Plasmid targeting and destruction by the DdmDE bacterial defence system

Summary

While eukaryotic Argonautes play a pivotal role in post-transcriptional gene regulation through nucleic acid cleavage, some short prokaryotic Argonaute variants (pAgos) rely on auxiliary nuclease factors for efficient foreign DNA degradation (1). Here, we elucidate the activation pathway of the DNA Defense Module DdmDE system, which rapidly eliminates small, multicopy plasmids from Vibrio cholerae Seventh Pandemic Strain (7PET) (2). Through a combination of cryo-EM, biochemistry and in vivo plasmid clearance assays, we demonstrate DdmE is a catalytically inactive, DNA-guided, DNA-targeting pAgo with a distinctive insertion domain. We observe that DdmD transitions from an autoinhibited, dimeric protein to monomers upon loading of single-stranded DNA targets. Furthermore, the complete structure of the DdmDE-guide-target handover complex provides a comprehensive view into how DNA recognition triggers processive plasmid destruction. Our work establishes a mechanistic foundation for how pAgo utilize ancillary factors to achieve plasmid clearance, and provides insights into anti-plasmid immunity in bacteria.

Main

Prokaryotic organisms are engaged in a constant evolutionary arms race against parasitic mobile genetic elements, which include phage and plasmids (3, 4). While plasmids can propel rapid bacterial evolution, the benefits they confer are often offset by the fitness costs imposed by their maintenance. The threat posed by plasmids is evident from the abundance of bacterial anti-plasmid immune systems (5–10), and the subsequent anti-defense genes encoded within the leading regions of plasmids themselves (11). A recent study elegantly investigated why the Vibrio cholerae seventh pandemic strain (7PET, originating in 1961) are intolerant of plasmids (2), finding that this immunity was attributed to two distinct defense systems unique to 7PET strains: DdmABC and DdmDE. DdmABC (a type II Lamassu system (12–14)) confers immunity against large, low copy number plasmids, while DdmDE rapidly degrades smaller, multicopy plasmids, regardless of their origin of replication (2). These two defense systems impart a fitness advantage to 7PET strains, enabling cells lacking plasmids to outcompete other Vibrio cholerae strains that tolerate and maintain plasmids (2), which may have contributed to the success of the 7PET lineage (11, 15, 16), which still causes millions of cases of cholera each year (17). DdmDE, a dual-component system consisting of a predicted helicase-nuclease (DdmD) and a putative prokaryotic Argonaute (DdmE), is found in a small fraction of genomes (0.6% of the 22,000 genomes in the RefSeq database (18)). It is encoded within the Vibrio pathogenicity island 2 (VPI-2) of 7PET, a defense island that also contains a Zorya type I system (13), type I and IV Restriction-Modification system and a WYL-domain transcription regulator (12, 19–21) (Extended Data Fig 1). However, the underlying molecular mechanisms of how these systems detect and remove plasmids remains unclear. Here, we reveal a plasmid targeting mechanism orchestrated by DdmE, involving the utilization of a 5’ monophosphorylated (5’-P) DNA guide to identify complementary DNA targets. Subsequently, an autoinhibited DdmD dimer is recruited, culminating in the formation of a DdmD-DdmE target handover complex, which we demonstrate is pivotal for immune activation and effective plasmid degradation.

DNA targeting by DdmE

While bioinformatic analysis of DdmE was unable to detect known domains (2), initial structural prediction using AlphaFold2 (AF2) (22) indicated that DdmE may resemble a prokaryotic Argonaute (pAgo) protein, particularly the MID and PIWI domains (23). pAgos are a large and diverse family of proteins associated with bacterial immunity and can be divided into two major categories – large pAgos (containing N, PAZ, MID and PIWI domains) and short pAgos (containing MID and PIWI domains, but lacking N and PAZ) (24). Since pAgos are capable of guide-dependent DNA- and RNA-targeting (1, 25, 26), we hypothesized that DdmE may function as the targeting module of the DdmDE system.

Using previously-established pAgo guide and target substrates (27), we tested the guide-dependent target binding of DdmE. We observed that DdmE bound a DNA target using a single-stranded DNA guide with a 5’ monophosphate (5’-P) but was incapable of binding RNA targets or using a DNA guide with 5’-OH (Fig 1a). No cleavage was observed for any substrate, as expected since DdmE lacks a typical pAgo catalytic DEDD tetrad in the PIWI domain (Extended Data Fig 1) (28).

Fig 1. DdmE is a prokaryotic Argonaute that uses 5’-P DNA guides to identify DNA targets.

a, Native gel shift assay to determine guide and target binding preferences of DdmE. b, 3.1 Å-resolution cryo-EM structure of DdmE in complex with 5’-P DNA guide and DNA target. c, Model of DdmE nucleoprotein complex, with corresponding protein domains shown below. d, Guide 5’-P binding. e, Interactions between DdmE sensor loop and the guide – target duplex. f, Comparison of Thermus thermophilus (Tt)Ago active site with corresponding region of DdmE. g, In vivo plasmid clearance assay, testing the importance of DdmE residues. Significance between DdmDE and other variants was determined by ordinary one-way ANOVA test. Data are mean ± s.d. of at three independent experiments. ***P = 0.0001, ****P < 0.0001. For gel source data, see SI Fig. 1.

We then determined a cryo-EM structure of DdmE in complex with its cognate guide and target at a global resolution of 3.1 Å (Fig 1c, Extended Data Fig 2). We observed well-resolved density for 13 base pairs of the guide – target duplex (corresponding to positions 2 – 14 of the guide), while the density for the downstream duplex and the N-terminal (N) domain were more diffuse, indicating flexibility (Extended Data Fig 3). The displacement of the N domain upon guide-target hybridization is reminiscent of human Ago2 in instances with extended miRNA – target pairing (29). Incidentally, additional duplex density was observed at the 3’ end of the target (Fig. 1c, left), likely due to serendipitous self-complementarity of the substrates used in previous studies. This supplementary duplex is accommodated within the channel formed between the MID and insertion (INS) domains. The INS domain itself consists of a small α-helical bundle located within the PIWI domain, and has no known structural homologues.

