Time to lysis determines phage sensitivity to a cytidine deaminase toxin/antitoxin bacterial defense system

Toxin-antitoxin (TA) systems are ubiquitous two-gene loci that bacteria use to regulate cellular processes such as phage defense. Here, we demonstrate the mechanism by which a novel type III TA system, avcID, is activated and confers resistance to phage infection. The toxin of the system (AvcD) is a deoxycytidylate deaminase that converts deoxycytidines (dC) to dexoyuridines (dU), while the RNA antitoxin (AvcI) inhibits AvcD activity. We have shown that AvcD deaminated dC nucleotides upon phage infection, but the molecular mechanism that activated AvcD was unknown. Here we show that the activation of AvcD arises from phage-induced shutoff of host transcription, leading to degradation of the labile AvcI. AvcD activation and nucleotide depletion not only decreases phage replication but also increases the formation of defective phage virions. Surprisingly, infection of phages such as T7 that are not inhibited by AvcID also lead to AvcI RNA antitoxin degradation and AvcD activation, suggesting that depletion of AvcI is not sufficient to confer protection against some phage. Rather, our results support that phage with a longer lysis time like T5 are sensitive to AvcID-mediated protection while those with a shorter lysis time like T7 are resistant.

CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.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 preprint this version posted February 10, 2023. ; https://doi.org/10.1101/2023.02.09.527960 doi: bioRxiv preprint Page | 7 the expression of its native promoter. Likewise, to determine whether AvcD protein levels 156 changed concurrently with changes in avcI RNA levels, we quantified the AvcD protein using a 157 Western blot with antibodies specific to a C-terminal 6xHis tag. Notably, the full length of the 158 avcI transcript is slightly smaller than the 300 ribonucleotide bases, which is longer than the 159 minimum functional length of AvcI of ~171 bases that was previously determined (Fig. S1) [23]. 160 The levels of AvcI rapidly decreased when cells were treated with the transcriptional inhibitor 161 rifampicin, showing that in the absence of new transcription, AvcI is degraded ( Fig. 2A). 162 Importantly, spectinomycin, which inhibits protein synthesis instead of transcription, did not 163 decrease avcI levels (Fig. 2B), indicating that the degradation was specific to transcriptional 164 shutoff. These results are consistent with our previous study showing that rifampicin but not 165 spectinomycin activated AvcD in V. cholerae [23]. 166 We also infected these cells with phages T5 and T7 and quantified AvcI and AvcD over 167 time ( Fig. 2C-D). The half-life of the avcI transcript ranged from 1.5 min with T7 to 6.7 min with 168 T5 (Fig. S2). The fact that AvcI is degraded faster during T7 compared with T5 is a surprising 169 result given that AvcID provides protection against T5 phage but not T7. Concurrently, we also 170 found that AvcD protein levels did not change significantly in any of the conditions tested ( Fig.  171 2A-D). To determine whether AvcD is activated upon loss of AvcI, we measured the intracellular 172 abundance of dCTP and dCMP using UPLC-MS/MS before and after infecting the cells 173 containing active or inactive avcID with T5 or T7 phages. Surprisingly, both T5 and T7 infections 174 significantly decreased intracellular dCTP and dCMP in cells containing active avcID, 175 demonstrating that both phages activate AvcD (Figs. 3A-B). We also noted that T5 decreases 176 intracellular dCMP in cells containing inactive avcID (Fig. 3B). We hypothesize this result is due 177 to a T5 encoded 5' monophosphatase (dmp) which participates in the final stages of host DNA 178 degradation by dephosphorylating 5'-dNMP's substrates [28]. Collectively, these results suggest 179 that transcriptional shutoff coupled with the instability of the avcI RNA leads to the release of 180 . CC-BY 4.0 International license available 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 preprint this version posted February 10, 2023. ; https://doi.org/10.1101/2023.02.09.527960 doi: bioRxiv preprint Page | 8 existing AvcD from inhibition upon both T5 and T7 phage infection, but activation of AvcD is not 181 sufficient to protect against T7 phage infection. 