DdmABC-dependent death triggered by viral palindromic DNA sequences

SUMMARY

Defense systems that recognize viruses provide important insights into both prokaryotic and eukaryotic innate immunity mechanisms. Such systems that restrict foreign DNA or trigger cell death have recently been recognized, but the molecular signals that activate many of these remain largely unknown. Here, we characterize one such system in pandemic Vibrio cholerae responsible for triggering cell density-dependent death (CDD) of cells in response to the presence of certain genetic elements. We show that the key component is the Lamassu DdmABC anti-phage/plasmid defense system. We demonstrate that signals that trigger CDD were palindromic DNA sequences in phages and plasmids that are predicted to form stem-loop hairpins from single-stranded DNA. Our results suggest that agents that damage DNA also trigger DdmABC activation and inhibit cell growth. Thus, any infectious process that results in damaged DNA, particularly during DNA replication, can in theory trigger DNA restriction and death through the DdmABC abortive infection system.

Graphical abstract

In brief

Robins et al. report on the identification of cis-acting palindromic DNA sequences in plasmids and phages targeted by the V. cholerae Lamassu DdmABC defense system. A single mutation in the PriA DNA repair-replication protein blocks DdmABC activity. Moreover, DdmABC expression increases UV sensitivity, suggesting DNA repair is one signal for activation.

INTRODUCTION

Anti-viral defense systems have a long history of identification in prokaryotic systems through both bacteriophage and bacterial genetics.¹ Many of these anti-phage systems have recently been recognized to have eukaryotic counterparts that function in viral defense.^2,3 Recent advances in genomics and bioinformatics have facilitated the prediction of numerous putative anti-phage systems that have uncharacterized mechanisms of action.⁴ The genes for these anti-phage systems are often found positioned together on genetic elements that were likely horizontally transferred between bacterial strains.⁵ Frequently, these anti-phage systems have been experimentally shown to function in Escherichia coli^6–8; however, the signals that allow such systems to recognize phage-infected bacterial cells remain ill-defined.

Phages are the most abundant biological entities on planet Earth and are predicted to outnumber bacteria by a factor of 10-fold.⁹ Given this continuous predation, bacteria have evolved anti-phage systems and, in turn, phages have evolved escape strategies for these formidable defenses. Together these define an enormous ‘‘arms race’’ that shapes microbial ecology in all inhabited environments.^10–12 It is important to note that understanding how anti-phage defense systems work at a mechanistic level has provided numerous technological tools that now shape and drive biomedical and bioscience innovations.^13,14

Phage and mobile elements in bacterial pathogens are also highly relevant to diseases, including the lethal diarrheal syndrome called cholera.^15,16 For pathogenic Vibrio cholerae, the acquisition and chromosomal integration of the genome of a filamentous phage that encodes cholera toxin (CTX-ϕ) was a prerequisite for the emergence of strains responsible for all seven recognized cholera pandemics.¹⁷ However, the process that leads to the acquisition of CTX-ϕ involves multiple phage-like elements and a chromosomal island that encodes the toxin-co-regulated pilus (TCP), an adhesin and clumping factor that serves as both the receptor for the phage as well as a critical human intestinal colonization factor.^17–20

All pandemic strains of V. cholerae are genetically similar and classified as either classical or El Tor biotypes.²¹ Classical strains are believed to be responsible for the first six pandemics, and separate waves of El Tors variants account for the current ongoing 7th pandemic.^{15,16,22–24} Other genetically distinct strains have acquired both TCP and CTX-ϕ, but these strains are not responsible for widespread epidemics.^25–27 Thus, the unique repertoire of genetic elements in the 7th pandemic El Tor strains is believed to contribute to the fitness of this dominant pandemic clade of V. cholerae that emerged with the subsequent elimination of classical strains as the cause of cholera worldwide.^16,28

V. cholerae has been documented to be subject to phage predation that correlates with the collapse of cholera epidemics.^29–32 These observations have prompted many studies on phage resistance mechanisms in V. cholerae, as well as phage-encoded countermeasures that block some anti-phage systems.^33,34 For example, quorum regulation mediated by auto-inducers, molecules that accumulate typically under high-density growth conditions, was reported to promote phage resistance in 7th pandemic V. cholerae strains by an unclear mechanism.³⁵ Although V. cholerae cells grow to become dense biofilms on the intestinal mucosal surface,³⁶ phage predation still occurs in vivo and human victims shed copious amounts of phages as well as clumps of densely packed viable V. cholera cells.³⁷

All contemporary pathogenic V. cholerae strains have acquired two chromosomal islands termed the Vibrio 7th Pandemic Islands I (VSP-1) and II (VSP-2).²⁸ Gene products encoded by both islands have recently been shown to be capable of restricting phage growth and plasmid persistence by mechanisms including programmed cell death of infected cells.^6–8,38 These conclusions have been largely been based on the over-expression of VSP-1 or VSP-2 genes in the surrogate bacterial species E. coli. More recently, quorum-regulated VSP-2 genes in the ddmABC operon have been demonstrated to modestly inhibit the growth of a single identified vibriophage by two to three log10s but only when co-present with specific VSP-1 putative anti-phage genes.³⁸ These authors speculated that altered nucleotide pools by the expression of the VSP-1 deoxycytidine deaminase product AvcD might impair phage DNA replication and, consequently, we considered a model where the VSP-2 DdmABC proteins could recognize defective replication intermediates or DNA damage as signals. Importantly, such signals that activate any anti-phage and anti-plasmid defense system in V. cholerae remain ill-defined.

To discover the signals that activate these defense systems in V. cholerae, we analyzed a laboratory plasmid and two phages we identified that were specifically restricted in growth by VSP-2 genes ddmABC that encode the Lamassu-like anti-phage system.⁴ The transformed plasmid remarkably caused bacterial death only at high cell density and in a Lamassu-dependent manner. Using genetic approaches, we identified the cell-death-inducing signal as a palindromic sequence in the plasmid that is predicted to form a single-stranded DNA (ssDNA) stem-loop (or hairpin). We cloned sequences from vibriophage that triggered the Lamassu-dependent death of cells when grown at high density. We found that all these phage-encoded sequences were partially palindromic and predicted to form stem-loop hairpin structures. Additional mutations outside the plasmid that could suppress the density-dependent, hairpin-triggered cell death response mapped to the quorum-regulated ddmABC operon as well as an essential conserved host gene product (PriA) that recognizes stalled DNA replication forks and has roles in DNA repair.^39,40 DNA-damaging agents also triggered higher levels of cell death dependent on DdmABC but this was reduced in the PriA mutant independent of DdmABC. We present a model for how phage infection is detected by the DdmABC system and how that recognition likely triggers cell death through the activation of a nuclease domain encoded by the DdmA protein. If correct, this model predicts that the DdmABC system can detect not only phages and vectors (or plasmids) that encode palindromic sequences but also viral and host processes that cause chromosomal or viral DNA degradation.

RESULTS

Identification of a sequence in a plasmid that triggers V. cholerae cell death at a high cell density

El Tor 7th pandemic strains of V. cholerae have been reported to inhibit the maintenance of certain plasmids.⁸ To independently confirm this observation, we attempted to transform a derivative of a colE1 origin laboratory plasmid pWR1566 that encodes LacZ and carbenicillin (Carb) resistance into a lacZ mutant of a 7th pandemic El Tor V. cholerae strain (WPR2700). LacZ expression was monitored with the colorimetric substrate bromo-4-chloro-3-indolyl-β-d-galactopyranoside (XGAL), which forms a dark blue product when hydrolyzed by LacZ. Although LacZ expression was observed in Carb-selected transformants, colonies grew poorly and released copious amounts of LacZ as indicated by the appearance of the dark blue halos around colonies. These results suggested that cell death and lysis were occurring; a phenomenon previously described as ‘‘blue ghosts’’ in E. coli⁴¹ (Figure S1A). Surprisingly, WPR2700 cells carrying the plasmid grew well in liquid media containing Carb, but, when those liquid cultures were plated on agar, regardless of Carb, we observed cell death that was most apparent where cells were present at high density; accordingly, we called this phenomenon cell density-dependent death (CDD) (Figure S1B).

Because CDD occurred only at high cell density on a solid surface, we reasoned this may involve quorum-regulated gene products as well as recognition of a signal present on the plasmid that triggered cell death by abortive restriction. To isolate mutations that conferred resistance to the CDD signal, plasmid-transformed WPR2700 cells were spotted at high density on agar containing Carb and XGAL. While most cells died, we did observe exceedingly rare and mostly white colonies that arose at a frequency of approximately 10⁻⁷ after resuspending and plating on agar Carb-selective XGAL agar. Because the majority of these Carb^r clones no longer produced LacZ by colorimetric screening, we predicted that plasmid sequence deletions had occurred, and indeed sequencing of the purified mutant plasmids revealed deletions that removed a common 5′ region upstream of the lacZ gene (Figure S1C). Curiously, these deletions did not occur between short, directly repeated homologous sequences as previously observed in other studies of spontaneous deletion formation⁴² (Figure S2). We reasoned that this deleted region of pWR1566 must encode a cis-acting sequence that induces CDD. Examination of deleted sequences revealed a 79-nucleotide plasmid sequence originally derived from the primary replication origin of E. coli bacteriophage T7 including an RNase III site that had previously been present in the parental vector (Figures S1C and S1D). These sequences form hairpins as double-stranded RNA (dsRNA) that can be processed by RNases to regulate transcript stability and gene expression.⁴³ There is no known role for these secondary structures in DNA, although they are energetically favored to form in ssDNA because of their palindromic nature. Plasmids lacking this sequence due to a constructed deletion (see below) or related plasmids that lacked this same T7-derived sequence were maintained at high density.