The first base of the DNA guide is flipped out and does not base-pair with the target, and instead folds back on itself to coordinate a single Mg²⁺ ion through the phosphate groups of the first and second nucleotides (Fig 1d). The first guide nucleobase is accommodated through sequence-independent interactions, including stacking with Y363 in the MID domain, K405 interacting with the 5’-P, and R382 interacting with the second phosphate and rigidifying the flipped-out position one nucleobase (Fig 1d).

In a previously-determined structure of Thermus thermophilus Ago (TtAgo, PDB 4NCB), the central region of the guide – target duplex is contacted by a short alpha-helix from the L1 domain (Extended Data Fig 3). In place of this helix, DdmE contains a short unstructured loop that connects the L1 and L2 domains (Fig 1e). This ‘sensor loop’ lines the minor groove from positions 6 – 9. Within the sensor loop, two basic residues penetrate the minor groove (K230, R232). By intercalating within the minor groove of the guide-target duplex the sensor loop may confer specificity for B-form duplex, providing a structural rationale for the inability of DdmE to utilize RNA guides or targets which would have A-form duplexes. Alignment of the PIWI domains of TtAgo and DdmE reveals (23)that three of the four residues that correspond to the catalytic tetrad are positive or hydrophobic (R677, M428 and P488), while the final equivalent residue is completely absent, preventing coordination of divalent catalytic ions (Fig 1f), providing a structural basis for the catalytic inactivation of DdmE.

To test the importance of various DdmE residues, we adapted an in vivo plasmid interference assay (Extended Data Fig 1) (30). In short, co-expression of DdmD and DdmE causes plasmid clearance, which translates into fewer colonies after antibiotic selection (Extended Data Fig 1). Presence of both DdmD and DdmE resulted in a ~10,000-fold reduction in transformation efficiency, while expression of either protein alone or expression without antibiotic selection had no effect. Notably, the Y363A point mutation reduces the plasmid elimination efficiency by ~1,000-fold (Fig 1g), confirming its importance in guide recognition, while neither of the two residues within the dPIWI domain that contact the guide-target duplex appear to be critical for in vivo plasmid clearance (Fig 1g). Mutation of either sensor loop residues (K230, R232) reduced the plasmid interference in vivo by ~500- and 1000-fold, respectively (Fig 1h). Thus, despite a lack of discernable sequence similarity to other previously-characterized pAgo proteins (25, 31), our structural and biochemical data demonstrate that DdmE is a short pAgo that uses a 5’-P DNA guide to target DNA.

DdmD nuclease and ATPase activity

Having established that DdmE uses a 5’-P DNA guide to target complementary DNA sequences, we next sought to characterize DdmD. DdmD contains a PD-(D/E)-xK nuclease domain which is also found in type II restriction enzymes, which typically use Mg²⁺ or an alternative other divalent cation as a co-factor (32, 33). We tested the metal-dependent cleavage activity of DdmD using a 5’ FAM-labelled DNA substrate and observed that MnCl₂ was essential for cleavage, while MgCl₂, NiCl_2, and CaCl₂ did not support cleavage (Fig 2a).

Fig 2. DdmD is a dimeric helicase-nuclease.

a, Dependence of ssDNA cleavage by DdmD on divalent cations. b, DdmD cleaves ssDNA, but not duplexed DNA. Minor cleavage products are observed for a DNA duplex with a 5’ overhang. c, Stimulation of DdmD ATPase activity by DNA substrates as used in b. Significance between ATPase stimulation by ssDNA compared to no DNA was determined by ordinary one-way ANOVA test (n=3 technical replicates). ****P < 0.0001. Error bars correspond to standard error of the mean. d, 3.2 Å-resolution cryo-EM structure of DdmD, and corresponding model. Domains are colored as in the below domain schematic. e, 3.0 Å-resolution cryo-EM structure of DdmD in complex with short overhang duplex. The second monomer of DdmD is shown as transparent density. f, 3.0 Å-resolution structure of DdmD in complex with a DNA substrate with longer overhangs. g, Y194 caps the 3’ end of the ssDNA within RecA channel. h, In vivo plasmid clearance assay. Data are mean ± s.d. of three independent experiments started from separate colonies. i, Structural changes to DdmD Linker domain upon DNA binding. j, Schematic of DdmD activation. For gel source data, see SI Fig. 1.

We then tested the impact of DNA structure on nuclease activity of DdmD. While DdmD degraded single-stranded (ss)DNA after 1h of incubation at 37°C, it was unable to degrade double-stranded DNA (Fig 2b). We also observed incomplete degradation of a partially duplexed substrate with a 10 nucleotide 3’-overhang, but no cleavage of a substrate with a 5’-overhang of the same length (Fig 2b). This indicates that DdmD preferentially targets ssDNA.

Since DdmD contains RecA helicase domains, which typically couple DNA translocation with ATP hydrolysis, we next tested how these substrates may stimulate the ATPase activity of DdmD. We performed an ATPase assay in the absence of Mn²⁺ to prevent nucleic acid cleavage. We observed strong stimulation of ATPase activity (~20,000-fold) in the presence of ssDNA, but this effect was severely reduced in the presence of fully or partially double-stranded DNA (Fig 2c).

The inability of DdmD to degrade dsDNA suggests a mechanism to prevent non-specific DNA cleavage and cellular toxicity. Therefore, this immune system may be able to confer broad protection without risking autoimmunity with DdmD identifying the DNA structural context once DdmE has identified a target sequence, as has been demonstrated for other defense systems (8, 34).

DdmD autoinhibition and activation

Next, we determined a cryo-EM structure of DdmD at a global resolution of 3.2 Å, enabling us to build a complete atomic model de novo (Fig 2d, Extended Data Fig 2). Surprisingly, DdmD forms a pincer-shaped dimer, with two copies of DdmD stacked upon each other via the RecA2 and linker domains. This assembly has cyclic 2-fold (C2) symmetry and a buried surface area of ~2,300 Ų. The DdmD dimer was unexpected, as this arrangement positions the DNA exit sites of the typical RecA translocation channel of each monomer in opposite orientations, preventing processive DNA translocation. This suggests that the DdmD dimer is an autoinhibited state.

