HORMA domain proteins and a Trip13-like ATPase regulate bacterial cGAS-like enzymes to mediate bacteriophage immunity

Bacterial HORMA domain proteins bind and activate second messenger synthesis by their associated CD-NTases

CBASS systems encoding putative HORMA domain proteins and an ortholog of the AAA+ ATPase Pch2/TRIP13 (hereon termed Trip13) fall into two major classes, one of which encodes a single HORMA domain protein (here termed HORMA1) and the second of which encodes two such proteins (HORMA2 and HORMA3) (Figure 1A and Figure S1A) (Burroughs et al., 2015). The HORMA1 and HORMA2/3 operons also encode diverged Trip13 ATPases, suggesting that the HORMA and Trip13 genes have co-evolved within each family of operons (Figure S1A–C) (Tromer et al., 2019). Bacterial CD-NTases have been grouped into different clades based on their sequence similarity (Whiteley et al., 2019), and we find that HORMA1 operons mainly encode clade C01 CD-NTases (hereon termed CdnC), while HORMA2/3 operons encode clade D05 CD-NTases (termed CdnD). To address the function of these operons, we chose a CdnC+HORMA1+Trip13 operon from E. coli strain MS115-1, and a CdnD+HORMA2/3+Trip13 operon from Pseudomonas aeruginosa strain ATCC27853, for further study (Figure 1A). The E. coli operon is found in several patient-derived E. coli strains, and the P. aeruginosa operon is found in over a quarter of patient-derived P. aeruginosa strains. Both operons also encode related putative endonuclease effectors, termed REase by Burroughs et al. (Burroughs et al., 2015) and here termed NucC (Nuclease, CD-NTase associated).

As a first step, we performed yeast two-hybrid analysis with the E. coli MS115-1 (Ec) proteins, and found that Ec CdnC and Ec HORMA bind one another (Figure 1B and Figure S2A). When co-expressed in E. coli, Ec CdnC and Ec HORMA form two complexes with both 1:1 and 2:2 stoichiometry (Figure S2B). In a yeast three-hybrid experiment, a stabilized hexamer mutant of Ec Trip13 (Walker B motif residue Glu159 to Gln; Trip13^EQ) binds to an Ec CdnC:HORMA complex (Figure 1B and Figure S2A), and in vitro reconstitution reveals that this ternary complex has a stoichiometry of 6 TRIP13^EQ:1 CdnC:1 HORMA (Figure S2C). Similar yeast two-hybrid analysis with the P. aeruginosa ATCC27853 (Pa) proteins showed that Pa HORMA2 and HORMA3 interact with one another, and that both HORMA proteins interact with Pa CdnD (Figure S2E). In vitro reconstitutions reveal that these proteins form a CdnD:HORMA2:HORMA3 complex with 1:1:1 stoichiometry (Figure S2F–H). Finally, both Ec NucC and Pa NucC self-interact in yeast two-hybrid assays, suggesting that NucC forms homodimers or larger oligomers (Figure S2A and Figure S2E).

Both Ec CdnC and Pa CdnD were previously found to be inactive in vitro (Whiteley et al., 2019), leading us to test if their activity depends on their cognate HORMA and Trip13 proteins. For this purpose, we tested purified proteins in an in vitro second messenger synthesis assay based on prior work with a related CD-NTase, E. cloacae CdnD02 (Figure 1C) (Whiteley et al., 2019). We first tested for activity of Ec CdnC and the Ec CdnC:HORMA complex (both 1:1 and 2:2 complexes). As the related cGAS and OAS enzymes both require nucleic acid binding for activation (Hornung et al., 2014), we also included double-stranded DNA as a potential activator. We find that while CdnC is inactive on its own (Figure 1C, sample i), the 2:2 complex of Ec CdnC and HORMA synthesizes two phosphatase-resistant products in the presence of DNA (Figure 1C, sample v). Curiously, the 1:1 CdnC:HORMA complex is inactive (Figure 1C, sample iv). CdnC+HORMA is activated by long double-stranded DNAs, but not by single-stranded DNA or short (40 bp) double-stranded DNA (Figure S3A). CdnC activity is disrupted by mutation of the putative active site residues Asp72 and Asp74, demonstrating that CdnC uses a similar biosynthetic mechanism as related CD-NTases (Figure 1C, sample vi). Finally, we found that addition of sub-stoichiometric amounts of Ec Trip13 (1:60 molar ratio of active Trip13 hexamers to CdnC+HORMA) strongly reduced second messenger synthesis (Figure 1C, sample vii). Together, these data support a model in which Ec CdnC is activated by binding its cognate HORMA protein and DNA, and inactivated by Trip13, potentially through direct disassembly of the CdnC:HORMA complex (Figure 1F).

We next used liquid chromatography coupled to tandem mass spectrometry (LC MS/MS) to analyze the major and minor products of Ec CdnC. We identified the major product as cyclic tri-AMP (cAAA; Figure 1D–E), and the minor product as cyclic di-AMP (cAA; not shown). Ec CdnC synthesizes only cAAA/cAA even in the presence of both GTP and ATP, indicating that the enzyme uses only ATP as a substrate (not shown). Thus, Ec CdnC synthesizes a cyclic trinucleotide second messenger upon activation by its cognate HORMA domain protein and double-stranded DNA.