Identification of V. cholerae suppressors of CDD

Using the same selective approach, we identified V. cholerae chromosomal mutants that could maintain and replicate pWPR1566 and also grow as healthy blue colonies on agar containing Carb and XGAL. Next-generation sequencing of each of these mutants was used to identify CDD suppressor alleles in the V. cholerae chromosomal genome and also estimate the copy number of pWR1566 by comparing the level of reads that mapped to both the plasmid and chromosomes. Of 24 mapped suppressor mutations in blue colonies that were chromosomal, five were found to affect plasmid copy number based on mapping depth (Table 1). All five mutations decreased plasmid copy 8- to 20-fold and four mutations were located in the gene encoding PcnB (PAP-I), a protein known to decrease the copy number of plasmids with a Col-E1-type replicon, including our vector^44,45; the fifth mutation was located in a known functional domain of priA, a gene encoding PriA, a DNA-binding ATPase/helicase protein that promotes the assembly of the primosome and replication restart proteins and also essential for replication of Col-E1-origin plasmids.^39,46 This result suggested the reduction in plasmid-encoded signal permitted survival. A majority (18 of 24) of the suppressor mutations mapped to three adjacent genes in the VSP-2 genomic accessory island that define the ddmABC operon.⁸ These include missense, nonsense, frameshifts, deletions, and also insertions in these genes. Thus, we speculated that the DdmABC proteins were part of the CDD system responsible for the plasmid-induced cell death that was triggered by growth at high cell density. Other suppressor alleles that were not mapped to ddmABC, pcnB, and priA included a mutation in luxO (Table 1); the luxO mutation is consistent with the observation that the ddmABC operon is quorum regulated.³⁸

Table 1.

Identified genomic suppressors and their effects on plasmid copy number and VIB04 phage titers

Strain	pWR1566 copy number	Phage titer (VIB04)
WPR2700	N/A	<10
ΔVSP-2	N/A	5.6 × 108
ΔVSP-2 priA Met462Arg	N/A	6.6 × 108
ddmA Ile237::OrfAB	36	<10
ddmA Asp250Tyr	42	<10
ddmA Δ287–317frameshift	80	<10
ddmB Trp21stop	80	<10
ddmB Pro94frameshift	76	<10
ddmB Arg100stop	55	<10
ddmB Arg160frameshift	44	<10
ddmC Cys50Phe	43	<10
ddmC Leu125Pro	27	<10
ddmC Arg148frameshift	72	<10
ddmC Lys159stop	10	<10
ddmC Gln162Lys	54	<10
ddmC Asn197frameshift	38	<10
ddmC Ile255Ser	80	<10
ddmC Lys412Glu	33	<10
ddmC Thr513::OrfAB	25	<10
ddmC Asp523Tyr	82	6.3 × 108
ddmC Phe571Cys	9	6.8 × 108
luxO Met87Arg	40	5.8 × 108
pcnB Δ287–457frameshift	1	<10
pcnB ILe34del	12	<10
pcnB Asp230frameshift	5	<10
pcnB Trp364stop	0.25	<10
priA Met462Arg	5.5	7.4 × 108

In addition to VSP-2 and non-VSP-2 mutants, strains WR2700 and WR2700 ΔVSP-2 are included as controls permissive and restrictive for phage growth. N/A, not applicable.

Predicted structures and functions of VSP-2 proteins

By sequence homology alone, ddmABC gene products belong to the Lamassu anti-phage system.⁴ A hidden Markov model analysis we performed and a previous in silico structural analysis identified a conserved CAP4 nuclease domain in DdmA that is also present in an anti-phage nuclease effector protein AbCap4 that requires a cyclic tri-oligonucleotide for its activation by binding a domain called SAVED^8,38,47; however, DdmA does not have a SAVED domain but instead a C-terminal domain 7 motif (Pfam PF20283: CDT7) belonging to ABC-three-component (ABC-3C) systems (Figure 1A). DdmB is predicted to be a small globular protein with no predicted homologs with known functions but does possess an ABC-3C system middle component motif (Pfam PF20131: MC3). We detected a strong predicted structural similarity between DdmC and the well-studied eukaryotic Rad50 and bacterial SbcC-like proteins (Figure S3). Features found in other Rad50/SbcC-like proteins (a conserved Walker A and an atypical Walker B NTP-binding motif) were independently verified in previous work,⁸ and are located in the head domains at the N and C termini of DdmC, respectively. The three proteins closely resemble other members of the recently identified ABC-3C complex. These complexes typically include a toxic effector enzyme that attacks an essential cellular process when its cognate partner proteins recognize danger signals such as those presented by an invading DNA virus.⁴⁸

Figure 1. Alphafold2 predictions of the three proteins encoded in the DdmABC operon.

(A) Diagram of DdmABC comparing ABC-3C systems.

(B) Diagram of DdmA, DdmB, and DdmC and predicted domains and motifs. Mutations identified in this work are indicated.

(C) Cartoon and surface rendering of DdmC monomer.

(D) Cartoon and surface rendering of DdmC dimer.

(E) Cartoon and surface rendering of DdmA₁B₁C₁ complex.

(F) Cartoon and surface rendering of DdmA₂B₂C₂ complex.

(G) Cartoon rendering of DdmA monomer.

(H) Cartoon rendering of DdmA in DdmA₂B₂C₂.

(I) Cartoon rendering of DdmC from DdmA₂B₂C₂ complex showing predicted structural locations of plasmid and phage-permissive mutations in this work. Phage-permissive mutations are indicated (*).

(J) Cartoon side-view rendering of DdmC dimer interface in DdmA₂B₂C₂ showing mutations.

(K) Cartoon and bottom-view rendering of DdmC dimer in DdmA₂B₂C₂ showing mutations.

No experimentally derived structure has yet been solved for the Lamassu anti-phage-type system. The tertiary structures of structural maintenance of chromosome (SMC)-family proteins including Rad50-Mre11, SbcCD, and the anti-phage Wadjet system show proteins in the complex assemble as dimers.^49,50 We employed the AlphaFold2 in silico structure-prediction program to infer and juxtapose structures of monomeric, dimeric, and heteromultimeric complexes.⁵¹ We then mapped mutations onto these structures to compare those selected for plasmid tolerance with those that abolished phage restriction.

The Alphafold2-predicted structure for DdmC resembles other SMC proteins as it possesses two conversed globular regions with predicted ATPase motifs in the C-terminal and N-terminal (CTD/NTD) domains. These globular head domains are separated by long alpha-helices that form coiled-coil structures separated by a central hinge (Figures 1B and S3).^49,52 Rather than one, two hinged regions in the coiled-coil domain of DdmC protein appear to facilitate conformational changes in the complex (Figure 1C). For the DdmC dimer, AlphaFold2 predicts significant intermolecular interactions between the head domains and regions of the coiled-coil domains (Figure 1D). The interactions between DdmA and DdmB with DdmC vary when DdmC monomer is compared to the dimer (Figures 1E vs. 1F). In the DdmA₂B₂C₂ complex, the position of DdmB appears more displaced with very minor interaction with the central hinge (Figure 1F). Here, DdmA forms a dimer that interacts with the secondary hinged region of the DdmC₂ dimer. There are significant conformational differences in the DdmC dimer compared to the dimer in DdmA₂B₂C₂ complex (Figures 1D vs. 1F). In the hetero-oligomeric complex, the DdmC coiled-coil extended domains are compressed downward, and interactions within the globular head domains are altered (Figure 1F).

The predicted structure for the DdmA monomer differs from that in the DdmA₂B₂C₂ complex. The conformation of the NTD CAP4 nuclease domain alone, including the catalytic residue (K56), is unchanged in either but the CTD is extended and altered in the monomer, possibly opening the protein to DNA (Figures 1G vs. 1H). The majority of selected mutations in DdmA are nonsense or frameshift except for two missense mutations and all are within the CTD (Figure 2B). In ABC-3C proteins, the effector complex is inactive and conformational changes in these domains assist in extending and activating associated effectors.⁴⁸ DdmB is predicted to be a small globular protein and all mutations are premature stop or frameshift mutations. Using Alphafold2 predictions, we cannot speculate how DdmB mutations influence proteins in the complex. In contrast, a majority of mutations in DdmC are missense mutations. Most are located within the two head domains proximal to the predicted Walker A and B motifs (Figures 1I–1K). Three mutated residues (Q162, D523, and F571) are located within the inter-protein interface between DdmC proteins. Both Q162 and D523 are parts of secondary structures that form the dimer interface. Both F571 residues in cognate proteins are predicted to interact with one another (Figure 1K).