To understand how DdmD is activated, we incubated the complex with the non-hydrolyzable ATP analogue adenylyl-imidodiphosphate (AMP-PNP) and a forked dsDNA substrate (18-bp duplex, and 15-nt 5’ and 3’ overhangs) and determined a structure at a global resolution of 3 Å resolution (Fig 2e). This structure is also dimeric and largely resembles the apo complex, but the linker domain is displaced, and the gap between the nuclease domains increases from 48 to 55 Å. 11-nt of the 3’ overhang of the forked DNA substrate were well-resolved in the RecA channel of each monomer. Within the RecA channel, DdmD makes numerous non-specific charge-based contacts, which are essential for in vivo plasmid clearance (Extended Data Fig 4).

Examination of the 2D class averages of the short fork DNA-bound complex revealed a minor population of monomeric DdmD (Extended Data Fig 4), which yielded a highly anisotropic 3D reconstruction (Extended Data Fig 5h). Based on this observation, we hypothesized that the autoinhibited DdmD dimer may disassemble and become activated upon loading of longer DNA substrates. To test this, we determined a structure of DdmD bound to a forked dsDNA substrate with longer overhangs (30-nt). The resulting 2D class averages were predominantly monomeric, with a small population (~7%) of poorly resolved, flexible dimers. We resolved a 3.0 Å structure of the DNA-bound DdmD monomer (Fig 2g). In this structure, the linker domain is partially disordered, and the top of RecA2 appears flexible.

Within the short overhang structure, the 3’ end of each DNA strand leads to the center of the complex, but rather than clash, each strand is capped by a single aromatic residue, Y194 (Fig 2h). In the long overhang monomeric structure, ssDNA has passed through the channel, suggesting that Y194 may act as a sensor for ssDNA loading onto DdmD, preventing passage of a substrate from one DdmD to another. The Y194A mutation significantly reduced the plasmid clearance activity of DdmDE in vivo (Fig 2h).

Structural superposition of a monomeric DdmD subunit from apo and DNA-bound states highlights the conformational change of the Linker domain of up to ~10 Å upon DNA binding (Fig 2i). This structural shift induces the ~7 Å opening between the nuclease domain ‘pincers’ (Fig 2d & e), abrogating the Y194 blockade, causing the dimer to disassemble and enabling translocation and passage of ssDNA.

We tested DNA cleavage by DdmD in the presence of ATP, MgCl₂ and MnCl₂ (essential co-factors for ATP hydrolysis and nuclease activity, respectively). While ssDNA was efficiently degraded, a partial degradation product was observed for a substrate with a 5’ overhang, which would enable trimming by DdmD but not DNA unwinding and translocation (Extended Data Fig 5). However, when supplemented with ATP, complete degradation was observed, confirming that the helicase activity of DdmD is necessary to generate single-stranded DNA for degradation. We also observed that DdmD unwinds forked dsDNA substrates in an ATP-dependent manner, providing further evidence that DdmD is a processive RecA-like helicase (Extended Data Fig 5c & d), similar to with other previously-characterized SF2 helicases (35–38).

These data suggest a model where DdmD is an autoinhibited dimer and loading of ssDNA triggers a structural switch to facilitate disassembly into monomers. Since the DNA exit channel of DdmD is blocked within the dimeric assembly and translocation-coupled unwinding appears to be essential for complete DNA degradation, we propose a model where DNA loading triggers a structural change in DdmD that promotes a dimer-to-monomer transition (Fig 2j). The monomeric DdmD is then able to processively unwind and translocate along one duplex strand, while the now-separated second DNA strand is degraded by the nuclease domain. Size-exclusion chromatography, native PAGE and negative stain EM analysis and further suggests that DNA loading triggers a dimer-to-monomer transition (Extended Data Fig 5e, f, g, h). This mechanism of autoinhibition and substrate-induced activation provides a rationale for the inability of DdmD to clear plasmids in the absence of DdmE, potentially mitigating the toxic consequences of constitutive expression of a non-specific nuclease. While phage and plasmids also contain single-stranded regions during replication, these regions may be protected from DdmD by single-stranded DNA-binding (SSB) proteins.

DdmDE handover complex structure

We hypothesized that since both DdmE and DdmD are required for plasmid clearance in vivo, they may assemble into a handover complex to ensure targeted DNA targeted by DdmE is subsequently degraded by DdmD. Using size-exclusion chromatography, and found that while DdmD and DdmE proteins alone did not form a complex, preincubation of DdmE with 5’-P guide and complementary target with a 15-nt poly-T 5’ and 3’ overhangs prior to addition of DdmD enabled us to purify a complete handover complex (Fig 3a), which we subsequently vitrified and imaged using cryo-EM.

Fig 3. DdmE recruits DdmD upon target recognition for plasmid degradation.

a, Size-exclusion chromatography analysis of DdmD – DdmE – guide – target complex reconstitution, with corresponding fractions monitored by SDS-PAGE (right). *Denotes peak fraction used for cryo-EM preparation of DdmDE handover complex. b, 2.5 Å-resolution cryo-EM structure of DdmDE handover complex. c, Top-down view of DdmDE handover complex, showing path of DNA target as it is transferred from DdmE to DdmD. Top DdmD monomer has been removed for viewing clarity. Y194 caps the 3’ end of the DNA target. For gel source data, see SI Fig. 1.

During 3D classification of the DdmDE handover complex cryo-EM dataset, we observed a mixture of oligomeric species including Ddm(D)₂(E)₂ and Ddm(D)₂(E)₁, which we resolved at 3.0 and 2.5 Å, respectively (Extended Data Fig 2 & 6). An additional population of a 1.3 MDa dodecameric (DdmD₆E₆) species was also resolved (Extended Data Fig 5d), reflecting the emerging theme of supramolecular oligomeric assembles in bacterial immunity (including CBASS (39, 40), CRISPR-Cas (41), Gabija (42) and RADAR (43, 44)). While it is tempting to speculate that DdmD₆E₆ oligomers may provide a mechanism of locally concentrating DdmDE complexes to cellular foci to efficiently target multicopy plasmids and their replicating intermediates, we focused out attention on DdmD₂E₁ since this was the best resolved assembly (Fig 3b).

