Biology and Evolution 30, 772–780 (2013).
76. S. Will, T. Joshi, I. L. Hofacker, P. F. Stadler, R. Backofen, LocARNA-P: accurate boundary prediction and
improved detection of structural RNAs. RNA 18, 900–914 (2012).
77. E. P. Nawrocki, S. R. Eddy, Infernal 1.1: 100-fold faster RNA homology searches. Bioinformatics 29,
2933–2935 (2013).
78. E. Rivas, RNA structure prediction using positive and negative evolutionary information. PLOS Computa-
tional Biology 16, e1008387 (2020).
79. C. Camacho, G. Coulouris, V. Avagyan, N. Ma, J. Papadopoulos, K. Bealer, T. L. Madden, BLAST+: archi-
tecture and applications. BMC Bioinformatics 10, 421 (2009).
80. M. N. Price, P. S. Dehal, A. P. Arkin, FastTree 2 – Approximately Maximum-Likelihood Trees for Large
Alignments. PLOS ONE 5, e9490 (2010).
81. I. Letunic, P. Bork, Interactive Tree Of Life (iTOL) v5: an online tool for phylogenetic tree display and an-
notation. Nucleic Acids Res 49, W293–W296 (2021).
82. E. S. Wright, DECIPHER: harnessing local sequence context to improve protein multiple sequence align-
ment. BMC Bioinformatics 16, 322 (2015).
83. G. E. Crooks, G. Hon, J.-M. Chandonia, S. E. Brenner, WebLogo: A Sequence Logo Generator. Genome
Res. 14, 1188–1190 (2004).
84. A. Hecht, J. Glasgow, P. R. Jaschke, L. A. Bawazer, M. S. Munson, J. R. Cochran, D. Endy, M. Salit, Mea-
surements of translation initiation from all 64 codons in E. coli. Nucleic Acids Research 45, 3615 (2017).
ACKNOWLEDGEMENTS
We thank S. Pesari and Z. Akhtar for laboratory support, A. Bernheim and J. Bondy-Denomy for helpful
guidance on phage infection experiments, E. Semenova for helpful advice on ssDNA sequencing library
preparation, A.W.P. Fitzpatrick for insightful discussions on Neo function, B. Wiedenheft for suggesting
the Neo acronym, M. Laub for generously providing T5 phage, L.F. Landweber for qPCR instrument ac-
cess, and the JP Sulzberger Columbia Genome Center for NGS support.
FUNDING
S.T. was supported by an NIH Medical Scientist Training Program grant (T32GM145440) and an NIH
Ruth L. Kirchstein Individual Predoctoral Fellowship (F30AI183830). L.C.T. and M.W.G.W. were sup-
ported by a National Science Foundation Graduate Research Fellowship. J.T.G. was supported by a Hu-
man Frontier Science Program postdoctoral fellowship (LT001117/2021-C). L.E.B. was supported by
the Schaefer Research Scholars Program, the Hirschl Family Trust, and NIH grant R35GM124633. M.J.
was supported by the NIH (R01AG071869 and R01HG012216), NSF (Award 2224211), and a Columbia
University start-up package. This research was supported by a Pew Biomedical Scholarship, an Irma T.
Hirschl Career Scientist Award, and a generous start-up package from the Columbia University Irving
Medical Center Dean’s Office and the Vagelos Precision Medicine Fund (to S.H.S.).
AUTHOR CONTRIBUTIONS
S.T. and S.H.S. conceived of and designed the project. S.T. performed most experiments and experimental
analyses. V.C. performed cell growth and plaque assays, time course experiments, and helped with clon-
ing. D.J.Z. performed initial phylogenetic analyses and selected DRT2 homologs, generated the ncRNA
covariance model, performed cell growth and plaque assays, and helped with cloning. R.Z. performed
RT-qPCR and cell growth assays, and helped with cloning. G.D.L. performed and helped with the anal-
ysis of Nanopore sequencing experiments. T.W. performed phylogenetic analyses and bioinformatically
identified Neo homologs. L.C.T. prepared samples and performed mass spectrometry experiments. M.W.
performed cell growth assays. M.W.G.W. helped with the design and analysis of Neo protein expression
data. J.T.G. contributed to the interpretation of rolling-circle reverse transcription results. L.E.B. advised
on the design and interpretation of Northern blotting experiments. M.J. advised on the design and inter-
pretation of mass spectrometry experiments. S.T. and S.H.S. discussed the data and wrote the manuscript,
with input from all authors.
COMPETING INTERESTS
Columbia University has filed a patent application related to this work. S.H.S. is a co-founder and scien-
tific advisor to Dahlia Biosciences, a scientific advisor to CrisprBits and Prime Medicine, and an equity
holder in Dahlia Biosciences and CrisprBits.
DATA AND MATERIALS AVAILABILITY