We next tested the activity of Pa CdnD in the presence of Pa HORMA2, Pa HORMA3, and Pa Trip13. While Pa CdnD shows lower activity than Ec CdnC in vitro and requires Mn²⁺ instead of Mg²⁺ as a cofactor, we find that CdnD also primarily synthesizes cAAA (Figure S3B). Pa CdnD requires HORMA2 for activity, but is not activated by HORMA3 (Figure S3B), perhaps reflecting functional specialization of the two HORMA domain proteins in this operon. As with the E. coli proteins, addition of Pa Trip13 to a reaction with Pa CdnD and HORMA2 strongly inhibits second messenger synthesis (Figure S3B). Finally, we find that in contrast to Ec CdnC, DNA is not required for Pa CdnD activity, nor does addition of DNA further stimulate cAAA production (Figure S3B). Thus, while Ec CdnC and Pa CdnD are regulated differently, the two enzymes share a common dependence on HORMA domain protein binding for activation, and generate the same cyclic trinucleotide second messenger.

Structures of HORMA-associated CD-NTases

To address the structural and biochemical mechanisms by which Ec CdnC and Pa CdnD synthesize cAAA in a regulated manner, we first purified and determined high-resolution crystal structures of both enzymes (Figure 2A–B). Both Ec CdnC and Pa CdnD adopt a polymerase β-like nucleotidyltransferase fold, with two lobes sandwiching a large active site. Ec CdnC and Pa CdnD adopt similar overall structures, with a 3.95 Å Cα r.m.s.d. (over 250 residues), and are more structurally related to one another than to either DncV (CD-NTase clade A) or CdnE, both of which synthesize cyclic dinucleotide second messengers (Figure 2C).

In Ec CdnC, the active site is occupied by one molecule of ATP that was co-purified with the protein (Figure 2A); we were unable to remove this tightly-bound nucleotide. We could purify Pa CdnD without a bound nucleotide, and found by isothermal titration calorimetry that the enzyme binds strongly to ATP, but not to GTP (Figure S4A). These data agree with our findings that both Ec CdnC and Pa CdnD primarily synthesize cAAA. We determined structures of Pa CdnD in both Apo and ATP-bound states, and found that the two states differ by a 3.9° rotation of the enzyme’s N- and C-lobes with respect one another, with the ATP-bound state more “closed” than the Apo state (Figure 2B).

Our biochemical data show that Ec CdnC requires dsDNA for activation, similar to mammalian cGAS (Figure 1C). We modelled DNA binding to CdnC by overlaying our structure with a cGAS:DNA complex (Gao et al., 2013). In the resulting model, DNA contacts a surface of Ec CdnC that is strongly positively charged and highly conserved (Figure 2D and Figure S4B). To test the idea that this surface is responsible for DNA binding, we mutated several positively-charged residues on this surface of CdnC and measured both DNA binding and DNA-stimulated cAAA synthesis. The four mutants we tested had variable effects on DNA binding (Figure S4C), but all four eliminated cAAA synthesis by CdnC in the presence of HORMA+DNA (Figure 2E). These data support a role for this surface in DNA binding and activation of CdnC.

In mammalian cGAS, DNA binding to α-helices α7 and α8 restructures an “activation loop” in the enzyme’s active site to promote substrate binding and second messenger synthesis (Figure 2F) (Civril et al., 2013; Kato et al., 2013; Li et al., 2013; Zhang et al., 2014). In our structures of Ec CdnC and Pa CdnD, the conformation of the activation loop is equivalent to the activated state of cGAS (Figure 2G–I). Thus, while DNA binding is required for activation of Ec CdnC, the structural mechanism of activation is likely distinct from that of mammalian cGAS.

Pa CdnD does not require DNA for cAAA synthesis in vitro. Nonetheless, Pa CdnD shows a similar, albeit less pronounced, pattern of high conservation and positive charge on the surface equivalent to the cGAS DNA-binding surface (Figure S4D–E). We therefore tested DNA binding directly, but did not detect binding (Figure S4C). These data recall earlier findings with V. cholerae DncV, which possesses an equivalent conserved, positively charged surface, but whose synthesis of cGAMP is not stimulated by DNA (Kranzusch et al., 2014).

Bacterial HORMA domain proteins adopt canonical open and closed states

We next sought to determine whether the putative bacterial HORMA domain proteins share the canonical HORMA domain fold and mechanism of closure motif binding (Figure 3A). We first determined a structure of Pa HORMA2, and found that the protein adopts a canonical “closed” HORMA domain structure, with a central three-stranded β-sheet (β-strands 4, 5, and 6) backed by three α-helices (αA, B, and C) (Figure 3B). Characteristic of closed HORMA domains, the N-terminus of Pa HORMA2 is disordered, and the C-terminus wraps around the protein and forms a β-hairpin (β-strands 8′ and 8′′) against strand β5. Strikingly, the C-terminal safety belt of Pa HORMA2 embraces a peptide from the disordered N-terminus of a crystallographic symmetry-related molecule in the canonical closure motif binding site (Figure 3B). Thus, Pa HORMA2 represents a bona fide HORMA domain with a closure-motif binding C-terminal safety belt.