Within the DdmDE handover complex, the targeted DNA forms a duplex with the DdmE-bound guide for 14-bp, and the remaining 3’ end of the DNA target is channeled into the RecA domains of DdmD (Fig 3c). Unlike in the structure of monomeric DdmE described above (Fig 1), the N-terminal domain is ordered and well-resolved, indicating that extended target – guide base pairing dislodges the N domain (Extended Data Fig 3). The extreme 3’ of the target is capped by DdmD(Y194), suggesting this structure represents an intermediate where the target has been delivered to DdmD prior to activation. We therefore propose that our complex resembles a pre-initiation state, where target is being transferred to DdmD but DdmD dimer disassembly and ATP-driven translocation have not yet occurred.

DdmDE interface is required for activity

In the handover complex, DdmE exhibited lower local resolution, and required local focused refinement to improve the quality of the reconstruction (Extended Data Fig 2). This observation implies that DdmE is flexibly tethered to the DNA target during the handover process.

DdmD and DdmE have a bidentate interaction interface totaling ~1200 Ų, with a major interface – DdmD(RecA1) to DdmE(INS & L2) contributing ~700 Ų, while the minor interface – DdmD(Linker) to DdmE(MID) further stabilizes the complex and contributes an additional ~300 Ų of shared buried surface area (Fig 4a). Both interfaces are predominantly formed through electrostatic interactions (Fig 4c, Extended Data Fig 7). Charge-swap mutations of the three electrostatic contacts pairs significantly impaired in vivo plasmid clearance, confirming the importance of these contacts for activity (Fig 4e). Within the DdmD major interface, a single residue – R620 – stands out as highly conserved, and is inserted into a small pocket on the surface of DdmE between the L1 and L2 domains (Fig 4e, Extended Data Fig 7). The DdmD(R620A) mutation significantly reduced plasmid clearance in vivo and resulted in a reduced co-eluting DdmD-DdmE fraction as measured by size-exclusion chromatography (Extended Data Fig 5), confirming its role in handover complex formation.

Fig 4. DdmD-DdmE interface is essential for plasmid clearance.

a, Structures of DdmD and DdmE from handover complex (Fig 3), with complementary surfaces of DdmD and DdmE. Interface between DdmD(RecA1) and DdmE(INS – L2) are highlighted as blue and yellow, respectively. b, Electrostatic contacts between DdmD(RecA1) and DdmE(INS) at the major interface. c, An electrostatic contact at the minor interface between DdmD(linker) and DdmE (MID) domains. d, DdmD(R620A) interacts with DdmE within a conserved pocket. e, Plasmid interference assay to analyze DdmD-DdmE interactions. Significance between DdmDE and other variants was determined by ordinary one-way ANOVA test. ****P < 0.0001. Data are mean ± s.d. of three independent experiments started from separate colonies. DdmDE, DdmD and DdmE are the same data as shown in Fig 1h.

Discussion

Our work reveals the mechanism of anti-plasmid immunity by Vibrio cholerae DdmDE. We demonstrate that DdmE employs a 5’-monophosphophorylated DNA guide to selectively target complementary DNA sequences and subsequently recruit autoinhibited, dimeric DdmD. The process involves loading and translocating single-stranded DNA target into the DdmD RecA helicase channel, culminating in activation of the complex. Activation triggers the disassembly of the presumably autoinhibited DdmD dimer to facilitate processive DNA degradation. We propose that recruitment of dimeric DdmD ensures efficient plasmid DNA elimination by enabling each DdmD monomer to load a strand of plasmid DNA targeted by DdmE.

It is likely that DNA guides for DdmE are generated by the host RecBCD complex, as has been demonstrated for the DNA-guided, DNA-targeting Clostridium butyricum pAgo (45), based on the fact that RecBCD produces 5’-P ssDNA fragments and that plasmids characteristically lack Chi sites (46). This indicates that DdmDE may act to rapidly amplify multicopy plasmid degradation once RecBCD has identified a target. Further investigations are warranted to elucidate the guide generation process in the DdmDE system and other DNA-guided pAgos (1, 47).

The use of separate modules for target recognition and DNA degradation is reminiscent of type I CRISPR-Cascade systems, which employ the Cas3 nuclease-helicase and have been adapted as gene editing tools to introducing long-range genomic deletions (48, 49). By exploiting phage to deliver bespoke guides into virulent bacteria, it has been possible to redirect host CRISPR for self-targeting and confer antimicrobial protection (50, 51), and a similar self-targeting strategy could potentially be applied to DdmDE for treatment of Vibrio cholerae 7PET infections.

Since the DdmE(INS) domain is absent from all other short pAgo homologue structures determined to date (28), we propose that it functions as an adaptor module to recruit DdmD. Future structural engineering efforts could introduce the INS domain into other pAgo homologues to enable recruitment of DdmD and enhancement of nuclease activity. This modularity was recently demonstrated for a CBASS effector, where novel biosensors were developed through effector domain swapping (52). This structural understanding paves the way for future advancements in manipulating the DdmDE system for diverse applications.

Our DdmDE handover complex reveals how a catalytically inactive pAgo can recruit an ancillary factor for DNA destruction, expanding on the theme of diverse pAgos utilizing orthogonal effctors to achieve population-level immunity through abortive infection (53–55). Overall, these results underscore the diversity of pAgo immune strategies, and advances our understanding of host-plasmid interactions.

Methods

Expression and Purification of DdmDE

BL21(DE3) competent cells (Invitrogen) containing recombinant plasmids with DdmD-6xHIS and 6xHIS-MBP-DdmE, respectively, were cultured in LB media at 37°C until the OD₆₀₀ reached 0.6. Protein expression was induced by adding a final concentration of 0.5mM isopropyl-ß-D-1-thiogalactopyranoside (IPTG) for 18 hours at 18°C.