Next-generation sequencing data will be made available in the National Center for Biotechnology Infor-
mation (NCBI) Sequence Read Archive upon publication. Additional datasets generated and analyzed in
the current study are available from the corresponding author upon reasonable request.
SUPPLEMENTARY MATERIALS
Materials and Methods
Figs. S1 to S7
Tables S1 to S5 (provided separately)

2024 05 08 593200v1 Full

Uploaded by

Copyright:

Available Formats

2024 05 08 593200v1 Full

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2024 05 08 593200v1 Full

Uploaded by

Copyright:

Available Formats

bioRxiv preprint doi: https://doi.org/10.1101/2024.05.08.593200; this version posted May 8, 2024.

The copyright holder for this preprint (which

De novo gene synthesis by an antiviral reverse transcriptase

Rolling-circle reverse transcription generates concatenated cDNA products

Concatenated cDNAs encode a translated open reading frame (ORF)

Neo-encoded polypeptides induce cell dormancy

5. E. V. Koonin, K. S. Makarova, Y. I. Wolf, M. Krupovic, Evolutionary entanglement of mobile genetic ele-

7. T. Karvelis, G. Druteika, G. Bigelyte, K. Budre, R. Zedaveinyte, A. Silanskas, D. Kazlauskas, Č. Venclovas,

8. H. Altae-Tran, S. Kannan, F. E. Demircioglu, R. Oshiro, S. P. Nety, L. J. McKay, M. Dlakić, W. P. Inskeep,

9. C. Meers, H. C. Le, S. R. Pesari, F. T. Hoffmann, M. W. G. Walker, J. Gezelle, S. Tang, S. H. Sternberg,

10. D. Mayo-Muñoz, R. Pinilla-Redondo, N. Birkholz, P. C. Fineran, A host of armor: Prokaryotic immune

12. A. Millman, A. Bernheim, A. Stokar-Avihail, T. Fedorenko, M. Voichek, A. Leavitt, Y. Oppen-

13. L. Gao, H. Altae-Tran, F. Böhning, K. S. Makarova, M. Segel, J. L. Schmid-Burgk, J. Koob, Y. I. Wolf, E.

14. A. González-Delgado, M. R. Mestre, F. Martínez-Abarca, N. Toro, Prokaryotic reverse transcriptases: from

15. S. Silas, G. Mohr, D. J. Sidote, L. M. Markham, A. Sanchez-Amat, D. Bhaya, A. M. Lambowitz, A. Z. Fire,

19. M. R. Mestre, L. A. Gao, S. A. Shah, A. López-Beltrán, A. González-Delgado, F. Martínez-Abarca, J.

38. J. Vandierendonck, Y. Girardin, P. De Bruyn, H. De Greve, R. Loris, A Multi-Layer-Controlled Strategy for

47. A. V. Anzalone, P. B. Randolph, J. R. Davis, A. A. Sousa, L. W. Koblan, J. M. Levy, P. J. Chen, C. Wilson,

56. J. T. Robinson, H. Thorvaldsdóttir, W. Winckler, M. Guttman, E. S. Lander, G. Getz, J. P. Mesirov, Integra-

62. C. S. Hughes, S. Moggridge, T. Müller, P. H. Sorensen, G. B. Morin, J. Krijgsveld, Single-pot, sol-

63. R. Bruderer, O. M. Bernhardt, T. Gandhi, S. M. Miladinović, L.-Y. Cheng, S. Messner, T. Ehrenberger,

64. K. M. McKenney, R. P. Connacher, E. B. Dunshee, A. C. Goldstrohm, Chemi-Northern: a versatile chemi-

alignment. Nucleic Acids Research 33, 511–518 (2005).

68. J. Jumper, R. Evans, A. Pritzel, T. Green, M. Figurnov, O. Ronneberger, K. Tunyasuvunakool, R. Bates, A.

DATA AND MATERIALS AVAILABILITY

You might also like