The DarTG toxin-antitoxin system provides phage defense by ADP-ribosylating viral DNA

Abstract

Toxin-antitoxin (TA) systems are broadly distributed, yet poorly conserved, genetic elements whose biological functions are unclear and controversial. Some TA systems may provide bacteria with immunity to infection by their ubiquitous viral predators, the bacteriophage. To identify such TA systems, we searched bioinformatically for those frequently encoded near known phage defense genes in bacterial genomes. This search identified homologs of DarTG, a recently discovered family of TA systems whose biological functions and natural activating conditions were unclear. Representatives from two different subfamilies, DarTG1 and DarTG2, strongly protected E. coli MG1655 against different phages. We demonstrate that for each system, infection with either RB69 or T5 phage, respectively, triggers release of the DarT toxin, a DNA ADP-ribosyltransferase, that then modifies viral DNA and prevents replication, thereby blocking the production of mature virions. Further, we isolated phages that have evolved to overcome DarTG defense either through mutations to their DNA polymerase or to an anti-DarT factor, gp61.2, encoded by many T-even phages. Collectively, our results indicate that phage defense may be a common function for TA systems and reveal the mechanism by which DarTG systems inhibit phage infection.

Introduction

Bacteriophage, or phage, are the nearly ubiquitous viruses that infect bacteria. Their co-evolution with bacteria has led to an abundance of anti-phage defense systems. CRISPR and restriction-modification systems are two well-known systems that have been extensively characterized and famously co-opted as indispensable tools for molecular biology. In recent years, due to a recognition of the vast, unexplored biological potential of such systems, and the renewed interest in phage therapy as an alternative to antibiotics, there has been an explosion in the number of newly identified phage defense systems^1–4. However, in most cases, the mechanism of action of these phage defense systems remain unknown or incompletely elucidated.

One class of genetic elements increasingly implicated in phage defense are toxin-antitoxin (TA) systems, which are found in nearly all sequenced bacterial chromosomes, with some species encoding dozens of different systems^5,6. These systems typically feature a two-gene operon that encodes a growth-inhibiting toxin and a cognate, neutralizing antitoxin, which is often less stable than the toxin⁵. TA systems are categorized based on the nature of the antitoxin, with the four most common types featuring a small non-coding RNA that prevents toxin translation (type I), a protein that directly interacts with and neutralizes the toxin (type II), a non-coding RNA that directly interacts with the toxin (type III), or a protein that enzymatically reverses the activity of the toxin (type IV)⁵.

The biological functions of chromosomally-encoded TA systems have remained elusive and controversial^5,7–9. Cells expend significant resources in keeping these systems in an “off” state, in which toxin is neutralized by antitoxin¹⁰. Despite the seemingly high cost of their maintenance, the prevalence of TA systems suggests they are important for bacterial survival. TA systems are often postulated to be stress-response elements^8,11, but we previously found in Escherichia coli MG1655 that although stress can drive transcriptional induction of its 10 endoribonuclease toxins, active toxins do not get released¹⁰. Increasing evidence suggests that some TA systems function in phage defense, which may explain both their variability and ubiquity⁶. One of the best characterized examples are type III ToxIN systems, first identified in Pectobacterium atrosepticum, which feature an endoribonuclease toxin, ToxN¹². For a ToxIN system found in some E. coli strains, it was recently shown that toxin is liberated following phage-induced shutoff of host transcription and subsequent degradation of the unstable antitoxin¹³. The toxin then cleaves phage mRNAs to prevent translation of key structural components. Another well-characterized phage defense TA system is RnlAB, a type II system whose toxin is also an RNase¹⁴. For most other TA systems that function in phage defense, the toxin’s mechanism of action has only been studied by overexpression, not during infection, so how they disrupt the phage life cycle is unclear. However, given the remarkable diversity of biochemical functions ascribed to toxins beyond RNases, TA systems may block phage development at different stages in many different ways.

We set out to identify additional TA systems that provide phage defense by identifying systems frequently found near other phage defense elements. Phage defense systems are often co-located on bacterial chromosomes in so-called defense islands^15,16. Efforts to identify genes of unknown function that are frequently found in such genomic contexts has proven to be a fruitful strategy for identifying new phage defense systems^3,4. Applying this same approach specifically to TA systems led to the identification of two systems, DarTG1 and DarTG2, that can provide E. coli with potent defense against select phage. Previous work on DarTG systems demonstrated that DarT toxins can use NAD+ to ADP ribosylate DNA, and artificial overexpression of these toxins can disrupt chromosomal DNA replication^17–20. However, our work now demonstrates that under the natural activating conditions of phage infection, the DarT toxins in fact ADP-ribosylate phage DNA, which inhibits both viral DNA and RNA synthesis. Without new copies of their genomes to package, phages are unable to form progeny. Phages can evolve to overcome DarT activity by two different strategies, either mutating their DNA polymerase, likely to bypass ADP-ribosylation in the DNA, or by modifying an existing anti-DarT factor. In sum, our work demonstrates that DarTG systems can provide cells with strong defense against phage infection through the ADP-ribosylation of phage DNA. More generally, our work underscores the notion that the enzymatically diverse toxins of TA systems may equip bacteria with a diverse arsenal of phage defense mechanisms.

Results

DarTG systems provide defense against phage

We set out to examine the propensity of ten common type II TA systems to be present in defense islands, a property previously found as predictive for function in phage resistance³. To this end, we analyzed the genomic context of 202,402 toxin genes found in ~38,000 bacterial and archaeal genomes. For each type of toxin we calculated a ‘defense score’, calculated as the fraction of toxin homologs found within ten genes of known phage defense genes (Table S1). It was previously shown that defense scores > 0.4 are strongly predictive of anti-phage activity³. In our analysis, one TA family, the DarTG system, stood out with a defense score of 0.48, meaning that nearly half of the genes encoding DarTG homologs in this family are next to known defense genes in microbial genomes. Prior work had noted that darTG is sometimes found encoded within type I restriction-modification system operons¹⁷.

DarTG systems have not been previously shown to function in phage defense. To test if they can indeed provide defense against phages, we cloned two DarTG systems. In each case, we included the open reading frames encoding the toxin and antitoxin, as well as the native, upstream region encompassing the promoter. These systems were cloned into a pBR322 vector backbone and transformed into E. coli MG1655. Each system was then tested against a panel of 12 phages that can infect MG1655 in both fast and slow growth conditions (LB medium at 37 °C, fast growth; M9-glucose medium at 30 °C, slow growth) (Fig. 1a–c). Both systems provided robust defense against different phages under different conditions. DarTG1 prevented plaquing of RB69 and T5 in fast growth conditions, while DarTG2 provided robust phage defense against T5, SECϕ18, and Lust in slow growth conditions, with modest protection against T5 in fast growth conditions. We also tested whether a single, chromosomal copy of DarTG1 was sufficient to confer phage defense, and found that it did defend against RB69, albeit with less potency (Fig. 1d).

Figure 1. TA systems encoded near known phage defense elements also defend against phage.

a, Efficiency of plaquing (EOP) (indicated in green) for strains bearing the indicated DarTG systems infected with a panel of phage compared to a strain bearing an empty vector under two different growth conditions.

b-c, 10-fold serial dilution plaque assays of RB69 (b) or T5 (c) phage spotted on E. coli MG1655 harboring the indicated darTG system regulated by its native promoter (top), an empty vector (middle), or the darTG-bearing plasmid with a mutation in the predicted active site of the toxin (bottom).

d, EOP for RB69 phage on E. coli attHK::darTG1 compared to a wild-type MG1655 strain.

e, Genomic context for the darTG1 and darTG2 systems in (b-c), along with additional, selected darTG1-like and darTG2-like systems illustrating their frequent association with prophages and known phage defense elements.

f, Multiple sequence alignments of representative DarT toxins and DarG antitoxins.