The bacterial pellets were lysed via sonication in lysis buffer containing 20 mM Tris-HCl pH 7.5, 10% Glycerol, 1 M NaCl, 0.5mM (tris(2-carboxyethyl)phosphine) TCEP, 10 mM MgCl2, 1X DNAse, 2mM phenylmethylsulfonyl fluoride (PMSF) and Pearce protease inhibitor cocktils (ThermoScientific). Clarified lysate was loaded onto a HisTrapTM HP column, washed in a buffer containing 20 mM Tris-HCl pH 7.5, 10% Glycerol, 1 M NaCl, 0.5mM TCEP and eluted in buffer containing 20 mM Tris-HCl pH 7.5, 10% Glycerol, 1 M NaCl, 250 mM Imidazole pH 8.0, 0.5mM TCEP. Fractions containing either fusion protein (DdmD-6xHIS or 6xHIS-MBP-DdmE) were added to dialysis tubing and incubated in dialysis buffer containing 20 mM HEPES pH 7.5, 10% Glycerol, 150 mM KCl, 0.5 mM TCEP at 4°C for 16 hours. TEV protease was added to DdmE for affinity tag cleavage during dialysis. Dialyzed protein was then loaded onto a HiTrap Q HP anion exchange chromatography column, washed in a buffer containing 20 mM HEPES pH 7.5, 10% Glycerol, 150 mM KCl, 0.5 mM TCEP, and eluted in a buffer containing 20 mM HEPES pH 7.5, 10% Glycerol, 1 M KCl, 0.5 mM TCEP. Eluted Q column protein fractions were pooled, concentrated, and loaded onto a size exclusion SUPERDEX 200 g 16/500 in a buffer containing 20 mM HEPES pH 7.5, 10% Glycerol, 150 mM KCl, 0.5 mM TCEP. Size-exclusion eluent proteins were aliquoted, flash frozen in liquid nitrogen, and stored at −80°C.

Plasmid Interference Assay

BL21(DE3) competent cells containing recombinant plasmids for both DdmD and DdmE were grown up to an OD₆₀₀ ~0.6 at 37°C. The culture was then split into two equal volumes: one volume as an uninduced negative control and the other induced with a final concentration of 0.5mM IPTG. Both −/+ IPTG cultures were incubated for 16 hours at 18°C. A series of 8, 10-fold dilutions were made for both the −/+ IPTG cultures and were plated on dry LB agar plates containing antibiotics for both the DdmD and DdmE recombinant plasmids. The LB agar plates were incubated at 37°C overnight. Colonies were quantified from the dilution series to measure colony growth repression due to DdmDE targeting the recombinant plasmids and eliminating bacterial antibiotic resistance.

DdmDE handover complex in vitro reconstitution

20μM of purified DdmE was incubated with final concentrations of 10mM MgCl₂ and 25μM of an 18bp 5’P-DNA guide in a buffer containing 150 mM NaCl, 25 mM HEPES pH 7.5 at 37°C for 30 minutes. 33μM of DNA target was added to the reaction and incubated for an additional 30 minutes at 37°C. 20μM purified DdmD was incubated with a final concentration of 1mM Adenylyl-imidodiphosphate (AMP-PNP) in annealing buffer for 30 minutes at room temperature. Equal volumes of DdmE and DdmD reactions were mixed and incubated at 37°C for 30 minutes. The final DdmDE reaction was loaded onto a Superdex 200 g 16/600 and eluted in buffer containing 20 mM HEPES pH 7.5, 10% Glycerol, 150 mM KCl, 0.5 mM TCEP.

Cryo-EM sample preparation & data acquisition

Flash-frozen DdmE was incubated with 18-nt 5’-P DNA guide and DNA target in equimolar stoichiometries at a final concentration of 10 μM. 2.5 μL of this complex was applied to C-flat holey carbon grids (1.2/1.3, 400 mesh) which had been plasma cleaned for 30 seconds in a Solarus 950 plasma cleaner (Gatan) with a 4:1 ratio of O₂/H₂. Grids were blotted with Vitrobot Mark IV (Thermo Fisher) for 6 seconds, blot force 0 at 4°C & 100% humidity, and plunge-frozen in liquid ethane. Data were collected using a FEI Titan Krios cryo-electron microscope equipped with a K3 Summit direct electron detector (Gatan, Pleasanton, CA). Images were recorded with SerialEM v3.9.0 beta (56) with a pixel size of 0.81 Å. A total accumulated dose of 70 electrons/Ų during a 6 second exposure was fractionated into 80 frames. Motion correction, CTF estimation and particle picking was performed on-the-fly using cryoSPARC Live v4.0.0-privatebeta.2(57). All subsequent data processing was performed in cryoSPARC v4.3(58).

DdmD was prepared as above, at 10 μM in the presence of 1mM AMP-PNP and 2 μM short overhang target, in the hope that we could determine multiple structures of apo and DNA-bound DdmD from the same dataset. Grids were prepared as above. Data were collected on a FEI Glacios cryo-TEM equipped with a Falcon 4 detector. Data was collected in SerialEM, with a pixel size of 0.94 Å, a defocus range of −1.5 - −2.5 μm, and a total exposure time of 15s resulting in a total accumulated dose of 40 e/Ų which was split into 60 EER fractions. Subsequent data processing was performed as described above. Since the final 3.2 Å-resolution reconstruction contained no resolvable DNA density, we refer to this structure as the DdmD apo state.

For the short fork DNA-bound DdmD structure, 10μM DdmD was incubated with 20 μM DNA in the presence of 1 mM AMP-PNP for 1h at room temperature, and grids were prepared as described above. Data were collected on. Data were collected using a FEI Titan Krios cryo-electron microscope equipped with a K3 Summit direct electron detector (Gatan, Pleasanton, CA), and processed as described above.

For the long fork DNA-bound DdmD structure, 10μM DdmD was incubated with 20 μM DNA in the presence of 1 mM AMP-PNP for 1h at room temperature, and grids were prepared and imaged as described for DdmE.

For the DdmDE handover complex, 20 μM DdmE was pre-incubated with 25 μM guide at 37°C for 30 mins, before addition of 33 μM target DNA. The target DNA was designed to have no self-complementarity and contained an 18-nt region perfectly complementary to the guide sequence, flanked by 15-nt poly-T 5’ and 3’ overhangs. The final volume was 300 μl. 20 μl DdmD was incubated for 30 mins at 37°C with 1 mM AMP-PNP in 300 μl volume, and the DdmE(guide-target) and DdmD(AMP-PNP) reactions were subsequently co-incubated for 30 mins at 37°C before loading onto an S200 column. SEC peak fractions containing both DdmD and DdmE (as determined by SDS-PAGE analysis) were used to prepare grids as described above. The complex was imaged and data was collected as described for DdmE.