182 183 AvcID drives production of defective T5 phage 184 We found that AvcID provides resistance to T5 but not T7 (Figs. 1A-B infected with T5 had more viable CFUs than cells harboring inactive AvcID* after two hours of 191 infection (Fig. 4A). Furthermore, the AvcID-containing population generated ~100-fold fewer 192 PFUs than AvcID*-containing cells (Fig. 4B), supporting the notion that AvcID inhibits the 193 accumulation of functional T5 phages. Consistent with our liquid infection results described 194 above, AvcID did not impact the number of CFUs and PFUs when cells were infected with T7 195 Since a plaque assay only quantifies viable phages, we speculated that the total viral 197 particles produced could be underestimated if some of those virions were non-viable. To 198 determine total virions produced, we quantified the abundance of a specific phage gene for T5 199 and T7 in the phage particle samples using qPCR. Importantly, these samples had been treated 200 with DNase ensuring that only genomes protected by phage virions were quantified by this 201 assay. Our results indicated that the total number of T5 phage genomes decreased over time in 202 infected cultures of AvcID-containing cells compared to AvcID*-containing cells. However, the 203 magnitude of this decrease was less than that observed for the difference in PFUs between the 204 two samples (Fig. 4B, C). For example, at the two-hour time point, there was a difference of 205 ~100-fold in PFUs but only a difference of 20-fold in virions measured with qPCR. We observed 206 . CC-BY 4.0 International license available 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 preprint this version posted February 10, 2023. ; https://doi.org/10.1101/2023.02.09.527960 doi: bioRxiv preprint Page | 9 no significant difference in genome abundance between AvcID and AvcID* encoding cultures 207 when infected with T7 (Fig. 4F). 208 The greater magnitude difference for T5 of PFUs compared to phage genomes 209 suggested that the majority of virions produced from AvcID-containing cells contained genomes 210 that were defective to infect new cells and form plaques. We calculated the percentage of viable 211 phages by quantifying the ratio between PFUs and genome abundance of AvcID or AvcID* 212 infected cultures. Using this analysis, we estimated that by 30 min only 30% of T5 phage 213 derived from cells containing AvcID were functional, and the proportion of functional phage 214 generally decreased over time, suggesting that AvcID drives formation of defective T5 phage 215 virions (Fig. 4G). In contrast, a similar analysis for T7 indicates that nearly all the T7 phages 216 were viable even when they were from cells encoding avcID (Fig. 4G). This indicates that AvcID 217 confers protection by both decreasing phage replication and increasing defective phage 218 production for T5 while T7 can overcome these negative effects of AvcID through an unknown 219 mechanism. 220 Consistent with our observation that defective T5 phage are generated from avcID-221 encoding cells, transmission electron microscopy (TEM) images of negatively stained samples 222 revealed that particles produced from cells containing active AvcID have more defective phage 223 capsids-i.e. broken particles, or capsids with aberrant morphology, containing no genome and 224 without attached tails (Fig. 5A), compared to T5 phage from cells containing AvcID* (Fig. 5B). 225 Images of negatively stained particles of T7 isolated from avcID encoding cells showed no such 226 defects and the virions looked normal (Fig 5C). Together, the TEM results corroborate our 227 previous experiments confirming that the AvcID system inhibits production of functional T5 228 phage particles. Ung function together to reduce phage infection, we infected E. coli MG1655 or Δung E. coli 241 NR8052 encoding avcID or avcID * with T5 phage, measured the relative phage titer, and further 242 tracked bacterial growth by OD 600 over time. We hypothesized that the ung mutant would not 243 exhibit as robust of protection from T5 as the WT E. coli. Contrary to our hypothesis, the OD 600 244 of both strain backgrounds carrying active AvcID exhibited similar protection, suggesting that 245 Ung is not required for AvcID to protect E. coli from T5 phage (Fig. 6A). When comparing 246 relative phage titer, we observed no difference in AvcID-mediated protection from T5 in the 247 presence or absence of ung (Fig. 6B). Finally, we infected E. coli MG1655 containing either 248 AvcID or AvcID* with T5 phage while overexpressing a dUTPase (dut) to reduce accumulation 249 of dUTP, thereby preventing potential incorporation of uracil into phage genomes. The 250 overexpression of Dut had no effect on the phage defense conferred by AvcID (Fig. 6C). 