INTRODUCTION
Recent researches on diverse bacterial anti-phage immune systems have revealed ancient evolutionary roots for many central components of eukaryotic immune systems
1,2. One example is the Toll/Interleukin-1 Receptor (TIR)-domain proteins, which play a crucial role in regulating cell death and innate immunity in animals and plants, and have also been found to be essential in various anti-phage systems
3. These proteins can act either as sensors for detecting phage infections, as seen in Thoeris systems
4, or as effectors that prevent viral replication by degrading nicotinamide adenine dinucleotide (NAD
+) and inducing bacterial cell death, as observed in CBASS
5, Pyscar
6, and some prokaryotic Argonaute systems
7. In Thoeris systems, TIR domains convert NAD
+ into ADP-ribose (ADPR) isomers, which function as signaling molecules to activate downstream effectors
8-11.
Another example is the caspase-like proteins belonging to the C14 family of clan CD cysteine peptidases
12, which are broadly distributed in animals, plants, bacteria and archaea
13. These peptidases cleave polypeptides preferentially after an acidic aspartate residue (C14A subfamily), or after a basic arginine or lysine residue (C14B subfamily, including metacaspases or paracaspases). Specifically, there are twelve genes encoding caspases in the human genome, some of which are involved in promoting inflammation and others regulating programmed cell death
14-16. Prokaryotic caspase-like proteases have recently been uncovered to be involved in different anti-phage immune processes, acting either as specific proteases targeting bacterial gasdermins
17 or as non-specific proteases activated by the type III CRISPR-Cas systems
18-20.
In our recent research on immune systems supervised/regulated by CRISPR-Cas, we uncovered an anti-phage system, which remarkably, comprises only a TIR-domain protein and a caspase-like protease
21. In a recently published paper, this system was named type IV Thoeris regarding the conserved TIR component
22. In the current work, we demonstrated that type IV Thoeris is widely distributed across different bacterial and archaeal phyla. Interestingly, the TIR domain produces a signaling molecule N7-cADPR, which activates the caspase-like protease by interacting with a previously unknown C-terminal domain. We discovered that, in the absence of stimulus, the caspase component normally exists as inactive dimers that can spontaneously stack into right-handed helical filaments at high concentrations. The TIR-produced signal N7-cADPR triggers reassembly of these oligomers into active, double-stranded left-handed filaments. We determined high-resolution structures of Thoeris caspases in both inactive and active states, with or without substrate-analogous peptide inhibitor binding. These structures, complemented by functional analyses, reveal an unprecedented caspase activation paradigm, that is, a small signaling molecule triggers transition from one higher-order assembly of effectors to a completely different type of oligomerization and activation. These findings also widely enhance our understanding of caspase-mediated immunity across all three life domains.
RESULTS
Three type IV Thoeris systems each provides anti-phage immunity
To investigate the distribution of type IV Thoeris systems across various prokaryotes, we previously conducted a thorough search for operons containing both TIR-domain proteins (Pfam08937) and caspase-like proteases (Pfam00656) within all bacterial and archaeal genomes listed in the NCBI database
21. Here, we further constructed a phylogenetic tree based on the multiple alignment of caspase-like proteins (Supplementary Fig. S1a). Remarkably, type IV Thoeris systems were found to be prevalent in hundreds of bacteria and archaea spanning a wide range of phyla, with the feature of horizontal gene transfer (HGT). Interestingly, some Thoeris IV operons are embedded within CRISPR-Cas loci (Supplementary Fig. S1a, b), implying a mechanistic link to the bacterial adaptive defense line.
We synthesized five Thoeris IV operons (Supplementary Table S1) and found three variants (illustrated in Supplementary Fig. S1b) showing potency in combating coliphages (Fig. 1a). When Escherichia coli BW25113 cells were exposed to phage T6 or EP02SG, expression of each operon resulted in a marked reduction (by 1–3 logs) in plaque forming units (PFUs), with the Neisseria elongata 431 operon demonstrating the most potent antiviral effect (Fig. 1a). Notably, these antiviral effects were nullified upon mutation of the catalytic residue of either the TIR (a glutamate) or caspase (a cysteine) domain, underscoring the crucial roles played by the enzymatic activities of both proteins. By monitoring the growth curve of BW25113 cells under different multiplicities of infection (MOIs) by EP02SG, we observed that these three operons effectively protected the bacterial culture at a low MOI (0.03), but not at a high MOI like 10 (Fig. 1b; Supplementary Fig. S1c, d). It is suggested that type IV Thoeris systems confer population-level immunity by inducing an abortive infection (Abi)-like response, which sacrifices the infected individuals to protect the rest of the community. Consistently, under an MOI of 10, premature collapse of the culture was observed in Thoeris-expressing cells compared to the control group containing an empty vector.
We also explored the specific phage component responsible for triggering the immune response of type IV Thoeris. It has been reported that several proteins have potency to activate TIR proteins in the published literature
23. Therefore, we selected these genes to test their triggering activity. By introducing a plasmid expressing a particular protein of phage EP02SG into
E. coli cells containing the
Actinobacillus equuli 3820 system (
AeThoeris), we observed that the expression of the major capsid protein (MCP) led to significant cell growth defects and resulted in very few colonies on selective medium (Supplementary Fig. S2a). This trend was also observed when this protein was introduced into cells with the other two Thoeris operons, causing similar growth impairments (Supplementary Fig. S2b). Our findings suggest that MCP may be the trigger for the type IV Thoeris systems investigated here, which instigates the suicide of infected cells to halt the release of phage progenies and prevent further infection within the population.
TIR produces N7-cADPR to induce the peptidase activity of caspase
TIR-domain proteins are known to degrade NAD
+ into signaling molecules such as ADPR or its isomers to activate downstream immune effectors
8-11, suggesting that the TIR component of type IV Thoeris may generate NAD
+-derived signals to activate the peptidase activity of its caspase partner. To identify candidate target proteins of
N. elongata 431 caspase (
NeCaspase), we first chemically labeled NH
2-groups of proteins present in lysates of T6-infected cells expressing the defense system and then subjected the labeled proteins to MS. This allowed us to identify new protein N-termini produced specifically in cells expressing the defense system but not in control cells harboring the empty vector. Cleavage events were observed in multiple essential proteins of
E. coli (Supplementary Table S2) including RpsB, a constitutive subunit of the
E. coli 30S ribosome, which is essential for cell viability
24. The results showed a strong preference for an arginine in the P1 position and a slight preference for serine in the P1’ position (Fig. 1c). To verify that RpsB is a target of
NeCaspase, we purified RpsB and
NeCaspase protein and mixed them. Further inclusion of lysates of T6-infected cells expressing
NeTIR but not its E80Q mutant into the mix, resulted in cleavage of RpsB with a visible product band (Fig. 1d). However, when the catalytic cysteine in
NeCaspase or R225 in RpsB were mutated to alanine, the band of the cleavage product was absent (Fig. 1d). Taken together, our results suggest that
NeCaspase cleaves a range of proteins immediately after arginine residues upon activation during phage infection.
