Reverse transcribed ssDNA derepresses translation of a retron antiviral protein

25 Retrons are bacterial immune systems that prevent the spread of phages by initiating a toxic 26 response within infected hosts. All previously characterized retrons produce high levels of 27 multicopy single-stranded DNA (msDNA) in the cell by reverse transcription, which acts as an 28 antitoxin in the absence of phage infection. However, we describe here a non-canonical mechanism 29 for Type VI retrons, which do not produce detectable msDNA in the absence of phage, yet still 30 provide phage defense. Focusing primarily on Retron-Vpa2, a Type VI retron from Vibrio 31 parahaemolyticus, we show broad defense against phages and identify triggers of the system within 32 phage recombination systems. Within the Retron-Vpa2 operon, we find a highly enriched, 33 structured transcript that we term a hybrid RNA (hyRNA), which contains both the retron’s reverse 34 transcription template and a translationally repressed toxic effector coding sequence. We find that 35 phage infection induces the accumulation of high levels of msDNA and that this msDNA is necessary 36 for derepressing translation of the antiviral toxin. These findings present key biological and 37 mechanistic insights into a distinct group of retrons while highlighting the diversity of systems that 38 participate in bacterial immunity. 39 40 41 42 43 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint

44 Bacteria have evolved an arsenal of antiviral defense mechanisms to protect themselves from 45 phages, including a class of defense systems called retrons1–3. Retrons operate via abortive infection 46 – upon sensing phage, they release a toxic effector that induces host cell death or dormancy before 47 the phage can propagate, sparing the uninfected bacterial population2–5. Retrons are typically three-48 part systems, consisting of a reverse transcriptase (RT), a noncoding RNA (ncRNA), and a toxic 49 effector1–3,6. The RT recognizes a highly structured msr region on the ncRNA and reverse 50 transcribes an adjacent msd region into multicopy single-stranded DNA (msDNA)7–13. The 51 accumulation of high levels of msDNA by ongoing reverse transcription has been considered a 52 hallmark of retrons12–16. In fact, the distinctive bands that the msDNA generates on a 53 polyacrylamide gel enabled the discovery of the earliest identified retrons in the 1980s and are 54 used as a signature for Retron-containing species to this day13–19. 55 56 In canonical retron systems, the msDNA forms a complex with the remaining ncRNA and RT, which 57 together interact with the effector protein as an antitoxin to neutralize its toxic effect3,20–27. While 58 toxin-antitoxin motifs are common in phage defense, the incorporation of reverse transcribed DNA 59 in the antitoxin is unique to retrons, enabling these systems to respond rapidly to phage-encoded 60 DNA interacting proteins, such as nucleases3,28–30, single-stranded binding proteins28, and 61 methyltransferases3,16,22. During infection, these phage proteins degrade, perturb, or modify the 62 msDNA, leading to release of the toxin and activation of the defense response3,16,22,28,29. 63 64 Retrons are a particularly diverse class of defense systems. A bioinformatics study encompassing 65 over 1900 retrons from multiple bacterial phyla classified retrons into thirteen types based on 66 different protein domains found in the highly variable toxic effector component6. Following this 67 bioinformatic work, an experimental census was conducted to survey retrons from all thirteen 68 types, including characterization of msDNA production15. This census revealed an unexpected trait 69 among a previously untested group: Type VI retrons failed to produce detectable msDNA. 70 71 Here, we present the first mechanistic study of Type VI retrons, finding that this type deviates 72 substantially from the canonical mechanism described in previously studied retrons. These retrons 73 defend against phage despite the lack of detectable msDNA at baseline. We reveal that the msr-msd 74 and effector coding sequence occupy a contiguous transcript which we call the hybrid RNA 75 (hyRNA). Further characterization of the operon shows that the hyRNA secondary structure and an 76 additional accessory protein result in translational repression of the toxic effector. Counter to all 77 other retrons, reverse transcription of the Type VI msDNA is induced during phage infection, 78 specifically by phage-encoded recombination-associated proteins. Rather than serving as an 79 antitoxin, the Type VI msDNA derepresses the translation of the effector. This inverted triggering 80 mechanism constitutes an intriguing departure from our current understanding of retrons, 81 prompting a reconsideration of the mechanistic range of bacterial immunity. 82 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint

83 Retron-Vpa2 confers broad phage defense without detectable msDNA 84 Type VI retrons have a characteristic operon that begins w ith a predicted ncRNA region containing 85 the msr-msd, followed by an ORF encoding a small protein (SP) of unknown function (Fig 1a). 86 Downstream of this is another unknown accessory protein containing a helix-turn-helix (HTH) 87 domain, followed by the retron reverse transcriptase (RT). 88 89 Type VI retrons form a well-supported monophyletic clade (clade 3) within the phylogeny of retron 90 RTs6 together with the Type IV (e.g. Retron-Eco62) and Type V retrons (e.g. Retron-Sen131,32). In a 91 previously conducted experimental census of all retron types, we found no msDNA production from 92 Type VI retrons when expressed in E. coli15. Here, we replicated that result (Fig 1c). Five of the 93 retrons were retested exactly as in the retron census, with expression of just the retron RT and 94 predicted ncRNA (Retrons-Psp1, -Bfr1, -Tde1, -Psp2, and –Cle1). For two of the retrons, we 95 synthesized and expressed the entire retron operon to test whether omission of a component other 96 than the RT and ncRNA in the previous census might have accounted for the lack of msDNA. Of the 97 two retrons tested as a full operon, one had been previously included in the census in the more 98 minimal form (Retron-Vpa2 from Vibrio parahaemolyticus), while the other was newly included to 99 check whether matching the host species might yield detectable msDNA (Retron-Eco12 from E. 100 coli). Yet, we again observed no msDNA for any Type VI retron tested in any format. 101 102 We next tested whether these retrons confer defense against phage infection. We expressed each 103 under their native promoters in an E. coli BL21-AI derivative with its endogenous Retron-Eco1 104 deleted (henceforth bSLS.114)33. Despite the lack of msDNA production, we observed robust 105 defense against a virulent strain of phage lambda with each of the Type VI retrons (Fig 1d). For 106 Retron-Vpa2, we then tested for defense in an expanded panel of 92 phages, including the BASEL 107 collection34 (Fig 1e). We observed broad defense against the Demerecviridae, Myoviridae, and 108 Drexlerviridae families of phages. 109 110 We selected several phages that induced strong defense phenotypes (Bas10, Bas34, Bas53, Muut, 111 and Stevie_ev116) and infected liquid cultures of Retron-Vpa2-expressing E. coli MG1655 at a range 112 of MOIs from 0 to 10 (Fig 1f). We observed a clear abortive infection phenotype in Retron-Vpa2-113 expressing cells for phages Bas10, Bas34, and Stevie_ev116, where the population persists through 114 phage infection in a MOI-dependent manner. For phages Bas53 and Muut, cells expressing the 115 retron grow equally well regardless of MOI. By contrast, cultures without the retron exhibit collapse 116 at every MOI. Overall, these data indicate that Retron-Vpa2 confers protection to the bacterial 117 population in the presence of phage. 118 119 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint 120 Figure 1 – Retron-Vpa2 confers broad phage defense without detectable msDNA. 121 A) Schematic of the Type VI retron operon. B) Location of Type VI retrons and their close relatives in clade 3 of a retron RT 122 phylogeny. C) PAGE analysis for msDNA abundance comparing several Type VI retrons with Retron-Eco1 serving as a positive 123 control for msDNA. A single-stranded DNA ladder with lengths marked in nucleotides is shown for reference. D) Spot assay 124 showing titration of phage lambdavir on Retron-Eco12- and -Vpa2-expressing strains of E. coli bSLS.114 relative to an empty 125 vector control. Efficiency of plating is quantified as pfu/mL in the adjacent bar graph (one-way ANOVA P=0.0152, Dunnett’s 126 test corrected for multiple comparisons versus empty vector: Retron-Eco12 P=0.0181; Retron-Vpa2 P=0.0179). E) Fold 127 defense conferred by Retron-Vpa2 against a panel of 92 phages, depicted by shaded boxes. Colored bars denote phage family, 128 subfamily, and genus. F) Growth curves of E. coli MG1655 expressing either Retron-Vpa2 or an empty vector, infected with 129 five different phages at MOIs of 0, 0.01, 0.1, 1, and 10. Additional statistical details in Supplementary Table 1. 130 131 132 133 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint 134 135 Recombination systems trigger Retron-Vpa2 in two phages 136 137 To identify factors responsible for triggering Retron-Vpa2 defense, we examined phage escapees 138 for lambdavir and Stevie_ev116. Whole genome sequencing of the lambdavir escapees revealed 139 various mutations in the single-stranded annealing protein (beta) and exonuclease (exo) genes of 140 the lambda Red operon, a known recombination system that also includes the RecBCD inhibitor 141 gam35–37 (Fig 2a). Many of these are early stop and frameshift mutations that result in loss-of-142 function of the proteins. Of the mutations that do not disrupt the open reading frame, the point 143 mutations in exo reside near its catalytic site and may influence its enzymatic activity (Fig S1a-c). 144 The point mutation in beta resides in its C-terminal domain which interacts with exo38 (Fig S1e-f). 145 Stevie_ev116 escapee mutations were primarily found in an exonuclease (exo) and recombinase 146 (rec) contained in the same operon (Fig 2b, S1d, S1g-h). Due to the presence of an ERF family 147 domain on the recombinase, we predict that this operon is also a recombination system39. 148 149 We experimentally validated our candidate trigger genes from lambdavir by co-expressing them 150 with Retron-Vpa2 in liquid cultures of E. coli bSLS.114 and measuring OD600 over 16 hrs. Co-151 expression of the retron with the full lambda Red operon resulted in strong growth inhibition 152 compared to a control strain lacking the retron, confirming that this operon is sufficient to trigger 153 the abortive infection phenotype (Fig 2c-d, S2a). We also introduced two mutations that we 154 observed in the escapees (beta T132fs and exo W163*) to the operon, which each resulted in partial 155 rescue of the growth suppression phenotype. Deleting the third component of the Red operon, gam, 156 which is a known trigger of Type IV and V retrons2,28, did not rescue growth. Given that the 157 combination of beta and exo is sufficient to fully trigger the system, we next tested them separately, 158 finding that neither affected cell growth when expressed individually (Fig 2e, S2b). Interestingly, 159 gam resulted in slight inhibition when expressed in the absence of beta and exo (Fig 2e, S2b). This 160 differs from the effect of gam on other clade 3 retrons, such as Type IV Retron-Eco6 and Type V 161 Retron-Sen1. For both these retrons, gam triggers them to the same degree as the full Red operon, 162 with no requirement for beta or exo (Fig 2f-g, S2c). This points to a clear distinction between the 163 triggering mechanism of Retron-Vpa2 and that of its close relatives. 