The genes encoding the DarTG1 and DarTG2 systems we cloned were not near other known defense systems, but were each within prophages in E. coli strains C7 and 2-460-02_S4_C3, respectively (Fig. 1e). Homologs of the DarT toxins were previously shown to be single-stranded DNA ADP-ribosyltransferases^17,19. A multiple sequence alignment of DarT homologs (often annotated as containing DUF4433), including the ones we cloned and those previously characterized biochemically, revealed high similarity across the entire length of the proteins (Fig. 1f, top, Extended Data Fig. 1). There was complete conservation of many residues including a glutamate (E152 and E147 in the cloned DarT1 and DarT2, respectively) (Fig. 1f, red highlight), known to be critical for catalysis of these ssDNA ADP ribosyltransferases^17,19. We mutated this conserved glutamate to an alanine in both DarT1 and DarT2 and found that phage defense was abolished in each case (Fig. 1b–c). Hereafter, we refer to these inactive, mutant versions of the TA systems as DarT*G1 and DarT*G2.

In contrast to the DarT toxins, a multiple sequence alignment of their cognate DarG antitoxins revealed two distinct families (Fig. 1f, bottom, Extended Data Fig. 2). DarG1 contains a putative YbiA-like fold (80% confidence, Phyre2 prediction), a domain predicted to function in ADP-ribose processing²¹, while DarG2 features a highly conserved, N-terminal macrodomain known to hydrolyze the ADP-ribose modifications introduced by their cognate toxins^17,22,23. The C-terminal region of both proteins, which has been implicated in binding directly to DarT2¹⁸, is conserved in both protein families.

DarTG-mediated phage defense functions by abortive infection

Phage defense often occurs via an abortive infection (Abi) mechanism in which the infected cell dies but no phage progeny are produced, thereby preventing spread of the virus in a population. Abi mechanisms are traditionally thought to result from a defense mechanism that directly kills the host cell, but can also arise if the defense mechanism targets the virus, with the host cell dying because the virus triggers irreversible damage, such as chromosome degradation. One key characteristic of Abi mechanisms is that when most cells are infected at a high multiplicity of infection (MOI), the growth of the bacterial population stops, while at lower MOIs, the uninfected bacteria can continue to grow. To test whether the DarTG systems trigger Abi, we infected cells harboring either the native DarTG1 or the inactive DarT*G1 system with RB69 phage at varying MOIs, and tracked bacterial growth by OD₆₀₀ over time in microtiter plates (Fig. 2a, Extended Data Fig. 3a). At MOIs of 10 and 1, cells harboring DarTG1 did not grow and the OD₆₀₀ of the cultures decreased over time, indicating that cells were lysing. Consistent with this interpretation, we found that no viable cells remained 30 min post-infection for either DarTG1 or DarT*G1 containing cells (Fig. 2c). At MOIs of 0.1 and 0.01, the OD₆₀₀ of cultures harboring DarTG1, but not DarT*G1, increased over time indicating that cell growth continued as DarTG1 prevents the phage infection from spreading throughout the population (Fig. 2a). Similar trends were seen for DarTG2 cells infected with T5 (Fig. 2b, 2d, Extended Data Fig. 3b). These data suggested that both DarTG1 and DarTG2 provide phage defense via an abortive infection mechanism.

Figure 2. DarTG systems provide phage defense via an abortive infection mechanism.

a-b, Growth curves for strains with an empty vector or the indicated plasmid-encoded TA system after infection with RB69 (a) or T5 (b) phage at varying multiplicities of infection (MOI). The mean and S.D. of 7–11 technical replicates are presented; data are representative of 2 independent experiments (Extended Data Fig. 3).

c-d, Survival of E. coli encoding the indicated DarTG systems as measured by colony forming units (cfu) after 15 min of infection with RB69 at MOI 5 (c) or 30 min of infection with T5 at MOI 20 (d). Data are from 3 independent biological replicates.

e-f, One-step growth curves, showing measurements of phage concentration (plaque forming units (pfu/mL)) over time in cultures of DarTG1- or DarT*G1-containing cells infected with RB69 (e) or DarTG2- or DarT*G2 containing cells infected with T5 (f) at MOI 0.01 during the first round of infection. Data are from 3 independent biological replicates.

g, Time-lapse microscopy of DAPI-stained RB69-infected E. coli encoding the indicated DarTG1 systems. Scale bars, 4 μM.

h, Time course microscopy of DAPI-stained T5-infected E. coli bearing plasmids with the indicated DarTG2 systems. The 85 min time point of DarT*G2 DAPI is displayed at a different scale to accommodate the high intensity of DAPI staining present in this sample. Scale bars, 4 μM.

We directly assessed the number of RB69 progeny produced in DarTG1 and DarT*G1-containing cells grown in shaking flasks during infection. For cells containing the inactive DarT*G1 system, the initial burst occurred about 20 min after infection and released ~50 phages (Fig. 2e). In contrast, when DarTG1 was present, no phage progeny were detected up to 45 min post-infection. We obtained similar results for DarTG2 infected with T5 (Fig. 2f) although the timing of the burst was slower. Taken all together, our results support an abortive infection mechanism for DarTG-containing cells, in which activation of the toxin effectively thwarts the production of new phage particles, but infected cells do not survive.

We also tracked infected cells by time-lapse fluorescence microscopy, using the cell-permeable DNA dye 4’,6-diamidino-2-phenylindole (DAPI) to stain both bacterial and phage DNA. Phage particles appear as extracellular puncta in the DAPI channel (Extended Data Fig. 4a) and thus phage infections can be visualized in real time, while DAPI staining of host DNA simultaneously allows tracking of bacterial cell lysis (Extended Data Fig. 4b). We found that cellular DNA appeared more compacted in the DarTG1-containing cells prior to lysis (Fig. 2g, 20’ time point, Extended Data Fig. 4c). In the DarT*G1-containing cells, the DNA appeared diffuse throughout the infection process, and following lysis, new phage particles appeared (Fig. 2g, Extended Data Fig. 4d). There are some DAPI-stained, extracellular puncta present around both DarTG1 and DarT*G1 cells prior to cell lysis, which are likely unadsorbed phage. The number of these extracellular puncta increased substantially following lysis of cells harboring DarT*G1, but not DarTG1.

We found that T5 infections did not proceed effectively under time-lapse microscopy conditions, so we sampled from liquid cultures of DAPI-stained, T5-infected cells and imaged them at various time points after infection (Fig. 2h). Similar to DarTG1, we saw fewer DarTG2-containing cells lysing combined with a dramatic difference in the appearance of new phage particles following infection of DarT*G2 versus DarTG2-containing cells. These microscopy experiments support an abortive infection mechanism for both DarTG systems and, at least for DarTG1, suggest that DarT may affect DNA.

Activated DarT inhibits DNA synthesis by ADP-ribosylating DNA

Previous studies demonstrated that ADP-ribosylation of chromosomal DNA by DarT, either following ectopic expression of DarT or the artificial depletion of DarG, inhibits DNA replication in E. coli and M. tuberculosis^17,18,20. However, we hypothesized that after phage infection, its more relevant biological function is to prevent the replication of phage genomes. To test this hypothesis, we first monitored the uptake of radiolabeled thymidine at various time points after RB69 infection. In DarTG1-containing cells, DNA synthesis rates did not substantially increase, particularly compared to cells with DarT*G1 where the levels of thymidine incorporation increased ~15-fold following phage infection (Fig. 3a). Differences in DNA synthesis rates were detected as early as 5 minutes post-infection, indicating a rapid activation of the DarT1 toxin following phage infection. DarTG2 similarly prevented an increase in DNA synthesis following T5 infection (Fig. 3b). To rule out that these effects of DarTG1 and DarTG2 were simply non-specific, or secondary, effects of an activated phage defense system, we also measured DNA synthesis rates in conditions where the E. coli RnlAB TA system, which contains an RNase toxin, is activated. We infected either wild-type or ΔrnlAB cells with T4 Δdmd, a variant of T4 susceptible to defense by the RnlAB system²⁴. The rate of DNA synthesis was similar in wild-type and ΔrnlAB cells, supporting the conclusion that DarTG1 and DarTG2 specifically affect phage replication (Fig. 3c).