To further validate the activity of
NeCaspase, we synthesized a six-residue peptide using the residues preceding the identified target cleavage site in RpsB (positions 220 to 225, TVREGR) fused to 7-amino-4-methylcoumarin (AMC), which is expected to release free AMC when the peptide cleavage occurs downstream of the arginine residue. The results showed that
NeCaspase is able to cleave the synthetic peptide when incubated with lysate of infected cells expressing
NeTIR but not its E80Q mutant (Fig. 1e). Furthermore,
NeCaspase can be activated by the lysate of T6-infected
E. coli cells expressing
EcTIR-RN587,
NeTIR,
AeTIR (reported in this study),
EcTIR-328 (from
E. coli strain 328) or
PsTIR (from
Pseudomonas sp. 1–7
22), using the above peptide as a substrate (Fig. 1e), while the activation effect of those expressing
AeTIR was much weaker than others for an unknown reason.
Next, we sought to characterize the caspase-activating signaling molecule produced by Type IV Thoeris. First, we incubated purified
AeTIR/
NeTIR/
EcTIR (RN587) with NAD
+ in vitro, and further high-performance liquid chromatography (HPLC) and nuclear magnetic resonance (NMR) characterization showed that the main product
in vitro was 1′′–2′ gcADPR for all the three TIR proteins (Supplementary Fig. S3). However, during the preparation of our manuscript, a recently published paper reported that the signaling molecule in Thoeris IV is ADP-cyclo[N7:1′′]-ribose (N7-cADPR), a molecule that has not been described before that study
22. To verify the identity of this molecule, we first incubated
NeCaspase with the lysate of infected cells expressing
NeTIR, and then further purified the
NeCaspase protein through desalting. The absorbance
260nm/
280nm ratio of the purified
NeCaspase increased from ~0.5 to ~1.2 after incubation with the cell lysate (Supplementary Fig. S4a). Further, proteolysis of the purified
NeCaspase sample released molecules that are able to activate both
NeCaspase and
PsCaspase (Supplementary Fig. S4b). Then we subjected the released molecules to liquid chromatography coupled with mass spectrometry (LC-MS), which showed that the molecule has the same retention time and fragmentation pattern as N7-cADPR, but is distinct from 1′′–2′ gcADPR (Fig.1f; Supplementary Fig. S4c, d). This suggests that for TIR proteins, the main product
in vitro is not the molecule that activates downstream immunity
in vivo, a similar case of which was also reported for
Streptococcus thermophilus Csm complexes
25. To investigate whether other reported TIR products display activating potential, we incubated
NeCaspase or
PsCaspase with known TIR products, which showed that only N7-cADPR is able to activate both
NeCaspase and
PsCaspase (Fig. 1g; Supplementary Fig. S5a). Moreover, HPLC assay showed that only N7-cADPR, but not other nucleotides, was bound by caspases from different species (Supplementary Fig. S5b–g). Taken together, upon phage infection, TIR proteins of type IV Thoeris produce N7-cADPR to activate the caspase component, which then degrades multiple cellular proteins to terminate phage replication.
Apo caspase forms dimers that tend to oligomerize
To investigate the activation mechanism of caspase by N7-cADPR, we set out to study the structures of apo and activated caspases using NeCaspase, AeCaspase, PsCaspase, and EcCaspase strain 328 (EcCaspase RN587 could not be purified to homogeneity). Finally, we succeeded in solving the structure of PsCaspase in different states through cryo-electron microscopy (cryo-EM), which will be presented below. Interestingly, size exclusion chromatography-multi-angle light scattering (SEC-MALS) analysis showed that PsCaspase itself mainly folds as a dimer but also shows oligomeric components (Supplementary Fig. S6a). We further confirmed that assembly of the oligomeric apo-PsCaspase filaments is concentration-dependent in vitro (Supplementary Fig. S6b). We finally succeeded in solving the structure of apoPsCaspase in its oligomeric state at 2.58 Å through enrichment of these components (Supplementary Fig. S7 and Table S3). Notably, this oligomer is characterized by two parallel right-handed helical filaments of stacked caspase dimers, connected to each other through one interface each turn (Fig. 2a–c). Each filament has a helical pitch of around 108.6 Å, comprising approximately 3.4 caspase dimers (Fig. 2d). Within the helical assembly, adjacent caspase dimers exhibit a significant rotation by approximately 105.6 degrees in the dihedral angle along the filament extension orientation (Fig. 2d), giving the filament a typical right-handed helical shape rather than a linear structure seen in most ligand-induced oligomerization of effectors.
The dimeric unit of
PsCaspase filament reveals a cross-over C2-symmetrical dimer, with each monomer comprising two domains: an N-terminal caspase domain (NTD, residues 1–228) and a C-terminal domain (CTD, residues 229–328) (Fig. 2e). Dali search
26 returned the para-caspase domain of the mucosa-associated lymphoid tissue lymphoma translocation 1 (MALT1), which is also a dimer, as the most similar entry for the NTD (Supplementary Fig. S8a). The NTD exhibits typical features of a caspase, comprising a central β-sheet of six β-strands, which is surrounded by two α-helices (α2 and α3) on one side and three α-helices (α1, α4 and α5) on the other side (Fig. 2e). Two caspase domains form an antiparallel homodimer through their respective β6 strands, resulting in the formation of a contiguous 12-stranded β-sheet for the caspase domain with the active sites positioned on opposite faces of the central β-sheet. This overall arrangement of the caspase domain is identical to that of known caspase homodimers
27. Beyond these structural similarities to canonical caspases, the Thoeris caspase also exhibits unique structural features that may be important for its function. The most prominent feature is the presence of the CTD, which contains a 3-stranded β-sandwich flanked by 4 α-helices (Fig. 2e). Residues 259–268 lack electron density in all the solved structures of
PsCaspase, suggesting the flexibility nature of this region. The two CTDs of
PsCaspase support the two caspase domains with domain–domain interactions from the bottom, also forming a symmetrical dimer themselves (Fig. 2e). For the CTD, only entries with low
Z scores (< 6.7) were returned, with the multiple antibiotic-resistance repressor (MarR) from
Listeria monocytogenes as the most similar one (Supplementary Fig. S8b). Nevertheless, the interaction mode of two CTDs in
PsCaspase is completely different from that of the similar region in MarR (Supplementary Fig. S8b).
Interestingly, the dimeric PsCaspase units within the helical filament are not the same but exhibit two different inter-protomer conformations, which recur alternately in the filament (Fig. 3a). Superimposition of the two conformations showed that while the NTD of both conformations aligns relatively well, the CTD exhibits apparently two distinct opening sizes (Fig. 3b). Therefore, we denoted these two conformations as “closed” and “open” forms for the CTD, respectively (Fig. 3a). Consistently, the dimer interface covers a buried area of 4,006.6 and 3,765.9 Å2 for the “closed” and “open” conformations, respectively, in which the buried areas of CTD are 786.6 and 572.8 Å2, respectively. This suggests that the inter-protomer flexibility within the dimeric unit, especially the CTD, might be an inherent characteristic of PsCaspase, which facilitates formation as well as maintenance of the right-handed helical assembly. In addition, the interaction between two adjacent filaments is mediated completely by the NTD, in a head-to-head pattern (Fig. 2a).