164 165 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint 166 Figure 2 – Recombination systems trigger Retron-Vpa2 in two phages. 167 A) Operon of the genes mutated in lambdavir escapees of Retron-Vpa2. Specific mutations are labeled in red. Mutations used 168 in trigger assays are in bold. B) Operon of the genes mutated in Stevie_ev116 escapees of Retron-Vpa2. Specific mutations are 169 labeled in red. C) Trigger assay showing growth curves of liquid cultures expressing Retron-Vpa2 with variants of the lambda 170 Red operon, relative to an empty vector with the Red operon, measured over 16 hrs. Each line is the mean of three biological 171 replicates, error bands are plotted in Fig S2a. D) Comparison of OD600 measurements from panel (C) at the 10 hr timepoint, 172 shown as a percentage of the empty vector control at the same timepoint. Bars show the mean of three biological replicates 173 with individual replicates plotted as white circles (one-way ANOVA P=0.0023, Dunnett’s test corrected for multiple 174 comparisons versus empty vector+Red: Retron-Vpa2+Red P=0.0017, Retron-Vpa2+Beta+Exo P=0.0029, all other conditions 175 ns). E) Trigger assay showing growth curves of liquid cultures expressing Retron-Vpa2 with Gam, Beta, or Exo, relative to an 176 empty vector with gam, measured over 16 hrs. Growth curve of Retron-Vpa2 with the Red operon from panel (C) shown in a 177 dotted red line for comparison. Each line is the mean of three biological replicates, error bands are plotted in Fig S2b. F) 178 Trigger assay showing growth curves of liquid cultures expressing either Retron-Eco6 or Retron-Sen1 with either the Red 179 operon or gam alone, relative to an empty vector with the Red operon, measured over 16 hrs. Each line is the mean of three or 180 more biological replicates, error bands are plotted in Fig S2c. G) Comparison of OD600 measurements using data from panels 181 (C), (E) and (F) at the 10 hr timepoint, shown as a percentage of the empty vector control at the same timepoint. The Retron-182 Vpa2 + Red and Retron-Vpa2 + Beta + Exo bars are repeated from (D) to permit direct comparison. Bars show the mean of 183 three or more biological replicates with individual replicates plotted as white circles (one-way ANOVA P=0.0002, Šídák's test 184 corrected for multiple comparisons versus empty vector+Red: Retron-Vpa2+Red P=0.0073, Retron-Vpa2+Beta+Exo P=0.0145, 185 Retron-Eco6+Red P=0.0119, Retron-Eco6+Gam P=0.0027, Retron-Sen1+Red P=0.0118, Retron-Sen1+Gam P=0.0298, all other 186 conditions ns; Šídák's test corrected for multiple comparisons Retron-Eco6+Red versus Retron-Eco6+Gam and Retron-187 Sen1+Red versus Retron-Sen1+Gam both ns ). Additional statistical details in Supplementary Table 1. 188 189 A hybrid RNA encodes a toxic effector and forms a complex with the RT and HTH 190 We next investigated each component of the Retron-Vpa2 operon. First, we sought to 191 experimentally validate the msr-msd region that we predicted based on its secondary structure15. 192 Transcripts from the ncRNA region of the retron operon are generally highly enriched in a cell2. We 193 therefore performed RNAseq on Retron-Vpa2-expressing E. coli bSLS.114 and found that 194 sequencing coverage was enriched not only in the predicted msr-msd region, but also in the 195 downstream SP coding sequence (Fig 3a). Performing the experiment with a catalytically dead RT 196 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint (YADD -> YAAA catalytic site mutation) had no effect on this enrichment (Fig S3). Because this RNA 197 contains both the msr-msd and the effector coding region in a contiguous transcript, we decided to 198 name it the hybrid RNA (hyRNA). 199 200 The Vpa2 hyRNA is approximately 470 nt in length. Its predicted secondary structure, based on 201 minimal free energy, resembles three arms extending from a central node (Fig 3b). The first arm 202 contains a highly structured region characteristic of a retron msr-msd7,15. The second arm contains 203 two stem-loops which incorporate the ribosomal binding site (RBS) and start codon of the SP. The 204 third arm contains the remaining coding sequence of the SP. The 3’ end of the transcript then wraps 205 back around and anneals to the first arm via a set of inverted repeats, which resembles the a1/a2 206 priming region of a canonical retron9. A covariance model of the hyRNA independently supports the 207 predicted secondary structure (Fig S4a) and reveals multiple areas of high sequence conservation, 208 including the stems in the first arm, the inverted repeats that link each arm to the central node, the 209 SP RBS and start/stop codons, and a set of direct repeats in the loop regions of the first and second 210 arm (Fig S4b). 211 212 Given the SP’s unique position within the hyRNA, we were curious about its role in this system. Its 213 predicted structure reveals a C-terminal domain with a hydrophobic core that is conserved across 214 multiple Type VI homologs (Fig S5a-d). When expressed independently from a synthetic operon, we 215 observed that the SP impairs cell growth at a similar level to triggering the retron with the lambda 216 Red operon (Fig 3c-d, S6a). Introduction of an early stop mutation (K15*) in the SP eliminated the 217 retron’s toxicity upon triggering. Additionally, introduction of three point mutations (I39R, F43R, 218 F45R) in the SP’s conserved hydrophobic core significantly reduced toxicity under the same 219 conditions. From these results, we conclude that the SP alone is the toxic effector of this retron and 220 that its C-terminal domain is essential for toxicity. 221 222 We then proceeded to mutate other components of the system in the context of the full operon (Fig 223 3e-f, S6b). We found that deleting the noncoding region of the hyRNA from the 5’ end up until the 224 RBS (Δnc-hyRNA) was highly toxic. Similarly, mutating a conserved residue in the HTH protein 225 (R7D) was also toxic. However, mutating the RT catalytic domain (YADD -> YAAA)40 was not toxic, 226 indicating that reverse transcription is not necessary to neutralize the SP, contrasting with 227 canonical retron mechanisms. 228 229 To understand which components of the system physically interact with each other, we next 230 engineered a version of Retron-Vpa2 with FLAG-tagged RT and HA-tagged HTH in a dSP (K15*) 231 background. We expressed the system in E. coli MG1655-DE3 and performed a pull-down on either 232 the RT or HTH, followed by a Western blot and TBE-Urea PAGE to detect co-immunoprecipitation of 233 the tagged proteins and hyRNA respectively. Pulling down on the RT led to co-immunoprecipitation 234 of the HTH and hyRNA (Fig 3g, S7a). Pulling down on the HTH led to co-immunoprecipitation of the 235 RT and hyRNA (Fig 3h, S7b). Thus, the RT, HTH, and hyRNA form a complex in vivo. We also checked 236 for msDNA in the co-immunoprecipitation but did not detect any in either pull-down (Fig S7c). The 237 co-immunoprecipitated RNA from the RT pull-down was confirmed to be the hyRNA through 238 sequencing (Fig S7d). 239 240 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint 241 Figure 3 – A hybrid RNA encodes a toxic effector and forms a complex with the RT and HTH. 242 A) RNAseq coverage (normalized to maximum coverage) of Retron-Vpa2 operon. B) Predicted secondary structure of the 243 hyRNA with msr-msd, inverted repeats (IR), ribosome binding site (RBS), SP start codon (AUG) and SP coding region labeled. 244 C) Growth curves of liquid cultures expressing SP alone, Retron-Vpa2 triggered with the Red operon, and two Retron-Vpa2 SP 245 mutants triggered with the Red operon, relative to an empty vector with the Red operon, measured over 16 hrs. Cultures were 246 induced at 2 hrs (vertical dotted line). Each line is the mean of three biological replicates, error bands are plotted in Fig S6a. 247 D) Comparison of OD600 measurements from panel (C) at the 10 hr timepoint, shown as a percentage of the empty vector 248 control at the same timepoint. Bars show the mean of three biological replicates with individual replicates plotted as white 249 circles (one-way ANOVA P<0.0001, Dunnett’s test corrected for multiple comparisons versus empty vector+Red: SP P<0.0001, 250 Retron-Vpa2+Red PYAAA) mutant, relative 252 to an empty vector, measured over 16 hrs. Cultures were induced at 2 hrs (vertical dotted line). Each line is the mean of three 253 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint biological replicates, error bands are plotted in Fig S6b. F) Comparison of OD600 measurements from panel (E) at the 10 hr 254 timepoint, shown as a percentage of the empty vector control at the same timepoint. Bars show the mean of three biological 255 replicates with individual replicates plotted as white circles (one-way ANOVA P<0.0001, Dunnett’s test corrected for multiple 256 comparisons versus empty vector: Retron-Vpa2 Δnc-hyRNA P<0.0001, Retron-Vpa2 HTH(R7D) P<0.0001, all other conditions 257 ns). G) Anti-FLAG pull-down of cultures expressing Retron-Vpa2, with and without a 3xFLAG-tagged RT. On the left are anti-258 HA and anti-FLAG Western blots of the pull-down and lysate fractions, with numbers showing protein weight in kDa. On the 259 right is a TBE-Urea PAGE analysis showing detection of hyRNA, with numbers showing length in nt. Complete gels are found 260 in Fig S7a. H) Anti-HA pull-down of cultures expressing Vpa2, with and without a HA-tagged HTH. On the left are anti-HA and 261 anti-FLAG Western blots of the pull-down and lysate fractions, with numbers showing protein weight in kDa. On the right is a 262 TBE-Urea PAGE analysis showing detection of hyRNA, with numbers showing length in nt. Complete gels are found in Fig S7b. 263 Additional statistical details in Supplementary Table 1. 264 265 msDNA production is repressed until Retron-Vpa2 is triggered 266 Given that we detected no abundant msDNA and the catalytic activity of the RT is not necessary for 267 toxin neutralization, we wondered whether the RT could play a non-catalytic role in this retron. We 268 tested phage defense against lambdavir with a version of Retron-Vpa2 carrying a catalytically dead 269 RT and found that inactivating reverse transcription eliminated phage defense (Fig 4a). Given the 270 necessity of reverse transcription for defense, the absence of msDNA was even more puzzling. Thus, 271 we decided to implement a new strategy for detecting msDNA, inspired by our previous work 272 developing a CRISPR-based molecular recorder33. We call this strategy Spacer-seq (Fig 4b). Spacer-273 seq harnesses the E. coli Type I-E CRISPR-Cas adaptation machinery, which mediates capture of 274 foreign DNA elements into CRISPR arrays41–44. This will “record” the presence of retron msDNA into 275 the host cell’s genome over time. In this CRISPR-Cas system, captured DNA is recorded as 33 nt 276 spacers via the integrase proteins Cas1 and Cas243,44. 