To identify preferred closure motif sequences for bacterial HORMA proteins, we performed phage display with purified Ec HORMA, Pa HORMA2, and Pa HORMA3, and a library of 7-mer peptides. We identified consensus binding motifs enriched in hydrophobic amino acids for both Pa HORMA2 and Ec HORMA (Figure S5A–B), and confirmed binding by coexpression and purification of peptide-HORMA complexes (Figure S5C–D). Thus, both Ec HORMA and Pa HORMA2 can likely adopt the closed conformation and bind specific closure motif peptides.

To determine whether Pa HORMA2 can also adopt an “open” HORMA domain state, we generated a truncated construct lacking the C-terminal safety belt (Pa HORMA2-ΔC). The crystal structure of Pa HORMA2-ΔC shows that in contrast to full-length Pa HORMA2, the N-terminus of Pa HORMA2-ΔC is well-ordered and packs as a β-strand (β1) against strand β5 (Figure 3C). This transition of the N-terminus from disordered to an ordered β1 strand is characteristic of open HORMA domains (Rosenberg and Corbett, 2015). Thus, bacterial HORMA domain proteins can adopt both the open and closed states, and bind closure motif peptides equivalently to their eukaryotic relatives.

We next determined a structure of Pa HORMA3, and found that this protein also adopts an open state with an ordered N-terminal β1 strand (Figure 3D). Based on the sequence of the Pa HORMA3 N-terminus and the predicted disorder of its poorly-conserved C-terminus (Figure 3D), we propose that this protein may be “locked” in the open state similarly to the eukaryotic HORMA domain protein Atg101 (Qi et al., 2015; Suzuki et al., 2015). Consistent with this idea, we were unable to identify Pa HORMA3-binding peptide sequences by phage display (not shown). Our data therefore suggests that in HORMA2/3 containing CBASS systems, HORMA2 can adopt the open and closed states and bind closure motif peptides, while HORMA3 adopts an open state and does not bind peptides.

We next determined a structure of a Pa HORMA3:HORMA2:peptide complex (Peptide 1: EVMEFNP). Strikingly, the Pa HORMA3:HORMA2 dimer bears no resemblance to dimers of eukaryotic HORMA domains, which assemble through their αC helices (Mapelli et al., 2007; Rosenberg and Corbett, 2015). Rather, Pa HORMA3 and Pa HORMA2 interact through both proteins’ αA and αB helices, plus the short α-helix in the C-terminal region of Pa HORMA3 (Figure 3E). The interface does not involve any elements of Pa HORMA2 that change conformation significantly between the open and closed states, and yeast two-hybrid analysis shows that the open Pa HORMA2-ΔC construct can also bind Pa HORMA3 (Figure S5E–G). Thus, while Pa HORMA3 apparently does not bind closure motif peptides or undergo conformational conversion, it can likely scaffold larger complexes by binding Pa HORMA2 and CdnD.

HORMA proteins bind their cognate CD-NTases specifically in the closed state

We next sought to understand how bacterial HORMA domain proteins regulate their cognate CD-NTases. We first purified and determined the structure of a complex containing Pa CdnD, HORMA2, and Peptide 1, and found that Pa HORMA2 binds the C-lobe of CdnD (Figure 4A). Compared to its Apo state, HORMA2-bound Pa CdnD shows a 6.6° rotation of the N-lobe toward the C-lobe, consistent with the idea that binding of HORMA2 may aid activation of CdnD. HORMA2 interacts with CdnD through its β8′ strand and αC helix (Figure 4B); as the β8′ strand does not form in the open state, this binding mode is likely dependent on HORMA2 adopting a closed conformation. Indeed, we found by yeast two-hybrid analysis that Pa HORMA2-ΔC, which adopts an open conformation, does not interact with Pa CdnD (Figure 4C). Consistent with our ability to reconstitute a CdnD:HORMA2:HORMA3 complex, we could model a structure of Pa HORMA2 bound to both CdnD and HORMA3 with no apparent clashes (Figure S6A).

We next purified and determined the structure of an Ec CdnC:HORMA-ΔN complex, which lacks the disordered N-terminal 11 residues of Ec HORMA. In this complex, the structure of Ec HORMA closely resembles Pa HORMA2 in the closed state, with a well-ordered C-terminal safety belt forming a β-hairpin (β8′-8″) against strand β5 (Figure 4D). Notably, in this structure we find that Ec HORMA-ΔN is bound to a short peptide from the N-terminal His₆-tag of a symmetry-related HORMA-ΔN molecule, resulting in a CdnC:HORMA-ΔN complex with 2:2 stoichiometry (Figure S6B). Unlike our structure of Ec CdnC alone, in which the protein is bound to a single ATP molecule co-purified from E. coli, the two copies of Ec CdnC in the CdnC:HORMA-ΔN structure contain either no ligand or AMP in the active site (Figure 4D). This suggests that in comparison to isolated Ec CdnC, the CdnC:HORMA complex may be more conformationally flexible and enzymatically active.