Cryo-EM data processing, model building and figure preparation

All cryo-EM data processing was done in CryoSPARC v4.3. For all datasets, particles were picked using blobpicker, and extracted particles were subjected to 2D classification. Suitable classes were selected and used for ab-initio reconstruction (3 classes) and heterogeneous refinement (using the ab-initio models as templates). After multiple rounds of ab-initio reconstruction and heterogeneous refinement, and a homogeneous subset of particles had been curated, non-uniform refinement was used to determine a high-resolution consensus reconstruction. To overcome the small size of DdmE (~75 kDa), local refinement of the entire complex with custom parameters (using pose/shift gaussian prior during alignment, with rotation search extent set to 5 degrees, and shift search extent set to 3 Å) was used, resulting in a 3.1 Å reconstruction. DdmD apo and DdmD short fork DNA-bound structures were refined with C2 symmetry. DdmD long fork DNA-bound structure was refined with C1 symmetry, and local refinement of the entire complex was used with default parameters to improve the quality of the reconstruction. In the DdmDE handover complex, a consensus reconstruction reached 2.5 Å resolution, and local refinement of DdmE was used to compensate for local flexibility. Phenix Resolve density modification was used to aid interpretation and model building of DdmE (59).

For structural modelling, AF2 structures of DdmD and/or DdmE were rigid body fitted into the corresponding maps, and Isolde v1.6 (60) was then used for flexible fitting. Nucleic acids were built using Coot v0.9.1 (61). Real-space refinement as implemented within Phenix v1.18.2 (62) was performed to optimize model geometry.

All structural figures and movies were generated using ChimeraX-1.6 (63, 64). Modevector arrows were created using the PDBarrows script described in Chaaban et al. (65).

Native electrophoretic mobility shift assays (EMSAs)

1 μM DdmE was pre-incubated with 2 μM 5’-P or 5’-OH DNA/RNA guides in 50 mM Tris-HCl pH 7.5, 250 mM KCl, 10 mM MgCl₂ at 37°C for 30 mins prior to subsequent incubation with 5 nM 5’-Cy5 labelled DNA/RNA targets, followed by another 30 min incubation at 37°C. A total of 10 μl of each sample was mixed with 2 μL 6× loading buffer (30% v/v glycerol, 5% Ficoll 400, 50 mM NaCl, 10 mM HEPES pH 8, 5 mM EDTA, 0.002% w/v bromophenol blue). Electrophoresis was carried out on a non-denaturing 1.5% agarose gel for 90 min at 100 V in 1 × Tris-borate-EDTA buffer. Gels were imaged using a fluorescence scanner. Uncropped and unprocessed scans are provided in the Source Data file.

Cleavage assay

DdmE cleavage: samples prepared as for the binding assay were quenched by the addition of 1 mM EDTA and mixed with SDS-containing loading buffer. 5 μl of each sample was loaded onto a 15% TBE Urea-PAGE gel, and electrophoresis was carried for 30 mins at 180 V in 1xTBE buffer. Gels were imaged using a fluorescence scanner.

DdmD cleavage: 2 μM DdmD was incubated with 500 nM 5’-FAM labelled DNA target in 25 mM HEPES pH 7.5, 150 mM NaCl buffer, supplemented with a final concentration of 10 mM of various divalent cations, for 1h at 37°C. Reactions were quenched as described above and resolved on a 15% TBE Urea-PAGE gel as described above. One it was determined that MnCl₂ was an essential co-factor for DdmD cleavage, reactions were performed in 25 mM HEPES, 150 mM NaCl, 10 mM MnCl₂ buffer. In Ext. Data Fig 5a & b, reactions were supplemented with 10 mM MgCl2 and 1 mM ATP.

Fig 5. Model of targeted plasmid degradation by DdmDE.

DdmE (or multiple DdmE) uses a 5’-P DNA guide to recognize multicopy plasmids. Target recognition results in recruitment of a dimeric, autoinhibited DdmD, and the DNA is loaded within the RecA channel. Initiation of ssDNA-activated ATP hydrolysis results in the processive translocation of DdmD along the DNA, with the dimer splitting. As DdmD translocate, the non-target DNA strand may be passed to the nuclease active site for non-specific cleavage.

ATPase assay

ATPase assays were as previously performed (8, 66). Briefly, 1 μM DdmD was incubated with 2 μM DNA substrates in 25 mM HEPES pH 7.5, 150 mM NaCl, 10 mM MgCl₂ buffer, and 1 mM ATP. ATP hydrolysis was monitored using the EnzChek Phosphate Assay Kit, using a Synergy HTX multimode reader (BioTek Instruments, Inc.). Reactions were performed in triplicate in a 384-well flat bottom clear microplate (Greiner Bio-one #E190239B), at 37°C for 45 minutes, and phosphate release was quantified by measuring absorbance at 360nm. Single exponential curves were fitted for each substrate using KinTek (67), and fold increase in ATP hydrolysis for each DNA substrate compared to DdmD alone are reported in Fig 2c. Significance between ATPase stimulation by ssDNA compared to no DNA was determined by ordinary one-way ANOVA test (n=3).

DNA unwinding assays

For the real-time unwinding assay, 2 μM DdmD was incubated with 10 nM duplex of 5-FAM-labelled DNA annealed to BlackHole Quencher-labelled partially complement strand, resulting in a 30-nt single-stranded fork and a 25-bp duplex. After mixing all components apart from ATP, an unlabeled DNA strand fully complementary to the FAM-labelled oligo was added at a final concentration of 100 nM to prevent re-annealing of the quencher-labelled strand. Reactions were performed in 25 mM HEPES pH 7.5, 150 mM NaCl, 10 mM MgCl₂ buffer at 37°C with varying concentrations of ATP, and increase in FAM fluorescence was monitored over time, using a CLARIOstar Plus multi-detection plate reader (BMG Labtech).