251 Together, these data suggest that accumulation of dUTP or incorporation of uracils into the 252 phage genome does not contribute to AvcID-mediated protection. Our results demonstrated that both T5 and T7 activate AvcID, but this system only 256 protects against T5 infection. When considering this discrepancy, we noted that T7 has a much 257 . CC-BY 4.0 International license available 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 preprint this version posted February 10, 2023. conditions. We interpret this result to mean that after a delay in the lysis of the initial MMS-267 treated T7, AvcID was unable to protect against subsequent rounds of phage infection. 268 To further explore the impact of lysis time on AvcID-mediated protection, we infected E. AvcID*-containing cells (Fig. 7D). Additionally, the total number of T7 412 genomes quantified 277 using qPCR decreased over time in infected cultures of AvcID-containing cells compared to the 278 inactive variant (Fig. 7E). Using these results, we estimated the proportion of functional T7 412 by 279 30 min was only approximately 2% of the total phage virions (Fig. 7F). These data indicate that 280 AvcID confers protection against T7 412 by increasing defective phage production and generating 281 non-functional phage, similar to our results for T5. 282 283 . CC-BY 4.0 International license available 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 preprint this version posted February 10, 2023. ; https://doi.org/10.1101/2023.02.09.527960 doi: bioRxiv preprint

Relationship between phage replication time and AvcID protection 284
Our results suggested that lysis time is a key factor that determines phage sensitivity or 285 resistance to AvcID. To further explore this idea, we measured the association between phage shorter lysis times. We found no clear relationship between the phage genome size and their 295 respective lysis time (Fig. S5). Together these results suggest that lysis time, but not genome 296 size, contributes to phage sensitivity to AvcID. 297

DISCUSSION 299
Phage predation is a constant evolutionary pressure that shapes the diversity and fitness 300 of bacteria that has driven the evolution of multiple antiphage defense systems. The underlying 301 mechanisms of certain antiphage defenses such as RMs, which utilize DNA modifications to 302 distinguish host and foreign DNA, are well-characterized [2, 32, 33]. On the contrary, the 303 mechanism of activation of the cyclic nucleotide-based systems (i.e., CBASS) in response to 304 phage infection is generally not understood [3, 34]. Moreover, although many novel phage 305 defense systems have been recently identified, it is typically unclear why they provide protection 306 against some phage but not others [7, 23, 24, 35, 36]. Here, we reveal the mechanism of how 307 the AvcID TA system is activated in response to phage infection and its impact on the phage's 308 morphogenetic pathway. We also determined that lysis time is an important factor that drives 309 . CC-BY 4.0 International license available 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 AvcD, although the mechanism for this activation was not known [23,24] . This work is the first 316 to show that this activation is due to degradation of the AvcI sRNA antitoxin. The AvcI antitoxin 317 is produced at high abundance compared to AvcD [23]. In other TA systems, there is typically a 318 Rho-independent terminator located between the toxin and antitoxin genes [39,40]. However, 319 sequence analysis did not predict such a terminator between avcI and avcD, and thus how the 320 AvcI sRNA is formed and produced at much higher levels than AvcD requires further study. 321 Though AvcID did not provide protection against T7, avcI transcripts were rapidly 322 degraded upon T7 infection, implying that avcI levels decrease when cells undergo 323 transcriptional inhibition regardless of the cause. However, the activation of AvcD does not have 324 any detrimental effect on the viability of T7, in contrast to T5, implying that T7 has evolved to 325 disregard any detrimental effects inflicted by AvcID or similar phage defense systems. Given 326 that we observed no difference in viable phage from functional AvcID versus AvcID* containing 327 cells (Fig. 4), we suggest that even in the presence of AvcID and depleted dCTP, T7 replicates 328 enough genomes to fully package all the capsid heads that are produced. Phage can synthesize 329 more genomes than capsid heads, consistent with this interpretation [41,42]. It should be noted 330 that examination of AvcD orthologs from E. coli expressed in a heterologous E. coli host did 331 show protection against T7 using a plaquing assay [24]. This result highlights the high degree of 332 specificity in phage protection for each defense system. The reason for this difference between 333 these studies is not clear, but we speculate it may be due to the specific molecular features of 334 the two AvcD systems being studied or differences in the T7 phage that were tested. 