Within the dimer interface, NTD dimerization involves interactions from mainchain atoms of β6 strands of both protomers (Fig. 3c). Moreover, Q201 in the NTD forms hydrogen bonds with H318 and Y319 of the CTD of the dimer-related protomer. The sidechains of Q208 and R209 also form hydrogen bonds and hydrophobic interactions with the mainchain and sidechain of V216 of the dimer-related protomer (Fig. 3c). To assess the importance of the caspase dimer interface, we performed phage defense assay and caspase cleavage assay to evaluate the effect of interface mutants on caspase activation. SEC-MALS confirmed that the mutant Q201A/Q208A/R209G only folds as a monomer in solution (Supplementary Fig. S6c). The same mutation abolished both its caspase activity in the presence of N7-cADPR (Fig. 3d) and IV Thoeris-mediated resistance to T6 phage (Fig. 3e). This indicates that perturbation of the dimer interface can significantly affect enzymatic activity, underscoring the critical role of dimerization in maintaining caspase function. Taken together, our structural and functional data demonstrate that inactivated Thoeris caspase forms a cross-over dimeric unit, which could oligomerize into two inter-connected parallel right-handed helical filaments, and notably, its dimerization is essential for N7-cADPR-induced enzymatic activity and antiphage function.
Assembly of apo caspase filament
To investigate the molecular mechanism driving filament formation by apo PsCaspase, we comprehensively analyzed the interaction interfaces between the PsCaspase dimers within the filament. To describe the interface more conveniently, we select one dimer in the open form (Caspase B1/B2) and describe its interfaces with the two adjacent closed-form dimers (Caspase A1/A2 and C1/C2), respectively (Supplementary Fig. S9a). Each interface is divided into two parts, based on the involvement of NTD or CTD of Caspase B1/B2. In the interface between Caspase A1/A2 and Caspase B1/B2, Caspase B1 NTD interacts with both Caspase A1 CTD and Caspase A2 NTD (Supplementary Fig. S9b), while CTD of both Caspase B1/B2 interact with both NTD and CTD of Caspase A2 (Supplementary Fig. S9c). In the interface between Caspase B1/B2 and Caspase C1/C2, Caspase B2 NTD interacts with both Caspase C1 NTD and Caspase C2 CTD (Supplementary Fig. S9d), while Caspase B1 CTD interacts with both NTD and CTD of Caspase C1 (Supplementary Fig. S9e). Due to the low resolution of the map of the parallel helical filaments (6.72 Å), which we could only dock the structure of single filaments into, detailed interactions could not be convincingly determined (Fig. 2a). Taken together, the right-handed helical assembly is maintained by interactions involving both NTD and CTD of two adjacent PsCaspase dimers while the connection between the two filaments is maintained by interactions between NTDs of PsCaspase dimers from each filament.
Apo caspase is self-inhibited
Closer inspection of the
apo PsCaspase dimer and comparison with human MALT1 reveal that the
PsCaspase, albeit in its dimeric form, is in an inactive state: loop L4 (residues 130–148, the loop regions are numbered according to the strands that they follow, L1–L5, respectively
28), which follows the active-site cysteine C129 and forms the linker between the large and small subunits in canonical caspases
27, appears flexible in
PsCaspase with residues 135–139 completely disordered in the electron density (Fig. 3f). The segments of L5 (residues 158–168) that could be unambiguously placed in the electron density adopt a conformation similar to the one observed in auto-inhibited MALT1 (Fig. 3f)
29. In this “inactive form”, the S
γ of the active-site cysteine, C129 in
PsCaspase, is not in its catalytically competent position and unable to form the proper catalytic dyad with the conserved H83, both of whose mutations also abolished the caspase activity in the presence of N7-cADPR (Fig. 3d) as well as IV Thoeris-mediated resistance to T6 phage (Fig. 3e). Furthermore, the corresponding S1 pocket of
PsCaspase is occupied by the L5 loop, a loop that carries many residues that are crucial for substrate recognition in active caspases. Notably, this self-inhibited conformation does not result from its oligomerization into helical filament or the interaction between two parallel filaments, but an inherent nature of the
PsCaspase dimer. First, this loop is away from the interface between two dimers in the filament (Supplementary Fig. S9f). Second, the conformation of the L5 loop is exactly the same between the structures of the two forms of dimer (Supplementary Fig. S9g). Third, free
PsCaspase dimers are available in solution (Supplementary Fig. S6a) but display no activity in the absence of N7-cADPR (Supplementary Fig. S5a), suggesting that
PsCaspase is also self-inhibited in its dimeric form. Taken together, the
apo PsCaspase dimer presents itself as an inactive protease with a dislocated nucleophile, a disabled substrate recognition site and an occupied, autoinhibited S1 pocket.
N7-cADPR induces the assembly of caspase dimers into left-handed double helices
To further understand how N7-cADPR is recognized by caspase and mediates subsequent caspase activation, we continued to study the structure of PsCaspase in its active state. Strikingly, N7-cADPR-bound PsCaspase displays a completely different assembly of filament compared to that of the apo form, with the longest length exceeding 380 nm (Fig. 4a). Then we successfully determined the cryo-EM structure of the PsCaspase bound to N7-cADPR at a resolution of 2.60 Å (Supplementary Fig. S7 and Table S3). The filament is composed of two left-handed helical “chains” formed by PsCaspase dimers through “side-by-side” interaction of their NTD, respectively, which are intertwined with each other to form a “double-helix” through an N-terminal “head-to-head” interaction (Fig. 4b). The “double-helix” has a helical pitch of around 466.9 Å, comprising approximately 26 caspase dimers, 13 in each chain, and a diameter of ~150 Å in the radial direction (Fig. 4c). Within the single left-handed helical assembly, adjacent caspase dimers exhibit a rotation of approximately 26.9 degrees (Fig. 4d), but in a typical left-handed helical manner (Fig. 4b) rather than the right-handed shape of the apo form. Negative-stain EM of the lysates from E. coli cells expressing the Thoeris IV system from Pseudomonas sp. 1–7 confirmed the in vivo production of left-handed helices upon T6 phage infection (Supplementary Fig. S10).
N7-cADPR binds within a pocket formed between two CTDs
The cryo-EM map allows unambiguous identification of density for one N7-cADPR molecule within the pocket formed between two CTDs of a dimer, but not between two dimeric caspase domains (Fig. 5a). Specifically, the adenine base of N7-cADPR is sandwiched between the sidechain of W301 from both CTDs and stabilized by π–π and π–alkyl interactions (Fig. 5b). Furthermore, the phosphate groups are bound by electrostatic interactions from the sidechain of Q284 and hydrogen bonds from the mainchain nitrogen of W301. Moreover, the hydroxyl group of the ribose moiety is also stabilized by hydrogen bonds from E255 sidechains (Fig. 5b). These residues are highly conserved among Thoeris caspases except that W301 is replaced by tyrosine which can play similar functions here in other homologs (Supplementary Fig. S11). The structure provides a detailed representation of the precise binding mode between caspase and N7-cADPR, but not other reported products of TIR domains. Notably, the symmetrical nature of the pocket within the two CTDs might be more suitable for accommodating a ligand with a nearly symmetrical conformation. Among the known products of TIR enzymes, N7-cADPR is likely the most symmetrical one with its two ribose carbon 1 atoms attached to nitrogen atoms N7 and N9 in the adenine base, respectively (Fig. 5a). Interestingly, cADPR, which is most structurally similar to N7-cADPR among the known TIR products cannot be bound by caspase homologs (Supplementary Fig. S5c). This might be attributed to a less stable stacking of the adenine moiety of cADPR from W301, compared to simultaneous presence of π–π and π–alkyl interactions between N7-cADPR and W301. This structural insight provides a clear explanation for our observations that the Thoeris caspase is selectively activated only by N7-cADPR, but not the other known TIR products (Fig. 1g).