Figure 3. DarTG inhibits phage DNA replication by ADP-ribosylating viral DNA.

a-c, DNA synthesis rates as measured by ³H-labeled thymidine incorporation at the indicated time points after infection of strains encoding the indicated TA systems infected with RB69 at MOI 5 (a), T5 at MOI 20 (b), or T4 Δdmd at MOI 5 (c). Graph depicts 3 independent biological replicates.

d, Fraction of RB69 phage versus host-derived DNA, as determined by Illumina sequencing, for DNA extracted from DarTG1 or DarT*G1 cells infected with RB69 at an MOI of 5.

e, Sequencing read coverage from experiment in (d). One representative sample is shown.

f-g, Same as (d-e) but for T5 infection of DarTG2 and DarT*G2 cells at an MOI of 20.

h, Overview of an ELTA assay, wherein DNA extracted from infected E. coli cells is incubated with OAS1 protein activated with the synthetic dsRNA analog poly(I:C) (orange line), and ³²P-dATP. Incorporation of ³²P-dATP onto ADP-ribose modifications is measured by scintillation counting.

i, ELTA measurements (top) and dot blots with an anti-ADP-ribose antibody (bottom) for a ssDNA DNA fragment ADP-ribosylated in vitro compared to unribosylated DNA extracted from E. coli. Data from 4 independent replicates are shown.

j-k, ELTA measurements of DNA from E. coli encoding the indicated DarTG systems after infection with RB69 at MOI=5 and 15 min post-infection (j) or T5 at MOI=20 (k) at 20 or 40 min post-infection (k). Asterisks indicated p < 0.01 (two-sided t-test). Lower panel of (j) depicts a dot blot of corresponding DNA (DarTG1 cells, left and DarT*G1 cells, right) probed as in (i). Each dot on the graph represents an independent biological replicate for ELTA assays, and the dot blots are representative of at least 2 independent experiments.

We also deep sequenced DNA extracted from cells post-infection. For RB69 infection of cells harboring DarTG1, ~30% of the total DNA was phage derived, whereas for cells with DarT*G1, phage DNA was ~70% of the total (Fig. 3d). Taken together with the thymidine incorporation result for RB69 infections (Fig. 3a), this result indicates that active DarT1 prevents the replication and accumulation of new phage DNA. We also examined sequencing coverage across the RB69 genome, finding a non-uniform distribution of reads for cells with DarTG1, but not DarT*G1, suggesting that the toxin disrupts DNA replication elongation, not initiation (Fig. 3e).

For DarTG2 cells infected with T5, DNA sequencing indicated that only ~20% of reads were phage-derived at 20 min post-infection, but > 90% of reads were viral by 40 min (Fig. 3f). There were no major differences in the percentage of reads derived from T5 when comparing the DarTG2 and DarT*G2 cells, likely because T5 triggers rapid and complete degradation of the host chromosome in both cases²⁵. Nevertheless, the thymidine incorporation assays indicated that phage replication was strongly inhibited in DarTG2 cells (Fig. 3b). Consistent with this interpretation, agarose gel electrophoresis of DNA extracted from infected cells demonstrated a significant decrease in the total amount of DNA in DarTG2 cells compared to DarT*G2 cells (Extended Data Fig. 5). From the DNA sequencing we found that, as with DarTG1, there were also significant differences in read coverage indicating that DarT2 likely also disrupts phage DNA replication elongation (Fig. 3g). These results are consistent with an Abi mechanism in which the primary target of DarT is phage DNA, not host DNA.

DarT family toxins have been shown to ADP-ribosylate ssDNA in vitro and when overexpressed in bacterial cells¹⁷. We therefore hypothesized that RB69 and T5 infection activates the DarT1 and DarT2 toxins, respectively, and that activated toxin blocks replication of either dsDNA phage genome by ADP-ribosylating DNA in the cell. To assess whether DarTG-dependent ADP-ribosylation of DNA occurs in vivo following phage infection, we adapted a recently developed technique for measuring ADP-ribosylation of proteins to measure ADP ribosylation of DNA. This assay, enzymatic labeling of terminal ADP-ribose (ELTA), leverages the enzymatic activity of the innate immune protein, OAS1, to covalently attach dATP to ADP-ribose moieties²⁶ (Fig. 3h). This assay has not previously been used to assess the ADP-ribosylation of DNA. Therefore, we first confirmed that purified ADP-ribosylated ssDNA produced robust signal in an ELTA assay (Fig. 3i). We also confirmed ADP-ribosylation in a dot blot using an antibody recently shown to specifically detect ADP-ribose modifications on DNA (Fig. 3i, bottom)¹⁹.

We infected cells harboring darTG1 or darT*G1 with RB69, isolated DNA and then added OAS1 and ³²P-dATP to label ADP-ribose groups on the DNA (Fig. 3h). We detected a 6.5-fold increase in ADP-ribosylation of DNA extracted from DarTG1-containing cells 20 min post-infection compared to cells containing the inactive DarT1* variant by ELTA (Fig. 3j). We also detected a robust signal in an anti-ADP-ribose dot blot on the DNA extracted from RB69-infected cells containing DarTG1, but no signal for an equal amount of DNA extracted from infected cells harboring DarT*G1 (Fig. 3j, bottom). We also measured ADP-ribosylation of DNA following T5 infection of cells harboring darTG2, and found a similar, 3.5-fold increase in ADP-ribosylation in DarTG2 vs DarT*G2-containing cells after 20 min of infection (Fig. 3k). We also observed an ~5.7-fold increase in ADP-ribosylation of DNA after 40 min of infection when our sequencing indicated that virtually all DNA was of phage origin (Fig. 3f, k). Taken all together, our results indicate that DarT toxins are rapidly activated following phage infection and ADP ribosylate phage DNA to disrupt its replication.

DarT blocks RNA synthesis and alters phage protein production

We also asked whether the DarTG systems impact RNA and protein synthesis. RNA synthesis rates were assayed by monitoring radiolabeled uridine uptake following infection. RNA synthesis was significantly reduced in both DarTG1- and DarTG2- containing cells following infection with RB69 and T5, respectively (Figs. 4a–b). To assess how DarTG1 and DarTG2 affect protein synthesis, we pulse-labeled cells pre- and post-infection and then measured the incorporation of radiolabeled cysteine and methionine, using SDS-PAGE to resolve individual proteins. In contrast to DNA and RNA synthesis, protein synthesis rates did not differ substantially between DarTG1 and DarT*G1-containing cells infected with RB69 (Fig. 4c). However, there was a shift in which proteins were being synthesized, starting around 10 minutes post-infection. In particular, late-appearing species in DarT*G1-infected cells (Fig. 4c) or empty vector-infected cells (Extended Data Fig. 6a) were either not seen or were significantly reduced in DarTG1-infected cells, with many early-appearing species persisting throughout the 20 min time course. In contrast, when transcription is shutoff by treatment of cells with rifampicin, there was no shift in which proteins are being produced, only a reduction in band intensity (Extended Data Figure 6b). Together, these results demonstrate that phage infecting cells containing DarTG1 can initiate their gene expression program, but do not progress normally to the production of late gene products.

Figure 4. Activated DarT inhibits RNA synthesis and the timing of phage protein production.

a-b, RNA synthesis rates as measured by ³H-labeled uridine incorporation at various time points after infection of strains encoding the indicated TA systems infected with RB69 at MOI 5 (a) or T5 at MOI 20 (b).

c-d, Protein synthesis rates as measured by ³⁵S-labeled cysteine and methionine incorporation at various time points after infection for E. coli encoding the indicated TA systems and infected with either RB69 at MOI 5 (c) or T5 at MOI 20 (d) and resolved by SDS-PAGE. Data shown are representative of 2 independent biological replicates.

For DarTG2 and DarT*G2-containing cells infected with T5, the overall protein synthesis rates were again not substantially different (Fig. 4d). However, in this case, the banding pattern of DarTG2 and DarT*G2 cells remained more similar throughout the 40 min time course, though at the latest time points, DarTG2 cells appear to still be producing more middle gene products and less late gene products, as with DarTG1. Taken together, our results indicate that upon phage infection, both DarT toxins rapidly block DNA synthesis, and reduce RNA synthesis. Protein synthesis rates are not significantly affected, but the timing and identities of proteins synthesized are misregulated. We conclude that DarT activation disrupts the proper development of mature virions. Even if phage capsids were produced, the inhibition of DNA synthesis (Fig. 3) means no new phage genomes are available for packaging.