Figure 2. Sequences from restricted phages cloned in place of the T7 RNase III sequence cause CDD.

(A) Diagram showing the cloning of random phage DNA fragments into a plasmid and screening for CDD.

(B–D) Sequences identified in phage VIB04 by position and predicted secondary structure. A colorimetric confidence scale for base-pairing prediction is shown.

(E) One VIB04 sequence is homologous to the RNase III sites in related bacteriophage T7.

Identification of VSP-2 restricted phages and phage-associated signals that trigger CDD

We further reasoned that CDD is a response associated with abortive infection responses triggered by DNA sequences such as the T7 phage-derived sequence we identified in pWR1566. These sites might be recognized by the quorum-regulated DdmABC complex and PriA if such palindromic sequences also stall DNA replication processes. To test this hypothesis, we screened 48 V. cholerae bacteriophage samples in our collection using both wild-type (WT) and a constructed VSP-2 deletion mutant (ΔVSP-2) to identify those that were restricted by VSP-2. Two vibriophages (VIB04 and VIB05) were identified as being restricted and both plated as clear large plaques at least 10 million-fold better on a confluent lawn of the WR2700 ΔVSP-2 strain (Table 1). The genome sequence of these two vibriophages revealed that VIB04 is highly similar to coliphage T7/T3-like viruses and originally isolated as the El Tor V. cholerae typing phage classified as phage X as vibriophage N4.⁵³ VIB05 is a member of the Tawavirus genus similar to vibriophages Peru2 and JSF7.^54,55 Both VIB04 and VIB05 form morphologically similar particles with short non-contractile tails (i.e., podophages) and can be taxonomically assigned by genomic sequence to the Autographiviridae family phages that encode a self-transcribing RNA polymerase.⁵⁶ We were unable to isolate spontaneous mutants in either phage that overcame apparent VSP-2 restriction using high-titer cesium-purified stocks with concentrations exceeding ~10⁹ plaque-forming units (PFU)/mL.

We reasoned that, because both of these phages are efficiently restricted by VSP-2, their genomes are likely to have one or more DNA signals such as palindromic sequences that trigger CDD, and that these phage sequences may be essential for viral replication or propagation. We determined the genome sequences of VIB04 and VIB05 and reasoned that both encoded numerous putative palindromic sequences of various lengths. To test if these or other phage sequences were indeed capable of inducing CDD, we chose to focus on phage VIB04. Purified phage DNA was sheared to fragments of 400–600 nucleotides in length and then randomly shotgun cloned into a derivative of our LacZ⁺ screening vector pWR1566, which lacked the T7 phage-derived palindromic sequence that triggered CDD in a VSP-2-dependent fashion. Liquid cultures of V. cholerae carrying these libraries were plated and colonies that displayed the blue-ghost CDD death phenotype were identified in a small fraction of each shotgun library; plasmids recovered from these colonies were then sequenced. The cloned sequences from VIB04 that induced the blue-ghost death phenotype were highly similar to one another and all encode palindromic sequences that would form hairpins in single strands of DNA (Figure 2A); some but not all of these were highly similar to the RNase III site in T7 originally identified in plasmid pWR1566 (Figure S4, structures and positions indicated in Figures 2B–2D). Mapping these cloned palindromic sequences to the VIB04 genome showed that they were dispersed across the genome. We concluded that these phage-encoded palindromic sequences represent at least part of the signal that triggers CDD and that the presence of these in the VIB04 phage genome could explain why this phage is restricted by VSP-2.

Role of VSP-2 independent suppressors, including PriA

We plated both VSP-2 restricted phage VIB04 on the entire set of selected and sequenced V. cholerae chromosomal suppressors of CDD to determine if the suppressors obtained by our pWR1566 plasmid selection also permitted phage growth. A minority of ddmABC (two) and non-ddmABC (two) suppressors permitted a significant increase in the titer of phage VIB04 (Table 1). For example, two point mutations in the CTD of the DdmC protein (D523Y and F571C) improved the plating of VIB04 > 10⁸-fold and this was comparable to a full deletion of VSP-2. Most other selected mutations in the ddmABC genes failed to improve VIB04 plating. We reason that the majority of plasmid-selected point suppressor mutations in the ddmABC operon likely do not completely abolish the activity of this system against phage. Thus, the individual proteins of the DdmABC system are likely critical components for the CDD phenotype, but certain proteins or domains identified in our plasmid-based selection appeared to be more essential and specific for restricting the VIB04 phage.

Mutations in pcnB that reduced the copy number of pWR1566 displayed low phage titers that were similar to WT, indicating that pcnB mutations did not interfere with the ability of VSP-2 gene products to block VIB04 replication. In contrast, a strain with a single missense mutation present in priA (M462R) and luxO (M87R) permitted titers of phage as high as those seen in ΔVSP-2 (Table 1). Although PriA is also required for the replication of plasmids possessing a ColE1 origin of replication and pas sites that form higher-order DNA structures such as hairpins,⁴⁶ there is no evidence this protein plays any role in the replication of T7-like phages. The ability of a PriA mutation to alleviate DdmABC-dependent VIB04 phage restriction suggests it plays a more critical role in either assisting or interfering with DdmABC activity.

Orthologues of the PriA protein in other bacteria recognize the junctions of separated dsDNA stalled replication forks and DNA hairpin sequences during replication restart.^57–59 The priA M462R CDD suppressor mutation we isolated is positioned within a highly conserved domain that contains a β-hairpin/strand-separation ‘‘pin element’’ that is located in a zinc-binding pocket that is conserved in RecQ helicases.⁵⁹ Mutations within this pin element reportedly eliminate PriA-mediated DNA unwinding, function, and interactions with primosome protein PriB.⁵⁹ This mutation introduces an additional positive charge in this domain and could increase affinity for DNA. The higher titers of VIB04 on the PriA M462R mutants compared to the low titers on WT support the importance of the PriA β-hairpin/strand-separation pin element in phage restriction by the CDD system and suggests that PriA may directly recognize critical DNA structures or DNA-protein complexes triggered by hairpin sequences. We were unable to construct complete priA deletion mutants in V. cholerae to compare the loss of PriA to the activity of PriA-M462R mutant. In E. coli, priA null mutants have poor viability and form filamentous cells.^57,60,61 We reason either increased or decreased PriA activity of this mutant interferes with DdmABC restriction of phage and plasmid DNA.

Mutations in a hairpin encoding palindromic sequence abrogate CDD

To more precisely examine the role of palindromic sequences that might encode ssDNA hairpins, we inserted a minimal short synthetic oligonucleotide that encodes a T7 RNase III-like sequence derived from VIB04 phage into a vector devoid of a T7 RNase III sites. This VIB04 palindrome was designated Seq2 and was compared with an oligonucleotide that was mutated at three nucleotides predicted to disrupt base pair interactions in the stem that could drive hairpin formation (designated Seq3) (Figures 3A and 3B). We performed a quantitative CDD assay to assess the relative potency of these various palindromic sequences. In brief, we measured viable bacteria enumerated as colony-forming units (CFUs) when plated at high density on agar plates containing Carb for 12 h; cells either carried a plasmid encoding palindromic sequences Seq2, Seq3, or Seq2 in reverse orientation (Seq2-rev). We found that the Seq2 sequence was highly toxic in this quantitative CDD assay; its toxicity was comparable to the original T7 RNase III sequence we identified in pWR1566 (Figure 3C). Seq2-rev, the reverse orientation of Seq2, was also highly toxic. However, Seq3 displayed 25-fold less toxicity in this CDD assay, suggesting that a stable, double-stranded DNA stem is essential as a signal triggering CDD either directly or through its secondary effects on processes such as DNA replication or repair.

Figure 3. Mutations that disrupt the hairpin sequence eliminate cell-dependent death.

A) Schematic showing the position of the original sequence and its replacement by a synthetically designed sequence seq2. The hairpin-forming nucleotides in both the original and seq2 are boxed in green and purple.

(B) The predicted hairpin of the original and seq2 sequences and their derivatives are shown. Mutations are boxed to indicate potential disruptions to hairpin formation.

(C) Screening assay showing the presence or absence of CDD after 12 h. Cells recovered after 12 h for each plasmid in both WT and ΔVSP-2 mutants. Significance is indicated. Sample size n = 3, standard error of mean shown as bars, *p < 0.05 and ****p < 0.0001. Ordinary one-way ANOVA test was used to compare multiple samples.