To validate the critical role of the N7-cADPR-binding pocket, we introduced alanine substitutions at key residues involved in N7-cADPR interactions individually and tested their effects on caspase and phage defense activities. Notably, all the mutations of these residues almost abolished binding to N7-cADPR as revealed by HPLC assay (Fig. 5c), as well as the caspase activities of PsCaspase in the presence of N7-cADPR (Fig. 5d). As expected, these mutations also disrupted type IV Thoeris-mediated resistance to T6 phages (Fig. 5e). Interestingly, the PsCaspase or NeCaspase CTD alone does not bind N7-cADPR in HPLC either (Fig. 5c; Supplementary Fig. S5h), which might result from its monomer but not dimeric state on its own, as supported by SEC-MALS analysis of NeCaspase CTD (Supplementary Fig. S6d, the amount of PsCaspase CTD did not reach the requirement for SEC-MALS analysis), suggesting that maintenance of the dimeric state of the CTD within Thoeris caspase requires dimerization potential of the NTD. Taken together, these results highlight the essential roles of the CTD residues in N7-cADPR binding and the NTD residues in dimerization.
N7-cADPR-binding closes the CTD cleft of caspase dimers
To understand the molecular mechanism driving the transition from apo to N7-cADPR bound form of PsCaspase, we performed a comprehensive analysis of the conformational changes that occur in PsCaspase upon N7-cADPR binding. Notably, in the N7-cADPR-bound filament, all dimeric units adopt the same conformation. Structural comparison showed that N7-cADPR binding induces substantial conformational changes in the dimeric structure of PsCaspase, especially when compared to the open form. Compared to the apo open form, upon interaction with N7-cADPR, most notably, sidechain and mainchain of W301 of α9 engages in stacking and hydrogen bonding interactions with N7-cADPR, respectively. This interaction causes α9 to move an approximately 4.0 Å shift towards the adenine moiety of N7-cADPR (Fig. 5f). In addition, movement of other secondary structures harboring N7-cADPR-binding residues together results in “close” of the overall CTD (Fig. 5g). Interestingly, the conformation of N7-cADPR bound form more resembles that of the apo “closed” form in the CTD (Fig. 5h). That is, binding of N7-cADPR seemingly “unifies” the CTD of PsCaspase into the closed form. As mentioned above, CTD of the open-form PsCaspase dimer plays a key role in maintaining the right-handed helical assembly (Supplementary Fig. S9c, e). We propose that N7-cADPR binding-induced “close” of the CTD would influence many interfaces within the apo helical assembly (Fig. 5i), which likely results in its disassembly.
Assembly of the left-handed double-helical filament of caspase
To understand the molecular mechanism driving the left-handed double-helical filament formation upon N7-cADPR binding, we moved on to analyze the interfaces within the filament, the conformational changes that occur upon N7-cADPR binding, and the facilitating mechanisms for new type of oligomerization. The structure of the PsCaspase-N7-cADPR filament shows that individual dimeric PsCaspase units also form the basic repeating units of the double-helical filament (Fig. 6a). Within each single left-handed helical filament, interactions between PsCaspase units are dominated by hydrophobic and electrostatic interactions. Notably, W206, which is in the loop linking α5 and β6 of Caspase A1, is deeply inserted into a pocket surrounded by K3, A4, G43, K45 and L68 of Caspase B1, meanwhile, W206 of Caspase B2 is inserted into the same pocket in Caspase A2 (Fig. 6b, c). This residue functions as an anchor to interlock each two adjacent PsCaspase dimers throughout the axial direction. Moreover, R34 and R218 of Caspase B1 interact with the carboxyl of the C-terminus I328 and sidechain of D72 of Caspase A2 (Fig. 6c), respectively, and vice versa.
Like base-pairing in double-stranded DNA, the double-helical structure of two filaments is maintained by the “head-to-head” interaction between two PsCaspase dimers, also in a C2-symmetry to each other, in the radial direction. Specifically, R58 and R62 of Caspase A2 form electrostatic interactions with D103 and E87 of Caspase D1, respectively, and Y92 of Caspase A2 also forms hydrogen bonds with E109 of Caspase D1, and vice versa (Fig. 6d). Meanwhile, the same set of mutual interactions are also formed between Caspase A1 and D2. To assess whether the double-helical filament formation is critical for enzymatic activity of PsCaspase, we generated several PsCaspase mutants with alanine substitutions for residues involved in either axial or radial interactions. All these mutations resulted in loss of function both in the caspase activity (Fig. 6e) and in bacterial resistance to infection by T6 phages (Fig. 6f).
Next, we analyze how N7-cADPR binding promotes double-helical filament formation. First, since maintenance of the apo right-handed helical assembly relies on the alternating stacking open and closed form of PsCaspase, unified “closing” of the CTD by N7-cADPR would likely disrupt the assembly. Second, the buried surface areas between NTD and CTD in apo open, apo closed and N7-cADPR-bound PsCaspase dimer are 883.83, 1,035.6 and 1,095.3 Å2, respectively, suggesting that N7-cADPR binding further stabilizes the two domains with each other. Since N7-cADPR binding site is located far away from the NTD, allostery is involved in N7-cADPR-induced activation of PsCaspase. However, how N7-cADPR binding triggers conformational change in the caspase domain to facilitate double-helical filament formation remains not fully known. It might be that conformational selection is involved in this process. Taken together, N7-cADPR binding restricts the structural flexibility of PsCaspase that is in favor of the right-handed helix as observed in the apo form and facilitates formation of more ordered double-stranded helical assembly.
Left-handed double-helical assembly relieves caspase self-inhibition
To elucidate how double-helical filament formation activates PsCaspase activity, we performed a comparative analysis of the structures of PsCaspase in its apo and N7-cADPR-bound states. Since structures of the NTD are basically the same for the open and closed apo form, the open form is used here due to its density for more residues of the L4 loop. A series of conformational changes exist between the NTDs of apo and N7-cADPR-bound PsCaspase. Among them, the most notable change occurs in the region ranging residues 81–91 (β3A–β3B hairpin), which is raised up in the double-helical assembly and displays a ~14.6-Å shift for its middle region (Fig. 6g), resulting from the interaction from the NTD of the symmetry-related molecule in the other strand of the filament (Fig. 6d). Consequently, the β3A strand, which forms a long, bent strand ranging from residues 75–81 in the apo structure, cannot maintain its intact strand structure and is split into two strand parts linked by a short loop (Fig. 6g). The structural change of β3A–β3B region results in multiple consequences. First, the C-terminus of β4, which stacks with β3A in the sheet and harbors the catalytic C129, moves along with β3A. This results in a 6.4-Å shift for the Sγ atom of C129, thus rearranging the active site together with the movement of H83, which is right in the newly formed “short loop” within β3A (Fig. 6g). Second, the structural change of β3A–β3B hairpin will cause severe clashes to the L4 loop, whose density was not observed in the N7-cADPR-bound structure (for residues 131–142), suggesting its disordered nature after its structure cannot be maintained by the β3A–β3B hairpin as observed in the apo form (Fig. 6g). Third, the movement of C129 also results in a severe clash to the “self-inhibited” L5 loop, whose density was also not observed in the N7-cADPR-bound structure (for residues 159–168), suggesting its disordered nature in the N7-cADPR-bound form (Fig. 6g). Last, loop L1 (residues 11–19) also turns disordered in the N7-cADPR-bound structure probably due to the lack of stabilizing interactions from L4/5 as observed in the apo form. Taken together, compared to the “self-inhibited” apo form, PsCaspase in the left-handed double-helical assembly displays a rearrangement of the nucleophile and disorder of the “self-inhibited” loop, priming the PsCaspase for reacting with the substrate.