RB69 can escape DarTG1 defense by mutating 61.2 (adfA)

To gain additional insight into how DarTG systems function, we asked how phages can evolve to escape or overcome DarTG-mediated defense. To evolve resistant populations of phage, we serially passaged RB69 on both DarTG1 and DarT*G1-containing cells and monitored the pooled phage population for changes in susceptibility to DarTG1 (Fig. 5a)²⁷. By pooling together phages propagated on cells containing DarT*G1, we were able to increase the diversity of the population, which cannot normally replicate on DarTG1-containing cells. We isolated a DarTG1-resistant clone from each of five different, independently evolved populations (Fig. 5b) and sequenced their genomes. We also evolved, in parallel as a control, a population of RB69 only on DarT*G1-containing cells and sequenced the genome of this population. We then identified mutations present in the DarTG1-resistant populations and not the control population. For one clone, we could not identify any mutations, but the other four DarTG1-resistant clones each had a mutation in the same codon of the uncharacterized gene 61.2, which is predicted to encode a 212 amino acid protein. In three cases, the mutation results in a substitution of arginine-164 with a histidine, while in the fourth arginine-164 becomes a serine (Fig. 5b).

Figure 5. RB69 and SECϕ18 escape DarTG-mediated defense by two distinct mechanisms.

a, Overview of phage evolution experiment. DarTG and DarT*G-producing cells seeded in 96-well plates were infected with 10-fold serial dilutions of phage, with the highest well infected at an MOI of ~10. Following incubation, all cleared and partially cleared wells (dashed red lines) were pooled and the resulting phage used to re-infect cells. This process was repeated until phage were able to clear an increasing number of wells of DarTG-containing cells.

b, EOP of ancestral and evolved RB69 on DarTG1-containing cells. The mutations identified in gene 61.2 following genome sequencing are indicated on the right. The isolates shown were derived from 6 independently evolved populations of RB69.

c, Sequence alignment of the region surrounding residue 164 of gp61.2 for 20 homologs.

d, EOP of wild-type RB69 on E. coli attHK::darTG1 either bearing an empty vector or producing the wild-type RB69 gp61.2, the evolved RB69 gp61.2(R164H), or the wild-type T4 gp61.2. Lower panel depicts wild-type MG1655 bearing an empty vector. Data are representative of 2 independent biological replicates.

e, ELTA-based measurements of ADP-ribosylation of DNA extracted from E. coli encoding darTG1 or darT*G1 and infected with an evolved RB69 clone. Data of wild-type RB69 infection (Fig. 3j) is reproduced here for comparison. Data are from 3 independent biological replicates.

f, Colony forming units measured in cells bearing a vector with DarT1 either repressed (+glucose (glu)) or induced (+arabinose (ara)) along with an empty vector, a vector expressing the wild-type RB69 61.2, or the evolved RB69 61.2(R164H) variant.

g, Plaque assays showing the increased resistance of five independently evolved populations of SECϕ18 selected to overcome DarTG2 defense, with the mutations identified by whole-genome sequencing of each population.

h, Expression of the wild-type and an evolved allele of mga47 gene in cells containing DarTG2 or DarT*G2. Data are representative of 2 independent biological replicates.

i, ELTA-based measurements of ADP-ribosylation of DNA extracted from E. coli encoding darTG2 or darT*G2 and infected with the parental SECϕ18 strains or an evolved SECϕ18 clone. Data are from 2 independent biological replicates.

Intriguingly, a multiple sequence alignment of 124 homologs of gp61.2, which are encoded in both phage and bacterial genomes, revealed that histidine and serine, as well as asparagine, occur naturally at this position (Fig. 5c). Two phages (T4 and T6) closely related to RB69 both encode homologs of gp61.2 containing a histidine at this position of gp61.2. Additionally, we noted that 61.2 is close within the RB69 genome to dmd, which encodes an inhibitor of the RnlA toxin of the RnlAB toxin-antitoxin system²⁴. Together, these observations suggest that 61.2 homologs encode inhibitors of DarT toxins with the identity of position 164 strongly influencing the specificity of inhibition. To test our hypothesis, we asked whether the ectopic production of gp61.2(R164H) from evolved RB69 or the gp61.2 homolog from T4 could restore the ability of wild-type RB69 to infect cells containing DarTG1 on the chromosome. Indeed, producing either of these constructs improved the EOP of RB69 compared to cells harboring an empty vector or producing the wild-type gp61.2 from RB69 (Fig. 5d). We also assessed DarT1 toxin activity when cells were infected with evolved RB69 phages by measuring DNA ADP-ribosylation levels, reasoning that if DarT1 is inhibited by the gp61.2(R164H) variant, the toxin should be unable to ADP-ribosylate DNA. Consistent with this idea, there was no detectable DNA ADP-ribosylation in cells infected with the evolved RB69 producing the gp61.2(R164H) variant (Fig. 5e). We also tested whether the evolved gp61.2 could inhibit DarT1 in the absence of phage. Indeed, the gp61.2(R164H) variant, but not the wild-type gp61.2, could restore growth to cells producing DarT1 (Fig. 5f). Finally, we reasoned that a DarT1 inhibitor should not be essential to phage in the absence of darTG1. Using T4, which is amenable to genetic manipulation, we successfully made phage in which 61.2 contained a premature stop codon, or had a large region deleted, and found that in both cases, the phages had no deficiency in plaquing on cells lacking darTG1 (Extended Data Fig. 7). Taken all together, these data suggested that 61.2 encodes a DarT inhibitor and that the escape mutants we isolated improve the ability of the RB69 gp61.2 to inhibit DarT1. Based on these results, we have renamed 61.2 as adfA for anti-DarT factor A.

SECϕ18 can escape DarTG2 defense by mutating DNA polymerase

We also tried to evolve T5 phages to escape DarTG2-mediated defense, but were unable to identify escape mutants, even after several independent populations were passaged for 15 cycles each. We therefore tried to evolve SECϕ18 phage, which DarTG2 also defended against (Fig. 1b). In this case, resistant SECϕ18 populations were readily obtained, with increased infectivity emerging after 5 rounds of passaging (Fig. 5g). Full genome sequencing of the evolved escape phage populations and a control population passaged only on cells containing DarT*G2 revealed that each resistant population had accumulated one of five different mutations in mga47, which encodes the SECϕ18 DNA polymerase (Fig. 5g). The mutations in DNA polymerase did not obviously cluster when mapped onto a homology modelled structure of the protein.

There are three possibilities for how these mutations allow SECϕ18 phages to escape the DarTG2 system: (1) SECϕ18 DNA polymerase normally activates the toxin, with the mutant variant having lost the ability to do so; (2) the mutant, but not wild-type, variant of DNA polymerase neutralizes the toxin; or (3) the mutant DNA polymerase overcomes or somehow circumvents the activity of the toxin. We ruled out possibility 1 by overexpressing mga47 in cells containing the DarTG2 system, and found that it did not induce toxicity, indicating that mga47 is not sufficient to activate DarT2 (Fig. 5h). We addressed possibilities (2) and (3) by asking whether the DarT2 toxin was still able to ADP-ribosylate DNA following infection with the SECϕ18 escape phage. If ADP-ribosylation of DNA still occurs in the presence of the evolved phage, it would indicate that the DarT2 toxin is still active but that phages can replicate despite its activity. Indeed, we found that ADP-ribosylation of DNA occurs at a rate similar to that seen with wild-type SECϕ18 (Fig. 5i). This is in contrast to what we had observed when evolved RB69 infected DarTG1 cells (Fig. 5e), further underscoring that these phages have overcome DarTG-mediated defense by different mechanisms. These results strongly support a model in which SECϕ18 modifies its DNA polymerase to accommodate the ADP-ribose modifications on the DNA made by DarT2.

Discussion

DarTG-based phage defense

Like most TA systems, the DarTG system had previously only been characterized through the artificial overexpression of DarT toxin or the depletion of its antitoxin^17,18,20. This prior work elucidated the biochemical function of DarT toxins as ADP-ribosyltransferases that specifically target DNA and demonstrated that cells lacking DarG mount a DNA damage response as the ADP-ribosylation of DNA by DarT likely leads to replication fork stalling²³. However, the physiological function and native triggers of DarTG systems has been unclear. A Mycobacterium tuberculosis strain lacking darTG was reported to have a modest increase in growth after 15 days, though whether the growth defect of the wild-type results from ADP-ribosylation of DNA was not shown and why cells harbor a system that slows growth is not obvious. Here, we demonstrated that DarTG1 and DarTG2, taken from two strains of E. coli and representing the two different, major classes of DarTG systems, can provide host cells with potent defense against phages (Fig. 1).