The interaction of Ec HORMA with Ec CdnC closely resembles that of Pa HORMA2 with Pa CdnD (Figure 4D). Ec HORMA binds the CdnC C-lobe, adjacent to CdnC’s putative DNA-binding surface. Ec HORMA binds CdnC through a similar interface as the one used by Pa HORMA2, binding CdnC largely through its αA helix (Figure 4D–E). Ec HORMA’s β8′ strand does not contact CdnC, yet truncation of the protein’s C-terminus (Ec HORMA-ΔC, lacking residues 141-172) disrupts CdnC binding, suggesting that the Ec HORMA-CdnC interaction likely also depends on the closed HORMA domain state (Figure 4F).

In our structure of CdnC:HORMA-ΔN, Ec HORMA binds a symmetry-related HORMA N-terminus as a short β-strand, but with an N-to-C polarity opposite that observed in our structure of Pa HORMA2 bound to Peptide 1 (and indeed opposite that of all known eukaryotic HORMA:closure motif complexes). We interpret this switched polarity of closure motif binding as an artifact of a non-biological complex assembly, rather than an indication that this HORMA domain binds closure motifs in an unique manner. We modelled binding of a consensus closure motif sequence identified by phage display (HGKILLT) with the canonical N-to-C polarity, revealing that the hydrophobic residues in the Ec HORMA-binding consensus would all fit into hydrophobic pockets on Ec HORMA (Figure S6C). This supports the idea that our phage display experiment identified sequences that bind Ec HORMA as a closure motif.

The observation of a 2:2 complex in the structure of Ec CdnC:HORMA-ΔN suggests that the 2:2 complex we observe when purifying Ec CdnC:HORMA may be assembled through reciprocal binding of HORMA N-termini to opposite HORMA protomers. To test this idea, we purified an Ec CdnC:HORMA-ΔN complex in which the N-terminal His₆-tag was removed, and found that this complex forms only 1:1 complexes and is only minimally active (Figure 4G and Figure S6D). When we co-purified the same complex with a consensus closure motif peptide (HGKILLT), we found that it too forms a 1:1 CdnC:HORMA complex, but is now active in vitro (Figure 4G and Figure S6D). These data explain why in our earlier assays, only 2:2 CdnC:HORMA complexes were active for cAAA synthesis: these complexes’ HORMA proteins are bound to closure motif-mimicking peptides, while the HORMA proteins in 1:1 complexes are unbound. Thus, these data support the idea that activation of Ec CdnC requires a closure motif-bound HORMA protein.

Structure of a bacterial Trip13 engaged with its CD-NTase:HORMA substrate

The Trip13-like ATPase encoded by bacterial HORMA-containing operons negatively regulates the production of cAAA by the HORMA-CD-NTase complex (Figure 1C). To address the structural mechanism of this activity, and to compare this mechanism to that of eukaryotic Pch2/TRIP3 in control of the eukaryotic HORMA domain protein Mad2, we determined the structures of both an isolated bacterial Trip13 hexamer (Figure S7A–D) and of an Ec CdnC:HORMA:Trip13^EQ complex (Figure 5A; the Trip13 E159Q mutation stabilizes hexamer assembly and substrate interactions by disallowing ATP hydrolysis (Ye et al., 2015, 2017)). The crystal structure of Ec CdnC:HORMA:Trip13^EQ reveals a mode of substrate binding and engagement for bacterial Trip13 that is similar to eukaryotic Pch2/TRIP13, with the high resolution of 2.6 Å providing unprecedented insight into substrate engagement and unfolding by AAA+ ATPases.

A number of recent studies have reported moderate-resolution (3.2-5.0 Å) cryo-electron microscopy structures of AAA+ ATPase remodelers engaged with their substrates, including mammalian TRIP13 bound to a p31^comet:Mad2 complex (Alfieri et al., 2018; Deville et al., 2017; Gates et al., 2017; Han et al., 2017; Puchades et al., 2017; Ripstein et al., 2017; White et al., 2018; Yu et al., 2018). These structures have revealed that active AAA+ remodelers adopt a right-handed spiral or “lock-washer” conformation, with 4-5 nucleotide-bound subunits forming a tight helical spiral and engaging an extended substrate peptide with their “pore loops”. In the case of TRIP13, the structures revealed that the “adapter” protein p31^comet binds the top surface of the TRIP13 hexamer, thereby positioning the substrate Mad2 such that its extended N-terminus drapes into the hexamer pore for engagement and remodeling (Alfieri et al., 2018; Ye et al., 2017). In our structure of Ec CdnC:HORMA:Trip13^EQ, the Ec Trip13^EQ hexamer adopts a right-handed spiral conformation, with four subunits (monomers A-D) bound to ATP, and two subunit (monomers E and F) unbound (Figure S8A–B). The four ATP-bound subunits overlay closely, while monomers E and F show rotation of the large and small AAA subdomains with respect to one another (Figure S8C). Monomer F is positioned at the “bottom” of the spiral, and as such shows the most significant conformational differences compared to the other five subunits. Overall, the structure of the Ec Trip13^EQ hexamer closely resembles that of other substrate-engaged AAA+ ATPases, including mammalian TRIP13 (Alfieri et al., 2018).