277 278 We co-expressed Retron-Vpa2 with Cas1-Cas2 in E. coli bSLS.114 (which contains an endogenous 279 CRISPR array but no endogenous Cas proteins). After 24 hrs of induction in liquid culture, we 280 harvested cells and PCR amplified the CRISPR locus. We sequenced these amplicons, identified 281 newly acquired spacer sequences, and mapped them to the Retron-Vpa2 operon. From this, we 282 observed a 63 nt region of enriched coverage (normalized to total spacers acquired) that is absent 283 in a dRT control (Fig 4c). Encouragingly, this region covers a long stem loop in the predicted msr-284 msd portion of the hyRNA. We further validated this method by performing the same experiment on 285 Retron-Eco6 and Retron-Sen1, where we similarly saw enriched coverage of their known msd 286 regions (Fig S8a-b). No other peaks of this magnitude were found when mapping coverage across 287 the entire plasmid or E. coli genome (Fig S8c). Thus, there is some ongoing msDNA production from 288 Retron-Vpa2, which does not accumulate to a level we can visualize on a polyacrylamide gel. 289 290 To identify other factors that might influence msDNA production, we proceeded to perform Spacer-291 Seq on dSP (K15*) and dHTH mutants of Retron-Vpa2 (Fig 4d). While mutating the SP did not 292 change the distribution of coverage relative to wild-type, mutating the HTH (in a dSP background to 293 avoid toxicity) enhanced coverage by nearly five-fold. We then looked at msDNA coverage of Vpa2 294 dSP in the presence of different components of the lambda Red operon (Fig 4e-h). Notably, we 295 observed a rightward extension of the spacer distribution peak when Gam is expressed (Fig 4e, h), 296 meaning that this version of the msDNA contains an additional 23 nt at its 5’ end, totaling to a 297 length of 86 nt. This extension also occurs in a ΔrecB background, indicating that it is linked to 298 Gam’s biological function of RecBCD inhibition36,37 (Fig S7a). 299 300 Given the increase in Spacer-seq coverage yielded by the dHTH mutation, we attempted once again 301 to visualize msDNA on a gel under similar conditions. Expressing Retron-Vpa2 with a dHTH 302 mutation revealed a series of bands between 60 nt and 90 nt, which are not present in a dRT 303 control or the background dSP strain (Fig 4i, S7b). Additionally, expressing Retron-Vpa2 dSP in the 304 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint presence of the lambda Red operon yielded a single band between 80 nt and 90 nt, which increases 305 in intensity with a dHTH mutation. Thus, the HTH appears to repress reverse transcription. 306 Sequencing the product with the single band confirmed that it is the same 86 nt product identified 307 in Spacer-seq when Retron-Vpa2 was expressed in the presence of Gam (Fig 4i). We consider this 308 sequence as the full-length msDNA and refer to the 63 nt sequence observed in the absence of Gam 309 as the truncated msDNA. 310 311 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint 312 Figure 4 – msDNA production is repressed until Retron-Vpa2 is triggered. 313 A) Spot assay showing titration of phage lambdavir on Retron-Vpa2 wild-type vs. dRT expressed in E. coli bSLS.114, relative 314 to an empty vector control. Efficiency of plating is quantified as pfu/mL in the adjacent bar graph. B) Schematic of Spacer-315 seq. C) Per base coverage (normalized to total spacers acquired) of spacers aligning to the Retron-Vpa2 operon for the wild-316 type retron compared to a dRT control. Position of aligned spacers is also depicted as raw per base coverage on the predicted 317 secondary structure of the msr-msd region of the hyRNA. Data is a pool of three biological replicates. D) Normalized per base 318 coverage of spacers aligning to the msr-msd region of Retron-Vpa2 for various mutants, compared to the wild-type retron 319 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint from panel (C). Data is a pool of three biological replicates. E-H) Per base coverage (normalized to maximum coverage) of 320 spacers aligning to the msr-msd region of Retron-Vpa2 for a Retron-Vpa2 dSP mutant with Gam, Beta, Exo, or the Red 321 operon, compared to Retron-Vpa2 dSP alone from panel (D). Data is a pool of three biological replicates. I) PAGE analysis for 322 msDNA production of Retron-Vpa2 dSP with various mutants, as well as with co-expression of the Red operon. Retron-Eco6 323 band shown as reference. Ladder on the leftmost lane is single-stranded DNA with increments marked in nucleotides. 324 Complete gel is found in Fig S7b. Position of aligned reads from sequencing the Retron-Vpa2 dSP + Red sample is depicted as 325 raw per base coverage on the predicted structure of the hyRNA msr-msd. 326 327 SP is translated in the presence of triggers and msDNA 328 The final step of Retron-Vpa2 phage defense is release of the toxic SP to arrest growth in the 329 infected host. We next aimed to investigate the mechanism of SP neutralization/activation. Because 330 the RBS and start codon of the SP are sequestered within stem loops of the hyRNA, we hypothesized 331 that the toxin is translationally repressed until the retron is triggered. To measure its production in 332 vivo, we designed a split GFP complementation assay45 where SP abundance is detected through a 333 fluorescence-based readout. In this assay, an optimized superfolded GFP (sfGFP) is split into two 334 fragments which must self-assemble to reconstitute a fluorescent protein. The large fragment 335 contains beta sheets 1-10 (sfGFP1-10), which we express from an independent plasmid. The small 336 fragment contains beta sheet 11 (sfGFP11), which we fuse to the C-terminus of a truncated SP 337 (containing only the first 31 residues) within the Retron-Vpa2 operon. This way, the total length of 338 the hyRNA is preserved and the SP is rendered nontoxic (Fig S10). In E. coli bSLS.114, we co-339 expressed sfGFP1-10 with SP::sfGFP11 in a Δnc-hyRNA mutant of Vpa2 (which we previously saw 340 was toxic with a wild-type SP, thus we expect SP::sfGFP11 to be produced). However, we did not 341 observe fluorescence signal in this condition, nor in a condition where sfGFP11 is not fused to the 342 SP (Fig 5a). We decided to redesign our constructs, this time adding the first 31 residues of the SP 343 to the N-terminus of the sfGFP1-10 as well, with the rationale being that if there existed natural 344 interactions between SP monomers, it would improve the chance of self-assembly. Tagging both 345 sfGFP fragments with part of the SP resulted in a fluorescence signal well over 10000-fold higher 346 than a negative control without sfGFP11 (Fig 5b). 347 348 We used this assay to measure SP production in the context of a Retron-Vpa2 dHTH mutant and 349 observed high fluorescence (Fig 5c), indicating that the SP is translated when the HTH is 350 inactivated. Since this HTH mutant was previously shown to result in the accumulation of 351 detectable msDNA, we next checked whether SP production in the dHTH background requires 352 msDNA by mutating the RT so that no msDNA can be produced. We found that mutating the RT 353 dramatically reduced fluorescence under these conditions, demonstrating that msDNA is necessary 354 for translation of the SP. Meanwhile, co-expression of the system with variants of the lambda Red 355 operon only resulted in fluorescence in the Red Beta + Exo and wild-type Red conditions, 356 recapitulating our earlier trigger assay results which demonstrated the necessity of beta and exo for 357 complete retron activation (Fig 5d-e). Interestingly, the maximum fluorescence achieved when 358 triggering the retron with Red was roughly 10-fold less than in the Δnc-hyRNA and dHTH 359 conditions, but still 1000-fold higher than a control where the system is co-expressed with an 360 empty vector. Thus, the SP is translationally repressed, but this repression is alleviated when co-361 expressing phage triggers, removing the nc-hyRNA, or mutating the HTH, as long as the RT is intact 362 and msDNA can accumulate. 363 364 So far, we have observed parallels between the conditions that lead to msDNA production and those 365 that lead to SP production. We wondered whether the two processes are directly linked, where the 366 msDNA molecule itself is derepressing the translation of the SP. To test this, we co-expressed our 367 split GFP system with an engineered version of Retron-Eco1 that produces either a scrambled 368 mDNA sequence, the truncated version of Retron-Vpa2’s msDNA, or the full Retron-Vpa2 msDNA. 369 We observed clear upregulation of SP production in the condition with full Retron-Vpa2 msDNA 370 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint compared to the scrambled or truncated version (Fig 5f). This shows that abundant Retron-Vpa2 371 msDNA produced in trans from a different retron is sufficient to induce translation of the SP, and 372 that the stem-loop at the 5’ end of the msDNA is essential. 373 374 375 Figure 5 – SP is translated in the presence of triggers and msDNA. 376 A) Measurement of relative fluorescence units (RFU) for E. coli bSLS.114 liquid cultures expressing a split GFP system over 16 377 hrs, with induction at 0 hrs. No fluorescence is observed for sfGFP1-10 co-expressed with either sfGFP11 or SP::sfGFP11 (in 378 the context of Retron-Vpa2 Δnc-hyRNA), compared to a control without sfGFP11. Alphafold3-predicted structures for sfGFP1-379 10 and SP::sfGFP11 are also depicted. B) Measurement of RFU for E. coli bSLS.114 liquid cultures expressing a split GFP 380 system over 16 hrs, with induction at 0 hrs. Solid lines indicate the mean and error bands indicate the standard deviation of 381 three biological replicates. Strong fluorescence is observed for SP::sfGFP1-10 co-expressed with SP::sfGFP11 (in the context of 382 Retron-Vpa2 Δnc-hyRNA), compared to the systems tested in panel (A). Alphafold3-predicted protein structure shows 383 reconstitution of SP::sfGFP1-10 and SP::sfGFP11. C) Measurement of RFU for E. coli bSLS.114 liquid cultures expressing a split 384 GFP system in the context of Retron-Vpa2 Δnc-hyRNA or dHTH mutants over 18 hrs, with induction at 2 hrs (vertical dotted 385 line), compared to a control system lacking sfGFP11. Solid lines indicate the mean and error bands indicate the standard 386 deviation of three biological replicates. D) Measurement of RFU for E. coli bSLS.114 liquid cultures expressing a split GFP 387 system in the context of Retron-Vpa2 over 18 hrs, with induction at 2 hrs (vertical dotted line), co-expressed with either Gam, 388 Beta, or Exo, compared to co-expression with an empty vector. Solid lines indicate the mean and error bands indicate the 389 standard deviation of three biological replicates. E) Measurement of RFU for E. coli bSLS.114 liquid cultures expressing a split 390 GFP system in the context of Retron-Vpa2 over 18 hrs, with induction at 2 hrs (vertical dotted line), co-expressed with 391 variants of the Red operon, compared to co-expression with an empty vector. Solid lines indicate the mean and error bands 392 indicate the standard deviation of three biological replicates. F) Measurement of RFU for E. coli bSLS.114 liquid cultures 393 expressing a split GFP system in the context of Retron-Vpa2 over 18 hrs, with induction at 2 hrs (vertical dotted line), co-394 expressed with Retron-Eco1 engineered to produce either scrambled msDNA, truncated Retron-Vpa2 msDNA, or full-length 395 Retron-Vpa2 msDNA. Solid lines indicate the mean and error bands indicate the standard deviation of three biological 396 replicates. 397 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint

398 In this work, we uncovered a new mechanism of phage defense by a Type VI retron from Vibrio 399 parahaemolyticus, which we named Retron-Vpa2 (Fig 6). This system defends against a broad range 400 of phages, from which we characterized escapees of two phages and identified their recombination-401 associated trigger genes. This retron, and likely other members of its type, differs from others 402 through its inverted triggering mechanism, where the msDNA acts as the agent for toxin 403 proliferation, rather than inhibition. We determined that Retron-Vpa2 msDNA, which is 404 undetectable at baseline, accumulates 1) in the presence of phage infection, 2) with the expression 405 of phage triggers, or 3) if the HTH protein is mutated. The accumulation of msDNA is necessary for 406 translation of the toxic SP, which is transcribed as part of the hyRNA, but translationally repressed 407 in the absence of msDNA. Finally, the msDNA activates translation of the SP even if produced from 408 another retron in trans within the same cell. 409 410 411 Figure 6 – Model of Retron-Vpa2 phage defense mechanism 412 The core of this system is a complex comprising the RT, HTH, and hyRNA. The hyRNA contains the template for reverse 413 transcription and the coding sequence for SP translation. In its untriggered state, reverse transcription is inhibited by the 414 HTH and the SP is translationally repressed. Phage infection and expression of trigger proteins, such as the Red system of 415 lambdavir, induce accumulation of msDNA, which leads to translation of toxic SP. 416 417 Given the complexity of this mechanism, there remain several aspects of it that will prompt future 418 investigation. For example, we identify in this study that recombination-associated genes in phage 419 lambda and phage Stevie_ev116 trigger Retron-Vpa2. Still, it is uncertain exactly how this occurs. 420 We hypothesize that the triggers stabilize the msDNA, such that accumulation of msDNA leads to 421 translation of SP. This is supported by the fact that lambda beta and Stevie_ev116 rec are single-422 stranded annealing proteins (SSAPs) that bind ssDNA, while exo also interacts with DNA35,39. 423 Furthermore, it is encouraging that we can identify homologs of these proteins in the majority of 424 phages that Retron-Vpa2 defends against. In the case of lambda phage, the presence of gam would 425 also promote msDNA accumulation by inhibiting E. coli RecBCD. We think RecBCD, and potentially 426 prophage-associated nucleases, mediate turnover of msDNA from basal levels of reverse 427 transcription. It is also possible that repression of reverse transcription is alleviated by interactions 428 of phage triggers with the HTH. 429 430 Through our efforts to understand the role of reverse transcription in this system, we establish 431 Spacer-seq as a discovery-based tool for identifying retron msDNA. Here, it provided evidence that 432 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint reverse transcription does occur in the absence of triggers, but due to reduced efficiency and/or 433 high turnover rate, msDNA does not accumulate in the cell, and thus cannot be detected through 434 PAGE analysis. We speculate that much of the msDNA captured by the integrases are degradation 435 products of RecBCD and host nucleases, a known source of prespacers for CRISPR adaptation46,47. 436 Furthermore, the appearance of the 23 nt 5’ region in the msDNA when RecBCD is inhibited could 437 indicate that the site is normally shielded from the CRISPR integrases by this complex or some 438 other protein, or that it is not synthesized at all. 439 440 Perhaps the most intriguing aspect of this retron is its repression of the SP, which requires msDNA 441 production for translation. We think the underlying mechanism involves a type of riboregulator. 442 Notably, the SP RBS and start codon are located within stem loops in the hyRNA, which would 443 sequester them from ribosomal access. In addition, we notice a set of 7 nt direct repeats in the 444 hyRNA that are highly conserved in sequence and structure across the Type VI retrons (Fig S4a-b). 445 One repeat is in a loop region of the msd and the other is in a loop region of the predicted 446 riboregulator. When reverse transcribed, the repeat in the msd would be complementary to that in 447 the riboregulator, potentially leading to hybridization and subsequent release of the SP RBS and 448 start codon. Importantly, this 7 nt sequence is on the 5’ end of the msDNA, which we demonstrated 449 is specifically necessary for SP translation (Fig 5f). Future work will seek to further understand the 450 specific sequence and structural elements necessary for the msDNA to activate SP translation. 451 452 Finally, although the mechanism of SP toxicity remains elusive, we have made some important 453 insights into its biology. We demonstrated that maintaining hydrophobicity of its C-terminal alpha 454 helix is necessary for toxicity (Fig 3b). Furthermore, our results showing augmented fluorescence of 455 split GFP when both fragments are tagged with the first 31 residues of the SP suggest that the N-456 terminus of the protein mediates oligomerization or colocalization (Fig 5b). 457 458 Overall, our study into the biology and mechanism of Retron-Vpa2 reveals a novel retron system 459 where the toxic effector is translationally repressed within a hyRNA transcript and requires phage-460 induced accumulation of msDNA for expression. These insights will inform future studies on the 461 previously uncharacterized Type VI group of retrons while introducing a new perspective in our 462 understanding of Retron-mediated phage defense. 463 464 465 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint

466 467 Bacterial strains and growth conditions 468 The E. coli strains used in this study were NEB 5-alpha (NEB C2987) for cloning, MG1655 for the large panel 469 phage defense assays, MG1655-DE3 for the pull-down assays (see Pull-down assays for details on strain 470 construction), and bSLS.114 (derivative of BL21-AI lacking Retron-Eco1)48 for all other experiments. A ΔrecB 471 derivative of bSLS.114 (bKAZ020) was constructed using lambda Red recombineering49 for the experiment 472 shown in Supplementary Figure 9a. 473 474 For experiments with liquid cultures, bacteria were grown in lysogeny broth (LB) at 37°C, shaking at 250 475 rpm. Where appropriate, the LB was supplemented with antibiotics and inducers at the following working 476 concentrations: 200 μg/mL ampicillin, 100 μg/mL carbenicillin (GoldBio C-103), 25 μg/mL chloramphenicol 477 (GoldBio C-105), 35 μg/mL kanamycin (GoldBio K-120), 25 μg/mL spectinomycin (GoldBio S-140), 1 mM 478 isopropyl β-d-1-thiogalactopyranoside (IPTG, GoldBio I2481C), and 2 mg/mL L-arabinose (GoldBio A-300). 479 480 Phage strains and propagation 481 Experiments with phage lambda used a strictly lytic strain (lambdavir) that was generously provided by 482 Luciano Marraffini. To propagate the phage, a saturated culture of E. coli bSLS.114 was diluted 1:100 in 3 mL 483 LB supplemented with 0.1 mM MnCl2 and 5 mM MgCl2 (MMB) and grown to an OD600 of 0.25. At this point, the 484 culture was infected with phage at a MOI of ~0.2 and allowed to grow for an additional 16 hrs. After this time, 485 the culture was centrifuged for 10 min at 3,434g and the supernatant containing the phage lysate was filtered 486 through a 0.2-μm filter. The titer of the lysate was determined via plaque assay (see Plaque assays). Lysates 487 were stored at 4°C for further use. 488 489 For other phages used in the large panel defense screen, phages were isolated by using sterile inoculation 490 loop to streak from the original stock (obtained from various sources, see Supplementary Table 3) onto a 491 lawn of E. coli MG1655 on LB agar. The plate was incubated overnight at 37°C. Then, singe plaques were 492 picked and used to infect a 3 mL LB culture of E. coli MG1655 at an OD600 of ~0.2. The culture was then grown 493 for ~5 hrs at 37°C. After this time, the culture was clarified via centrifugation and the supernatant containing 494 the phage lysate was filtered through a 0.45 μm filter. 495 496 Plasmid construction 497 Plasmids containing Retron-Vpa2 (pMRM27) and Retron-Eco12 (pMRM38) were designed to incorporate the 498 wild-type operon (including the native promoter) into a high-copy pET-21 backbone under a T7/lac inducible 499 promoter. These constructs were synthesized by Twist Bioscience. Plasmids containing other retrons tested 500 for msDNA expression were taken from the previous retron census manuscript15. 501 502 For plasmids containing lambdavir trigger genes, the wild-type trigger genes were PCR amplified from phage 503 lysate and cloned using Gibson assembly (NEB E2621) into a pBR322 backbone under an arabinose-inducible 504 promoter (araBAD). Plasmids containing mutations to the trigger genes were cloned via Q5 site-directed 505 mutagenesis (NEB E0552). 506 507 Plasmids containing mutations to Retron-Vpa2 (e.g. Δnc-hyRNA, dRT, dHTH, dSP, dSP) or introducing protein 508 tags to retron components (e.g. HTH-HA, RT-FLAG) were similarly cloned using site-directed mutagenesis. 509 510 For Spacer-seq, the plasmid expressing the Cas1-Cas2 genes under a lac promoter (pSCL565) in a pCDF 511 backbone was taken directly from a previous manuscript50. 512 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint 513 For the split GFP complementation assay, optimized sfGFP1-10 (sequence taken from pAGM22082 in 514 Püllmann et al.51, Addgene #153515) was synthesized via Integrated DNA Technologies (IDT) and cloned into 515 a p15a backbone under a lac promoter using Gibson assembly. Site-directed mutagenesis was used to add the 516 first 31 amino acids of the SP to the N-terminus of this plasmid (SP-sfGFP1-10). Site-directed mutagenesis 517 was also used to fuse sfGFP11 (5’-RDHMVLHEYVNAAGIT-3’) with the SP gene (SP::sfGFP11) on a Retron-518 Vpa2-expressing plasmid. To clone the plasmid used for Retron-Eco1 expression of scrambled msDNA, the 519 retron operon under a T7/lac inducible promoter was amplified from a Retron-Eco1-based editor plasmid 520 (operon is rearranged with the RT preceding the ncRNA and the effector deleted) from a previous 521 manuscript52 and cloned into a pCDF backbone using Gibson assembly. Site-directed mutagenesis was used 522 on this plasmid to clone versions that expressed either the truncated or full version of the Retron-Vpa2 523 msDNA. 524 525 All oligos used in this study were synthesized by IDT. Details for all plasmids are listed in Supplementary 526 Table 4. 