Caspase filament bound to the substrate-analogous inhibitor
To uncover the structural basis for substrate binding and cleavage by the N7-cADPR-activated caspase filament, we determined the cryo-EM structure of activated caspase filament bound to a substrate-analogous peptide inhibitor z-VRPR-fmk (short for VRPR hereafter) at a resolution of 2.21 Å (Supplementary Fig. S7 and Table S3), which mimics the properties of the covalent intermediate of the reaction
30. The structure of the
PsCaspase-N7-cADPR-inhibitor complex shows an overall arrangement similar to that of the
PsCaspase-N7-cADPR complex (Fig. 7a). Within each
PsCaspase monomer, density corresponding to one inhibitor is clearly observed, which is deeply embedded, with the catalytic C129 (S
γ of C129) irreversibly alkylated by the inhibitor (Fig. 7b, c). Notably, the sidechain of P1-Arg of the inhibitor is stabilized by a negatively charged deep pocket formed by D24 from α1, D127 from β4 and E164 from L5 (Fig. 7b, c). Moreover, the mainchain atoms of P1-Arg are stabilized by H83, G84, C129 and A162. The mainchain and sidechain atoms of P3-Arg are also stabilized by E164. Consistently, mutations of D24 and E164 separately also led to a marked reduction in both caspase activity and phage resistance (Fig. 7d, e). These results underscore the critical role of the negatively charged pocket to accommodate the sidechain of P1 residue of the substrate in both caspase activity and antiphage function of the Thoeris caspase.
To entirely elucidate the mechanism of N7-cADPR-activated caspase activity, we performed a comparative analysis of the structures of PsCaspase in its apo and inhibitor-bound states. As discussed above, the structural elements that undergo conformational changes between apo and N7-cADPR-activated states (Fig. 6g) will also be focused on here. First, the L4 loop that follows C129, which is disordered in the N7-cADPR-activated state, now turns completely ordered due to the stabilization by the substrate-analogous inhibitor (Fig. 7f). However, it adopts a completely opposite relative position with the β3A–β3B hairpin compared to that in the apo form. That is, in the apo form the β3A–β3B hairpin is covered by L4, however, L4 is in turn covered by the β3A–β3B hairpin in the N7-cADPR-VRPR-bound structure (Fig. 7f). Second, substantial conformational change occurs to the “self-inhibited” L5 loop, which turns fully ordered but moves away from the S1 pocket and interacts with the inhibitor instead in the N7-cADPR-VRPR-bound structure. Notably, structural comparison between the apo and inhibitor-bound structure further confirms the “self-inhibited” role of L5, which covers the substrate binding channel like a lid in the apo structure (Fig. 7f). Specifically, Y161 within L5 lies in the same position as the P1-Arg in the apo structure. Moreover, C129 and H83 also show movement upon binding of the inhibitor when compared to either the N7-cADPR-bound (Fig. 7g) or the apo structure (Fig. 7f). That is, compared to the N7-cADPR-bound structure, two of the disordered loops, L4 and L5, both become ordered upon the inhibitor binding, while L1 remains disordered (Fig. 7g).
To prove that N7-cADPR binding is the prerequisite for substrate binding by PsCaspase, we set up an experiment in which PsCaspase was first incubated with the inhibitor in the presence or absence of N7-cADPR, and then half of each sample was collected. Afterwards, the remaining samples were buffer exchanged to remove unbound inhibitors, followed by another round of sample collection. Finally, N7-cADPR was added to both sets of samples to test the activation of the activity (Fig. 7h). The results showed that PsCaspase pre-incubated with the inhibitor can still be activated by N7-cADPR after buffer exchanging, but that pre-incubated with both N7-cADPR and the inhibitor completely lost potential to be activated (Fig. 7i). Taken together, present structural and biochemical data support an activation model for the PsCaspase dimer: in the apo structure, the L5 loop acts as a lid, covering the channel to accommodate the sidechain of P1-Arg in the active site of the enzyme (Fig. 3f). Upon N7-cADPR binding, the double-helical left-handed filament structure induces the L5 loop to be disordered and rearrangement of the active site, completely opening the active site and allowing substrates to enter (Fig. 6g). Finally, substrate binding stabilizes the L5 loop to accommodate itself in the active site (Fig. 7f).
DISCUSSION
Growing evidence supports that many central components of the innate immunity are shared across the tree of life
1,2. A significant example is TIR domain proteins, which govern cell death and innate immunity in animals, plants, and prokaryotes through their enzymatic activities
3. Recently, it has also been discovered that caspase-like proteases, known for their roles in eukaryotic innate immunity and programmed cell death, are involved in bacterial defenses against phages
17-20. Bacterial type IV Thoeris, which comprises a TIR domain protein that converts NAD
+ to N7-cADPR and a caspase-like protease that is activated by this signal molecule, unprecedentedly establishes a mechanistic connection between these two broadly distributed immune components. Notably, based on sequence similarity, we identified a multi-domain protein (Tci_124240) encoded by the plant
Tanacetum cinerariifolium, which encompasses both TIR and caspase domains, along with a domain similar to the CTD of Thoeris caspase (Supplementary Fig. S12a). Using AlphaFold3, we predicted its structure (Supplementary Fig. S12a) and observed significant structural homology of these three domains to those of IV Thoeris (Supplementary Fig. S12b). This finding supports a potential mechanistic link of TIR signaling to caspase-like proteases in plants.
The structures of caspase in its
apo, signaling-molecule-bound, and substrate analog-bound states, reported in our present work, provide insights into the mechanism by which caspase effectors sense and respond to antiviral nucleotide second messenger signals, ultimately cleaves multiple substrate proteins in cells (Fig. 7j). In the absence of phage infection and nucleotide signals, the caspase itself exists as dimers that might assemble into right-handed helical oligomers in some cases when the local concentration is high. In the
apo state, the active site of caspase is hindered by L5 loop, which blocks access to the active site. Upon phage infection, specific components of the phage, likely the major capsid protein, are sensed by the TIR-domain protein, which degrades cellular NAD
+ molecules, producing N7-cADPR. This signal molecule then binds to the CTD of the caspase-like protease, causing a significant conformational change in the caspase dimers, dissociating the right-handed helical oligomers if any and inducing formation of a double-stranded left-handed filament by caspase dimers. Filament assembly induces a conformational change in the L5 loop to make it disordered, which exposes the active site residues cysteine and histidine dyad, allowing peptide substrates to bind and be cleaved. These events result in an abortive infection response that prevents phage proliferation (Fig. 7j). Structural comparison between the
apo and N7-cADPR-bound form clearly indicates that right-handed helical oligomer of caspase dimers cannot be maintained upon N7-cADPR binding. Time-dependent incubation of N7-cADPR with
apo caspase also supports the disassembly of the right-handed filaments and then formation of the left-handed filaments (Supplementary Fig. S10C). Interestingly, the activation mechanism of
PsCaspase, that is, the formation of left-handed filaments, also bears some similarity to that of the type I Thoeris system
8. In type I Thoeris, the ThsA also forms a left-handed double-stranded filament-like structure, while the binding ratio and mode of the signaling molecule, as well as the activation mechanism are different between the type I and IV Thoeris systems. Together, our data reveal an unprecedented model of how the TIR-generated signal N7-cADPR drives dissociation of a parallel right-handed helical assembly, re-oligomerization into a double-stranded left-handed filament and restructures the catalytic center of the bacterial anti-phage caspase, converting it from a self-inhibited state into an active filament capable of substrate cleavage.