For gel-based unwinding assay, reactions were performed as described above, but without the BlackHole Quencher label on the strand annealed to the FAM-strand. After 30 minutes incubation at 37°C, reactions were quenched by addition of final concentration 1 mM EDTA and 1.2 units of Proteinase K (Thermo Fischer). After a further 30 minutes of proteinase K digestion at 37°C, samples were resolved on a 10% polyacrylamide TBE-PAGE gel and imaged for FAM fluorescence.

Native gel shift assay

To test the influence of DNA on DdmD oligomeric state 2 μM DdmD was incubated with 5 μM ssDNA or 30-nt overhang fork substrate for 30 minutes at 37°C, and resolved on a 10% polyacrylamide TBE-PAGE gel. Coomassie staining was performed to image for DdmD.

Extended Data

Extended Data Fig 1. DdmDE denfence system.

a, Representation of VPI-2 defense island. Annotation was performed by PADLOC. b, Schematic representation of in vivo plasmid clearance assay. In brief, DdmD and/or DdmE were transformed into E. coli BL21 DE3 cells. c, representative dilution series of cultures either induced or uninduced. d, Quantification of transformation-fold reduction for DdmDE co-transformed, DdmD and DdmE alone, DdmDE co-transformed in the absence of antibiotic selection, and DdmD – DdmEΔINS domain. Significance between DdmDE and other variants was determined by ordinary one-way ANOVA test. Data are mean ± s.d. of at three independent experiments started from separate colonies. e, Denaturing urea-PAGE gel of DNA and RNA targets bound by DdmE (same samples as in figure 1a). No cleavage was observed. Representative of three independent experiments. For gel source data, see Supplementary Information Figure 1.

Extended Data Fig 2. Resolutions of cryo-EM structures.

Gold-standard Fourier shell correlation (FSC) curves, Three-dimensional FSC curves and maps colored by local resolution for DdmE; FSC 3.1 Å (a), DdmD apoprotein; FSC 3.2 Å (b), DdmD + short overhang DNA; FSC 3.0 Å (c), DdmD monomer with long overhang DNA; FSC 3.0 Å (d), DdmDE consensus reconstruction; FSC 2.5 Å (e), DdmDE local refinement of DdmE; FSC 2.6 Å (f).

Extended Data Fig 3. Conformational changes during substrate handover.

a, Comparison of DdmE structure (colored) with DdmE in the context of the DdmDE handover complex. In the handover complex, the N-terminal domain is ordered and present (red). The structures are otherwise identical (RMSD <2 Å). b, 6 Å-low pass filtered map of DdmE with model fitted, showing the flexible density for the extended guide – target duplex. The position of the N-terminal domain is shown as a red box. c & d, comparison of TtAgo (PDB ID 4NCB) with DdmE.

Extended Data Fig 4. Binding of DNA by DdmD.

a, Cryo-EM 2D class averages of DdmD in the absence and presence of DNA substrates with different overhang lengths. White arrow denotes monomeric DdmD. Structural changes between DdmD apo, short fork DNA and long for DNA. b, Conformational changes in DdmD dimer upon short fork DNA loading. c, Interactions between DdmD and DNA. Residues highlighted are tested in panel c. d, Plasmid interference assay to analyse DdmD-DdmE interactions. Significance between DdmDE and other variants was determined by ordinary one-way ANOVA test. ****P < 0.0001 Data are mean ± s.d. of at three independent experiments started from separate colonies. DdmDE, DdmD and DdmE are the same data as shown in Fig 1h and ED Fig 1 c & d.

Extended Data Fig 5. DdmD is helicase-nuclease.

a & b, TBE-Urea PAGE gel analysis of DNA cleavage by DdmD in the presence and absence of ATP. A 3’-FAM-labeled DNA substrate was incubated with DdmD as ssDNA (left), annealed to a partially complementary strand (creating a forked duplex with a 30-nt overhang and a 25-bp duplex), and a 5’ overhang substrate (through annealing to a 25-nt complementary strand). Within the structures of DdmD bound to DNA, the 3’ end occupies the RecA helicase channel. ssDNA and forked DNA substrates are degraded, while a larger, incomplete degradation product for the 5’ overhang substrate is observed, since only the ssDNA overhang itself can be cleaved by DdmD. This indicates that unwinding and translocation is essential for full duplex degradation by DdmD. Representative of three independent experiments. b, Nuclease assay as in a, but the FAM is on the 5’ single-stranded end of the forked substrate. Since ssDNA is readily cleaved by DdmD, complete degradation is observed in the absence of ATP. Representative of three independent experiments. c, DdmD DNA unwinding assay, where a fluorophore-quencher pair (FAM and BlackHole Quencher) are on each strand of the forked substrate. DdmD unwinding is ATP-dependent. Data are mean ± s.d. of at three technical replicates d, Gel-based unwinding assay. ATP and DdmD are required for duplex unwinding, as monitored using native TBE 10% PAGE gel. e, Visualization of DNA-bound dimeric (left) and monomeric (right) complex by cryoEM. The monomeric DNA-bound DdmD suffers from severe preferred orientation. Representative of three independent experiments. f, Native PAGE gel analysis of DdmD oligomeric state in the absence and presence of DNA substrates. Apo DdmD runs as a dimeric species, which shifts to a mixture of monomer and dimer with ssDNA, and predominantly monomer with the 30-nt overhang forked DNA substrate used for structural analysis in Fig 2. Representative of three independent experiments. g, Negative stain EM 2D classes of DdmD in complex with DNA substates used for panel f. h, Visualization of DNA-bound dimeric (left) and monomeric (right) complex by cryoEM. The monomeric DNA-bound DdmD suffers from severe preferred orientation. i, SEC chromatogram of DdmDE handover complex (as shown in Fig 3A, green), and handover complex reconstituted with DdmD(R620A) point mutant (pink). A_280nm absorbance has been normalized to the size of the largest peak. The DdmD(R620A)E HC peak (peak i) is much smaller than wild-type, and unbound, dissociated DdmE is present (peak ii). Free DNA is in peak iii*, and has an A_260nm/A_280nm ratio of 1.8, while peaks i and ii had ratios of 1.2 and 1.1, respectively. SDS-PAGE analysis of DdmD(R620A)E HC SEC fractions is shown below. For gel source data, see Supplementary Information Figure 1.