The identification of phages that trigger these DarTG systems enabled us to characterize the activity of the toxins in vivo under growth conditions that naturally, and rapidly, liberate them. Our results indicate that within minutes after phages adsorb to the cell and inject their DNA, DarT toxins are active (Fig. 6). The liberated DarT then ADP-ribosylates phage DNA leading to a near complete cessation of DNA synthesis and an inhibition of RNA synthesis. These effects of the DarT toxins also impacted the program of phage gene expression and limited the production of late proteins (Fig. 4). Even if procapsids did assemble, DNA packaging would not occur as the phage genome is not fully replicated in cells with active DarT. Our results indicate that DarTG functions through an Abi-like mechanism in which infected cells die, but without producing progeny phage, thereby sparing uninfected neighbor cells. Although DarT can also target the host cell’s chromosome and is toxic when overexpressed, T5 and likely RB69 trigger chromosome degradation so there is likely little host DNA present for DarT to modify ^28,29. Even though the DarT toxin primarily targets viral DNA, phage inhibition occurs after key phage processes (e.g. chromosome degradation) have been initiated and thus the infected cells are unable to survive. Additionally, we isolated T5 escape mutants producing a variant phage DNA polymerase that enables replication of the phage genome despite its ADP-ribosylation by DarT and the subsequent release of mature virions. If DarT were aborting infection by modifying the host chromosome, such mutant phages would still be inhibited and not escape DarTG. Thus, we conclude that DarTG does not ultimately kill the host cell as in a conventional Abi mechanism, and instead acts to thwart phage replication directly.

Figure 6. Model for DarTG-mediated defense against phage.

When DarTG is not present (top), phage adsorb, inject their DNA and begin producing mRNAs and proteins. These proteins enable them to degrade the host chromosome, replicate their DNA, and produce the structural components required to make new phage particles. These components are assembled into capsids into which newly replicated genomes are packaged prior to lysis of the host cell. When DarTG is present (bottom), phage infection triggers release of the DarT toxin. The toxin ADP-ribosylates phage DNA, thereby inhibiting DNA synthesis and reducing transcription. Because some transcription occurs, particularly early on, phage are able to degrade the host chromosome and the host cell does not recover. However, due to the absence of newly replicated phage genomes and reduced late protein production, new phage particles are not produced.

Activation of the DarTG system by phage infection

How DarT becomes activated following phage infection is not yet clear. As noted above, DarT is active within 5 minutes post-infection, as measured by differences in DNA synthesis rates between DarTG and DarT*G cells (Fig. 3a), but without significant transcriptional induction of the system. Recent work on an E. coli ToxIN system demonstrated that the antitoxin toxI is intrinsically unstable and so must be continuously produced to bind and neutralize ToxN¹³. Phage-induced shutdown of host transcription following T4 infection leads to the liberation of ToxN, an endoribonuclease, ~20 min post-infection. DarT activation could work similarly. However, although the antitoxins of TA systems are often more unstable than their cognate toxins, we recently found that many are not unstable enough to produce pools of active toxin on the time-scale of a phage infection simply by blocking transcription¹⁰. Thus, we favor the possibility that a particular phage protein or factor somehow triggers activation of DarT. The involvement of a specific phage protein would potentially explain why DarTG1 and DarTG2 protect against different phages (Fig. 1a) and it might also explain why different growth conditions are required for their activity. Activation may require particular growth conditions if a host factor, such as a chaperone or protease, is required for releasing the toxin, but that factor is less abundant or less available in particular growth states.

If a specific phage protein is required to trigger DarT, it could stimulate separation of DarT from its cognate DarG antitoxin. However, the DarG antitoxins may not only bind and inhibit their cognate DarT toxins, as with canonical type II TA systems. Prior work on the enteropathogenic E. coli DarTG system indicated that these systems may also rely on a type IV mechanism in which the antitoxin indirectly antagonizes the toxin’s activity¹⁸. Notably, DarG1 contains structural homology to a YbiA-like domain that is predicted to be an ADP-ribose processing enzyme and DarG2 contains a classic macrodomain typically associated with ADP-glycohydrolases known to remove ADP-ribose modifications^21,22. Thus, in principle, DarT may always be active, but with sufficient DarG activity in uninfected cells to offset its toxicity. A phage product could inhibit, sequester, or degrade DarG, enabling the rapid accumulation of DarT activity.

Phage evasion of DarTG-mediated defense

Phages and their hosts are locked in an arms race wherein hosts acquire or evolve protection from phage infection, and phages evolve mechanisms to overcome these defenses. By evolving RB69 to escape the defense offered by DarTG1, we identified adfA (formerly 61.2) as a likely phage-encoded anti-DarT factor. The AdfA encoded by RB69 likely enables this phage to inhibit a different, but closely related DarTG system. The single substitutions in AdfA we identified enabled RB69 to then overcome the DarTG1 system we introduced. Notably, the mutations selected for in AdfA convert arginine 164 to histidine or serine, which occur naturally at the equivalent position in the AdfA homologs of T4 and T6. The presence of adfA homologs in T4, T6, RB69, and other T-even phages suggests that these phages have all been exposed to and selected to produce anti-DarT proteins, underscoring the notion that DarTG systems are critical and likely common phage defense systems in E. coli and possibly many other species. The adfA homologs are all found in similar genomic positions in the genomes of T-even phages and relatively close to dmd homologs that encode for RnlA toxin inhibitors. Thus, this region may represent an anti-TA system island, similar to the anti-CRISPR islands documented in mobile genetic elements of Enterobacteriaceae and Pseudomonas³⁰.

For SECϕ18, the mutants that escape defense by DarTG2 all harbored point mutations in mga47, which encodes the phage DNA polymerase (Fig. 5g). Our ELTA measurements indicated that the DNA of escape phages is ADP ribosylated to a similar extent as wild-type phages (Fig. 5i) suggesting that the mutations in mga47 are not leading to a disruption of DarT2 activity. These mga47 mutations could, in principle, enable phage replication to happen faster and before DarT2 activity accumulates. More likely, they somehow promote the replication of modified DNA, but precisely how this occurs remains to be studied.

TA systems in phage defense

Our bioinformatic screen indicated that DarTG homologs are often found in phage defense islands and the initial report on DarTG noted that homologs are sometimes inserted within type I restriction-modification systems. These results further supports the notion that proximity to known phage defense elements is a powerful means of identifying new phage defense systems^3,15. Intriguingly, the RnlAB family, which is known to consist of phage defense systems RnlAB and LsoAB, had the lowest phage defense score measured²⁴ (Table S1). This could indicate that only a subset of the RnlAB family are involved in phage defense. The two DarTG systems we examined are not encoded near any known or predicted phage defense system and instead are encoded within prophages (Fig. 1e). Prior work has found that prophages often harbor phage defense systems^31–33. The presence of TA systems on prophages has also been documented, and in some cases these systems may stabilize prophages within the bacterial chromosome, analogous to the role of some TA systems in stabilizing plasmids^34–36. However, the primary, or perhaps additional function, of these prophage-associated TA systems may be to exclude other phage, which in threatening the host cell, also jeopardize their survival and propagation.

Identifying additional TA systems that are associated with defense islands or encoded within prophages could help to identify additional systems that function in phage-defense. These systems may provide cells a diverse arsenal of anti-phage elements that complement restriction-modification and CRISPR systems. Importantly, those latter systems can only block phages by degrading phage DNA or RNA, whereas TA systems employ toxins with a wide range of enzymatic activities, including the DarT toxins we found here to ADP-ribosylate phage DNA.

Methods

Analysis of TA system association with defense islands

Protein sequences of all genes in a set of 38,167 bacterial and archaeal genomes were downloaded from the Integrated Microbial Genomes (IMG) database³⁷ in October 2017. These proteins were clustered using the ‘cluster’ option of MMseqs2 (release 2-1c7a89 ref³⁸), with default parameters. Clusters were further aggregated into larger clusters using four additional cycles of clustering, in which—in each cycle—a representative sequence was taken from each cluster using the ‘createsubdb’ option of MMseqs2 and representative sequences were clustered using the ‘cluster’ option with the ‘–add-self-matches’ parameter. For the first additional clustering cycle, the ‘cluster’ option was run with default parameters; for the additional cycles 2–4, clustering was run with sensitivity parameter ‘-s 7.5’, and for the additional cycle 4, the ‘–cluster-mode 1’ parameter was also added.

