Bacterial strains and plasmids are listed in Supplementary Table S1. E. coli strains were grown in Luria-Bertani (LB) medium at 37°C, A. tumefaciens strains in 523 medium at 25°C, and D. dadantii 3937 in LB medium at 30°C. Antibiotics were added when necessary: for E. coli, 30 μg/mL of gentamicin and 250 μg/mL of streptomycin and for D. dadantii and A. tumefaciens, 50 μg/mL of gentamicin.

Plasmid pJN105 was used for the expression of toxin genes, and the pTrc200 plasmid was used for the expression of immunity genes. PCR was performed with KAPA HiFi Hot Start DNA Polymerase (Roche, Switzerland) using the genomic DNA of A. tumefaciens 1D1609. The primers used in the study are listed in Supplementary Table S2. Site-directed mutagenesis was performed using either DpnI or Gibson Assembly (New England BioLabs, USA), ligated using T4 ligase (Takara), and transformed into chemically competent E. coli DH10B. Plasmid DNA was isolated using the QIAprep Miniprep Kit (Qiagen, Germany). All constructs were confirmed by sequencing. The in-frame deletion mutants in A. tumefaciens were generated using a suicide plasmid via double crossing over by electroporation or by conjugation, as described previously (Santos et al., 2020).

The orthologs of V2c were analyzed using amino acid sequences after the DUF4150 domain via BLASTP search. Multiple sequence alignment was performed using MUltiple Sequence Comparison by Log-Expectation (MUSCLE) (Edgar, 2004). Full-length sequences of the V2c orthologs were analyzed to obtain domain architectures from the Pfam database (Mistry et al., 2021). The E-value threshold was set at 10⁻⁵. HHpred (Soding et al., 2005) and Swiss Model (Waterhouse et al., 2018) were used for homology-based protein structure prediction. Protein structure was predicted by AlphaFold (Jumper et al., 2021; Varadi et al., 2022) from the UniProt database. Secondary structure prediction was performed using default settings in PSIPRED 4.0 Protein Structure Prediction Server/DMPfold 1.0 Fast Mode (Buchan and Jones, 2019; Greener et al., 2019).

The growth inhibition assay was performed as described previously (Ma et al., 2014). In brief, overnight cultures of the E. coli DH10B strain with empty vectors or the derivatives were adjusted to an OD₆₀₀ of 0.1 in LB medium. A measure of 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG, Amresco) was used to induce the expression of the putative immunity protein. After 1 h of IPTG induction, L-arabinose (Ara, Sigma-Aldrich) was added to the final concentration of 0.2% to induce the expression of the toxin. Cell growth was recorded every hour at OD₆₀₀. Empty vectors were used as controls. After 8 h, cells were harvested to quantify the number of colony-forming units (log₁₀ CFU/mL) by automatic diluter and plater, easySpiral Dilute (Interscience, France).

In vivo plasmid DNA degradation analysis was performed as described previously (Ma et al., 2014). In brief, overnight cultures of the E. coli DH10B strain with empty vectors or the derivatives grown in LB broth were harvested and adjusted to OD₆₀₀ of approximately 0.3 in LB medium. E. coli cultures were induced with 1 mM IPTG at 0 h for v3c expressed from the pTrc200 plasmid, followed by L-arabinose (0.2%, final concentration) induction at 1 h to induce v2c from the pJN105 plasmid, and cultured for 2 h. Equal amounts of cells were used for plasmid DNA extraction, and an equal volume of extracted DNA was resolved in agarose gel, followed by ethidium bromide staining, and visualized using Gel Doc XR+ UV Gel Documentation Molecular Imager Universal Hood II (Bio-Rad, USA).