We next determined if Seq2 and Seq3 palindromic sequences could trigger cell death of the WT (VSP-2⁺) parental strain and ΔVSP-2 mutant strains. Bacterial cells were harvested by a standardized protocol and then viable cells were enumerated on agar plates that contained Carb (Figure 3C). In WT cells, approximately 25-fold more viable, Carb-resistant cells were recovered when the plasmid encoded the mutant Seq3 palindrome compared to the Seq2 palindrome. However, this difference became negligible in ΔVSP-2 mutant strains. We conclude that the VSP-2-encoded DdmABC system in combination with PriA most efficiently attacks cells carrying a plasmid encoding by the Seq2 palindromic sequence and that just three mutations localized to the stem of this palindrome (i.e., corresponding to Seq3) are sufficient to suppress this plasmid restriction by approximately 96%.

We wondered if both cell death and plasmid loss was occurring during our standard CDD assay. Accordingly, we performed a CDD assay and then determined how many viable cells were present and the level of plasmid that we could detect in either live or dead cells during the assay. Plasmids encoding Seq2 and Seq3 were compared in these assays and plasmid DNA present in bacterial cells was measured by Southern hybridization using the vector plasmid as a probe. Viable cells that could be recovered from agar were quantified at 0, 60, and 180 min on media with or without Carb. The CFU recovered on plates containing Carb define viable, plasmid-containing cells, while those that were observed on plates that lacked Carb define both cells that had lost as well as those that retained the plasmid. When agar-grown cells were plated on Carb-containing medium, we observed ~30-fold less CFU by 180 min for the Seq2 plasmid compared to the Seq3 plasmid when cells were VSP-2⁺ (Figure 4A). This difference in the recovery of viable Seq2 vs. Seq3 plasmid-carrying cells was not apparent in the ΔVSP-2 background (Figures 4A and 4B). Furthermore, this loss of viable counts of cells carrying the Seq2 plasmid was not apparent if cells were enumerated on agar medium that lacked Carb (Figure 4B). These data suggest that plasmid loss contributed to the observed restriction of the plasmid encoding Seq2. Indeed, when we quantitated plasmid levels in cells recovered from agar across this same time course, we observed a much higher loss of the Seq2-encoding plasmid compared to the Seq3-encoding plasmid in VSP-2⁺ cells. Because we cannot determine whether the Seq2 plasmid loss is caused by its degradation or by replication inhibition within this 180-min window of observation in the CDD assay, we conclude that both mechanisms could be responsible for our observed results (schematically modeled in Figure 4D). No significant loss of either plasmid was observed in priA M462R or ddmC F571C, confirming both single mutants are as defective as the ΔVSP-2 strain (Figures 4E and 4F).

Figure 4. DdmABC-based plasmid elimination and cell survival are shaped by antibiotics.

Strains are indicated and distinct plasmid sequences Seq2 and Seq3 are shown in Figure 4.

(A) Cells recovered and enumerated after high-density bacterial incubation on LB (Luria Broth) agar in the presence of carbenicillin (50 μg/mL).

(B) Cells recovered and enumerated after high-density bacterial incubation on LB agar in the absence of carbenicillin.

(C–E) Southern blot probing plasmid DNA from DNA extracted from each strain and its respective plasmid and time point during the 180-min incubation shown in (A) and (B). (D) Model showing loss of plasmid based on cloned sequence. (E) priA M462R and ddmC F571C cells recovered and enumerated after high-density bacterial incubation on LB agar in the presence of carbenicillin (50 μg/mL).

(F) priA M462R and ddmC F571C cells recovered and enumerated after high-density bacterial incubation on LB agar in the absence of carbenicillin. Sample size n = 3, standard error of mean shown as bars. Ordinary one-way ANOVA test was used to compare multiple samples.

We also were curious whether programmed cell death triggered by Seq2 in combination with VSP-2 could be responsible for this depletion of the Seq2 plasmid. We hypothesized that, if DdmABC was driving programmed cell death, then the DdmA nuclease might be responsible. Therefore, we measured the stability of V. cholerae chromosomes under these high-density growth conditions in WT and the VSP-2 deletion mutant. By pulsed-gel electrophoresis, we found that the two chromosomes of V. cholerae indeed undergo degradation in the presence of the Seq2 plasmid compared to cells lacking plasmid (Figure S5). This degradation was dependent on VSP-2 and suggests that Seq2-induced CDD likely causes cell death in part by triggering chromosome degradation.

CDD causes phage restriction and abortive infection

We characterized the role of VSP-2 in restricting the growth of phage VIB04 in both high-density liquid cultures and when spotted on agar surfaces. We measured phage adsorption and growth in WT and a ΔVSP-2 strain to establish phage infectivity growth parameters as well as the effects of VSP-2 on cell viability. When both strains are infected in liquid culture at a low multiplicity of infection, but high cell density, only the ΔVSP-2 mutant strain continues to produce infectious particles (Figure 5A). In contrast, WT strains incubated with phage continue to reduce the number of infectious particles over time. After 150 min, ΔVSP-2 had nearly 100,000-fold more PFUs present in supernatant fluids than WT. Cells carrying VSP-2 failed to produce an increase in infectious phage when compared to the ΔVSP-2 strain.

Figure 5. Phage production and cell viability during phage infection are impacted by VSP-2.

(A) Plaque-forming units (PFU) representing the production of infectious phage in both WT and ΔVSP-2 mutant strains in liquid culture for 150 min when infected at MOI 0.4.

(B and C) Cells were infected with VIB04 phage at MOI 0.4 and spotted at high density on LB agar to mimic CDD conditions. Both phage (PFU) and viable bacterial (CFU) were enumerated for 6 h post infection.

We surmised WT cells are also infected and fail to produce phage but did not know whether infected cells are killed by an abortive infection mechanism or instead only restrict phage growth but remain viable. To simultaneously measure the production of phage (PFUs) and viability of bacteria (CFUs), we infected cells to allow phage to adsorb and then collected the cells and plated these as a concentrated spot to mimic conditions that induced plasmid-dependent CDD. Relative to the WT, the ΔVSP-2 mutant grew to a higher density after 4–6 h in the absence of phage, but CFU recovery was modestly reduced when phage infected (Figure 5B). The difference in the production of phage indicated by an increase in PFU was highly significant between WT and the ΔVSP-2 mutant (Figure 5C). Consistent with our measurements in liquid growth, only V. cholerae ΔVSP-2 mutants are successfully infected and continue to produce infectious particles. At high plating density, the numbers of both WT and ΔVSP-2 mutant cells were found to decrease. Thus, it is likely that phage infection can induce VSP-2-dependent cell death of the WT strain and thus eliminate phage particles and their progeny; in contrast, phage simply kills the ΔVSP-2 mutant through viral replication in these phage-permissive cells.

Quorum sensing regulates the DdmABC operon

We identified a suppressor mutation in the luxO gene selected to maintain the plasmid that also alleviated the restriction of the VIB04 phage. Previous studies demonstrated the DdmABC protein complex is expressed at high cell density,³⁸ consistent with the CDD phenotype identified in this work. To test the function of LuxO M87R protein, we measured the expression of hapA, a gene normally repressed by the constitutively active LuxO mutant,⁶² and ddmA relative to housekeeping gene rpoB. Expression was measured at low and high cell density because LuxO is normally rendered inactive via quorum sensing when auto-inducer molecules accumulate in media.⁶³ Inactivated LuxO permits the expression of HapR, a transcriptional regulator necessary for transcriptional expression of the HapA protease.⁶⁴ At low cell density in liquid culture (optical density [OD] 600 = 0.3), the relative expression of hapA and ddmA genes is only 0.1% and 0.01% of rpoB (Figure S6A) in both WT and luxO mutant. At high OD (OD600 > 1.8) in the WT strain, the relative expression of hapA and ddmA is 2- to 4-fold that of rpoB, a 1,000–−10,000-fold induction (Figure S6B). For the luxO M87R mutant strain, the relative expression is also increased, but 10- to 20-fold lower than the WT strain (Figure S6B). We conclude this luxO allele is mimetic to the constitutive form and thus inhibition of ddmABC gene expression is sufficient to inhibit phage and plasmid restriction.

Expressed DdmABC complex in cell extracts degrades DNA, is DdmA nuclease dependent, and is inhibited by PriA protein

We were unable to affinity purify the expressed DdmABC complex from E. coli by using a 6X-HIS tag in the NTD and CTD of DdmC. However, we determined T7-RNAP-based expression of DdmABC proteins with a catalytically active nuclease will degrade foreign DNA in crude cell extracts. We selected a commercial E. coli B-derived T7-expression strain defective in type I (ecoBr-m-), type IV (mrr-), and modified cytosine restriction systems (mcrABC-) to reduce possible interference of foreign DNA degradation independent of DdmABC. Furthermore, we found the addition of V. cholerae cell lysates of strains deleted for the ddmABC genes increased the efficiency of DdmABC restriction at least 100-fold. We interpret this as additional evidence for the role of other V. cholerae gene products, and this is consistent with previous work. Using these lysate cocktails, we probed the activity of DdmABC on plasmid and phage DNA and the role of PriA.