527 528 Phylogenetic analysis 529 Homologs of Retron-Vpa2 were identified using the RT amino acid sequence (WP_069548583.1) as a BLASTP 530 search query of the NR protein database (max target sequences = 5000). The resulting sequences were 531 aligned with MAFFT53, and a Maximum-Likelihood phylogenetic reconstruction was made with 532 VeryFastTree54 with default parameters. Classification of the different retrons was done by annotating the 533 genomic neighborhoods with PADLOC55 and used to color the tree. 534 535 msDNA expression and gel analysis 536 Plasmids containing retrons were transformed into bSLS.114. Individual colonies were picked and grown in 3 537 mL LB + carbenicillin at 37°C until saturation. Each culture was diluted 1:100 in 25 mL LB + carbenicillin, 538 grown for ~2 hrs at 37°C (until OD600 reaches ~0.5), then induced with IPTG and arabinose, after which it was 539 grown for an additional 5 hrs. 540 541 DNA was isolated from the cells using the Qiagen Plasmid Plus Midi Kit (Qiagen #12945) and eluted in 150 μL 542 of molecular biology grade water. The eluted DNA was mixed with Novex TBE-Urea Sample Buffer (Invitrogen 543 LC6876) to reach a sample buffer concentration of 1X. This was then heated to 98 C for >=5 min and 15 µL 544 was loaded onto a Novex 15% TBE-Urea gel (Invitrogen EC6885). ss20 DNA Ladder (Simplex Sciences) was 545 also loaded in a separate well as a marker. The gel was run at 200 V for 45 min in preheated (>75°C) TBE 546 running buffer, then stained with SYBR Gold Nucleic Acid Gel Stain (Invitrogen S11494) and imaged on a 547 ChemiDoc Imaging System (Bio-Rad). 548 549 Plaque assays 550 Small-drop plaque assays were performed similarly to Mazzocco et al.56, which was used to both titer phage 551 stocks and test for defense. For titering, plaque assays were performed with E. coli bSLS.114. For defense 552 experiments with lambdavir, a plasmid containing Retron-Vpa2 (or an empty vector control) was first 553 transformed into E. coli bSLS.114. Individual colonies were picked and grown in 3 mL LB + carbenicillin until 554 saturation. This culture was then diluted 1:100 in MMB + carbenicillin and grown for 2-3 hrs at 37°C. For each 555 plaque assay performed, 200 μL of the passaged culture was mixed with 2 mL of MMB top agar (MMB with 556 0.75% agar) and poured onto a single well of a rectangular 4-well MMB agar plate (MMB with 1.5% agar). 557 After the top agar has solidified, tenfold serial dilutions in MMB of a phage lysate was spotted onto the plate 558 with 2 μL spots in technical triplicates. After the spots had completely dried, plates placed in an incubator at 559 37°C overnight. Plaque-forming units (pfus) were quantified using the formula: pfu count * dilution factor / 560 mL of lysate spotted = pfu/mL. 561 562 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint For the large panel defense assay, the protocol is similar to above, but plasmids were instead transformed 563 into E. coli MG1655 and colonies were grown in LB + ampicillin. Top agar was spiked with 2% of bacterial 564 culture directly from a saturated overnight and phage lysate was spotted onto the plate with 5 μL spots in 565 technical triplicates. 566 567 Phage infection growth curves 568 A plasmid containing Retron-Vpa2 (or an empty vector control) was transformed into E. coli MG1655. 569 Individual colonies were picked and used to start an overnight culture in LB + ampicillin. This saturated 570 culture was then diluted 1:100 into fresh LB + ampicillin and grown to an OD600 of ~0.2. 180 μL of this culture 571 was transferred to a 96-well microtiter plate and infected with phage. All wells received 20 μL of phage lysate 572 that was pre-diluted to various concentrations to achieve final multiplicities of infection (MOIs) of 0, 0.01, 0.1, 573 1, and 10. The plate was incubated in a plate reader (BioTek Synergy H1 Multimode Reader) at 37°C with 574 shaking at 200 rpm for 6 hrs, with OD600 measurements taken every 5 min. 575 576 Phage escapee isolation and sequencing 577 578 Phage lambdavir 579 580 Isolation of lambdavir escapees generally follows the protocol in Millman et al.2. 20 uL of lambdavir lysate (at a 581 titer of ~108 pfu/mL) was used to infect 200 µL of Retron-Vpa2-expressing E. coli bSLS.114 culture at an 582 OD600 of 0.3, then incubated for 15 min at room temperature. The infected culture was then mixed into 2 mL 583 of MMB top agar, poured onto a single well of a rectangular 4-well MMB agar plate, and grown overnight at 584 37°C. The next morning, individual plaques were picked and resuspended in 90 μL of phage buffer (50 mM 585 Tris pH 7.4, 100 mM MgCl2, 10 mM NaCl). Plaques were left in the buffer for 1 hr at room temperature, with 586 occasional vortexing to release the phage from the agar. This lysate was used to infect a 3 mL MMB + 587 carbenicillin culture of Retron-Vpa2-expressing E. coli bSLS.114 at an OD600 of ~0.2, which was grown 588 overnight at 37°C. The next morning, the culture was centrifuged at 4347g for 10 min and filtered through a 589 0.2-μm filter. Phage genomic DNA was isolated from 1 mL of this lysate using Norgen’s Phage DNA Isolation 590 Kit (Norgen Biotek 46800). 591 592 The purified DNA was prepped for nanopore sequencing on a MinION device using the standard Oxford 593 Nanopore Technologies (ONT) workflow for the Ligation Sequencing Kit (SQK-LSK109) with the Native 594 Barcoding Expansion (EXP-NBD196) on flow cell version R9.4.1 (FLO-MIN106D). The ancestral phage strain 595 was also sequenced alongside the escapees. Sequencing data was base called and aligned to a reference 596 genome (accession J02459.1) using Guppy (v6.1.7 from ONT). Aligned reads were used to generate a 597 consensus sequence for each phage using SAMtools57 (v1.18). Escapee-specific mutations were identified 598 through alignment of escapee phage genomes with the ancestral phage genome using Geneious. 599 600 Phage Stevie_ev116 601 602 To isolate Stevie_ev116 escapees, phage lysate was streaked onto a top agar lawn of 2% Retron-Vpa2-603 expressing E. coli MG1655 supplemented with ampicillin. After overnight incubation at 37°C, resulting 604 plaques were picked and used to infect a LB + ampicillin culture of Retron-Vpa2-expressing E. coli MG1655 at 605 an OD600 of ~0.2. The culture was incubated overnight at 37°C to propagate the phage, and phage lysates 606 were harvested the next morning. This process was repeated for three additional rounds to enrich the phage. 607 608 Phage DNA was purified from filtered lysates using a combination of enzymatic digestion and silica column-609 based extraction. To degrade contaminating bacterial nucleic acids, 450 μL of filtered phage lysate was 610 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint incubated with 50 μL of DNase I 10× buffer, 1 μL of DNase I (1 U/μL), and 1 μL of RNase A. Samples were 611 incubated at 37 °C for 1.5 hours without shaking. Enzymes were subsequently inactivated by adding 20 μL of 612 0.5 M EDTA. To digest remaining proteins, 1.25 μL of proteinase K was added, and samples were incubated at 613 56 °C for 1.5 hours without shaking. DNA was extracted using the DNeasy Blood & Tissue Kit (QIAGEN) 614 following a modified protocol. 450 μL of Buffer AL was added to each sample and mixed thoroughly by 615 vortexing, followed by incubation at 56 °C for 10 minutes. Next, 450 μL of 96-100% ethanol was added, and 616 the mixture was vortexed to ensure homogeneity. The entire sample was then transferred to a DNeasy Mini 617 spin column placed in a 2 mL collection tube and centrifuged at approximately 8,000 rpm for 1 minute. The 618 spin column was placed in a new collection tube, 500 μL of Buffer AW1 was added, and centrifugation was 619 repeated at 8,000 rpm for 1 minute. With the column in a new collection tube, another 500 μL of Buffer AW2 620 was added and samples were centrifuged at 8,000 rpm for 3 minutes. The spin column was then transferred 621 to a clean microcentrifuge tube for DNA elution. DNA was eluted by adding 30 μL of nuclease-free water 622 directly to the center of the membrane, incubating at room temperature for 1 minute, and centrifuging at 623 8,000 rpm for 1 minute. A second elution was performed by adding an additional 20 μL of nuclease-free 624 water, incubating for 1 minute, and centrifuging as before. Purified DNA was stored at 4 °C until use. 625 626 The extracted DNA was quantified using the Qubit dsDNA High Sensitivity Assay Kit (Thermo Fisher 627 Scientific). Sequencing libraries were prepared using the Nextera XT DNA Library Preparation Kit (Illumina), 628 following the manufacturer's protocol with minor modifications. 1 ng of input DNA per sample was subjected 629 to enzymatic tagmentation, which simultaneously fragmented the DNA and tagged it with adapter sequences. 630 Tagmented DNA was then amplified via limited-cycle PCR using Nextera XT barcoded index primers. Post-631 PCR clean-up was performed using AMPure XP magnetic beads (Beckman Coulter). Final library 632 concentrations were measured using Qubit and normalized to equimolar concentrations before pooling. 633 Libraries were submitted to Novogene (Beijing, China) for high-throughput sequencing. Sequencing was 634 performed on an Illumina platform using paired-end 2×150 bp chemistry. Adapter trimming and removal of 635 low-quality base calls were performed by Novogene using their in-house quality control pipeline, and 636 demultiplexed reads were returned in FASTQ format for downstream analysis. 637 638 Trigger assays 639 A plasmid containing the retron being tested (or an empty vector control) was co-transformed with a plasmid 640 containing the trigger gene into E. coli bSLS.114. Individual colonies were picked and grown until saturation in 641 3 mL LB + carbenicillin + chloramphenicol at 37°C. Cultures were then diluted 1:100 in 200 uL of fresh LB + 642 carbenicillin + chloramphenicol in a transparent 96-well microtiter plate ( Corning 3596). The plate was 643 covered with an air-permeable seal (Breathe-Easy, Diversified Biotech BEM-1) and incubated in a microplate 644 reader (Molecular Devices SpectraMax i3) with shaking at 37°C for 16 hrs with OD600 measured every 20 min. 645 At the 2 hr timepoint, IPTG and arabinose inducers were added to all cultures. Raw OD600 values were adjusted 646 by first multiplying by the empirically determined correction factor of 3.4967 , then subtracting the 0 hr 647 measurement for each sample. 648 649 RNA-seq 650 RNA-seq was performed similar to described in Millman et al.2, here using a Retron-Vpa2-expressing strain of 651 E. coli bSLS.114. A saturated culture of this strain was diluted 1:100 in 5 mL LB + carbenicillin and grown at 652 37°C to an OD600 of 0.6. The culture was centrifuged at 4347g for 10 min at 4°C. The supernatant was discarded, 653 and the pellet was retained. The pellet was then treated with 100 μL of 2 mg/mL lysozyme solution (in 10 mM 654 Tris, 1 mM EDTA, pH 8.0) and incubated for 15 min at 37°C in a 1.5 mL tube. At this time, 1 mL of TRI-reagent 655 (Sigma-Aldrich 93289) was added and mixed by pipetting, then 200 μL of chloroform was added and mixed by 656 pipetting. The sample was incubated at room temperature for 5 min for phase separation. The sample was then 657 centrifuged at 18,213g for 30 min at 4°C, after which the upper phase was carefully removed and transferred 658 to a fresh tube. This was mixed with 750 μL of cold isopropanol, sodium acetate (NaAOc) to a final concentration 659 of 0.3 M, and linear acrylamide (Invitrogen AM9520) to a final concentration of 10 μg/mL. The sample was 660 frozen either at - 20°C overnight or - 80°C for 1 hr , then centrifuged at 18,213g for 30 min at 4 °C. The 661 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint supernatant was discarded, and the pellet was washed twice with 500 μL of freshly prepared, cold 70% ethanol. 662 For each wash, the ethanol was carefully added without disturbing the pellet, then centrifuged for 15 min at 663 18,213g at 4°C, and then the supernatant was discarded. Pellet was air-dried for 5 min, then resuspended in 50 664 μL of molecular biology grade water and incubated at 56 °C for 10 min to elute . RNA concentration was 665 quantified via Nanodrop (Thermo Scientific) and Qubit fluorometer (Invitrogen). 