Extended Data Fig 6. Structures of different DdmD – DdmE oligomers.

a, DdmD₂E₂ complex, colored by local resolution. b & c, FSC and Three-dimensional FSC curves for DdmD₂E₂ complex; FSC 3.0 Å. d & e, DdmD₆E₆ and DdmD₂E₂ complexes, with DdmD colored beige and DdmE colored blue. 3- and 2-fold symmetry axes are annotated.

Extended Data Fig 7. Conservation analysis.

Conservation analysis of DdmE (a) and DdmD (b). For DdmD, the RecA helicase channel, nuclease active site and dimer interface are highly conserved. c, Conservation of DdmD – DdmE handover complex interface. DdmD(R620) is conserved and is buried within a similarly conserved pocked of DdmE. d, Electrostatics of DdmDE handover complex.

Extended Data Table 1.

Cryo-EM data collection and model validation statistics.

	DdmE	DdmD apo dimer	DdmD dimer + short fork DNA	DdmD monomer + long fork DNA	DdmDE handover complex
Data collection and processing
Voltage (kV)	300	200	300	300	300
Electron exposure (e-/Å2)	80	40	80	80	80
Defocus range (μm)			−1.5 to −2.5
Pixel size (Å)	0.83	0.94	0.83	0.83	0.83
Symmetry imposed
Initial particle images (no.)	14,617,722	1,388,925	1,514,189	5,182,824	4,447,772
Final particle images (no.)	594,426	90,513	299,440	235,731	1,497,134
Map resolution (Å)	3.08	3.19	3.04	2.99	2.53
FSC threshold	0.143	0.143	0.143	0.143	0.143
	<2.5 - >4.5	<2.5 - >4.5	<2.5 - >4.5	<2.5 - >4.5	<2.5 - >4.5
Refinement
Initial model used (PDB code)	N/A	N/A	N/A	N/A	N/A
Model resolution (Å)	3.4 0.5	3.3 0.5	3.2 0.5	3.2 0.5	2.9 0.5
FSC threshold
Map sharpening B factor (Å2)	120.1	104.6	134.3	104.6	95.7
Model composition
Non-hydrogen atoms	5657	18728	17761	9091	24990
Protein residues	569	2321	2143	1099	2952
Nucleotides	48	0	22	12	55
Ligands	Mg: 1	0	0		Mg: 1
Water
Mean B factors (Å2)
Protein	97.01	75.2	112.05	59.21	90.37
Nucleotides	125.12	0	106.13	53.63	93.72
Ligand	76.03	0	0	0	93.40
R.m.s. deviations
Bond lengths (Å)	0.005	0.021	0.015	0.007	0.008
Bond angles (°)	0.65	1.345	1.102	0.749	0.821
Validation
MolProbity score	2.01	1.83	1.5	1.68	1.63
Clashscore	10.01	7.59	6.33	7.96	5.00
Poor rotamers (%)	0	0	0.25	0	0
Ramachandran plot
Favored (%)	92.04	93.84	97.16	96.31	96.75
Allowed (%)	7.96	5.99	2.77	3.50	3.11
Disallowed (%)	0	0.17	0.18	0.81	0.14
Map CC (mask)	0.72	0.81	0.87	0.83	0.86

Extended Data Table 2.

List of oligonucleotides used in this study.

DNA guide 5'OH	GTTAGACTTTAAGTCAAT
DNA guide 5'P	5'P-GTTAGACTTTAAGTCAAT
RNA guide 5'OH	GUUAGACUUUAAGUCAAU
RNA guide 5'P	5'P-GUUAGACUUUAAGUCAAU
DNA target	5'-FAM-TTTATCAAAAAGAGTATTGACTTAAAGTCTAACCTATAGGATACTTACAG
RNA target	5'-FAM-UUUAUCAAAAAGAGUAUUGACUUAAAGUCUAACCUAUAGGAUACUUACAG
Short fork strand 1	ATGAGTATTCAACATTTTTTTTTTTTTTTT
Short fork strand 2	TTTTTTTTTTTTTTTATGTTGAATACTCAT
Long fork strand 1	ATG AGT ATT CAA CAT TTT TTT T*T*T* T*T*T* T*T*T* T*T*T* T*TT TTT TTT TTT
Long fork strand 2	TTT TTT TTT TTT TT*T* T*T*T* T*T*T* T*T*T* T*T*T TTT ATG TTG AAT ACT CAT
pUC19 guide	GGAAATGTTGAATACTCA
pUC19 target for SEC	TTTTTTTTTTTTTTTTGAGTATTCAACATTTCCTTTTTTTTTTTTTTT
DNA clevage top strand	5'-FAM-AGC TGA CGT TTG TAC TCC AGC GTC TCA TCT TTA TGC GTC AGC AGA GAT TTC TGC T
DNA cleavage bottom strand	AGCAGAAATCTCTGCTGACGCATAAAGATGAGACGCTGGAGTACAAACGTCAGCT
5' blocking oligo	AGACGCTGGAGTACAAACGTCAGCT
3' blocking oligo	AGCAGAAATCTCTGCTGACGCATAA
Unwinding 3’ FAM	AGCTGACGTTTGTACTCCAGCGTCTCATCTTTATGCGTCAGCAGAGATTTCTGCT-FAM
Unlabelled unwinding complement	AGCAGAAATCTCTGCTGACGCATAAATTATTTTTTTATTTTATTTTTTTAATTTA
Unwinding 5’quencher strand	5’-lowa Black FQ - AGCAGAAATCTCTGCTGACGCATAAATTATTTTTTTATTTTATTTTTTTAATTTA

^*Denotes phosphorothioate-modified DNA bases.

Data Availability

Structures of the DdmE in complex with guide and target, DdmD monomer in complex with ssDNA DNA, DdmDE handover complex, DdmD dimer, and DdmD dimer in complex with ssDNA have been deposited in the EMDB with accession codes EMD-41781, EMD-41790, EMD-41865, EMD-41782, and EMD-41785, respectively. Associated atomic coordinates are deposited in the PDB with accession codes 8U0J, 8U0W, 8U3K, 8U0S, and 8U0U, respectively. Source data for Figs. 1–4 and Extended Data Figs. 1,4,5 are provided with this paper.