To investigate a role for PriA, we added a CTD-6X-HIS affinity tag to the chromosomal priA gene and ran cell lysates through a Ni-NTA agarose column to remove endogenous PriA from V. cholerae cell lysates. We found purified bacteriophage T7 and plasmid DNA were eliminated nearly a million-fold in 2 h in our lysate cocktail in a DdmABC-dependent manner when DNA was quantified using qPCR-based detection (Figures 6A and 6B). Phage and plasmid DNA were added to a lysate cocktail using the catalytically dead DdmA nuclease mutant (K57A) and DdmBC was largely unrestricted even after 4 h. Diluted DNA was similarly restricted by nuclease-active DdmABC independent of concentration after 2 h (Figures 6C and 6D). When expressed PriA protein was added back, DcmABC-dependent restriction of phage and plasmid DNA was inhibited in a PriA dose-dependent manner for both the mutants and WT PriA M462R protein (Figures 6E and 6F).

Figure 6. Use of expressed DdmABC and PriA in cell extracts to measure DNA degradation.

(A) Experimental design to express DdmABC or PriA in bacteria and generate extracts to measure the degradation of purified phage and plasmid DNA using qPCR. DNA concentration is measured in molarity using a standard range.

(B) Time course of 240 min showing degradation of T7 phage genomic or plasmid pWR1566 in cell extract from cells expressing DdmABC protein.

(C) Degradation of T7 phage genomic DNA cell after 120 min using an extract from cells expressing DdmABC+ and DdmA 57A DdmBC proteins. Phage DNA was diluted 10-fold through 1:1,000.

(D) Degradation of pWR1566 plasmid DNA cell after 120 min using extract from cells expressing DdmABC+ and DdmA 57A DdmBC proteins. Phage DNA was diluted 10-fold through 1:1,000.

(E) Degradation of T7 phage genomic DNA cell using an extract from cells expressing DdmABC+ supplemented with a dilution series of PriA+ and PriA M462R proteins in cell extract after 60 min.

(F) Degradation of pWR1566 plasmid DNA cell using an extract from cells expressing DdmABC+ supplemented with a dilution series of PriA+ and PriA M462R proteins in cell extract after 60 min. Sample size n = 3, standard error shown as bars. Ordinary one-way ANOVA test was used to compare multiple samples.

Sensitivity to DNA-damaging agents

Our data indicate that the cis-acting palindromic sequences in both phage and plasmid DNA are lethal signals to bacteria in both a DdmABC- and possibly a PriA-dependent manner. How cell death is triggered likely involves activation of the CAP4 nuclease domain in DdmA because a defective nuclease mutant was previously demonstrated to abolish plasmid elimination that depended on ddmABC.⁸ The significant hairpin-forming structures identified as CDD-inducing signals may stall replication forks that then recruit PriA as suggested in earlier studies.^57,59,65 We hypothesized that DNA damage may induce similar DNA repair responses that require PriA and could trigger CDD. Therefore, we tested WT and CDD-defective strains for differences in their sensitivity to DNA damage. We found that, at inhibitory ultraviolet (UV) radiation doses, both single and double ΔVSP-2 and priA Met462Arg mutants showed much better growth and survival than their WT parental strain (Figure S7). At doses above 48 J/cm², only strains with either or both ΔVSP-2 and priA mutations were recovered. Thus, the PriA mutant protein reduces this sensitivity to UV independent of DdmABC.

DISCUSSION

We show the Lamassu system of V. cholerae encoded by the VSP-2 ddmABC operon restricts some vibriophages and plasmids by recognizing a signal associated with phage-encoded palindromic DNA sequences including, but not confined to, RNase III sites. The expression of ddmABC genes is quorum regulated and an isolated mutation in LuxO, likely constitutively active, alleviates the restriction of phage and plasmids at high cell density. The palindromic signal was first identified in a laboratory plasmid vector as a coliphage T7-derived phage sequence. However, non-RNase III palindromic sequences present in vibriophages also restricted plasmid DNA in a DdmABC-dependent fashion. DdmABC activity triggered by one of these phage-derived palindromic sequences could be abolished by only three nucleotide substitutions that are predicted to disrupt the stem of the DNA hairpin it can form.

The distinct classes of mutations selected in each component of DdmABC inform about the mechanistic action when a signal is sensed. We reason the DdmA nuclease effector is activated, probably by conformational changes and possible dissociation, by a DNA signal recognized by DdmC in the complex. All isolated suppressor mutations in DdmA map outside of the nuclease domain and conceivably prevent the active effector conformation. We show that expressed catalytically inactive DdmA protein fails to degrade phage and plasmid DNA, but our genetic screen may not be saturated enough to anticipate this class of mutation. Most of the isolated suppressor mutations in DdmC are instead missense and map to the two head domains. We speculate these alter the critical interaction with the DNA signal and prevent the propagation of signal-dependent conformational changes that activate the nuclease effector or both. Only two selected mutations in DdmC in our CDD plasmid-based screen completely abolished phage restriction, and these residues are positioned where the globular head domains are predicted to interact.

We discovered that at least one class of mutations in DdmABC prevents the restriction of phage and plasmid, but most permit the maintenance and replication of the plasmid while still restricting phage. The Wadjet anti-plasmid system belongs to a different anti-plasmid defense family with four proteins. One encodes an SMC-like protein that recognizes closed circular DNA, extrudes a loop, and then cleaves DNA using an activated nuclease effector.⁵⁰ Thus, two Wadjet proteins appear to confer independent activities during plasmid restriction. For this Lamassu defense system, we speculate that distinct mutations in DdmC may alter DNA binding and signal recognition, conformational changes in the complex, activation of the nuclease, or even other processes that manifest differently in the context of phage infection or plasmid replication. Our Alphafold2 structural prediction reveals significant changes in monomeric DdmA when dissociated from the complex. We cannot yet determine the mechanism of DdmA activation by the complex in the context of plasmid or phage DNA recognition, but our cell-extract-based assay reveals the nuclease is required for DNA degradation.

Proteins that resemble DdmC and other SMC proteins often play essential roles in DNA tethering, repair, replication, and recombination, usually as dimers associated with larger multimeric complexes.^66,67 SMC-like proteins possess ATP and nucleotide-binding domains as well as an extended coil-coil domain capped by a dimerization domain.^68–70 One of the best studied, DNA repair protein Rad50, also forms MRE11/NSB1 complexes that respond to viral invasion in eukaryotic cells and specifically recognize DNA double-stranded breaks (DSBs) and stalled replication forks.⁷¹ SMC bacterial repair protein SbcC forms a complex with its nuclease partner SbcD to recognize and cleave palindromic sequences that can form hairpins during DNA replication.^72,73 We speculate that DNA damage or instability during plasmid or phage replication may be recognized similarly. Some plasmids and phages naturally possess pas sites that form secondary structures in ssDNA and are recognized by PriA and are operative in replication⁴⁶; thus, such signals may be recognized as foreign DNA. The increased resistance of ΔVSP-2 strains to UV and DNA-damaging agents is supportive of this model.

Palindromic sites cause genomic instability and have been shown to induce single and DSBs via SbcCD recognition; DSB repair typically requires RecBCD recombination and the re-established replication fork protein PriA.⁷⁴ Thus, the identified suppressors that map to DdmABC suggest that the signal likely involves DNA lesions, breaks, and stalled replication forks induced by palindromic sequences. Although DdmABC proteins can resemble others that help coordinate DNA repair, recombination, replication, and maintain chromosomal stability, our results suggest that the V. cholerae DdmABC system instead acts to block replication of and restrict foreign DNA by recognizing impaired DNA replication or damage induced by these foreign genetic elements. The foreign DNA likely triggers DdmABC-mediated abortive infection that eliminates both foreign DNA and the infected cell, likely through chromosomal degradation. Thus, viral or plasmid elements that have invaded the cell are also eliminated before they have a chance to amplify and infect other nearby sister cells.

The priA M462R allele renders cells as permissive for VIB04 and VIB05 phage growth as those deleted in ddmABC. These data clearly implicate PriA in the process that influences DdmABC restriction of the growth of VIB04 and VIB05 phages. Curiously, PriA has been shown to recognize stalled replication complexes generated when palindromic sequences interfere with DNA replication processes.⁵⁸ PriA is universally conserved in bacteria but revealing a role for this protein in phage defense is a novel finding of the work presented here. We speculated that PriA may either interact with and recruit DdmABC or instead its activity competes and interferes with DdmABC DNA signal recognition or restriction. The increased resistance of priA M462R strain to UV suggests an increased efficiency of DNA repair process independent of DdmABC. Furthermore, DdmABC restriction of plasmid and phage DNA in extracts is reduced in a PriA-dependent manner. We hypothesize that PriA M462R better recognizes key signals in DNA resulting in blocked recognition of such sites, its DNA replication-repair mechanisms are more efficient in repairing a signal before being sensed by DdmABC, or both (Figure 7).