Distinct from C14A and C14B(M) metacaspase subfamilies, paracaspases C14B(P) do not require autocleavage or cleavage by other proteases to become fully activated
28. Similar to MALT1, the only member with detailed structural and mechanistic studies, the Thoeris caspases also do not undergo autocleavage to become activated. During activation, MALT1 undergoes first a monomer-to-dimer transition and then a substrate binding to be fully activated
29,31. However, our data uncover a unique activation mechanism, in which the Thoeris caspases are constitutively dimers and oligomers but in a self-inhibited state themselves. Substrates cannot directly bind the Thoeris caspases alone, but N7-cADPR-induced oligomerization of the caspase dimers into double-stranded left-handed filaments fully relieves the self-inhibition state and opens the substrate-binding channel of the caspase. While caspase enzymes have been extensively and well-studied, the discovery that these enzymes also respond to and are regulated by signaling molecules to be activated is unprecedented. In conclusion, our findings unravel the ancient mechanistic link between TIR-based immune signaling and the protease activity of caspase-like proteins in bacterial anti-phage immunity, potentially enhancing our knowledge of innate immunity across various organisms, including animals and plants.
MATERIALS AND METHODS
Bacterial strains and growth conditions
Escherichia coli DH5α was used for plasmid construction, E. coli BL21 (DE3) for protein expression and purification, and E. coli BW25113 as the host strain for phage assays. All strains were cultured at 37 °C in LB medium (10 g/L tryptone, 5 g/L yeast extract, 10 g/L NaCl). Phages T6 and EP02SG were propagated in liquid LB cultures. After removing bacterial cells by centrifugation, the supernatants were filtered through 0.22-µm membranes. Phage titers were determined by spot assay.
Plasmid construction and transformation
The plasmids, oligonucleotides, and synthetic genes used in this study are listed in Supplementary Table S1. Double-stranded DNA fragments were amplified using Phanta Super-Fidelity DNA polymerase (Vazyme) and assembled into predigested plasmids using the Trelief® Seamless Cloning Kit (Tsingke) following a Gibson-assembly strategy. All constructed plasmids were confirmed by DNA sequencing.
Plasmids were transformed into E. coli DH5α and E. coli BL21 (DE3) using commercial competent cells. Competent E. coli BW25113 cells were prepared using the Ultra-Competent Cell Preps Kit (Sangon). A total of 200 ng of plasmid DNA was mixed with 100 µL of competent cells, recovered in 500 µL LB at 37 °C for 1 h, serially diluted, and plated on selective agar plates containing appropriate antibiotics. Transformation efficiency was calculated from colony counts and dilution factors. Each experiment included three biological replicates.
Plaque assay and infection dynamics in liquid culture
Phage stocks were serially diluted 10-fold in LB and plated onto double-layer agar plates of E. coli BW25113 harboring either an empty vector or defense systems. For infection dynamics, 180 μL of early-log phase cultures were transferred into wells of 96-well plates containing 20 μL of phage diluent to achieve a final MOI of 0.03 or 10 for phage EP02SG, or 20 μL of LB for uninfected controls. Infections were performed in triplicate using cultures prepared from three independent colonies. OD600 was measured every 15 min for 6 h at 37 °C using a Stratus 600 nm real-time plate reader (Cerillo) with continuous shaking.
Identification of product of TIR in vitro
40 mL reaction mixture of NeTIR with NAD+ was used to isolate 1′′–2′ gcADPR. After the solution was concentrated by rotary evaporation, High-performance liquid chromatography (HPLC) was used to purify product 1′′–2′ gcADPR. HPLC analysis was performed on a Shimadzu LC-20AT system using a C18 analytical column (AQ-C18, 250 × 4.6 mm, particle size 5 μm; Velch Technologies).
For the first separation step, the mobile phase consisted of water containing 10 mM ammonium formate (pH 8.5, solvent A) and acetonitrile (solvent B). The gradient was as follow: 0 min, 3% B; 7 min, 3% B; 12 min, 70% B; 15 min, 70% B; 17 min, 3% B. Flow rate: 1 mL/ min. For the second separation step, the mobile phase consisted of water containing 0.1% formic acid (solvent C) and acetonitrile containing 0.1% formic acid (solvent D). The gradient was as follow: 0 min, 5% D; 7 min, 5% D; 11 min, 100% D; 13 min, 100% D; 16 min, 5% D. IAD was separated using similar steps.
LC-MS analysis was performed on a Shimadzu UPLC / QTOFMS system using a C18 analytical column (AQ-C18, 250 × 4.6 mm, particle size 5 μm; Velch Technologies). The gradient was as follow: 0 min, 3% B; 7 min, 3% B; 12 min, 70% B; 15 min, 70% B; 17 min, 3% B. Flow rate: 1 mL/ min.
About 8 mg of purified 1′′–2′ gcADPR was concentrated by rotary evaporation and then dissolved in 300 µL of D
2O. The sample was transferred to a 5 mm NMR tube rated for 500 MHz. The Bruker Avance Ⅲ 500 MHz NMR spectrometer was utilized to acquire
1H,
13C,
1H -
1H-
13C HSQC, and
1H-
13C HMBC spectra at 298 K. The chemical structure of each compound was determined by assignments of
1H and
13C peaks and correlations, especially those linking two ribose rings (Supplementary Fig. S3). Similarly, the structure of IAD was confirmed by NMR. HMBC spectra shows both H1’A-C8A and H1’I-C2I cross-peaks
9.
Protein expression and purification
pET28a-His6-SUMO constructs encoding NeCaspase, PsCaspase, AeCaspase, EcCaspase strain 328, TIR proteins from different species, and their respective mutants were transformed into E. coli BL21 (DE3) for recombinant protein expression. The PsCaspase mutants were generated by two-step PCR and were subcloned, overexpressed and purified in the same way as the wild-type protein. Cultures were grown in 1 L LB medium supplemented with 50 μg/mL kanamycin at 37 °C until the OD600 exceeded 0.8, followed by induction with 0.2 mM IPTG and incubation at 16 °C for 12 h. Cells were harvested by centrifugation at 2,350× g for 30 min at 4 °C, resuspended in 30 mL binding buffer (50 mM Tris, pH 8.0, 0.2 M NaCl, 20 mM imidazole), and lysed by sonication on ice. The lysates were clarified by centrifugation (15,000× g, 4 °C, 1 h) and loaded onto a pre-equilibrated His-Trap HP column. After washing with 20 mM imidazole, the SUMO tags were removed on-column using Ulp1 protease at 16 °C for 3 h, and the proteins were eluted with buffer containing 20 mM imidazole.
EcRpsB and its mutants were expressed from pRSFDuet-EcRpsB in E. coli BL21 (DE3) under the same induction conditions. Bound proteins were eluted with 300 mM imidazole.
When necessary, proteins were further purified by size-exclusion chromatography using a Superdex 200 10/300 GL column (Cytiva). Purified proteins were flash-frozen and stored at ‒80 °C.