335 . CC-BY 4.0 International license available 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 preprint this version posted February 10, 2023. The growth of several well-known phages is inhibited when their DNA contains dUMP 343 and Ung is present in the host cells. For example, to counter this negative effect T5 encodes its 344 own dUTPase for reducing the dUTP level such that dUMP is limited in its genome [43] . 345 However, the presence of the AvcID system prevents this T5 infection even in the absence of 346 Ung, indicating that dUMP incorporation into the phage genome may not be the cause for the 347 phage viability defect. Rather, our evidence suggests it is the depletion of the dC pool that is 348 responsible for the reduction in functionality of T5 phage particles. Remarkably, the depletion in 349 the dC pool has no effect on T7 viability even though its G/C content is approximately 52% [44]. of T4 might also be due to competition for dC nucleotides since T-even phages are known to 356 possess enzymes that can methylate deoxycytosine-containing bases to evade bacterial RMs 357 [46]. Whether this is also a strategy to resist AvcID is under investigation. 358 We obtained two lines of evidence that linked phage lysis time to AvcID sensitivity. First, 359 treatment of phage virions with MMS, which is known to delay lysis time [31], enhanced 360 sensitivity to AvcID (Fig. S2). Secondly, the resistance of T7 to AvcID can be completely 361 . CC-BY 4.0 International license available 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 preprint this version posted February 10, 2023. [48]. This result studying avcID in its native genome context and host is consistent with our 373 conclusion that lysis time is an important driver of AvcID sensitivity and resistance [48]. 374 Prior work on Type III TA systems suggests they are associated with abortive infections 375 (Abi), which is defined as the host committing altruistic suicide to prevent phage replication [38]. 376 However, overexpression of AvcD does not lead to cell death but does impair genome 377 replication, and this effect can be rescued by inhibiting expression of the toxin or overexpressing 378 avcI in trans [23]. Given that avcI is degraded, subsequently releasing existing AvcD to 379 deaminate dC pools upon phage infection, we propose that protection conferred by AvcD is not 380

through abortive infection. This conclusion is supported by our observation that infection of 381
AvcID containing cells with a high MOI does not enhance killing of the host cells (Fig. 1). 382 Similar to the AvcID system, bacterial dGTPases protect against phage infection by 383 dephosphorylating dGTP to dG to inhibit phage DNA replication and that this system is also 384 activated upon phage-induced transcriptional shutoff [24]. It is, however, unclear whether the 385 dGTPase system is a TA system. While other types of TA systems have been demonstrated to 386 have antiphage properties, whether they are activated in a similar mechanism as the Type III 387 . CC-BY 4.0 International license available 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 TA system, exerts its effect via modification of cellular target Prs, which is involved in nucleotide 390 biosynthesis, though the ParST system has not been demonstrated to be involved in phage 391 defense. The mechanism of AvcID bears a resemblance to both DarTG and ParST but is 392 distinct from both in terms of the mechanism for toxin function and activation. This suggests that 393 manipulating nucleotide pools is a conserved function of many TA systems and antiphage 394 defense mechanisms. (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

RNA Extraction for Northern Blot Following Phage Infection 447
RNA isolation and qRT-PCR analysis were carried out as previously described [51]. 448 Briefly, triplicate overnight cultures of E. coli carrying pAvcI-AvcD-6xHis were subcultured 1:100 449 in LB and grown to an OD 600 of 0.3. 1 mL of each replicate was pelleted and flash-frozen by the 450 ethanol-dry ice slurry method. RNA was extracted using TRIzol ® reagent following the 451 manufacturer's directions (Thermo Fischer Scientific TM ). RNA quality and quantity were 452 determined using a NanoDrop spectrophotometer (Thermo Fischer Scientific TM ). 453

RNA Probe Synthesis and Purification 455