Each cluster with 20 or more genes was annotated with the most common pfam, COG, and IMG product annotations in the cluster. For each toxin belonging to one of the analyzed toxin families, genes from all clusters annotated as containing this toxin were aggregated. For each of the toxins families, the fraction of genes that have known defense genes in their genomic environment spanning 10 genes upstream and downstream the inspected gene was recorded. Defense score for each toxin family was calculated as previously described, except that genes belonging to TA systems were eliminated from the positive set³. A representative DarTG system from two distinct families were identified in E. coli strains for further analysis: DarTG1 is encoded in E. coli C7 (accession number CP010240.1) and DarTG2 is encoded in E. coli 2-460-02_S4_C3 (accession number NZ_JNRF00000000.1).

Analysis of DarTG1 and DarTG2 context and sequence

A blastp search was seeded with either DarT1 or DarT2. Resulting hits with e-values < 10⁻⁵⁰ were selected and associated nucleotide sequences were obtained. Prophage regions were identified as those with > 50% of genes having annotated phage elements (e.g. genes predicted to encode integrases, recombinases, phage tail or phage capsid proteins) or by using the Phaster phage identification web tool (www.phaster.ca). Phage defense elements were identified by referencing Genbank annotations with known phage defense systems (e.g. restriction-modification, abortive infection).

Representative examples of DarTG1 systems were identified through a blastp search seeded with DarG1 with e-values < 10⁻⁵⁰. Because of conserved gene order (darT always preceding darG), the protein sequence of the protein upstream was used for the DarT alignments. For DarTG2, systems were identified from previous studies^17,18,20, or through a similar strategy described for DarTG1. Protein sequences were aligned in Geneious v. 2020.0 (Biomatters) using the Clustal Omega algorithm and the resulting image files were exported. Similarity was determined using a Blosum62 matrix with threshold set to 1. In each column, the largest group of similar residues were identified and colored according to the indicated legend; any residues outside of this group are not colored. Annotations for DarG protein domains were identified through a BLAST (macrodomain) or Phyre2 structural prediction (YbiA-like, 80% confidence).

Strains and growth conditions

All bacterial and phage strains are listed in Table S2. Escherichia coli was grown at 37 °C in LB medium for routine maintenance and cloning. Phages were propagated by infecting E. coli MG1655 or E. coli MG1655 ΔrnlAB (T4 Δdmd only) cultures of OD ~0.1–0.3 at an MOI of 0.1 and incubated with aeration at 37 °C. Following clearing, any remaining cells were pelleted by centrifugation and lysates were filtered through a 0.22 μM filter. SECϕ18 was concentrated for ELTA experiments by centrifugation of cleared lysates for 2 hrs at 10,000 × g and resulting pellets were resuspended in ~100x less volume.

All RB69 infection experiments were performed 37 °C in LB medium, whereas experiments with T5 and SECϕ18 were performed at 30 and 20 °C, respectively, in M9 medium (6.4 g/L Na₂HPO₄-7H₂O, 1.5 g/L KH₂PO₄, 0.25 g/L NaCl, 0.5 g/L NH₄Cl medium supplemented with 0.1% casamino acids, 0.4% glucose, 2 mM MgSO4, and 0.1 mM CaCl₂), and overnight cultures for experiments were prepared in the same media used for the experiment. Media for selection or plasmid maintenance were supplemented with carbenicillin (100 μg/mL), chloramphenicol (20 μg/mL), or kanamycin (30 μg/ml) as necessary unless otherwise indicated. Induction of ectopic expression were effected with anhydrous tetracycline (100 ng/μL), arabinose (0.2% w/v), or vanillate (100 μM), as necessary.

Plasmid construction

All primer and synthesized gene sequences are listed in Table S3. DNA encoding candidate DarTG1 and DarTG2 system open reading frames as well as 200 bp upstream and overlapping ends to the pBR322 vector (MLR1 and MLR2, respectively) were commercially synthesized by Integrated DNA Technology as gBlocks and assembled into a promoter-less backbone of pBR322 amplified with primers MLR3 and MLR4 by Gibson assembly. The pBR322-DarTG1^E152A, and pBR322-DarTG2E^147A variants were inserted into the corresponding wild-type plasmids by site-directed mutagenesis with inverse PCR using primers MLR5 and MLR6, or MLR7 and MLR8, respectively. To generate the chromosomal insertion of darTG1, the darTG1 region containing its native promoter was amplified with primers MLR9 and MLR10, and cloned into pAH144 vector linearized with primers MLR11 and MLR12 by Gibson assembly. The mga47 open reading frame was amplified from wild-type or evolved SECϕ18 with primers MLR13 and MLR14 and cloned via Gibson assembly into a version of pBAD33 with a kanamycin resistance cassette linearized with MLR15 and MLR16. The 61.2 open reading frame was amplified from either the wild-type RB69 or an evolved clone from population 1 with primers MLR17 and MLR18, or from T4 phage with primers MLR19 and MLR20 and cloned into pKSV45-P_van-kan linearized with primers MLR21 and MLR22. The pEXT20-61.2^RB69 and pEXT20-61.2^RB69 (R164H) vectors were generated by Gibson assembly in which the pEXT20 vector was linearized with primers MLR29 and MLR30 and the insert was amplified from the wild-type RB69 DNA or the evolved phage using primers MLR31 and MLR32. The p-DarT1 construct, in which the toxin is expressed from an arabinose promoter, was produced by amplifying DarT1 using primers MLR33 and MLR34 and digested with SacI and KpnI, and ligated into the pJB37 vector digested with the same enzymes³⁹. All plasmids were confirmed by Sanger sequencing of the inserts.

Strain construction

Plasmids described above were introduced into MG1655 by TSS transformation or electroporation⁴⁰. A single copy of darTG1 was inserted onto the MG1655 chromosome at the HK022 attachment site using the CRIM system⁴¹, using the pAH144-darTG1 vector with the pAH69 helper plasmid. A scarless, unmarked deletion of rnlAB was constructed by two-step allelic exchange. The sacB-neoR cassette from pIB279 amplified with ML23 and ML24 and inserted using lambda Red recombinase^42,43. The deletion oligo (MLR25) was subsequently transformed into lambda Red-containing cells and counterselection was performed on 5% sucrose plates as previously described. The deletion of rnlAB was verified by PCR amplification and sequencing of sucrose-sensitive, kanamycin-sensitive clones.

Phage gene deletions

T4 mutants were generated using a CRISPR-Cas system for targeted mutagenesis⁴⁴. Sequences for RNA guides to target Cas9-mediated cleavage were designed using the toolbox in Geneious Prime 2021.2.2 targeting adfA but nowhere else in the T4 genome. The guides were inserted into the pCas9 plasmid as previously described ⁴⁵ using oligos MLR27 and MLR28, and resulting pCas9-61.2-cr6 plasmid was transformed into MG1655 and tested for its ability to target the T4 genome by restricting T4. Escape plaques were isolated, and two mutated variants with disruptions of the 61.2 gene were identified by sequencing: a premature stop codon at position 68 and a large deletion of residues 65–196 (Extended Data Fig. 7).

Plaque assays, phage titering, and efficiency of plating assays

Overnight cultures of the indicated bacterial cells were mixed 1:200 with melted 0.5% agar made with either LB or M9 medium. The agar was then overlaid onto a 1.2% agar plate composed of the corresponding medium. Ten-fold serial dilutions of the indicated phage were spotted on top. For RB69 experiments, plates composed of LB medium were incubated at 37 °C, whereas for T5 and SECϕ18 experiments, plates were composed of M9 medium and incubated at 30 and 20 °C, respectively. All experiments were performed independently at least 3 times.

Growth curves

E. coli bearing the indicated DarTG plasmids were grown to an OD₆₀₀ of 0.3 and then mixed with phages at the indicated MOIs. 8–12 replicate 100 μL aliquots of the resulting mixtures were seeded into 96-well plates, and growth was measured at 15 min intervals with orbital shaking on a plate reader (Biotek) at 37 °C for DarTG1-RB69 experiments and 30 °C for DarTG2-T5 experiments. Outliers were removed and data presented are the mean and standard deviation of 6–12 plate replicates. Experiments were replicated independently at least 2 times.