As noted above, eukaryotic TRIP13 recognizes its HORMA domain substrate Mad2 through the adapter protein p31^comet (Alfieri et al., 2018; Ye et al., 2017). In the Ec CdnC:HORMA:Trip13^EQ complex, CdnC serves as the adapter, binding the top face of the Trip13 hexamer at the interface of monomers E and F (Figure 5A). CdnC-Trip13 binding positions Ec HORMA over the Trip13 hexamer pore, with its extended N-terminus draping into the pore and interacting with the pore loops from Trip13 monomers A-E (Figure 5B). These pore loops form a tight spiral and engage the extended HORMA N-terminus with two-residue periodicity: Trip13 pore loop 1 residues Gly119 and Val121 in monomers A-E engage even-numbered residues from HORMA (Ser2, Tyr4, Tyr6, Val8, and Glu10) through main-chain hydrogen bonding (Figure 5C–D). At the same time, conserved histidine residues (His173) in pore loop 2 of Trip13 monomers A-D form a “ladder” with each histidine side-chain forming two hydrogen bonds with the side-chains of serine/threonine residues at odd-numbered positions in the HORMA N-terminus (Ser3, Ser5, and Thr7) (Figure 5C–D). In keeping with these highly specific interactions, we observe a striking pattern of conservation in the N-termini of HORMA1 proteins (which includes Ec HORMA), with conserved serine/threonine residues in every second position (Figure 5E). This conservation pattern extends to Ec HORMA residue 17, strongly suggesting that Ec Trip13 can undergo at least 3-4 cycles of ATP hydrolysis and translocation before disengaging, sufficient to partially unfold helix αA and disrupt both HORMA-CdnC binding (which primarily involves helix αA) and HORMA-closure motif binding (by destabilizing the C-terminal safety belt). HORMA2 proteins do not possess this pattern of conserved serine/threonine residues in their N-termini, and Trip13 proteins in HORMA2/3 operons lack a conserved pore loop 2 histidine residue (Figure S7E–F), suggesting a different mode of HORMA-Trip13 recognition in these operons. Thus, our structure of the Ec CdnC:HORMA:Trip13^EQ complex supports a model in which Trip13 disassembles a CD-NTase:HORMA complex by remodeling the HORMA domain protein, and provides the highest-resolution view to date of a AAA+ ATPase hexamer engaged with its substrate.

Our biochemical data shows that addition of Trip13 inactivates second messenger synthesis by Ec CdnC+HORMA+DNA, with Trip13 effectively inhibiting second messenger synthesis at a 1:60 ratio of Trip13 hexamer to CdnC+HORMA (Figure 1C, sample vi). This data is consistent with the idea that Trip13 catalytically disassembles and inactivates the CdnC:HORMA complex, likely through engagement and unfolding of the HORMA N-terminus. In agreement with this idea, we found that Trip13 is unable to effectively inhibit second messenger synthesis when ATP binding is disrupted by a mutation to the Walker A motif (K87A; Figure 5F, sample iii). Further, Trip13 is unable to inhibit second messenger synthesis by an Ec CdnC:HORMA-ΔN complex, which lacks the extended N-terminus of Ec HORMA and thus cannot be engaged by Trip13 (Figure 5F, samples iv-v). These data support a model in which bacterial Trip13 enzymes attenuate signaling by their cognate CD-NTases by disassembling an active CD-NTase:HORMA complex.

NucC is a cAAA-activated DNA endonuclease

Bacterial CBASS systems encode one of several putative effector proteins, which include proteases, phospholipases, and at least three families of putative endo- and exonucleases (Burroughs et al., 2015). Both the E. coli MS115-1 and P. aeruginosa ATCC27853 systems encode a restriction endonuclease-related protein we term NucC (Nuclease, CD-NTase associated). We purified E. coli MS115-1 NucC and tested its ability to degrade purified plasmid DNA in the presence of second messengers including cyclic and linear di-AMP, and the cyclic trinucleotides cAAA and cAAG. We found that NucC degrades DNA to ~50-100 bp fragments in the presence of low-nanomolar concentrations of cAAA, and is also activated to a lesser extent by cyclic di-AMP and cAAG (Figure 6A). NucC is apparently insensitive to DNA methylation: the enzyme is equally active on unmethylated PCR product, plasmid purified from a K-12 based E. coli strain (NovaBlue), and plasmid purified from strain MS115-1, which harbors the CBASS system and several predicted DNA methylases not found in K-12 strains (Figure S9A). Thus, NucC is DNA endonuclease activated by cAAA, the second messenger product of its cognate CD-NTases.