Figure 7. Proposed model of DdmABC recognizing signals in DNA generated by damage and stalled replication.

Interference by PriA protein binding, repair, or both is shown.

Here we also show that the DdmABC system is highly active in V. cholerae against certain phages that encode CDD activation signals. Furthermore, DdmABC proteins may reduce the fitness of bacteria that encounter various stresses that damage DNA and elicit such responses. We discovered that the DdmABC system increases sensitivity to both UV radiation and chemical agents that were previously demonstrated to induce damage.^75–77 Thus, this phage resistance mechanism in V. cholerae 7^th pandemic strains may impose a fitness cost for environmental strains of V. cholerae (which typically lack the DdmABC system) if these strains experience strong ambient sunlight and UV flux or exposure to mutagenic molecules in aquatic habitats or other environments. Because DdmABC is found to be quorum regulated like other anti-phage systems,³⁵ we further conclude that this anti-phage system is most active when V. cholerae is at high density within the infected host^31,36 or within clumps of bacteria shed from cholera victims.³⁷ This is relevant because lytic phages are commonly isolated in cholera stool during epidemics.^29,30

Recently published mechanistic work on anti-phage defense systems continues to reveal striking similarities between bacteriophage abortive infection mechanisms and eukaryotic innate immunity processes.^2,3,47 Previous work indicates RAD50-MRE11-NBS1 DNA damage machinery components (which are orthologous to DdmA and DdmC) are regulated and disrupted by several adenovirus-encoded proteins during infection and this is critical for maintaining viral DNA replication.^78,79 Both RAD50 and MRE11 are implicated in the recognition of cytoplasmic dsDNA and activation of the innate immune genes including STING and IRF3.⁸⁰ RAD50-CARD9 signaling complexes in dendritic cells have been demonstrated to activate nuclear factor κB (NF-κB) and the production of IL-1β when a dsDNA vector was transfected or when dendritic cells were infected with a DNA virus.⁸¹ Thus, insights gained on the mechanism of phage recognition through the DdmABC/PriA system may be highly relevant to the innate immune recognition of eukaryotic DNA viruses or the use of DNA vectors in human gene therapy.

Of note, DdmABC also possesses similarities to studied anti-phage defense systems encoded on bacteriophages. The overcoming lysogenization defect (Old) protein encoded by bacteriophage P2 kills E. coli recB and recC mutants and interferes with phage lambda growth by killing cells unless the lambda-encoded recombination genes are deleted.⁸² The CTD structure of Old protein also bears resemblance to Rad50 protein but possesses a fused nuclease domain similar to MRE11 at its N terminus.⁸³ Thus, even phages have apparently acquired and adapted Rad50/SbcC-like proteins to sense recombination-based signals and then utilize those signals to degrade the chromosome of the host bacterium. We propose RAD50 and many other SMC-like proteins, which were previously classified as first responders for DNA damage recognition and repair, are also likely to be major sensors of damaged DNA that may represent pathogen-associated molecular patterns (PAMPs) indicative of the replication of foreign DNA. How this anti-phage and anti-plasmid system has shaped the evolution of V. cholerae is an important consideration in future studies focused on understanding the emergence of this highly successful pandemic pathogen. These insights may also be important to recognizing how mammalian cells respond to DNA viruses or vectors used in therapeutic applications.

Limitations of the study

This study leverages a genetic selection to identify different classes of mutations in the Lamassu DdmABC and other V. cholerae proteins that permit the maintenance of an incompatible plasmid. A cis-acting sequence responsible for plasmid incompatibility is identified. Purification of the DdmABC complex in our laboratory has proved difficult for biochemical and structural studies and we speculate this may be due to the inherent DNA-binding or DNA-damage-sensing activities that may result in proteins dissociating from the complex. Catalytically, DdmABC is toxic when overexpressed in E. coli and the catalytically dead DdmA may still be released when DdmC binds DNA. Future directions will focus on methods to express and purify DdmABC to validate predicted structures and study its dynamic interactions with DNA signals including hairpins, replication forks, and sites of DNA damage using established biochemical in vitro assays.

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

Queries and requests for resources, strains, and reagents should be directed to and will fulfilled by the main contact John Mekalanos (john_mekalanos@hms.harvard.edu).

Materials availability

Resources and materials generated in this study are available from the lead contact.

Data and code availability

The published article includes all data and analysis generated in this study and does not include original code. Information required to reanalyze any data in this study is available from the lead contact upon request.

METHOD DETAILS

Bacterial strains and plasmids

Strains and plasmids are listed in Table S1. Strain TND0652 was a generous gift from Ankur Dalia and published elsewhere.⁸⁴ WR2700 was first constructed to test the functionality of the V.cholerae O395 CRISPR-CasI-E system in a derivative of the El Tor strain E7946. To construct WPR2700, we first inserted a 1.4kb DNA fragment containing the LacI gene and an inducible tac promoter in front of the CRISPR-Cas complex in O395 at nt position 2827729 using suicide vector pRE107⁸⁷ and 500 nts of flanking sequence. The DNA from O395 was prepared, transformed, and integrated into the chromosome TND0625 and we selected for uptake using ampicillin (100μg/ml). In this strain, the CRISPR-CAS-I-E merodiploid sequence was crossed out using sucrose counter-selection as previously described.⁸⁷ Insertion of LacIq-ptac-CRISPR/Cas-IE was confirmed by PCR and sequencing. A 37,850 nucleotide region from O395 from PurD (Chromosome I position 311908) through Fis genes (Chromosome I position 349559) replaced the 20,417 nucleotide region E7946 derivative this introduced the entire CRISPR-CAS-I-E operon and several adjacent genes both up and downstream of O395.

pWR1566 is derived from pUC18-miniTn7-LacZ and has a deletion of the Gentamycin resistance gene between flanking FRT sites generated by co-transformation with the pFRT2 vector expressing flip recombinase.⁸⁶

The sequence of ddmABC operon and a ribosome binding site upstream of ddmA was PCR amplified and cloned into the XbaI and EcoRI sites of pET21a to construct pWR1589. The C terminus of DdmC was modified to with an in-frame 6X-His tag. The region between the XbaI and AvrII sites in the wildtype ddmA gene was replaced to introduce a translated K57A mutation in plasmid pWR1590. The V.cholerae priA and priA M462R gene sequences were PCR amplified and cloned into the XbaI and SacI sites of pET21a. The C terminus of PriA was modified with an in-frame 6X-His tag to construct pWR1592 (priA) and pWR1593 (priA M462R).

To construct strain WR2742, the priA gene with a 6X-His tag in the CTD was cloned into suicide vector pRE107 with ~500 nts flanking on either side of the gene. This was mated into pWR2700 using the MFD-pir donor strain⁸⁵ and conjugates that acquired this sequence as chromosomal recombinants were selected on ampicillin (100μg/ml). The merodiploid sequence was crossed out using sucrose counter-selection as previously described⁸⁷ and the chromosomal insert was validated by sequencing the PCR product of PriA and its adjacent sequences.

CDD assay and selection and sequencing of suppressors

pWR1566 was purified from laboratory E.coli strain NEB 10-beta (New England Biolabs) and used to transform V.cholerae strain WR2700 using electroporation. Bacteria were grown in Luria Broth (LB) in liquid or on agar except when indicated otherwise. Transformants were selected on LB agar supplemented with Carbenicillin (100μg/ml) and XGAL (150μg/ml). Colonies were resuspended in liquid no more than 12 h after plating and grown in LB media supplemented with Carbenicillin (100μg/ml) at 37°C.

For screening of CDD and selection of survival mutants, transformed strains were grown to high OD600 (>1.5) in shaking tubes or flasks at 37°C (RPM = 300). 1.0mL of the culture was centrifuged at 3000 RFC for 4 min in a microcentrifuge. The supernatant was decanted and cells were resuspended in 100ul LB. This was serially diluted 4-fold 12 times in a 96-well plate row. 3ul of each dilution was spotted on LB agar (Carb 100ul/ml +XGal 150μg/ml) and allowed to dry for 10 min before growth at 30°C. Growth on plates was observed 16–20 h post-plating.

For the selection of survivors from CDD, we plugged spotted agar areas where cell growth was absent. These plugs were resuspended in 1.0 mL LB, vortexed, and plated on LB agar (Carb 100ul/ml +XGal 150μg/ml). Colonies that grew were scored for blue/white and purified on LB (Carb 100ul/ml +XGal 150μg/ml). Plasmid and chromosomal DNA were prepared using the Zymo Quick-DNA Isolation kit (Zymo D3024). Mutants were sequenced using Illumina Miseq and libraries were prepared using the DNA Ultra II library preparation kit (New England Biolabs E7645). Mutations and depth were mapped using CLC-Bio Workbench v8.