Preparation of TIR metabolites
E. coli BW25113 carrying pUC19-TIR was streaked onto LB agar plates containing ampicillin and incubated overnight. A single colony was inoculated 1:100 into 100 mL LB supplemented with 100 µg/mL ampicillin and 500 µM IPTG, and cultured at 37 °C until OD600 reached 0.5–0.8. Cells were then infected with T6 phage at an MOI of 1 and incubated for an additional 25–60 min.
Cells were harvested by centrifugation at 6,000 rpm for 10 min at 4 °C, washed once with cold water, and centrifuged again. Pellets were resuspended in 500 µL pre-chilled buffer (25 mM Tris-HCl, pH 7.4, 100 mM NaCl) and lysed by sonication for 5–10 min. Lysates were cleared by centrifugation (13,000 rpm, 4 °C, 10–20 min). The flow-through with molecular weight lower than 3-kDa (TIR metabolites) was collected by ultrafiltration (3-kDa cutoff; 13,000 rpm, 4 °C, 30–40 min) and stored at ‒20 °C.
Substrate cleavage assay
Protein cleavage reactions were carried out in buffer containing 25 mM Tris-HCl (pH 7.4) and 100 mM NaCl. EcRpsB (30 µM) was incubated with 10 µM NeCaspase or its mutants together with lysates of cells expressing wild-type or E80Q NeTIR. The reactions were incubated at 37 °C for 1 h and terminated by adding SDS loading buffer followed by boiling. Reaction products were analyzed by SDS–PAGE.
Binding-release assay
Recombinant NeCaspase pre-incubated with NeTIR-derived metabolites was concentrated using a 3-kDa ultrafiltration device and washed 5–6 times with ice-cold buffer (25 mM Tris-HCl, pH 7.4, 100 mM NaCl) to remove unbound small molecules. The retentate was treated with Proteinase K (50 µg/mL) at 4 °C for 4 h, and protein degradation was confirmed by 12% SDS–PAGE. The flow-through with molecular weight lower than 3-kDa was collected by centrifugation through a 3-kDa filter (14,000× g, 4 °C, 30–40 min).
For activity assays, 30 µL of the flow-through was added to a 100-µL fluorescence reaction mixture containing 25 mM Tris-HCl (pH 7.4), 100 mM NaCl, 1 mM DTT, 20 µM AMC-IVINQR substrate, and 10 µM NeCaspase. Fluorescence was recorded at 37 °C every 30 s for 20 min (Ex/Em = 341/441 nm). NeCaspase without TIR metabolites served as the negative control.
Caspase activity assay with TIR metabolites
Fluorescent caspase assays were carried out in 100 µL of 25 mM Tris-HCl (pH 7.4), 100 mM NaCl, and 1 mM DTT. The reaction contained 20 µM AMC-IVINQR substrate and NeCaspase at a final concentration of 10 µM, with or without the caspase inhibitor Z-VRPR-FMK (20 µM, twice the enzyme concentration). 10 µL of TIR metabolite was added. Reactions were incubated at 37 °C, and fluorescence was measured every 30 s for 20 min using a microplate reader (excitation 341 nm, emission 441 nm). Reactions without TIR metabolites served as negative controls. For assays using individual small molecules (including N7-cADPR, ADPR, cADPR, 1′′–2′ gcADPR, 1′′‒3' gcADPR, pRib-AMP, 2',3'–cAMP, 2',3'‒cGMP, and His-ADPR), the same reaction conditions were used, except that the TIR metabolites in NeCaspase reactions were replaced with the corresponding small molecule at a final concentration of 20 µM. Fluorescence acquisition and control settings were identical to those described above. All assays were performed with at least three independent biological replicates.
Incubation of NeCaspase and PsCaspase with cyclic ADPR analogs or TIR-derived metabolites and SDS-PAGE analysis
NeCaspase and the inactive mutant NeCaspaseC131A were diluted to 20 μM in assay buffer (25 mM Tris, pH 7.4, 100 mM NaCl). 1′′–2′ gcADPR and N7-cADPR were added to a final concentration of 100 μM. Reaction mixtures were incubated at 37 °C, and 10 μL aliquots were taken at 0, 12, and 24 h. Each aliquot was immediately mixed with 4× Laemmli sample buffer containing SDS and a reducing agent to stop the reaction. Samples were heated at 100 °C for 5 min, briefly centrifuged, and resolved on 12% SDS-PAGE alongside molecular-weight markers. Gels were stained with Coomassie Brilliant Blue G-250 and imaged.
For PsCaspase, the wild-type protein and the inactive mutant PsCaspaseC129A were diluted to 20 μM in the same assay buffer. EcTIR strain 328-derived metabolites produced by the wild-type or E84Q mutant enzyme were added to the reaction mixtures. Incubation, sampling, reaction termination, and SDS-PAGE analysis were performed under the same conditions as those described for NeCaspase.
LC-MS analysis of NeCaspase-bound metabolites
Purified NeCaspase protein was incubated with N7-cADPR at 4 °C for 1 h. The reaction mixture was then washed three times with buffer (25 mM Tris-HCl, pH 7.4, 150 mM NaCl) using a 3-kDa ultrafiltration device to remove unbound metabolites. The retained protein fraction was digested with proteinase K on ice for 30 min, and the released small-molecule products in the filtrate were subjected to liquid chromatography-mass spectrometry (LC-MS) analysis.
Chromatographic separation was performed on a Waters ACQUITY UPLC HSS T3 column (1.8 µm, 2.1 × 100 mm). The mobile phase consisted of solvent A (10 mM ammonium acetate adjusted to pH 9 with ammonium hydroxide) and solvent B (methanol). The flow rate was 0.30 mL/min. The LC gradient was programmed as follows: the run started at 95% A and 5% B and was held for 1 min, followed by a transition to 15% A and 85% B at 4 min. At 4.5 min, the composition was changed to 0% A and 100% B and maintained until 7.5 min. The system was then returned to 95% A and 5% B at 7.6 min and equilibrated until 9 min.
Eluted metabolites were analyzed using electrospray ionization mass spectrometry operated in negative-ion mode. Metabolite identification was based on comparison of retention times and MS/MS fragmentation patterns with authentic standards.
HPLC-based binding analysis
Purified NeCaspase, AeCaspase, EcCaspase strain 328, or PsCaspase (30 μM each) was incubated with 20 μM N7-cADPR or its analogs on ice for 2 h. The mixtures were centrifuged through 3-kDa ultrafiltration devices to remove unbound metabolites, and the filtrates were collected. NeCaspase mutants were processed in the same way. The filtrates were analyzed by HPLC using an Bonnasil-BS AQ C18 (4.6 × 250 mm, 100 A). Chromatographic separation was performed at a flow rate of 1.0 mL/min under isocratic elution with 97% 20 mM potassium phosphate buffer (pH 6.0) and 3% acetonitrile. Detection was carried out at 273 nm.
SEC-MALS
Static light scattering experiments of NeCaspase, its mutants, truncation variants, and AeCaspase were performed in 10 mM Tris-HCl (pH 8.0), 200 mM NaCl, and 5 mM DTT using a GE Healthcare Superdex 200 Increase 10/300 GL size-exclusion column coupled to a Wyatt DAWN HELEOS laser photometer and a Wyatt Optilab T-rEX differential refractometer. Data were analyzed using Wyatt ASTRA 7.3.2 software.