The E. coli MS115-1 CBASS system confers bacteriophage immunity through an abortive infection mechanism

Recent work has demonstrated that two CBASS systems with different architectures provide immunity against a range of bacteriophages, acting through an abortive infection mechanism to kill cells prior to the completion of phage replication (Cohen et al., 2019). Our structural and biochemical data suggest that in HORMA+Trip13 containing CBASS systems, HORMA domain proteins recognize specific peptides from a bacteriophage, then bind and activate their cognate CD-NTase to produce cAAA, which in turn activates a defensive response by the effector nuclease NucC. We tested the resistance of E. coli MS115-1 to bacteriophage λcI- (lacking the lysogeny gene cI), and found that in contrast to a laboratory strain of E. coli (JP313; (Economou et al., 1995)), E. coli MS115-1 is immune to bacteriophage λcI- infection as judged by plaque formation (Figure 6B–C). We next inserted the entire E. coli MS115-1 operon (including CdnC, HORMA, Trip13, and NucC) into E. coli JP313 on a plasmid vector, and found that the operon confers robust immunity to bacteriophage λcI- (Figure 6B–C). Immunity was compromised when we disrupted second messenger synthesis by CdnC (CdnC D72N/D74N mutant), HORMA-CdnC binding (HORMA-ΔC or R32A/R35A mutants), or NucC’s DNA cleavage activity (NucC D73N mutant; based on sequence alignments with type II restriction endonucleases), demonstrating that immunity depends on both the sensor and effector modules in this pathway (Figure 6B–C). Disruption of Trip13’s ATP hydrolysis activity (Trip13^EQ mutant) did not compromise immunity, consistent with the idea that Trip13 is a negative regulator of this pathway (Figure 6B–C).

To determine whether the E. coli MS115-1 CBASS system functions by abortive infection, we tracked the growth of bacterial cultures infected at high multiplicity of infection (MOI) with bacteriophage λcI-. In cells lacking the CBASS operon infected at an MOI of 2.5, we observed complete culture collapse beginning ~90 minutes post-infection due to phage-induced cell lysis (Figure 6D). In contrast, cells harboring the CBASS operon showed culture collapse beginning ~60 minutes post-infection, fully 30 minutes prior to phage-induced cell lysis. At an MOI of 0.25, we observed a sharp drop in growth rate at ~60 minutes post-infection in CBASS-harboring cells, followed by recovery of growth, while cells lacking CBASS showed culture collapse beginning at ~90 minutes post-infection (Figure 6E). Introduction of the NucC D73N active-site mutation eliminated these effects (Figure 6D–E). Together, these data support the idea that the E. coli MS115-1 CBASS system functions by abortive infection, causing the death of infected cells prior to the completion of phage replication. The requirement for NucC further suggests that cell death is caused by NucC-mediated host cell genome destruction.

LEAD CONTACT AND MATERIALS AVAILABILITY

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Kevin D. Corbett (kcorbett@ucsd.edu). All unique/stable reagents generated in this study are available from the Lead Contact with a completed Materials Transfer Agreement.

EXPERIMENTAL MODELS AND SUBJECT DETAILS

All proteins were produced in E. coli strain Rosetta2 pLysS (EMD Millipore). Cells were grown in standard media (2XYT broth) with appropriate antibiotics, and standard temperatures (37°C for growth, 20°C for protein expression induction). For cellular assays with E. coli MS115-1 (BEI Resources #HM-344) and JP313 (Economou et al., 1995), cells were grown in LB broth at 30°C with appropriate antibiotics.

METHOD DETAILS

Crystallization and structure determination

P. aeruginosa HORMA3:HORMA2:Peptide 1

A complex of Pa HORMA3 (residues 1-133), HORMA2, and Peptide 1 (EVMEFNP) was obtained by co-expressing and purifying His₆-MBP-tagged Peptide 1 with untagged HORMA3 and HORMA2, with the tag cleaved prior to crystallization. Crystals were obtained by mixing protein (10 mg/mL) in crystallization buffer 1:1 in hanging drop format with well solution containing 100 mM HEPES pH 7.5, 0.5 M sodium acetate, 31% PEG 3350, and 1 mM ATP. Crystals were flash-frozen in liquid nitrogen directly from the crystallization drop. We collected a 2.0 Å diffraction dataset on SSRL beamline 9-2, processed the data using the autoxds pipeline, and determined the structure by molecular replacement with PHASER, using the structures of Pa HORMA3 and Pa HORMA2. Initial models were manually rebuilt with COOT and refined using phenix.refine.

DNA Binding

For DNA binding assays, a 40 bp double-stranded DNA was produced by annealing 5′-FAM Oligo #1 and Oligo #2 (see Key Resources Table). Binding reactions (30 μL) contained 50 nM dsDNA probe and 0.39-50 μM protein in binding buffer (25 mM Tris-HCl pH 8.5, 50 mM NaCl, 5 mM MgCl₂, 1 mM DTT, and 5% glycerol), and were incubated 30 minutes at room temperature for binding. Pre-treated nitrocellulose (top; GE Healthcare) and nylon (GE Healthcare Hybond-N; bottom) membranes were assembled into a Bio-Rad bio-dot 96-well vacuum filtration apparatus. Samples were passed through the filtration apparatus, washed three times with 200 μL binding buffer, then imaged using a Bio-Rad Chemi-Doc gel imager. For K_d value calculation, images were quantified with ImageJ (Schneider et al., 2012), then binding curves were fit in Prism v. 8 (GraphPad Software).