666 667 15 μg of the eluted RNA was treated with TURBO DNase (Invitrogen AM2238). Then, TRIzol precipitation and 668 washes from the previous paragraph was performed once again on the treated RNA. This time the sample was 669 eluted in 15-20 μL of molecular biology grade water and quantified via TapeStation (Agilent) to ensure a RIN > 670 5. Ribosomal RNA depletion was performed on 1000 ng of the sample using the Illumina Ribo -Zero rRNA 671 Removal Kit (Illumina 20040526) following the kit protocol and using RNAClean XP beads (Beckman Coulter 672 A63987) for the cleanup steps. Synthesis of cDNA and Illumina TruSeq adapter ligation was performed on 100 673 ng of depleted RNA using the NEBNext Ultra II Directional RNA Library Prep Kit (NEB E7760) following the kit 674 protocol and using AMPure XP beads (Beckman Coulter A63880) for the cleanup steps . The prepared RNA 675 library was sequenced on an Illumina NextSeq 2000 system. Sequencing reads were aligned to the retron 676 operon using Geneious. 677 678 hyRNA secondary structure prediction 679 The hyRNA sequence from RNA-seq was input into RNAfold58, which predicts secondary structure by 680 computing minimum free energy, and visualized with forna59. The resulting structure (see Fig 3b) was 681 independently supported by covariance modeling (see below and Fig S3a). 682 683 hyRNA covariance modeling 684 Homologs of Retron-Vpa2 were identified using the RT amino acid sequence (WP_069548583.1) as a BLASTP 685 search query of the NR protein database (max target sequences = 100). Flanking nucleotide sequences 1 kb 686 upstream and downstream of RT genes were extracted, clustered at 99.9% sequence identity to remove 687 replicates with CD-HIT60 (v4.8.1), and aligned with MAFFT53 (v7.525). The resulting alignment was trimmed 688 at the 5′ and 3′ ends to the boundaries of the Retron-Vpa2 hyRNA as determined by RNA-seq. These putative 689 hyRNA homologs were clustered at 95% sequence identity with CD-HIT and realigned with mLocARNA61 690 (v2.0.1). The resulting structure-based multiple sequence alignment was used to build and calibrate a 691 covariance model (CM) of the hyRNA using the Infernal suite62 (v1.1.5). 692 693 To identify other hyRNA homologs across additional Type VI retron loci using this CM, a database of diverse 694 Type VI retron RTs was generated as previously described (see Phylogenetic analysis). Briefly, a broad RT 695 Pfam profile (PF00078) was searched against the ClusteredNR protein database. Hits were retrieved and 696 aligned against the HMM profile using hmmalign –trim from the HMMER suite63, and all sequences in the 697 alignment shorter than 150 amino acids were removed with SeqKit64. The alignment of remaining RTs (n = 698 306,351 members) was used to build a tree with VeryFastTree54. Tree members were annotated with the best 699 hit from an MMseqs2 easy-search65 against a custom reference database of known RTs, revealing a clade of 700 Type VI retron RTs (n = 779 members) that contained Retron-Vpa2. 701 702 The 1-kb nucleotide flanking regions of these Type VI retron RTs were extracted and clustered at 95% 703 sequence identity using CD-HIT. The CMsearch function of Infernal was then used to scan the previously built 704 CM through these sequences to identify additional hyRNA homologs, and the final hits (n = 81 Type VI retron 705 loci, including Retron-Vpa2) were evaluated for statistically significant co-varying base pairs with R-scape66 706 (v1.4.0) at a default E-value threshold of 0.05 (Fig S3a). To visualize conserved RNA motifs, a sequence logo 707 was generated from the structure-based multiple sequence alignment of these hyRNA hits using WebLogo67 708 (v3.7.9) (Fig S3b). 709 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint 710 Toxicity assays 711 A plasmid containing wild-type or mutant Retron-Vpa2 (or an empty vector control) was transformed into E. 712 coli bSLS.114. Individual colonies were picked and grown until saturation in 3 mL LB + carbenicillin at 37 °C. 713 Cultures were then diluted 1:100 in 200 uL of fresh LB + carbenicillin in a transparent 96-well microtiter plate 714 (Corning 3596). The plate was covered with an air -permeable seal (Breathe-Easy, Diversified Biotech BEM-1) 715 and incubated in a microplate reader (Tecan Infinite 200 PRO) with shaking at 37° C for 16 hrs with OD 600 716 measured every 10 min. At the 2 hr timepoint, IPTG and arabinose inducers were added to all cultures. Raw 717 OD600 values were adjusted by first multiplying by the empirically determined correction factor of 718 6.230707455, then subtracting the 0 hr measurement for each sample. 719 720 Pull-down assays 721 Three plasmids were constructed containing the Retron-Vpa2 operon with either a C-terminal Flag tag on the 722 RT, a C-terminal HA tag on the HTH, or both tags. For assays using RT as bait, a non-tagged RT plasmid was 723 used as a negative control. Similarly, for assays using HTH as bait, a non-tagged HTH plasmid served as the 724 negative control. A modified E. coli MG1655 strain lacking bacterial defense systems and stably integrating 725 the λDE3 prophage (carrying an IPTG-inducible T7 RNA polymerase gene) was generated using the λDE3 726 Lysogenization Kit (Novagen). This strain is referred to as MG1655-DE3. 727 728 Plasmids were transformed into E. coli MG1655-DE3 and plated on LB agar supplemented with 100 µg/mL 729 ampicillin. Liquid cultures were grown at 37°C with shaking until reaching an OD₆₀₀ of 0.6. Expression of the 730 operon was induced with 0.2 mM IPTG for 4 hours at 37°C. After induction, OD₆₀₀ was measured and used to 731 normalize sample input across conditions. According to the OD600, cells were harvested at >3000 × g for 10 732 minutes at 4°C and resuspended in 0.8–2 mL of lysis buffer (20 mM Tris-HCl pH 7.5, 150 mM KCl, 1 mM 733 MgCl₂, 1 mM TCEP, 0.1% Triton X-100) to ensure that the total cell concentration is similar in all tubes. Lysis 734 was performed by sonication (3 pulses of 10 seconds at 25% amplitude). Soluble proteins were separated by 735 centrifugation at >16,000 × g for 30 minutes at 4°C. The clarified lysate (800 µL) was incubated with 50 µL of 736 anti-FLAG M2 affinity gel (Sigma-Aldrich, A2220) or anti-HA agarose beads (Thermo, 26181) prewashed with 737 lysis buffer, for 30 minutes at 4°C with gentle agitation. Beads were pelleted at 3000 × g for 3 minutes, and 738 the supernatant was removed. Bound complexes were washed four times with lysis buffer lacking Triton X-739 100. Beads were resuspended in ~60 µL of lysis buffer without triton X-100 and divided into three aliquots 740 (~20 µL each) for protein, RNA, and DNA analysis. 741 742 For Protein Analysis by SDS-PAGE and Western Blot, the samples were mixed with NuPAGE™ LDS Sample 743 Buffer (Invitrogen, NP0007) and denatured by boiling for 5–10 minutes at 95°C. Proteins were separated by 744 SDS-PAGE (NuPAGE™ Bis-Tris Mini Protein Gels 4 12–% Acrylamide, Invitrogen, 12030166) and transferred 745 to a nitrocellulose membrane (iBlot™ 3 Transfer Stacks, Invitrogen, IB33001). Entire pull-down samples (40 746 µL) were loaded, as well as 30 µg of total protein from the input lysate to assess expression levels. The 747 membranes were then blocked for 1 hour at room temperature in TBST (Tris-buffered saline with 0.1% 748 Tween-20) containing 0.5% powdered milk. Detection was performed using HRP-conjugated anti-HA (Anti-749 HA-Peroxidase, High Affinity, Roche) or anti-FLAG M" antibodies (Sigma, A8592), diluted 1:1000 in TBST with 750 0.5% powdered milk. Membranes were washed three times for 10 minutes in TBST without milk, and 751 chemiluminescence was visualized using SuperSignal™ West Pico PLUS Chemiluminescent Substrate (Thermo 752 Scientific, 34577). 753 754 For RNA analysis, samples were resuspended in urea loading dye (4M Urea, 1 mM Tris pH 7.5, 5 mM EDTA), 755 denatured by boiling for 10 minutes at 95°C, and cooled on ice for 2 minutes. DNA samples were pretreated 756 with ~ 100 μg/ml of RNAse A for 10 minutes at 37°C before denaturation. RNA and DNA were resolved on 757 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint 10% TBE-Urea gels (Invitrogen, EC68755BOX) run at ~180 V for 70–90 minutes. RNA and DNA was 758 visualized by staining with GelRed Nucleic Acid Stain (10000X DMSO, Millipore, SCT122) for 10 minutes prior 759 to imaging. 760 761 Extraction of RNA/DNA from pull-down samples for sequencing 762 Pull-down samples were prepared as described above. The entire sample (20 µL of beads) was resuspended 763 in 300 µL of nuclease-free ddH₂O and denatured by boiling at 95 °C for 5–10 minutes to release proteins and 764 nucleic acids from the beads. RNA was then purified using the phenol:chloroform extraction method, followed 765 by precipitation with sodium acetate and ethanol. Purified RNA was directly prepped for sequencing using 766 the NEBNext Ultra II Directional RNA Library Prep Kit (NEB E7760) and sequenced on an Illumina NextSeq 767 2000 system. Sequencing reads were aligned to the reference operon using Geneious. 768 769 Spacer-seq 770 Spacer-seq experiments were conducted similarly to the naïve CRISPR -Cas adaptation experiments in a 771 previous manuscript50. A plasmid containing wild -type or mutant Retron-Vpa2 was co -transformed with the 772 Cas1-Cas2-expressing plasmid into E. coli bSLS.114. For experiments involving trigger genes, a third plasmid 773 containing the trigger gene was also transformed. Three colonies (biological replicates) for each sample were 774 picked and grown until saturation in 0.5 mL LB + carbenicillin + spectinomycin (+ chloramphenicol when the 775 trigger-containing plasmid is present) in a 96-well deep well plate (Thermo Scientific 260251) covered with 776 an air -permeable seal (Breathe-Easier, Diversified Biotech BERM-2000) at 37°C with shaking at 1000 rpm. 777 Cultures were then diluted 1:100 in 0.5 mL LB + antibiotics + IPTG + arabinose in a new deep well plate and 778 grown for 24 hrs at 37°C. Cells were then harvested by boiling 25 μL of culture with 25 μL of molecular biology 779 grade water at 95°C for 10 min. 0.5 μL of each boiled lysate was then used as template for PCR amplification of 780 the CRISPR array . Specifically, the primers used in this PCR (Supplementary Table 5) amplified the region 781 between the leader sequence of the array and the second pre -existing spacer. The primers also introduce d 782 Illumina TruSeq adapters to the amplicons . Three versions of the forward primer were used , each with a 783 different barcode assigned to a different biological replicate, enabling pooling of replicates for each sample 784 prior to sequencing. Amplicons were then prepared for Illumina sequencing using standard workflows and 785 sequenced on a NextSeq 2000 system. 