Bacterial survival after phage infection

E. coli MG1655 containing either DarTG1 plasmids (pBR322-darTG1 or pBR322-darTG1) or DarTG2 plasmids (pMLR6, pMLR7) were grown to OD₆₀₀ of 0.3 and then phages were added at an MOI of 5 for the RB69 experiments and MOI of 20 for the T5 experiments. Suspensions were incubated without shaking for 5 min to allow for adsorption, then cultures were washed twice with an equal volume of media to remove unadsorbed phage. Infected cells were incubated at the appropriate temperatures for an additional 5–10 min before serial dilutions were plated on LB plates to determine bacterial colony forming units. Data presented are the mean and standard deviation of 3 independent biological replicates.

One-step growth curves

Bacterial cells with appropriate plasmids were infected at an MOI of 0.01 in LB medium at 37 °C (darTG1-RB69 experiment and rnlAB-T4Δdmd experiment) or in M9 medium at 30 °C (darTG2-T5 experiment). Samples were collected at regular time intervals, serially diluted, and immediately spotted on 0.5% LB top agar mixed with a wild-type MG1655 indicator strain to determine plaque forming units/mL. Data presented are the mean and standard deviation of 3 independent biological replicates.

Microscopy

Cells were grown under conditions described above to an OD₆₀₀ of 0.3, then stained with DAPI (1 μg/mL) for 10 min. Cells were then treated with 100 μg/mL carbenicillin or infected with phages at a MOI of 5 (RB69) or 20 (T5), as indicated in figure legends. For darTG1-RB69 experiments, the resulting suspension was washed once to remove unadsorbed phages and 1 μL was spotted onto a 1.5% agarose pad prepared with LB containing DAPI (1 μg/mL) and placed in a 35 mm glass-bottom dish with 20 mm micro-well #0 coverglass (Cellvis) sealed with parafilm. Phase-contrast and epifluorescence images were taken at 10 min intervals on a Zeiss Observer Z1 microscope using a 100x/1.4 oil immersion objective and an LED-based Colibri illumination system using MetaMorph software (Molecular Devices, LLC).

For darTG2 experiments, the DAPI-stained, infected cells were incubated with shaking at 30 °C and 1 mL aliquots were removed at the indicated time points. Cells were washed once to remove unadsorbed phage, spotted on 1.5% agarose pads prepared with M9 medium, placed on glass slides, and 3–5 fields were immediately imaged. Data presented are representative of at least 2 independent biological replicates.

Representative cells from the 20 min time point of the DarTG1 and DarT*G1 datasets were analyzed in Fiji using the “Plot Profile” tool and the measurements were normalized to cell length. Each cell profile were smoothed in Excel using the “Exponential Smoothing” tool with a damping factor of 0.7. Fluorescence profiles were visually inspected and categorized into Diffuse, Asymmetric, and Bimodal groups.

DNA extractions after phage infection

For bulk DNA assays, cells in a volume of 10–25 mL were grown and infected as described above for bacterial survival assays were incubated with aeration until the indicated time points. One milliliter of the bacterial culture was pelleted and washed once to remove unadsorbed phage. DNA used for sequencing and ELTA experiments were derived from 10 mL of culture harvested by vacuum filtration onto a 0.2 μM filter and washed once with an equal volume of media. Filters were placed in conical tubes and cells were removed by scraping into 1 mL of media and pelleting the resulting suspension. In all experiments, samples were flash-frozen in liquid nitrogen and stored at −80 °C. DNA was extracted using the Gentra Puregene Kit (Qiagen). For agarose gels, an equal volume of DNA was loaded for each sample and visualized on a 0.7% agarose gel stained with ethidium bromide. The amount of DNA was quantified in ImageJ (http://imagej.nih.gov/).

Incorporation assays

Bacterial cells were infected under the same conditions as described for bacterial survival assays and flasks were maintained with aeration at the appropriate temperature. For DNA and RNA synthesis measurements, an aliquot of the bacterial culture was removed prior to phage addition (T = 0 min) or at the indicated time points post-infection and transferred to a microcentrifuge tube containing methyl-³H thymidine (Perkin-Elmer) (40 μCi/mL) or uridine, 5–6[³H] (Perkin Elmer) (6 μCi/mL). Tubes were incubated at the same temperature as infected cultures for 1 min (37 °C growth) or 2 min (30 °C growth). Reactions were quenched by addition of non-radioactive thymidine or uridine (1.5 mM) and incubated an additional 1 or 2 min. Samples were added to ice cold trichloroacetic acid (TCA) (10% w/v) and incubated at least 30 min on ice to allow for precipitation. The resulting sample was vacuum filtered onto a glass microfiber filter (Whatman, 1820–024) that had been pre-wetted with 5% w/v TCA. Filters were washed with 35x volume of 5% w/v TCA, then with 5x volume of 100% ethanol. Air dried filters were placed in tubes with scintillation fluid and measured in a scintillation counter (Perkin Elmer). Three independent biological replicates are presented.

For protein synthesis assays, a 1 mL aliquot of bacterial culture was removed either prior to phage addition or rifampicin treatment (t = 0 min), or at indicated time points post-treatement, and incubated with EasyTag™ EXPRESS-³⁵S Protein Labeling Mix, [³⁵S] (Perkin Elmer) at 44 μCi/mL for 2 min (37 °C) or 10 min (30 °C). Labeling was quenched with addition of unlabeled cysteine and methionine at 3 mM, and proteins were precipitated by addition of ice cold TCA (13% w/v) and incubation on ice for at least 30 min. Samples were pelleted, washed twice with 100% acetone, then resuspended in resuspension buffer (100 mM Tris (pH 11.0), 3% w/v SDS). Samples were resolved by 4–20% SDS-PAGE, after which the gel was soaked in Gel Dry Drying Solution (Thermo Fisher), dried on a vacuum dry gel dryer, and exposed to a phosphorimaging screen for 1–4 days. The screen was imaged on a Typhoon scanner (GE Healthcare) at 50 μM resolution. A representative image of two independent experiments is presented.

DNA sequencing

DNA collected as described above was sheared in a Diagenode Bioruptor 300 sonicator water bath for 15 × 10 s cycles at maximum intensity. Samples were further purified with a Zymo PCR cleanup kit and concentrations were determined on a Nanodrop. Sequencing libraries were prepared by shearing DNA to 150–200 nt using a Bioruptor (Diagenon), then cleaning and size-selecting DNA using AMpure XP magnetic beads (Beckman Coulter). The resulting DNA was end-repaired with T4 DNA polymerase (NEB), Klenow DNA polymerase (NEB), and T4 polynucleotide kinase (NEB), AMPure bead purified, then 3’ adenylated using Klenow Fragment (3’–5’ exo) (NEB). Adaptors were ligated with blunt/TA ligase master mix (NEB) and amplified with barcoding primers for ~10 cycles with KAPA Hi-Fi Master Mix (Roche). The resulting libraries were cleaned and size selected with Ampure XP beads and run on a NextSeq500 at the MIT BioMicroCenter. Sequencing data is available at the NCBI Sequence Read Archive (BioProject PRJNA776027).

Reads were trimmed and mapped using Geneious 2020.0 (Biomatters) with the Geneious medium-sensitivity mapping to both MG1655 (accession number CP025268.1) and phage reference genomes (RB69 accession number NC_004928; T5 accession number NC_005859.1). The percent of reads mapping to the phage genome was calculated as (reads mapped to phage/total reads mapped*100) for each sample, and is the average of sequencing data generated from 2 or 3 independent biological replicates. Coverage maps were generated by dividing the number of non end-gap characters at each position by the sum of all positions and multiplied by 10⁹, then generating a moving average with an interval of 100. The resulting data were plotted as a function of position. Coverage data presented are from one replicate representative of 2–3 independent biological replicates.

ELTA assays

A positive control was generated in vitro via enzymatic ADP-ribosylation of ssDNA as described previously¹⁷. Briefly, a 10 μL reaction containing 30 ng/μL of DNA substrate containing a TCTC motif (MLR26), 50 mM Tris-HCl pH 8.0, 150 mM NaCl, 100 μM NAD⁺, and 30 ng/μL recombinant Taq DarT (from Thermus aquaticus) was incubated at 37 °C for 2 hr. The reaction was purified using the Monarch PCR and DNA cleanup kit (NEB) to remove unincorporated dATP. The successful incorporation of ADP-ribose was confirmed by visualization on 3% agarose gel after ELTA labeling with Cy5-dATP. The experiment in Fig. 3I was performed with 10 femtomole of the in vitro produced ADP-ribosylated DNA, while the control reaction with non-ribosylated DNA was performed using 1 μg of sheared DNA extracted from E. coli cells grown to mid-log.