The method for RNA probe production was modified from a previously described 456 protocol [23]. The AvcI DNA template for in vitro transcription was PCR amplified from pAvcI 457 using Q5 High-Fidelity DNA Polymerase (NEB TM ). To incorporate the T7 promoter into the final 458 AvcI DNA template, the forward primer included the T7 promoter sequence prior to the 459 homologous sequence for AvcI. Additionally, the first two residues of the reverse primer were 2-460 OMe modified to reduce 3-end heterogeneity of the transcript [52]. The PCR reaction was 461 analyzed using a 1% agarose gel, and the band corresponding to the AvcI DNA template was 462 . CC-BY 4.0 International license available 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 preprint this version posted February 10, 2023. Hybridization Buffer (Invitrogen TM ) with gentle shaking. Next, the pre-hybridization buffer was 484 removed, and hybridization buffer containing 1 nM of purified probe was added. The membrane 485 was hybridized for 12-16 hours at 60°C with gentle shaking. Next, the membrane was rinsed 486 twice every five minutes with 2x saline-sodium citrate (SSC) buffer, 0.1% SDS at 60°C and then 487 twice every 15 minutes with 0.1X SSC, 0.1% SDS at 60°C. The biotin-labeled probes were 488 . CC-BY 4.0 International license available 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 preprint this version posted February 10, 2023. ; https://doi.org/10.1101/2023.02.09.527960 doi: bioRxiv preprint a new tube, and the cell pellets were washed twice with equal volume of LB to remove 515 unadsorbed phage. For PFU measurements, the supernatants were serially diluted in MMB 516 medium (LB + 0.1 mM MnCl2 + 5 mM MgCl2 + 5 mM CaCl2) and 5 μL of each dilution was 517 spotted on a lawn of bacteria seeded in MMB agar plate (MMB + 0.5% agar). PFU plates were 518 then grown at RT overnight and plaques quantified the following day. For CFU measurements, 519 resuspended cell pellets were then incubated at 37°C for 5-10 minutes before being serially 520 diluted 10-fold in PBS and 5 μL of each dilution was spotted on LB plates. CFU plates were then 521 grown at 37°C overnight and colonies were quantified the following day. SDS, pH 6.8), and denatured for 10 min at 95°C. Denatured lysates were centrifuged at 15,000 535 x g for 1 min to pellet cellular debris, and the supernatant was used to quantify the total protein 536 concentration in the sample by using the DC protein assay (Bio-Rad) and a BSA standard curve 537 acetate. DNA quality and quantity were determined using a NanoDrop spectrophotometer 556 (Thermo Fischer Scientific). 557 For measuring phage genome abundance, 25 μL reactions consisted of 5 μL each 0.625 558 μM primers 1 and 2, 12.5 μL 2X SYBR master mix, and 2.5 μL of 2.5 ng/μL phage genomic 559 DNA. qPCR reactions were performed in technical duplicates for biological triplicate samples. 560 The relative abundance was calculated by comparing the C t values of phage infected E. coli 561 with AvcID to inactive AvcID* at each timepoints. 562 563 Alkylation of T7 phage 564 . CC-BY 4.0 International license available 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

Transmission Electron Microscopy (TEM) 573
High titer phage stocks were prepared using a 15 mL soft agar overlay with 150 μL of E. 574 coli DH10B and 150 μL of 2.26x10 10 PFU/mL (T5 or T7). Liquid cultures were then grown using 575 30 mL LB with 100 μL plus 0.5 mL overnight culture harboring a plasmid with active or inactive 576 avcID system (AmpR) and one plaque (either T5 or T7). The culture was grown at RT while 577 shaking at 200 RPM for 6 hours or until clear. 3mL of chloroform was added and the culture was 578 incubated for an additional 5 minutes before being centrifuged at 8,000 x g (~7,000 RPM in F14-579 14x50cy rotor) for 10 minutes at 4°C. The supernatant was then spun at 26,000 x g (~12,500 580 RPM in F14-14x50cy rotor) for 90 minutes at 4°C to pellet the phage. The pellet was 581 resuspended in 1.5 mL of phage buffer (10 mM Tris, pH 7.6, 10mM MgCl 2 ) by nutating overnight 582 at 4°C. 583 Approximately ~5 µL of phage samples were applied to freshly glow discharged (PELCO 584 easiGlow, 15 mA, 45 s) continuous carbon support film grids (Ted Pella, Cat. No: 1754-F) for 60 585 seconds, followed by washing with distilled water and then staining with 1% aqueous Uranyl 586 Acetate (Electron Microscopy Solutions, Cat. No: 22400-4). Grids were blotted dry with 587 Whatman filter paper. The phage samples were imaged at the RTSF Cryo-EM Core Facility at 588 Michigan State University using a Talos Arctica operated at 200 keV. Micrographs were 589 . CC-BY 4.0 International license available 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 preprint this version posted February 10, 2023. (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 preprint this version posted February 10, 2023. ; https://doi.org/10.1101/2023.02.09.527960 doi: bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted February 10, 2023. ; https://doi.org/10.1101/2023.02.09.527960 doi: bioRxiv preprint