786 787 To analyze sequencing data, Illumina FASTQ files containing sequencing reads for each sample were 788 demultiplexed using the custom barcodes introduced to each biological replicate. Demultiplexed files were then 789 processed using custom code ( https://github.com/Shipman-Lab/Spacer-Seq) developed in a previous 790 manuscript33, which handles read trimming and extraction of spacer sequences. From this, a FASTA file of all 791 newly acquired spacers was obtained for each biological replicate of each sample. These were again pooled into 792 a single FASTA file for each sample and aligned using Bowtie68 (v1.3.1) to a dictionary of references containing 793 the Retron-expressing plasmid, the Cas1 -Cas2-expressing plasmid, and the E. coli BL21-AI genome . Reads 794 aligning to multiple references were discarded. Coverage per nucleotide position was determined from aligned 795 reads using SAMtools57 (v1.18) and normalized either to the total number of newly acquired spacers for a given 796 sample, or to the maximum coverage observed in the Retron-expressing plasmid. Data analysis was performed 797 on Jupyter Lab69 (v4.0.7). Visualization of spacer coverage on the Retron-Vpa2 hyRNA secondary structure was 798 performed using VARNA 70 (v3-93). Visualization of spacer coverage across the full plasmid and genome as 799 shown in Fig S6c was performed using Circos71 (v0.69-9). 800 801 Direct msDNA sequencing 802 Following a protocol from a previous publication15, 79 μL of eluted DNA from a bacterial culture midiprep (see 803 msDNA expression and gel analysis) was treated with 10 μL of DBR1 (50 ng/μL in-house prep, see previous 804 manuscript23) and RNaseH (NEB M0297) in 10 μL of rCutSmart buffer (NEB B6004) and molecular biology 805 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint grade water to reach a total reaction volume of 100 μL. The reaction was incubated at 37 °C for 30 min. Then, 806 the product was cleaned up with Zymo’s ssDNA/RNA Clean & Concentrator kit (Zymo Research D7010 ) and 807 eluted in 15 μL of molecular biology grade water. 808 809 Poly(A)-tailing of cleaned up msDNA was performed by mixing 12.25 μL of molecular biology grade water, 2.5 810 μL of terminal transferase reaction buffer (NEB B0315), 0.25 μL dATP (of a 10 mM stock diluted from NEB 811 N0440), 3 μL of terminal transferase (TdT, NEB M0315), and 10 μL of msDNA. Reaction was incubated at room 812 temperature for exactly 60 sec and then stopped by heating to 70°C for 5 min (~25 adenosines should have 813 been added in this time). 814 815 After this, complem entary strand synthesis was performed using a poly (T) primer which also contains an 816 Illumina TruSeq adapter ( Supplementary Table 5) . The reaction was setup as follows: 28.9 μL of molecular 817 biology grade water, 8 μL of TdT reaction from the previous step, 5 μL of NEBuffer 2 (NEB B7002), 0.1 μL of the 818 poly(T) primer (of a 100 uM stock from IDT), and 5 μL of dNTP mix (of a 10 mM stock from NEB N0447) were 819 mixed and heated to 80°C, then allowed to cool to room temperature (facilitates primer annealing). Then, 3 μL 820 (15 units) of DNA Polymerase I, Large (Klenow) Fragment (NEB M0210) was added, and the reaction was 821 incubated at 37°C for 30 min. The reaction was purified using the QIAquick PCR Purification Kit (Qiagen 28104) 822 and eluted in 15 uL of molecular biology grade water. 823 824 Following this, oligos containing the second Illumina TruSeq adapter (Supplementary Table 5) were annealed 825 by mixing 10 μL of each oligo (top and bottom strand, from a 100 μM stock concentration) with 10 μL of 826 molecular biology grade water and 10 μL of NEBuffer 2, then heated to 95°C for 2 min and allowed to cool back 827 to room temperature. 1 μL of the annealed adapters were mixed with 4 μL of the extended msDNA product and 828 5 μL of Blunt/TA Ligase Master Mix ( NEB M0367). This was incubated for 15 min at room temperature, after 829 which the reaction was immediately cleaned up using AMPure XP Beads (Beckman Coulter A63880). Cleanup 830 was performed with 1.8x beads:DNA volume ratio and the product was eluted in 10 uL of molecular biology 831 grade water. Sample was prepared for Illumina sequencing following standard workflows and sequenced on a 832 NextSeq 2000 system. Sequencing reads were trimmed for adapter sequences and aligned to the Retron-Vpa2 833 operon using Geneious. Visualization of read coverage on the Retron-Vpa2 hyRNA secondary structure was 834 performed using VARNA70 (v3-93). 835 836 Split GFP complementation assays 837 In preliminary experiments developing this assay, a plasmid containing either sfGFP1-10 or SP-sfGFP1-10 with 838 a kanamycin-resistance cassette and a plasmid containing sfGFP11 (either on its own or fused to the SP in the 839 context of Retron-Vpa2 Δnc-hyRNA) with a carbenicillin-resistance cassette were co- transformed into E. coli 840 bSLS.114. In all subsequent experiments, a plasmid containing a version of Retron-Vpa2 SP::sfGFP11 with a 841 carbenicillin-resistance cassette and a plasmid containing SP-sfGFP1-10 with a kanamycin-resistance cassette 842 were co-transformed into E. coli bSLS.114. For experiments requiring trigger genes, a third plasmid containing 843 the trigger (or an empty vector control) with a chloramphenicol-resistance cassette was also co-transformed. 844 For experiments using Retron-Eco1 to produce msDNA in trans, a third plasmid containing the engineered 845 retron with a spectinomycin-resistance cassette was also co-transformed. 846 847 For each sample, three individual colonies were picked (biological replicates) and grown until saturation in 0.5 848 mL LB + antibiotics in a 96 -well deep well plate (Thermo Scientific 260251 ) covered with an air -permeable 849 seal (Breathe-Easier, Diversified Biotech) at 37°C with shaking at 1000 rpm. Cultures were then diluted 1:100 850 in 200 uL of fresh LB + antibiotics in a black, clear- bottom 96-well microtiter plate (Corning 3603). Samples 851 were arranged to have at least one empty well between each set of replicates to avoid bleed -through 852 fluorescence across samples. The plate was covered with an air -permeable seal (Bre athe-Easier, Diversified 853 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint Biotech BERM-2000) and incubated in a microplate reader (Tecan Infinite 200 PRO) with shaking at 37° C for 854 16 hrs with fluorescence measured from the bottom of the wells using an excitation wavelength of 485 nm and 855 emission wavelength of 520 nm . Measurements were taken every 10 min. At the 2 hr timepoint, IPTG and 856 arabinose inducers were added to all cultures. In preliminary experiments developing the assay, inducers were 857 instead added at the 0 hr time point and the experiment was run for 16 hrs. Fluorescence readings were 858 adjusted by subtracting the 0 hr measurement for each sample. 859 860

861 Work was supported by funding from the National Science Foundation (MCB 2509382), the Robert 862 J Kleeberg Jr. and Helen C. Kleberg Foundation, and the Gordon and Betty Moore Foundation. 863 K.Z. was supported by a National Science Foundation Graduate Research Fellowship and a UCSF 864 Discovery Fellowship. 865 A.C. was supported by a Lundbeck Foundation grant R380-2021-1448. 866 D.J.Z. was supported by the Fulbright U.S. Student Program, which is sponsored by the U.S. 867 Department of State and the Danish-American Fulbright Commission. 868 A.D.H was supported by a Ph.D. fellowship from Junta de Castilla y León and European Social Fund 869 Plus (EDU/1868/2022) and an EMBO Scientific Exchange Grant (11561). 870 G.M. is part of CPR, which is supported financially by the Novo Nordisk Foundation (NNF14CC0001, 871 NNF24SA0098829). This work was also supported by the ERC-AdG 101096548 (INTETOOLS), 872 NNF0024386, NNF17SA0030214, and NNF18OC0055061 grants to G.M, who is a member of the 873 Integrative Structural Biology Cluster (ISBUC) at the University of Copenhagen. 874 R.P.-R. was supported by a Lundbeck Foundation grant (R347-2020-2346) and a research grant 875 (VIL60763) from VILLUM FONDEN. 876 S.L.S. is a Chan Zuckerberg Biohub – San Francisco Investigator. 877 878 We thank Mylinh Bernardi and Felicia Miller of the Gladstone Genomics Core for their assistance 879 with sequencing co-immunoprecipitated RNA from our pull-down assays. 880 881 AUTHOR CONTRIBUTIONS 882 Author contributions follow the CRedIT taxonomy 883 (https://www.elsevier.com/researcher/author/policies-and-guidelines/credit-author-statement) 884 885 K.Z.: Methodology, Investigation, Validation, Formal Analysis, Writing – Original Draft, Visualization 886 M.R-M.: Methodology, Investigation, Validation, Writing – Review & Editing 887 D.P.: Methodology, Investigation, Validation, Writing – Review & Editing 888 J.M.K.: Methodology, Investigation, Validation, Writing – Review & Editing, Visualization 889 A.C.: Methodology, Investigation, Validation, Writing – Review & Editing, Visualization 890 D.J.Z.: Methodology, Software, Formal Analysis, Writing – Review & Editing, Visualization 891 M.R.M.: Methodology, Software, Formal Analysis, Writing – Review & Editing, Visualization 892 R.Z.: Methodology, Investigation, Validation 893 A.D-H.: Investigation 894 G.M.: Methodology, Writing – Review & Editing, Supervision, Funding Acquisition 895 R.P-R.: Methodology, Writing – Review & Editing, Supervision, Funding Acquisition 896 A.G-D.: Conceptualization, Writing – Review & Editing, Supervision 897 S.L.S.: C onceptualization, Writing – Review & Editing, Visualization, Supervision, Project 898 administration, Funding acquisition 899 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint 900 COMPETING INTERESTS 901 G.M. is a stockholder of Ensoma and has been consultant for Orbis Medicines. The remaining 902 authors declare no competing interests. 903 904 DATA AND CODE AVAILABILITY 905 Sequencing data associated with this study are available in the NCBI SRA (PRJNA1320719) 906 https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1320719 907 908 Custom code used to process or analyze data from this study is available at: 909 Spacer-seq analysis was performed using code sourced from https://github.com/Shipman-910 Lab/Spacer-Seq. 911 912 SUPPLEMENTARY MATERIALS 913 See supplementary files for Supplementary Figures S1-S10 and Supplementary Tables 1-5. 914 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint

915 1. Gao, L. et al. Diverse enzymatic activities mediate antiviral immunity in prokaryotes. Science 916 369, 1077–1084 (2020). 917 2. Millman, A. et al. Bacterial Retrons Function In Anti-Phage Defense. Cell 183, 1551-1561.e12 918 (2020). 919 3. Bobonis, J. et al. Bacterial retrons encode phage-defending tripartite toxin–antitoxin systems. 920 Nature 609, 144–150 (2022). 921 4. Bernheim, A. & Sorek, R. The pan-immune system of bacteria: antiviral defence as a community 922 resource. Nat. Rev. Microbiol. 18, 113–119 (2020). 923 5. Lopatina, A., Tal, N. & Sorek, R. Abortive Infection: Bacterial Suicide as an Antiviral Immune 924 Strategy. Annu. Rev. Virol. 7, 371–384 (2020). 925 6. Mestre, M. R., González-Delgado, A., Gutiérrez-Rus, L. I., Martínez-Abarca, F. & Toro, N. 926 Systematic prediction of genes functionally associated with bacterial retrons and classification 927 of the encoded tripartite systems. Nucleic Acids Res. 48, 12632–12647 (2020). 928 7. Hsu, M. Y., Inouye, S. & Inouye, M. Structural Requirements of the RNA Precursor for the 929 Biosynthesis of the Branched RNA-linked Multicopy Single-stranded DNA of Myxococcus 930 xanthus. J. Biol. Chem. 264, 6214–6219 (1989). 931 8. Inouye, S., Hsu, M.-Y., Eagle, S. & Inouye, M. Reverse transcriptase associated with the 932 biosynthesis of the branched RNA-linked msDNA in Myxococcus xanthus. Cell 56, 709–717 933 (1989). 934 9. Lampson, B. C., Inouye, M. & Inouye, S. Reverse transcriptase with concomitant ribonuclease H 935 activity in the cell-free synthesis of branched RNA-linked msDNA of Myxococcus xanthus. Cell 936 56, 701–707 (1989). 937 10. Hsu, M. Y., Eagle, S. G., Inouye, M. & Inouye, S. Cell-free synthesis of the branched RNA-linked 938 msDNA from retron-Ec67 of Escherichia coli. J. Biol. Chem. 267, 13823–13829 (1992). 939 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint 11. Shimamoto, T., Inouye, M. & Inouye, S. The Formation of the 2′,5′-Phosphodiester Linkage in the 940 cDNA Priming Reaction by Bacterial Reverse Transcriptase in a Cell-free System (∗). J. Biol. 941 Chem. 270, 581–588 (1995). 942 12. Shimamoto, T., Kawanishi, H., Tsuchiya, T., Inouye, S. & Inouye, M. In Vitro Synthesis of 943 Multicopy Single-Stranded DNA, Using Separate Primer and Template RNAs, by Escherichia coli 944 Reverse Transcriptase. J. Bacteriol. 180, 2999–3002 (1998). 945 13. Simon, A. J., Ellington, A. D. & Finkelstein, I. J. Retrons and their applications in genome 946 engineering. Nucleic Acids Res. 47, 11007–11019 (2019). 947 14. Yee, T., Furuichi, T., Inouye, S. & Inouye, M. Multicopy single-stranded DNA isolated from a 948 gram-negative bacterium, Myxococcus xanthus. Cell 38, 203–209 (1984). 949 15. Khan, A. G. et al. An experimental census of retrons for DNA production and genome editing. 950 Nat. Biotechnol. 43, 914–922 (2025). 951 16. Nakamura, K. L., Zhang, K., Mestre, M. R., Rojas-Montero, M. & Shipman, S. L. Phage defense and 952 genome editing using novel retrons sourced from isolated environmental bacteria. 953 2025.01.29.635429 Preprint at https://doi.org/10.1101/2025.01.29.635429 (2025). 954 17. Lampson, B. C. et al. Reverse Transcriptase in a Clinical Strain of Escherichia coli: Production of 955 Branched RNA-Linked msDNA. Science 243, 1033–1038 (1989). 956 18. Lim, D. & Maas, W. K. Reverse transcriptase-dependent synthesis of a covalently linked, 957 branched DNA-RNA compound in E. coli B. Cell 56, 891–904 (1989). 958 19. Lima, T. M. O. & Lim, D. A Novel Retron That Produces RNA-less msDNA inEscherichia coliUsing 959 Reverse Transcriptase. Plasmid 38, 25–33 (1997). 960 20. Carabias, A. et al. Retron-Eco1 assembles NAD+-hydrolyzing filaments that provide immunity 961 against bacteriophages. Mol. Cell 84, 2185-2202.e12 (2024). 962 21. Wang, Y. et al. Cryo-EM structures of Escherichia coli Ec86 retron complexes reveal architecture 963 and defence mechanism. Nat. Microbiol. 7, 1480–1489 (2022). 964 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint 22. Wang, Y. et al. DNA methylation activates retron Ec86 filaments for antiphage defense. Cell Rep. 965 43, (2024). 966 23. Palka, C., Fishman, C. B., Bhattarai-Kline, S., Myers, S. A. & Shipman, S. L. Retron reverse 967 transcriptase termination and phage defense are dependent on host RNase H1. Nucleic Acids 968 Res. 50, 3490–3504 (2022). 969 24. Wang, B., Hoffman, R., Hou, Y.-M. & Li, H. Structural Basis for Retron Co-option of Anti-phage 970 ATPase-nuclease. 2025.05.10.653283 Preprint at https://doi.org/10.1101/2025.05.10.653283 971 (2025). 972 25. Wang, C. et al. Disassembly activates Retron-Septu for antiphage defense. Science 0, eadv3344 973 (2025). 974 26. Li, X. et al. Architecture and mechanism of a dual-enzyme retron system in prokaryotic 975 immunity. 2025.05.23.655875 Preprint at https://doi.org/10.1101/2025.05.23.655875 976 (2025). 977 27. García-Rodríguez, F. M., Martínez-Abarca, F., Wilkinson, M. E. & Toro, N. Similar mechanisms of 978 retron-mediated anti-phage defense for different families of tailed phages. 2024.02.09.579579 979 Preprint at https://doi.org/10.1101/2024.02.09.579579 (2024). 980 28. Stokar- Avihail, A. et al. Discovery of phage determinants that confer sensitivity to bacterial 981 immune systems. Cell 186, 1863-1876.e16 (2023). 982 29. Pausch, P. et al. Retron Eco2 breaks tRNAs for antiphage defense. Preprint at 983 https://doi.org/10.21203/rs.3.rs-7196019/v1 (2025). 984 30. Wang, Y. et al. Structural basis of the RNA-mediated Retron-Eco2 oligomerization. Cell Discov. 985 11, 73 (2025). 986 31. Rychlik, I., Sebkova, A., Gregorova, D. & Karpiskova, R. Low-Molecular-Weight Plasmid of 987 Salmonella enterica Serovar Enteritidis Codes for Retron Reverse Transcriptase and Influences 988 Phage Resistance. J. Bacteriol. 183, 2852–2858 (2001). 989 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint 32. Pilousova, L. & Rychlik, I. Retron Se72 utilizes a unique strategy of the self-priming initiation of 990 reverse transcription. Cell. Mol. Life Sci. 68, 3607–3617 (2011). 991 33. Bhattarai-Kline, S. et al. Recording gene expression order in DNA by CRISPR addition of retron 992 barcodes. Nature 608, 217–225 (2022). 993 34. Maffei, E. et al. Systematic exploration of Escherichia coli phage–host interactions with the 994 BASEL phage collection. PLOS Biol. 19, e3001424 (2021). 995 35. Caldwell, B. J. & Bell, C. E. Structure and mechanism of the Red recombination system of 996 bacteriophage λ. Prog. Biophys. Mol. Biol. 147, 33–46 (2019). 997 36. Murphy, K. C. Lambda Gam protein inhibits the helicase and chi-stimulated recombination 998 activities of Escherichia coli RecBCD enzyme. J. Bacteriol. 173, 5808–5821 (1991). 999 37. Wilkinson, M. et al. Structural basis for the inhibition of RecBCD by Gam and its synergistic 1000 antibacterial effect with quinolones. eLife 5, e22963 (2016). 1001 38. Smith, C. E. & Bell, C. E. Domain Structure of the Redβ Single-Strand Annealing Protein: the C-1002 terminal Domain is Required for Fine-Tuning DNA-binding Properties, Interaction with the 1003 Exonuclease Partner, and Recombination in vivo. J. Mol. Biol. 428, 561–578 (2016). 1004 39. Iyer, L. M., Koonin, E. V. & Aravind, L. Classification and evolutionary history of the single-strand 1005 annealing proteins, RecT, Redβ, ERF and RAD52. BMC Genomics 3, 8 (2002). 1006 40. Lampson, B. C., Inouye, M. & Inouye, S. Retrons, msDNA, and the bacterial genome. Cytogenet. 1007 Genome Res. 110, 491–499 (2005). 1008 41. Datsenko, K. A. et al. Molecular memory of prior infections activates the CRISPR/Cas adaptive 1009 bacterial immunity system. Nat. Commun. 3, 945 (2012). 1010 42. Yosef, I., Goren, M. G. & Qimron, U. Proteins and DNA elements essential for the CRISPR 1011 adaptation process in Escherichia coli. Nucleic Acids Res. 40, 5569–5576 (2012). 1012 43. Nuñez, J. K. et al. Cas1–Cas2 complex formation mediates spacer acquisition during CRISPR– Cas 1013 adaptive immunity. Nat. Struct. Mol. Biol. 21, 528–534 (2014). 1014 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint 44. Nuñez, J. K., Harrington, L. B., Kranzusch, P. J., Engelman, A. N. & Doudna, J. A. Foreign DNA 1015 capture during CRISPR–Cas adaptive immunity. Nature 534, S13–S14 (2016). 1016 45. Cabantous, S., Terwilliger, T. C. & Waldo, G. S. Protein tagging and detection with engineered 1017 self-assembling fragments of green fluorescent protein. Nat. Biotechnol. 23, 102–107 (2005). 1018 46. Levy, A. et al. CRISPR adaptation biases explain preference for acquisition of foreign DNA. 1019 Nature 520, 505–510 (2015). 1020 47. Shiriaeva, A. A. et al. Host nucleases generate prespacers for primed adaptation in the E. coli 1021 type I-E CRISPR-Cas system. Sci. Adv. 8, eabn8650 (2022). 1022 48. Lopez, S. C., Crawford, K. D., Lear, S. K., Bhattarai-Kline, S. & Shipman, S. L. Precise genome 1023 editing across kingdoms of life using retron-derived DNA. Nat. Chem. Biol. 18, 199–206 (2022). 1024 49. Datsenko, K. A. & Wanner, B. L. One-step inactivation of chromosomal genes in Escherichia coli 1025 K-12 using PCR products. Proc. Natl. Acad. Sci. 97, 6640–6645 (2000). 1026 50. Lopez, S. C., Lee, Y., Zhang, K. & Shipman, S. L. SspA is a transcriptional regulator of CRISPR 1027 adaptation in E. coli. Nucleic Acids Res. 53, gkae1244 (2025). 1028 51. Püllmann, P. et al. A modular two yeast species secretion system for the production and 1029 preparative application of unspecific peroxygenases. Commun. Biol. 4, 562 (2021). 1030 52. Fishman, C. B. et al. Continuous multiplexed phage genome editing using recombitrons. Nat. 1031 Biotechnol. 43, 1299–1310 (2025). 1032 53. Katoh, K. & Standley, D. M. MAFFT Multiple Sequence Alignment Software Version 7: 1033 Improvements in Performance and Usability. Mol. Biol. Evol. 30, 772–780 (2013). 1034 54. Piñeiro, C., Abuín, J. M. & Pichel, J. C. Very Fast Tree: speeding up the estimation of phylogenies 1035 for large alignments through parallelization and vectorization strategies. Bioinformatics 36, 1036 4658–4659 (2020). 1037 55. Payne, L. J. et al. Identification and classification of antiviral defence systems in bacteria and 1038 archaea with PADLOC reveals new system types. Nucleic Acids Res. 49, 10868–10878 (2021). 1039 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint 56. Mazzocco, A., Waddell, T. E., Lingohr, E. & Johnson, R. P. Enumeration of Bacteriophages Using 1040 the Small Drop Plaque Assay System. in Bacteriophages 81–85 (Humana Press, 2009). 1041 doi:10.1007/978-1-60327-164-6_9. 1042 57. Danecek, P. et al. Twelve years of SAMtools and BCFtools. GigaScience 10, giab008 (2021). 1043 58. Lorenz, R. et al. ViennaRNA Package 2.0. Algorithms Mol. Biol. 6, 26 (2011). 1044 59. Kerpedjiev, P., Hammer, S. & Hofacker, I. L. Forna (force-directed RNA): Simple and effective 1045 online RNA secondary structure diagrams. Bioinformatics 31, 3377–3379 (2015). 1046 60. Li, W. & Godzik, A. Cd-hit: a fast program for clustering and comparing large sets of protein or 1047 nucleotide sequences. Bioinformatics 22, 1658–1659 (2006). 1048 61. Will, S., Joshi, T., Hofacker, I. L., Stadler, P. F. & Backofen, R. LocARNA-P: Accurate boundary 1049 prediction and improved detection of structural RNAs. RNA 18, 900–914 (2012). 1050 62. Nawrocki, E. P. & Eddy, S. R. Infernal 1.1: 100-fold faster RNA homology searches. 1051 Bioinformatics 29, 2933–2935 (2013). 1052 63. Eddy, S. R. Accelerated Profile HMM Searches. PLOS Comput. Biol. 7, e1002195 (2011). 1053 64. Shen, W., Le, S., Li, Y. & Hu, F. SeqKit: A Cross-Platform and Ultrafast Toolkit for FASTA/Q File 1054 Manipulation. PLOS ONE 11, e0163962 (2016). 1055 65. Steinegger, M. & Söding, J. MMseqs2 enables sensitive protein sequence searching for the 1056 analysis of massive data sets. Nat. Biotechnol. 35, 1026–1028 (2017). 1057 66. Rivas, E., Clements, J. & Eddy, S. R. A statistical test for conserved RNA structure shows lack of 1058 evidence for structure in lncRNAs. Nat. Methods 14, 45–48 (2017). 1059 67. Crooks, G. E., Hon, G., Chandonia, J.-M. & Brenner, S. E. WebLogo: A Sequence Logo Generator. 1060 Genome Res. 14, 1188–1190 (2004). 1061 68. Langmead, B., Trapnell, C., Pop, M. & Salzberg, S. L. Ultrafast and memory-efficient alignment of 1062 short DNA sequences to the human genome. Genome Biol. 10, R25 (2009). 1063 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint 69. Granger, B. E. & Pérez, F. Jupyter: Thinking and Storytelling With Code and Data. Comput. Sci. 1064 Eng. 23, 7–14 (2021). 1065 70. Darty, K., Denise, A. & Ponty, Y. VARNA: Interactive drawing and editing of the RNA secondary 1066 structure. Bioinformatics 25, 1974–1975 (2009). 1067 71. Krzywinski, M. I. et al. Circos: An information aesthetic for comparative genomics. Genome Res. 1068 (2009) doi:10.1101/gr.092759.109. 1069 1070 .CC-BY 4.0 International licenseavailable under a (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 The copyright holder for this preprintthis version posted October 22, 2025. ; https://doi.org/10.1101/2025.10.22.683967doi: bioRxiv preprint