For the in vitro DNase activity assay, the effector gene v2c was cloned in the pET28a plasmid to generate N-terminal His-tagged V2c, and the immunity gene v3c was constructed in the pTrc200 plasmid. Site-directed mutagenesis was performed by using overlap extension PCR. The plasmid for expressing effector protein, v2c or its variants, and the plasmid for expressing immunity protein, v3c, were transformed into E. coli BL21 (DE3) for IPTG-induced co-expression. The bacterial culture was incubated at 37°C until OD₆₀₀ reached 0.5 and then 0.1 mM IPTG was added, followed by 5-h expression at 37°C. Pellets were collected, sonicated, and expressed proteins were purified using Ni Sepharose 6 Fast Flow histidine-tagged protein purification resin (Cytiva, Germany). Briefly, cells were lysed in extraction buffer (50 mM Tris–Cl, pH7.5, 0.1 M NaCl, 20 mM imidazole), and proteins were eluted by using an elution buffer with up to 250 mM imidazole in 50 mM Tris–Cl, pH7.5, 0.1 M NaCl.

A 10-μL reaction mixture containing 200 ng of pTrc200HA plasmid DNA and 1× CutSmart Buffer (New England BioLabs, USA) with partially purified V2c or its variants normalized by the major V2c protein band was incubated at 37°C for 1 h. The digestion of DNA was determined by agarose gel electrophoresis and visualized by SYBR Safe DNA gel stain (Invitrogen, USA). The reaction samples of V2c or its variants were scaled up for Western blotting analysis. An anti-V2c antibody was generated in rabbits against two V2c peptides, PFYWDFPNSQVGRD and KRIAMLRGDPTRYD (indicated in Supplementary Figure S2), by Yao-Hong Biotechnology Inc. (Taiwan).

A Western blotting analysis was performed as described previously (Lai and Kado, 1998). In general, proteins were resolved by SDS-PAGE and transferred onto a PVDF membrane by using a transfer apparatus (Bio-Rad, USA). The membrane was probed with primary antibodies against V2c (1:4000), Hcp (1:2500), TssB (1:4000), and GFP (1:1000), followed by incubation with a horseradish peroxidase-conjugated anti-rabbit secondary antibody (ChemiChem) (1:25000) and visualized with the ECL system (Perkin-Elmer Life Science, Boston, USA).

Escherichia coli DH10B carrying the empty vectors or the derivatives were grown overnight in LB broth at 37°C. For the expression of the effector gene alone, overnight culture was diluted to OD₆₀₀ reached 0.2 and sub-cultured for 1.5 h prior to induction by 0.2% L-arabinose for 1 h. For the co-expression of effector and putative immunity genes, overnight culture was diluted to OD₆₀₀ reached 0.1 and sub-cultured for 1 h prior to induction by 0.2% L-arabinose and 1 mM IPTG for 3–4 h. Cultures were harvested and resuspended with 30 μL of PBS with 0.5% Tween 20 (PBST). Three μL of cells were placed on a 2% agarose pad and imaged. To check the DNA degradation, the same protocol was followed, and live cells were stained with 2 μg/mL of Hoechst (Life Technologies, USA) and 1 μg/mL of FM4-64FX, a fixable analog of FM 4–64 (Invitrogen, USA) for 5 min and washed with PBS to remove the excess dye.

For the co-culture of D. dadantii ∆imp-GFP (S65T) with various 1D1609 EI pair mutants, strains were mixed at 1:10 ratio on Agrobacterium kill-triggering (AK) agar and the AK medium (3 g K₂HPO₄, 1 g NaH₂PO₄, 1 g NH₄Cl, 0.15 g KCl, 9.76 g MES in 900 mL of ddH₂0, pH 5.5) (Yu et al., 2020) was solidified by 2% (w/v) agar. Cells from 16-h co-culture at 25°C were concentrated to an OD₆₀₀ of 10 and spotted on a 2% agarose pad for imaging.

Imaging was performed using an inverted fluorescent microscope (IX81, Olympus Japan) with the objective lens UPLSAPO 100XO (Olympus, Tokyo, Japan), DAPI (49,000; Chroma), GFP (49,002; Chroma), and Cy3 (49,005; Chroma) filter sets and a CCD camera (ORCA-R2; Hamamatsu, Japan). The images were acquired using Olympus cellSens Dimension software.