Phage Display

For phage display, we used a commercial assay kit (Ph.D.-7 Phage Display Peptide Library Kit, New England Biolabs) and followed the recommended protocol for “solution phase panning with affinity bead capture” with the following modifications. We used His₆-tagged Ec HORMA, Pa HORMA2, and Pa HORMA3 as bait proteins. For affinity purification, we diluted 2 μL of bait protein at 0.2 mM and 10 μL phage library into 200 μL volume in TBST (50mM Tris pH7.5, 150mM NaCl, plus 0.1% Tween (first round) or 0.5% Tween (subsequent rounds)). After a 20 minute incubation at room temperature, 50 μL of magnetic Ni-NTA beads (Qiagen; pre-blocked with blocking buffer (0.1M NaHCO₃, pH 8.6, 5mg/mL BSA, 0.02% NaN₃) and washed 3x with TBST) were added, and the mixture was incubated a further 15 minutes at room temperature. Beads were washed 10x with wash buffer (50 mM Tris-HCl pH 7.5, 300 mM NaCl, plus 0.1% Tween (first round) or 0.5% Tween (subsequent rounds)). For elution, 1 mL elution buffer (0.2M Glycine-HCl, pH 2.2, 1mg/mL BSA) was added, incubated for 10 minutes at room temperature, then neutralized the eluate by adding 150 μL Tris-HCl pH 9.1. Eluted phage were titered and amplified according to standard protocols, and the selection was repeated, Following four rounds of selection, phages were isolated and the variable peptide region sequenced for at least 20 individual clones.

Second messenger synthesis

Cyclic trinucleotide synthesis was performed essentially as described (Whiteley et al., 2019). Briefly, 2 mL synthesis reactions contained 1 μM (for Ec CdnC) or 3 μM (for Pa CdnD) CD-NTase or CD-NTase:HORMA complex and 0.25 mM ATP (plus 0.25 mM GTP for Pa CdnD) in reaction buffer with 12.5 mM NaCl, 20 mM MgCl₂, 1 mM DTT, and 10 mM Tris-HCl pH 8.5 (or HEPES-NaOH pH 7.5 for reactions with Pa Trip13). For reactions with DNA, plasmid DNA (pUC18) was used at 0.15 μg/mL, 4 kb linear DNA was used at 0.1 μg/mL, single-stranded DNA (sheared salmon sperm DNA; Invitrogen #AM9680) was used at 1 μg/mL, and 40 bp dsDNA (annealed Oligos #1 and #2; see Key Resources Table) was used at 13 μg/mL. For reactions with Trip13, Ec Trip13 was used at 100 nM monomer/16.67 nM hexamer and Pa Trip13 was used at 300 nM monomer/50 nM hexamer; this represents a 60-fold lower molar concentration of active Trip13 hexamers compared to CdnC/HORMA in the reactions. Reactions were incubated at 37°C for 16 hours, then 2.5 units/mL reaction volume (5 units for 2 mL reaction) calf intestinal phosphatase was added and the reaction incubated a further 2 hours at 37°C. The reaction was heated to 65°C for 30 minutes, and centrifuged 10 minutes at 15,000 RPM to remove precipitated protein. Reaction products were separated by ion-exchange chromatography (1 mL Hitrap Q HP, GE Life Sciences) using a gradient from 0.2 to 2 M ammonium acetate. Product peaks were pooled, evaporated using a speedvac, and analyzed by LC-MS/MS.

Mass spectrometry

To characterize the products of CD-NTase reactions, we performed liquid chromatography-tandem mass spectrometry (LC-MS/MS). LC-MS/MS analysis was performed using a Thermo Scientific Vanquish UHPLC coupled to a Thermo Scientific Q Exactive™ HF Hybrid Quadrupole-Orbitrap™ Mass Spectrometer, utilizing a ZIC-pHILIC polymeric column (100 mm x 2.1 mm, 5 μm) (EMD Millipore) maintained at 45°C and flowing at 0.4 mL/min. Separation of cyclic trinucleotide isolates was achieved by injecting 2 μL of prepared sample onto the column and eluting using the following linear gradient: (A) 20 mM ammonium bicarbonate in water, pH 9.6, and (B) acetonitrile; 90% B for 0.25 minutes, followed by a linear gradient to 55% B at 4 minutes, sustained until 6 minutes. The column was re-equilibrated for 2.50 minutes at 90% B.

Detection was performed in positive ionization mode using a heated electrospray ionization source (HESI) under the following parameters: spray voltage of 3.5 kV; sheath and auxiliary gas flow rate of 40 and 20 arbitrary units, respectively; sweep gas flow rate of 2 arbitrary units; capillary temperature of 275°C; auxiliary gas heater temperature of 350 °C. Profile MS1 spectra were acquired with the following settings; mass resolution of 35,000, AGC volume of 1×10⁶, maximum IT of 75 ms, with a scan range of 475 to 1050 m/z to include z=1 and z=2 ions of cyclic trinucleotides. Data dependent MS/MS spectra acquisition was performed using collision-induced dissociation (CID) with the following settings: mass resolution of 17,500; AGC volume of 1×10⁵; maximum IT of 50 ms; a loop count of 5; isolation window of 1.0 m/z; normalized collision energy of 25 eV; dynamic exclusion was not used. Data reported is for the z=1 acquisition for each indicated cyclic trinucleotide.