Phage DNA shotgun cloning

Viral DNA from VIB04 was isolated from CsCl-purified particles dialyzed into 10mM Tris-Cl pH8.0 1mM EDTA and adjusted to 500ul using 10mM Tris-Cl pH8.01mM EDTA using one phenol extracted (200μl, equilibrated with 50mM Tris-Cl pH8.0). This was briefly mixed using a vortex and the top layer was carefully removed and kept after separation in a 55°C heat block for 20–30 min. Three successive Phenol:Chloroform: Isoamyl (20% volume, 25:24:1, phenol equilibrated with 50mM Tris-Cl pH8.0) extractions were performed as above. DNA was precipitated using three volumes of 100% ethanol and the pellet was washed twice with 80% ethanol and dried. DNA was dissolved in water, and sheared to ~500nt using a (QSONICA Q800R3) for 5 min (30 s cycled ON/OFF at 20%). DNA was end-repaired using DNA Repair Mix (NEB E6050L) and shotgun cloned into pWR1566 that had been cut with SmaI (NEB R0141L) and AvrII (NEB R0174L), blunted using DNA Repair Mix, and dephosphorylated using Antarctic Phosphatase (NEB M0289L) using T4 Blunt T/A DNA ligase (NEB M0367L). Transformants were cultured and screened for CDD as described in this work.

Cloning of artificial hairpin sequences

Complementary oligonucleotides that form seq2 and seq2_mutated were ordered. Cloned hairpin sequences are flanked by PmlI and SmaI restriction sites (Table S1). Oligonucleotides were mixed in 100ul at a concentration of 5uM, heated to 95°C, and slowly cooled until 50°C to anneal. DNA was digested with PmlI and SmaI (New England Biolabs), ligated into pWR1566 that had been cut with SmaI (NEB R0141L) and AvrII (NEB R0174L), blunted using DNA Repair Mix, and dephosphorylated using Antarctic Phosphatase (NEB M0289L) using T4 Blunt T/A DNA ligase (NEB M0367L). Cells were transformed into laboratory strain E.coli DH5a first and then plasmid DNA was prepped from 1000s of pooled colonies. The uncut purified plasmid was transformed into WR2700. Transformants were cultured and screened for CDD as described in this work.

Phage screening, preparation, and infection assays

A panel of V.cholerae phage lysates was prepared on V.cholerae strain MAK757 and plated on both WR2700 and WR2700 ΔVSP-2. Two candidate phages were identified as VSP-2 restricted. Both phages were sequenced and one was previously identified as V.cholerae phage N4. To avoid confusion, we renamed the N4 phage ‘‘VIB04’’ in this work because another unrelated phage also called N4 has been studied previously in the field. The second phage (VIB05) was isolated from lysate samples kindly provided by Dr. Shah Farque as a member of the International Center for Diarrheal Disease Research, Dhaka, Bangladesh.

Phages were propagated on WR2700 ΔVSP-2. In brief, sufficient phages were co-plated in 8mL soft top agar (0.3%) with bacteria (9 × 10⁷ CFU from liquid mid-log culture) over 1.5% LB agar and incubated overnight at 30°C to produce a confluent lysed lawn of V.cholerae cells. This was done in triplicate using 150 mm × 15 mm plates to increase yields. 30mL LB was poured onto each soft agar overlay and then both soft top agar and LB were mixed using a sterile plate spreader and the slurry was poured into a centrifuge bottle. This mixture was centrifuged at 15,000 RFC for 20 min to remove solids. The supernatant was adjusted to 1.0M NaCl and polyethylene glycol (MW 8000). was added to 0.8% w/v. This solution was incubated on ice for between 4 and 8 h and then centrifuged at 8,000 RFC for 30 min (4°C). The supernatant was poured off and the pellet was gently resuspended in 50mM Tris-Cl pH8.0 and 10mM MgCl2 and incubated on ice for 1–2 h to suspend phage virions. This suspension was centrifuged at 5,000 RFC for 10 min (4°C) to remove large insoluble material and the supernatant was collected. The supernatant was overlayed onto a CsCl step gradient (ρ = 1.62, 1.53, and 1.62) and spun at 40,000 RPM for 90 min using an SW55.1 rotor on a Beckman Ultracentrifuge. The phage band was collected between ρ = 1.53 and ρ = 1.62 steps and dialyzed into T7 Buffer (100mM Tris-Cl pH7.5, 1.0M NaCl, 1mM EDTA).

Phages were titered on WPR2700 ΔVSP-2 to measure the concentration of infectious particles. For phage lysis and infected cell survival assays, phage was infected at low MOI (0.01) in liquid culture and for agar assays, infected cells were pelleted at 4,000 RFC and spot-plated on 1.5% LB agar. Bacteria were resuspended in 1.0 mL at selected time points and the supernatant was both titered on WPR2700ΔVSP-2 and plated on LB agar to determine phage and bacterial concentrations.

Cell viability and growth after UV and mutagenic exposure

To measure survival after UV exposure, 10⁸ bacterial cells from a growing late log culture (OD 1.4) were spotted on LB agar and exposed to a range of UV doses determined by time exposure settings on a UV Stratalinker Model 1800. Agar with spotted cells was removed as a plug and cells were resuspended in 1.0mL LB, serially diluted, and plated on LB agar to measure CFU.

Pulsed-field gel electrophoresis

Approximately 10⁶ cells were spotted on semi-soft LB agar in 4ul LB. At defined time points, cells were suspended in 1mL LB and then pelleted by centrifugation. Cells were resuspended and embedded in 1% Pulse Field Certified Agarose (BioRAD #1620137) and processed according to the CHEF DR-II Pulsed Field Electrophoresis System (BioRAD). Chromosomal DNA was run in a cast 1% Pulse Filed Certified Agarose gel for 60 h at 1.5V/cm² with a 30-min switch time.

Southern blot to probe plasmid DNA in cells

Total DNA was prepared from plasmid-transformed V.cholerae cells resuspended from cultures incubated on LB agar using the Quick-DNA kit (Zymo). Southern blotting was followed using published procedures for random hexamer-based probe approaches.⁸⁹ Briefly, chromosomal and plasmid DNA was separated on a 0.7% agarose gel and then denatured, neutralized, and transferred to a BrightStar-Plus Positively Charged Nylon Membrane (Invitrogen AM10100) in 10X SSC buffer. The DNA probe was prepared using purified pWPR1566 DNA by use of random primers and the incorporation of a non-radioactive dUTP nucleotide probe. Probe DNA was prepared using alkali-stable digoxigenin-dUTP (0.1mM) (Enzo Lifesciences) and was incubated with dGTP, dATP, and dCTP (0.2mM) and also dTTP (0.1mM), NEB buffer 2 (1X), 5 units Klenow Fragment (3′ → 5′ exo-), (NEB M0212), and random hexamers (35μM) (NEB S1230S). The probed nylon membrane was blocked and incubated with a conjugated anti-dioxigene antibody and detection substrate according to the manufacturer’s protocol (DIG Nucleic Acid Detection Kit, Roche #1117504191).

DNA folding prediction

Predictive folding of ssDNA was completed using ViennaRNA v2.5.18.⁸⁸ Fold algorithm options included minimum free energy and partition function to calculate the base pairing matrix in addition to the structure. Energy parameters were set for DNA with dangling energies on both sides of a helix in any case.

Protein structure prediction

Protein structures for DdmABC and PriA were predicted using Alphafold2 run locally on the Orchestra cluster at Harvard Medical School.⁵¹ Structures were visualized using Pymol (v2.3.4) and colored to specific confidence. Proteins identified as similar to DdmC were identified using HHPRED in the MPI toolkit.⁹⁰

T7 RNAP-based expression of DdmABC and PriA proteins in cell extracts

Plasmids pWR1589, pWR1590, pWR1592 and pWR1593 were transformed in to NEBExpress competent E.coli and maintained using ampicillin (100μg/mL). For protein expression, bacterial cells were grown in 500mL 2X YT media buffered by phosphate (+Ampicillin 100ug/ml) shaking at 37°C (300 RPM) in a 3 L flask. 2X YT media is made using 16g Bacto Tryptone, 10g Bacto Yeast Extract, and 5g NaCl per liter. After autoclaving, this was buffered with 0.022M KH₂PO₄ and 0.040M K₂HPO₄.

To induce protein expression, once the OD600 reached 1.5 units, IPTG was added to the culture at a final concentration of 1mM. The culture was maintained shaking at 300RPM at 37°C for 4 h. The cells were pelleted at 6000 x g for 20 min at 4°C. The supernatant was removed and cells were washed twice in ice-cold 300mL S30A buffer and pelleted as above after each step. The pellet was frozen at −80°C overnight, thawed on ice, and resuspended in 10mL S30A buffer.⁹¹ This resuspension was lysed using a Cell Disruptor (TS series, Constant Systems) set at 25 KPSI and the lysate was separated from the insoluble material by centrifugation at 10,000 x g at 4°C for 20 min. 1mL aliquots of lysate supernatant were frozen and stored at −80°C for DNA restriction assays. For WR2742, cells were grown as above without induction. The S30A resuspended lysate was run through an S30A-equilibrated Nickel agarose column (ThermoFisher #88225) to remove His-tagged PriA.