All the image analysis was performed using the Fiji analyze particles tool (Schindelin et al., 2012). The micrograph of FM 4–64 fluorescence was used to segment surfaces (i.e., individual bacterial cells) with background subtraction and thresholding. Feret diameter was used. The cell length (μm) was determined from the cumulative counts of three random frames per sample. For the quantification of cell degradation, the Hoechst intensity was normalized using the formula intensity/unit area of cells, where intensity is major × minor × 3.14 (π).

Bacterial competitions were carried out as described previously (Wu et al., 2019) with modifications. Briefly, an overnight culture of A. tumefaciens and D. dadantii 3937 ∆imp was mixed at a 10:1 attacker-to-prey ratio and spotted on AK agar plates (Yu et al., 2020). The prey, A. tumefaciens C58 or D. dadantii 3937 ∆imp, was transformed with pRL662-GFP (S65T) for gentamicin selection or microscopy imaging. After 18 h of incubation at 25°C, the co-cultured cells were collected, serially diluted, and plated on LB agar containing gentamicin to quantify surviving prey cells by counting CFU.

The statistical analyses were performed using GraphPad Prism version 6.01 for Windows, GraphPad Software, La Jolla, California, USA.¹ The number of technical replicates and independent biological replicates, p-values, and statistical tests performed are indicated in the figure legends. The mean of independent replicates was compared using the one-way ANOVA followed by Tukey’s multiple comparisons tests ² for statistical analysis. The error bars indicate the standard error of the mean (SEM).

In 1D1609, vgrGa-associated effector V2a appears to be the major antibacterial toxin against Escherichia coli prey (Santos et al., 2020). V2c is a putative PAAR (DUF4150)-linked specialized effector (Figure 1A), but the presence or absence of v2c did not affect the antibacterial activity of 1D1609 against E. coli prey (Santos et al., 2020). Aside from the N-terminal PAAR domain, V2c does not show any sequence similarity to V2a (Supplementary Figure S2). BLAST and CDD searches did not identify a known or conserved domain at its C-terminal region. However, multiple sequence alignments of V2c and V2c orthologs revealed conserved SHH and HNH motifs at the C-terminal region (Figure 1B). These V2c orthologs all harbor the C-terminal Tox-SHH domain, but their N-terminal region can be fused to other known T6SS domains, such as Found in the type sIX effector (FIX), the DUF4150/PAAR-like domain, and the Rhs core domain (Jana et al., 2019; Figure 1C). Using the protein homology detection program HHpred (Soding et al., 2005), we found that the C-terminal region of V2c encodes two parallel β-strands connected by an α-helix, which is a signature structure in His-Me finger nuclease (Wu et al., 2020). Consistent with the result of secondary structure prediction, 3D structure prediction using AlphaFold (Jumper et al., 2021; Varadi et al., 2022; Figure 1D) shows that this domain consists of two antiparallel β-strands connected with an Ω loop with a His (H384) at the C-terminus of the first β-strand and followed by an α-helix, featuring a HNH motif of His-Me finger endonuclease (Jablonska et al., 2017; Wu et al., 2020). These suggest that V2c belongs to the Tox-SHH (Pfam PH15652) clade of the His-Me finger superfamily.

To determine the antibacterial and DNase activities of V2c, we expressed it using an arabinose-inducible promoter to determine if V2c is a bacterial toxin functioning as a nuclease. Growth inhibition measured by optical density was observed when v2c is expressed by arabinose induction in E. coli, and this can be neutralized by the co-expression of its putative cognate immunity, v3c, as compared to the vector control (Figure 2A). Furthermore, the colony-forming unit (CFU) count shows the reduced CFU of v2c-expressed cells, and higher CFU could be recovered with the co-expression of v3c (Figure 2B). In addition, the plasmid degradation assay showed that the expression of v2c upon arabinose induction resulted in the degradation of plasmid DNA in E. coli, and this is abolished when v3c is co-expressed (Figure 2C). The data indicated that V2c exhibits DNase activity in vivo.