Nuclease Assays

For all nuclease assays, UC Berkeley Macrolab plasmid 2AT (Addgene #29665; 4,731 bp) was used. Ec NucC (10 nM unless otherwise indicated) and second messenger molecules were mixed with 1 μg plasmid DNA or PCR product in a buffer containing 10 mM Tris-HCl (pH 7.5), 25 mM NaCl, 10 mM MgCl₂, and 1 mM DTT (50 μL reaction volume), incubated 10 min at 37°C, then separated on a 1.2% agarose gel. Gels were stained with ethidium bromide and imaged by UV illumination. Second messenger molecules were either purchased from Invivogen or synthesized enzymatically by Enterobacter cloacae CdnD02 and purified as previously described (Lau et al., 2019; Whiteley et al., 2019).

Plaque Assays

To measure bacteriophage immunity, the four-protein CBASS operon from E. coli MS115-1 (See Methods S1 for sequence) was PCR-amplified and inserted into vector pLAC22 for IPTG-inducible expression, then modified by PCR-based mutagenesis to generate mutant/tagged protein variants. Vectors were transformed into E. coli strain JP313 (Economou et al., 1995) for plaque assays.

For bacteriophage infection plaque assays, a single bacterial colony was picked from a fresh LB agar plate and grown in LB broth containing proper selection antibiotic at 37°C to an OD₆₀₀ of 0.1-0.2. CdnC operon expression was induced through the addition of 0.2 mM IPTG, followed by further growth for one hour to an OD₆₀₀ of 0.6-0.7. For phage dilution assays, 500 μL of cells were mixed with 4.5 mL of of 0.35% LB top agar was added and entire sample was poured on LB plates containing appropriate selection and/or IPTG (0.1 mM). Plates were spotted with 5 μL of bacteriophage λcI- (cI gene deleted to inhibit lysogeny and promote clear plaque formation) diluted in phage buffer (10 mM Tris, pH 7.5, 10 mM MgCl₂, 68 mM NaCl) plus 1 mM CaCl₂ at six dilutions: ~2.4x10⁶ PFU/mL and five 10-fold dilutions thereof. Plates were incubated at 37°C for 18 hours, then imaged. For plaque numbers, 500 μL of cells at OD₆₀₀ 0.6-0.7 were mixed with 10 μL of bacteriophage λcI (stock concentration ~2.4x10⁷ PFU/mL as measured by infection of JP313, diluted 1000-fold prior to mixing) diluted in phage buffer (10 mM Tris, pH 7.5, 10 mM MgCl₂, 68 mM NaCl) plus 1 mM CaCl₂, and incubated at room temperature for 30 min. 4.5 mL of 0.35% LB top agar was added and entire sample was poured on LB plates containing appropriate selection and IPTG (0.1 mM) and incubated at 37°C. Plaques were counted after 18 hours of growth.

Bacterial Growth Assays

pLAC22 vectors were constructed encoding either the full CBASS operon from E. coli strain MS115-1 (see Methods S1) or a variant encoding the NucC D73N mutant under the control of an IPTG-inducible promoter. E. coli strain JP313 (a derivative of E. coli MC4100) (Economou et al., 1995) was transformed with these vectors and incubated in 5 mL LB broth with 0.1 mg/ml ampicillin in culture tubes at 30°C with rolling. OD at 600nm was measured after 3 hours and cells were diluted to an OD₆₀₀ of 0.1 in a reservoir containing LB with 100ug/ml ampicillin and 0.2mM IPTG. 100 μl of cells were loaded into a 96-well plate (four replicates per strain) and 5 μl of bacteriophage λcI- diluted in phage buffer with 1mM CaCl₂ was added (final multiplicity of infection either 0.25 or 2.5). The 96-well plate was shaken at 30°C between measurements of OD₆₀₀, as an average of 25 flashes every 2 minutes, by a Tecan Infinite M200 plate reader.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analysis of primary diffraction datasets was performed with standard measures including R_sym, R_meas, and CC_1/2 to determine data quality and resolution cutoffs. Statistical analysis of refined crystal structures used standard measures including R and R_free factors, and geometry analysis by the MolProbity server (Chen et al., 2010).

DATA AND CODE AVAILABILITY

All primary diffraction datasets are available at the SBGrid Data Bank (https://data.sbgrid.org) under the accession numbers noted in Table S3 and the Key Resources Table. All reduced diffraction datasets and refined molecular coordinates are available at the RCSB Protein Data Bank (https://www.rcsb.org) under the accession numbers noted in Table S3 and the Key Resources Table. The original imaging data for Figure 1 and Figure S1 is available at Mendeley Data (https://data.mendeley.com) under DOI 10.17632/dr8mjtwh2d.1.