For DNA restriction assays, a 1:3 mixture of PriA-depleted V.cholerae WR2742 lysate and E.coli lysate from cells expressing induced DdmABC from pWR1589 and pWR1590. For assays using E.coli-expressed PriA protein, the above mixture was mixed 4:1 with E.coli lysate from cells expressing PriA from pWR1592 or pWR1593. Diluted PriA lysate was prepared using cold S30A buffer.

qPCR and RT-qPCR assays

DNA concentration in cell extracts was monitored for both purified T7 and pWR1566 using qPCR (oligonucleotides in Table S1). Kapa Sybr Fast qPCR kit (Kapa Biosystems KK4601) was mixed with oligonucleotides and DNA according to the protocol provided by the manufacturer. Reactions were run and measured using the Eppendorf Mastercycler RealPlex2 system. DNA concentrations were calculated by using a standard of T7 and pWR1566 DNA over a range between 100p.m. and 1a.m.

For RT-qPCR-based expression of rpoB, hapA, and ddmA, we used KAPA SYBER FAST One-Step quantitative reverse transcription-PCR (qRT-PCR) kit (Kapa Biosystems KK4651). RNA was first purified from bacterial cells grown at low and high OD using the Zymo Direct-zol RNA miniprep kit (Zymo Z5227). Oligonucleotide sequences are provided in Table S1. Reactions were run and measured using the Eppendorf Mastercycler RealPlex2 system. Annealing temperatures were set at 55°C. Relative gene expression changes were calculated using the Livak method.⁹² The reference gene used was rpoB.Quantification and Statistical Analysis.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistics

In CDD assays that measured the toxicity of cloned hairpin sequences, and were performed in triplicate, we applied an ordinary one-way ANOVA in PRISM (v9.4.1) for Windows.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
V.cholerae strain MAK757	Our collection	MAK757
V.cholerae E7946 SmR, Ptac-tfoX, pilA S67C,	Dalia et al.84	TND0652
lacZ:lacIq, ΔVC1807:CmR
Classical Ogawa V.cholerae strain	Our collection	O395
O395 LacI and tac promoter inserted 5′ to CRISPR-CAS operon	This work	WR2699
TND0652 purD-fis::WR2699 purD-fis	This work	WR2700
ΔVSP-2 = WR2700 ΔVC0490-VC0516	This work	WR2701
WR2700 ddmA Ile237::OrfAB	This work	WR2702
WR2700 ddmA Asp250Tyr	This work	WR2703
WR2700 ddmA Δ287-317frameshift	This work	WR2704
WR2700 ddmB Trp21stop	This work	WR2705
WR2700 ddmB Pro94frameshift	This work	WR2706
WR2700 ddmB Arg100stop	This work	WR2707
WR2700 ddmB Arg160frameshift	This work	WR2708
WR2700 ddmC Cys50Phe	This work	WR2709
WR2700 ddmC Leu125Pro	This work	WR2710
WR2700 ddmC Arg148frameshift	This work	WR2711
WR2700 ddmC Lys159stop	This work	WR2712
WR2700 ddmC Gln162Lys	This work	WR2713
WR2700 ddmC Asn197frameshift	This work	WR2714
WR2700 ddmC Ile255Ser	This work	WR2715
WR2700 ddmC Lys412Glu	This work	WR2716
WR2700 ddmC Thr513::OrfAB	This work	WR2717
WR2700 ddmC Asp523Tyr	This work	WR2718
WR2700 ddmC Phe571Cys	This work	WR2719
WR2700 luxO Met87Arg	This work	WR2720
WR2700 pcnB Δ287-457frameshift	This work	WR2721
WR2700 pcnB ILe34del	This work	WR2722
WR2700 pcnB Asp230frameshift	This work	WR2723
WR2700 pcnB Trp364stop	This work	WR2724
WR2700 priA Met462Arg	This work	WR2725
WR2700 ΔVSP-2 priA Met462Arg	This work	WR2726
WR2700 priA::6XHis fusion on CTD	This work	WR2742
BL21 fhuA2 [lon] ompT gal sulA11 R(mcr73:miniTn10-TetS)2 [dcm] R(zgb-210:Tn10-TetS) endA1 Δ(mcrC-mrr)114:IS10	New England Biolabs	NEBExpress
MFD-λpir	Ferrières et al.85
Biological samples
El Tor Group X phage Vibriophage N4	ATCC B1352-B5; Chattopadhyay et al.53	VIB04
VIB05	This work, Isolated from Bangladesh sample provided Shah Faruque and ICDDR,B	VIB05
Chemicals, peptides, and recombinant proteins
alkali-stable dioxigene-dUTP	Enzo Lifesciences	ENZ-NUC113-0025
dNTPs, set of 4	New England Biolabs	N0446S
Klenow Fragment (3′ → 5′ exo-)	New England Biolabs	M0212
random hexamers	New England Biolabs	S1230S
SmaI	New England Biolabs	R0141L
PmlI	New England Biolabs	R0532L
AvrII	New England Biolabs	R0175L
Antarctic phosphatase	New England Biolabs	M0298L
T4 Blunt T/A DNA ligase	New England Biolabs	M0367L
Critical commercial assays
KAPA SYBER FAST One-Step quantitative reverse transcription-PCR (qRT-PCR) kit	Kapa	KK4651
KAPA SYBR FAST qPCR kit	Kapa	KK4601
Zymo Direct-zol RNA miniprep kit	Zymo	Z5227
DIG Nucleic Acid Detection Kit	Roche	1117504191
Zymo Quick-DNA miniprep	Zymo	D3024
DNA Ultra II library preparation kit	New England Biolabs	E7645
Oligonucleotides
GTGCTATAACATAAAGCACCTTACGCTTGGA	This work	T7_20000F
CGAGTCCATCGGGGTCGCTCAGTAA	This work	T7_25000R
CTATAGAGGGACAAACTCAAGGTCATTCGC	This work	pUC18LacZhairpin_for
GGATGTGCTGCAAGGCGATTAAGT	This work	pUC18LacZhairpin_rev
ACCTGAAGGTCCAAACATCG	This work	RpoB_F
CAAAACCGCCTTCTTCTGTC	This work	RpoB_R
ACGGTACAGTTGCCGAATGG	This work	HapA_F
GCTGGCTTTCAATGTCAGGG	This work	HapA_R
GAACACACAGGTAGTACATCAACTT	This work	VC0492_F
CCGATGTAATTTGTATGACTCTCGT	This work	VC0492_R
Recombinant DNA
pUC18mini-Tn7-LacZ	Choi et al.86	pWR1566
pUC18mini-Tn7-LacZ ΔGmR between FRT sites	This work	pWR1566
Suicide Plasmid pRE107	Edwards et al.87
Suicide Plasmid pRE107 with 1.4kb fragment containing LacI and pTac promoter with 500nt flanking nt position O395 2827729 cloned into MCS (ScaI-KpnI)	This work	pWR1560
Suicide Plasmid pRE107 with cloned insert containing PriA CTD with 6X-His tag and ~500nt flanking sequence.	This work	pWR1593
pWR1566 with phage DNA inserted from VIB04	This work	pWR1594
pWR1566 with phage DNA inserted from VIB04	This work	pWR1595
pWR1566 with phage DNA inserted from VIB04	This work	pWR1596
pWR1566 deleted between SmaI and blunted AvrII site	This work	pWR1570
pWR1570 Seq2 between SmaI and blunted AvrII site	This work	pWR1571
pWR1570 Seq2rev between SmaI and blunted AvrII site	This work	pWR1572
pWR1570 Seq2mutated between SmaI and blunted AvrII site	This work	pWR1573
pET21a	Novagen	pET21a
pET21a ddmABC	This work	pWR1589
pET21a ddmA-K57A ddmBC	This work	pWR1590
pET21a priA	This work	pWR1591
pET21a priA 462R	This work	pWR1592
Software and algorithms
CLC Genomics Workbench V8.0	CLC Genomics Workbench	https://digitalinsights.qiagen.com/products-overview/discovery-insights-portfolio/analysis-and-visualization/qiagen-clc-workbench-premium/
ViennaRNA Package 2.0	Lorenz et al.88	https://www.tbi.univie.ac.at/RNA/
Pymol	The PyMOL Molecular Graphics System	https://www.pymol.org/
Alphafold2	Yang et al.51	https://github.com/google-deepmind/alphafold
Prism v9.4.1 for Windows		https://www.graphpad.com/features
Other
Nickel agarose column, 1mL	ThermoFisher	#88225
BrightStar™-Plus Positively Charged Nylon Membrane	Invitrogen	AM10100

Highlights.

• DdmABC targeting and degradation of plasmids are dependent on certain cis-acting sequences

• DNA sequences that induce DdmABC-based plasmid instability form hairpins

• Palindromic sequences include those from related phages that are also restricted by DdmABC

• DdmABC is blocked by a PriA allele