We further investigated whether the SHH and HNH motifs are involved in DNase activity. The His-Me finger superfamily has a strictly conserved His residue and catalytic metal ion, which are both essential for nucleic acid hydrolysis. Based on this premise, we define a region spanning amino acid residues 380–409 as a His-Me finger domain core region (CR). Based on the alignment, H384 is the catalytic His and H408 is for metal ion binding, while H383 may function as a second residue for metal ion coordination. To verify their roles in antibacterial and DNase activity, a series of deletions as well as single and double amino acid substitution variants were generated. Strikingly, no growth inhibition or plasmid degradation were observed in any of the mutants expressing the CR deletion (V2c^CR), single (V2c^H383A, V2c^H384A, and V2c^H408A), or double (V2c^HAHA substitution in both H383 and H384) variants (Figures 2A,C,D). Higher CFU was also recovered from E. coli culture expressing these mutants alone or with v3c (Figure 2B). To further confirm the DNase activity of V2c, each V2c and its variants were fused with 6x-His and purified for in vitro DNase activity assay. The results showed that the purified V2c fraction is able to degrade DNA but not the eluate from E. coli expressing the vector (pET28a) (Figure 2D). Consistent with the in vivo plasmid degradation assay, no activity could be detected from V2c^H384A, V2c^H408A, or V2c^HAHA variants. However, the purified V2c^H383A fraction exhibited DNase activity by degrading the plasmid DNA substrate. The Western blotting analysis of the reaction samples shows whether V2c and its variants used for the assay were at similar protein levels. Taken together, the results of in vivo and in vitro DNA degradation assays show that the conserved H384 and H408 in the HNH motif are crucial for V2c DNase activity, whereas H383 in the SHH motif appears to be less critical for in vitro but may be required for in vivo nuclease activity.

We next examined the cell morphology and DNA integrity of E. coli cells expressing v2c and the two variants (v2c^CR and v2c^HAHA) that are defective in DNase activity and toxicity. Cell morphology was observed by phase contrast, while the bacterial cell membrane and DNA were stained by FM 4–64 and Hoechst, respectively, under the fluorescence microscope. Hoechst signals are significantly reduced when v2c is expressed by arabinose induction in E. coli, and this can be rescued by the co-expression of its cognate immunity v3c shown in the images and quantification (Figures 3A,B). Hoechst signals are diminished only in E. coli cells producing v2c but not v2c^HAHA, v2c^CR, and vector control, suggesting that cellular DNA degradation is caused by V2c DNase activity, which can be neutralized by its immunity protein V3c. In addition, we also observed cell elongation in E. coli expressing v2c, and co-expression of v3c rescues the cell elongation phenotype (Figures 3A,C). However, E. coli cells expressing v2c^CR and v2c^HAHA, as well as other single amino acid substitution variants (v2c^H383A, v2c^H384A, and v2c^H408A), do not abolish the elongated cell phenotype (Figures 3A,C; Supplementary Figures S3A,B).

A previous study showed that V2a is a Tox-AHH nuclease that serves as the major antibacterial weapon in 1D1609 (Santos et al., 2020). H385 and H386 are two of the predicted His residues in the AHH motif of V2a, corresponding to H383 and H384 in the SHH motif of V2c. The structure predicted by AlphaFold indicates that V2a also features a His-Me finger, having two α-helices with histidine residues (H430 and H456) followed by the β-strands (Figure 4A). Previous findings demonstrated that the expression of v2a but not v2a^H385A caused growth inhibition and plasmid degradation in E. coli (Santos et al., 2020). To gather further insight into the V2a AHH motif in the His-Me finger domain, we evaluated the growth and morphology of arabinose-induced v2a mutants. The results showed that the expression of v2a caused a reduction in Hoechst-stained DNA signals and resulted in cell elongation (Figures 4B–D). The expression of v2a^H385A has similar levels of Hoechst-stained DNA signals and cell length as compared to the vector control. These results suggest that the cell elongation phenotype observed was caused by V2a DNase activity, whereas V2c shows DNase activity-independent cell elongation. We also determined the toxicity of another putative effector V4b of 1D1609, a PAAR (DUF4150)-linked specialized effector but harboring an Rhs domain with structural similarity to the insecticidal toxin (Santos et al., 2020; Supplementary Figure S4A). Expression of v4b caused growth inhibition in E. coli, in which the growth inhibition can be rescued by the co-expression of its putative cognate immunity v5b to a level similar to vector control (Supplementary Figure S4B). However, no DNA degradation or cell elongation could be observed (Supplementary Figures S4C–F). These results show that V4b is a bacterial toxin and V5b is the cognate immunity protein, suggesting that the cell elongation phenotype may be specific to the toxicity of nuclease effectors.