DNA for all other ELTA assays was harvested as described above. For SECϕ18 experiments in Figure 5i cells were infected at an MOI of 1000 due to the low adsorption of this phage under the growth conditions (<1%) and infected cultures were incubated for 2 hours prior to harvesting. The DNA was sheared for 15 × 30 sec cycles on the BioRuptor Sonicator on the maximal intensity setting and even shearing was confirmed by agarose gel electrophoresis. The resulting samples were further purified with a Zymo DNA Clean and Concentrator kit (Zymo Research).

Each 20 μL ELTA reaction consisted of 10 mM Tris-HCl pH 7.5, 20 mM magnesium acetate, 2.5 mM DTT, 10 μCi (0.05 μCi/μL) dATP, [α-³²P] (Perkin Elmer), 2 μM OAS1, 100 μg/mL low molecular weight poly I:C (Invivogen), and 1 μg of DNA. These reactions were incubated at 37 °C for 2 hours after which the DNA was purified as for the control DNA. Half of the resulting sample was transferred to a 4 mL scintillation tube and counted as above. Data were normalized to a reaction run with no DNA input. This ADP-ribosylated ssDNA control was run alongside each set of reactions to confirm normal assay function. Two or three independent biological replicates are presented in each graph.

Anti-ADP-ribose dot blots

Dot blots were performed as described previously¹⁹. A 2 pmol aliquot of either the ADP-ribosylated ssDNA substrate or the corresponding unribosylated substrate (described above) (Fig. 3i), or 2 μg of DNA isolated from phage-infected cells (either DarTG1 or DarT*G1-containing, as described above) (Fig. 3j), was spotted onto a nitrocellulose membrane and cross-linked at 1200 J with a Stratalinker UV crosslinker. The membranes were probed with poly/mono-ADP ribose antibody (E6F6A) (Cell Signaling) at 1:1000. Membranes were developed with SuperSignal West Femto Maximum Sensitivity Substrate (Thermo-Fisher) and imaged on a ChemiDoc system (Biorad). One blot is depicted, which is representative of two independent experiments.

Phage evolution experiments

Each RB69 evolution experiment consisted of 8 wells of cells producing DarTG1 (ML3) and 8 wells of cells producing DarT*G1 (ML4) seeded with ~10⁶ cells/well from overnight bacterial cultures in a 96-well plate in LB medium. Cells were infected with 10-fold serial dilutions of phage cultures with the highest MOI at 10, and one well of each strain uninfected to control for cross-contamination. Plates sealed with a Breathe Easy plate seals (Sigma) and aerated by micro-orbital shaking in an ELTI plate shaker at 37 °C. Plates were incubated for 4–6 hours. The cleared and partially cleared wells were pooled, centrifuged to remove unlysed bacteria, and the resulting supernatants were used to infect the next evolution round. Resistance to DarTG1 was monitored by number of cleared wells as well as by EOP assays. Once resistance to DarTG1 was observed, single plaques were isolated in top agar plates and tested for resistance. One resistant clone from each of 5 independently evolved phage populations was propagated for further analysis and sequencing as described below. A control population that was only passaged on cells containing DarT*G1 was also analyzed.

The SECϕ18 evolution experiments were designed as above, except wells were seeded with E. coli cells producing DarTG2 (ML5) and DarT*G2 (ML6), propagated in M9-glucose medium, and plates were incubated at 20 °C for 16 hrs. Sequencing was performed on evolved populations as opposed to single clones described for the RB69 evolution.

To extract phage DNA, lysed bacterial cultures were centrifuged to remove bacterial cells, then 100 μL of supernatant containing phages was incubated with TURBO DNase I (0.1 mg/mL) and RNase A (0.1 mg/mL) for 45 min at 37 °C to remove residual nucleic acids from bacterial cell lysis. Nucleases were inactivated by addition of EDTA (10 mM) and incubation for 15 min at 37 °C. Proteinase K (0.2 mg/mL) was added and samples incubated at 50 °C for 30 min to disrupt phage capsids. Samples were centrifuged to 2 min at 21000 × g and DNA precipitated from supernatant by standard sodium acetate/ethanol precipitation. The resulting DNA was processed for Illumina sequencing as described for E. coli infection sequencing experiments and mapped to the RB69 reference genome (NC_004928) or SECϕ18 reference genome (LT960609). Only mutations that arose uniquely in the phages evolved on E. coli containing DarTG, that were not also found in the phages evolved on E. coli lacking active DarTG systems, were considered. For the SECϕ18 control population, only the two genes (mga32 and mga47) that had acquired mutations in the populations evolved in the presence of DarTG2, were PCR-amplified and Sanger sequenced. Mutations were only identified in mga32 for the control experiment, so only mga47 mutations were considered relevant to DarTG2 resistance. Mutations were identified using Geneious Variant Finder with the threshold set to > 20%.

Data Availability

The bioinformatic analysis was performed using protein sequences available in the Integrated Microbial Genomes (IMG) database (https://img.jgi.doe.gov/). DarTG1 and DarTG2 were identified in Escherichia coli C7 (NCBI Accession GCA_001901425.1) and Escherichia coli 2-460-02_S4_C3 (NCBI Accession GCA_000704545.1). Sequencing data are available on NCBI (BioProject PRJNA776027). Sequencing data were aligned to reference genomes of MG1655 (Accession number CP025268.1), RB69 (Accession number NC_004928), SECϕ18 (accession number LT960609), or T5 (accession number NC_005859.1). All other source data has been deposited to Mendeley Data, V1, doi: 10.17632/v9bmr549nf.1.

Extended Data

Extended Data Fig. 1. Sequence alignment of DarT homologs.

Full multiple sequence alignments of representative DarT homologs corresponding to Fig. 1f.

Extended Data Fig. 2. Sequence alignment of DarG homologs.

Full multiple sequence alignments of representative DarG homologs corresponding to Fig. 1f.

Extended Data Fig. 3. DarTG systems provide phage defense via an abortive infection mechanism.

Growth curves for strains with the indicated plasmid-encoded TA system after infection with RB69 (a) or T5 (b) phage at varying multiplicities of infection (MOI). The mean and S.D. of 6–12 technical replicates are presented; data are an independent biological replicate for data presented in Fig. 2a–b.

Extended Data Fig. 4. DAPI-staining detects phage particles and changes in DNA compaction in infected cells.

a, Snapshots taken of cultures of E. coli MG1655 cells containing DAPI 10–20 min after addition of the indicated phage. b, Time-lapse series of DAPI-stained E. coli MG1655 cells on agarose pads containing 100 μg/mL carbenicillin (top) or untreated (bottom). Lysed cells are indicated with red arrows. c-d, Fluorescence intensity profiles of individual cells are plotted for DarTG1 (c) or DarT*G1 (d) cells. Profiles for 18–20 representative cells are shown for each condition and are grouped into diffuse (left), bimodal (middle), or asymmetric (right). Scale bars, 4 μM.

Extended Data Fig. 5. Less DNA is recovered from infected cells when a DarTG2 system is present.

Total DNA extracted from cells containing darTG2 or darT*G2 and infected with T5 for the times indicated. The intensity of the band at each time post-infection relative to the pre-infection band is reported at the bottom. Two independent replicates are shown.

Extended Data Fig. 6. Activated DarT inhibits the timing of phage protein production.

Protein synthesis rates as measured by ³⁵S-labeled cysteine and methionine incorporation at various time points after infection for E. coli bearing the indicated plasmids and infected with either RB69 at MOI 5 (a) or treated with 300 μg/mL rifampicin (rif) (b). Data shown are representative of 2 independent biological replicates.

Extended Data Fig. 7. Gene 61.2 is not essential in T4.

Plaque assays of the wild-type T4, or variants with either a premature stop-codon in gp61.2 (L68*) or a large deletion of residues 65–196 (T4 61.2^Δ65–196) on E. coli with an active or inactive version of the DarTG1 system.