To correlate the DNA degradation and cell elongation phenotypes observed by the ectopic expression of v2c in E. coli in a more biologically relevant context, we further performed the interbacterial competition assay. We first selected A. tumefaciens strain C58, which possesses incompatible EI pairs with 1D1609 and has previously been demonstrated to be susceptible to 1D1609 T6SS killing (Wu et al., 2019), for intra-species competition. Each of the 1D1609 mutants lacking multiple EI pairs was co-cultured with a C58 strain harboring pRL-GFP (S65T) prey in vitro on AK agar (Yu et al., 2020) and the number of viable C58 prey cells (log₁₀ CFU/mL) was quantified. Similar to the previous killing assay using E. coli prey (Santos et al., 2020), we detected approximately 2 log of reduced prey cell survival by the 1D1609 WT attacker as compared to the ∆4EI mutant with the deletion of all four EI pairs (Figure 5A). Other double (∆EIbd) or triple deletion mutants (∆EIabd, ∆EIbcd) exhibited compromised but detectable antibacterial activity against C58, indicating that v2a or v2c alone is sufficient to exhibit antibacterial activity against C58.

Intrigued by the observation that v2c could exhibit antibacterial activity to the A. tumefaciens C58 prey but no significant effect against E. coli prey (Santos et al., 2020), we reasoned that E. coli may not be a physiologically relevant prey for A. tumefaciens although v2c is able to cause toxicity when directly expressed in E. coli (Figures 2A,B). We then selected D. dadantii as the prey cells for inter-species competition because it is a phytopathogen but also a close relative of E. coli, both belonging to Enterobacteriaceae. To avoid potential counterattack by D. dadantii T6SS (Koskiniemi et al., 2013), a T6SS-deficient mutant, ∆imp with deletion of imp operon, harboring pRL-GFP (S65T) (D. dadantii ∆imp-GFP (S65T)) was used as a prey. CFU counting of D. dadantii ∆imp-GFP (S65T) prey cells after co-culture show approximately 1.5 log of reduced prey cell survival by 1D1609 WT attacker, as compared to the ∆4EI mutant (Figure 5B). The presence of v2a only (∆3EIbcd) exhibited comparable antibacterial activity as the WT, while v2c only (∆3EIabd) showed weak but detectable antibacterial activity as compared to ∆4EI, indicating that v2c itself is able to kill D. dadantii but with much weaker activity than v2a. Strikingly, 1D1609 expressing both v2a and v2c (∆2EIbd) could result in stronger killing effects than WT. The use of pRL-GFP (S65T) with constitutive expression of GFP also allowed us to observe the D. dadantii ∆imp-GFP (S65T) prey cells under the fluorescence microscope after co-culture. We observed very few cells expressing GFP when competing with WT or ∆3EIbcd expressing v2a only. The lysed and elongated D. dadantii cells were also detected when co-cultured with 1D1609 with either strains expressing v2a only (∆3EIbcd), v2c only (∆3EIabd), or both ∆3EIbd (Figure 5C). We further quantified the cell length of GFP-expressing D. dadantii cells. The results show that the majority of cells were elongated after 18 h of co-culture with 1D1609 expressing either v2a (∆3EIbcd) or v2c (∆3EIabd), or a combination of both (∆3EIbd), as compared to those D. dadantii cells alone or co-incubated with ∆4EI (Figure 5D).

Taken together, these results suggest that V2c nuclease exhibits antibacterial activity at both intra-species and inter-species competitions but at different capacities. While V2a and V2c exert similar antibacterial activity against A. tumefaciens C58, V2c only exhibits weak antibacterial activity but functions synergistically with V2a when competing with D. dadantii prey, resulting in elongated and lysed cells.