Analytical ultracentrifugation (AUC)
AUC sedimentation velocity experiments were performed to analyze the oligomeric state of NeCaspase and its ligand-bound complex. NeCaspase was diluted to a final concentration of 25 mM in buffer containing 25 mM Tris-HCl (pH 7.4) and 200 mM NaCl. Where indicated, N7-cADPR was added at a molar ratio of 5:1 relative to NeCaspase. Samples were loaded into double-sector Epon centerpieces and centrifuged at 42,000 rpm at 20 °C using a Beckman Optima XL-A analytical ultracentrifuge equipped with an An-50 Ti rotor.
Sedimentation profiles were continuously recorded at 280 nm. The data were analyzed using the continuous c(s) distribution model in SEDFIT to determine sedimentation coefficients and assess the oligomerization state. Figures were generated using GUSSI.
Effect of N7-cADPR on inhibitor binding and activity of PsCaspase
To investigate whether inhibitor binding to PsCaspase requires prior activation by N7-cADPR, PsCaspase was incubated under three conditions: (1) PsCaspase + N7-cADPR + inhibitor, (2) PsCaspase + inhibitor. All components were mixed at a 1:1:1 molar ratio in a total volume of 20 μL using buffer containing 25 mM Tris (pH 7.4 and 500 mM NaCl, with PsCaspase at a concentration of 50 μM. After incubation, samples were either directly subjected to activity assays or further processed by buffer exchange.
Enzymatic activity was measured immediately after incubation by adding N7-cADPR together with the fluorogenic substrate AMC-EEKAR. Under these conditions, PsCaspase alone exhibited robust enzymatic activity, whereas PsCaspase pre-incubated with the inhibitor, either in the presence or absence of N7-cADPR, showed minimal activity.
The same samples were subjected to buffer exchange using 3-kDa molecular weight cutoff centrifugal filters. For each wash, 200 μL of buffer was added, and samples were washed six times to remove unbound small molecules. Following ultrafiltration, the retained protein was concentrated back to a final volume of 20 μL prior to activity measurements. After washing, PsCaspase pre-incubated with both N7-cADPR and inhibitor remained inactive, whereas PsCaspase pre-incubated with inhibitor alone or without any ligand displayed comparable enzymatic activity.
Phylogenetic analysis
To preliminarily estimate the occurrence frequency of defense system in microbial genomes, we downloaded bacterial genomic data from NCBI in 2020 and constructed a representative dataset by selecting one genome per species (n = 45,248 genomes). Protein sequences were aligned using MAFFT (version v7.487) and then converted into HMM profiles separately. HMMER (version 3.4) was employed to search each protein in the representative dataset against all HMM profiles, applying strict filtering parameters (expect threshold of 1e‒8) to minimize the inclusion of unrelated proteins. The resulting alignments were utilized to create a new set of signature gene profiles for the second iteration. Caspase-3 (NP_001341706.1) was used as an outgroup.
Bioinformatic analysis
Structures were predicted using AlphaFold3 Server. Domain analysis was performed using HHpred (toolkit.tuebingen.mpg.de/tools/hhpred).
In vivo cleavage assays and mass spectrometry
E. coli BW25113 expressing NeTIR-Caspase or empty vector was inoculated into 200 mL LB medium (50 μg/mL kanamycin and 0.5 mM IPTG) and cultured on a shaker at 37 °C. Cultures were grown to an optical density (OD600) of 0.6‒0.8 and then phage EP02SG (MOI = 2) was added. After 30 min, cells were collected by centrifugation (2,350× g, 30 min, 4 °C) and resuspended in 2 mL of PBS buffer (200 mM, pH7.4). The cells were then sonicated on ice and after centrifuged (12,000× g, 20 min, 4 °C) to obtain the supernatant. For MS analysis, E. coli cells were sonicated on ice and after centrifuged (12,000× g, 20 min, 4 °C) to obtain the proteins. After reduction and alkylation, the proteins were labeled with TMT tag according to the manufacturer’s procedures (ThermoFisher Scientific). and proteolytic digestion was performed with trypsin (Roche) with a 1:50 enzyme-to-protein ratio at 37 °C overnight. Digested samples were incubated with NHS-Activated Agarose (ThermoFisher Scientific) for 2 h at room temperature to enrichment the unlabeled N-terminal peptides. Then the agarose resin was transferred to 10-kDa Nanosep membrane (Pall Corporation, Puerto Rico, USA) centrifuging for 3 min at 12,000 rpm to collect the TMT-labelling N-terminal peptides. The collected components were vacuum-dried after desalted. The dried peptides samples were resuspended in a solvent of 0.1% formic acid (v/v) and subjected to an EASY-nLC 1200 interfaced via a Nanospray Flex ion source to an Orbitrap Exploris 240 mass spectrometer (Thermo Scientific). Samples were loaded onto a C18 trap column (3 μm particles, 150 μm ID, 3 cm length, Dr. Maisch GmbH) and separated using a C18 analytical column (1.9 μm particles, 150 μm ID, 20 cm length, Dr. Maisch GmbH) at a flow rate of 0.5 μL/min with a 100 min gradient composed of water containing 0.1% formic acid (solvent A) and 80% acetonitrile containing 0.1% formic acid (solvent B). The gradient was 5–9% B for 4 min, 9–32% B for 85min, 32–100% B for 11 min. Data were acquired in data dependent acquisition (DDA) mode, using a 2-seconds cycle time method. MS1 resolution was set to 60,000 (at 200 m/z), mass range of 300‒1500 m/z, normalized AGC target of standard and custom maximum injection time. for MS2 the first mass (m/z) was set as 105 and the resolution was set to 15,000 quadrupole isolation 1.4 m/z, normalized AGC target of 200 m/z, dynamic exclusion of 40 s and maximum injection set to custom. Raw data were processed with the Mascot search engine (v.3.0.0, 2024, matrixscience.com; Matrix Science Ltd) Peptide sequences were searched against a protein database generated from E. coli proteome downloaded from Uniprot. Quantitative comparisons were calculated based the quantification function of Mascot Demon for TMT 6Plex. Across all proteins, PsCaspase cleavage sites were selected based on the following criterion: > 2.5-fold median enrichment of peptides starting at the next position. These criteria selected 55 cleavage sites, 46 of which (83.6%) were found at arginine positions. These 55 sites were used to compute a sequence logo using WebLogo3.
Cryo-EM sample preparation and data acquisition
4 μL aliquots of protein at a concentration of 1.2 mg/mL were applied to discharged 300-mesh Quantifoil R1.2/1.3 grids (Quantifoil, Micro Tools GmbH). Grids were blotted for 7 s and plunged into liquid ethane using an FEI Mark IV Vitrobot operated at 8 °C and 100% humidity. PsCaspase in three states followed the same protocol.
Cryo-EM data of
PsCaspase in the
apo paired form were acquired on a 200 kV Glacios2 microscope equipped with a Falcon 4i direct electron detector. The defocus values ranged from ‒1.0 µm to ‒2.0 µm. Images were recorded automatically using E Pluribus Unum (EPU) software at a binned pixel size of 0.740 Å, with a total exposure of about 50 e
−/Å
2 fractionated over 40 frames. Frame alignment and dose-weighted summation were performed using the patch motion correction in CryoSPARC
32.
Cryo-EM data for PsCaspase in the apo form, PsCaspase in complex with N7-cADPR, and PsCaspase in complex with N7-cADPR and inhibitor were collected on a 300 kV Titan Krios G4 equipped with a Gatan K3 detector and a GIF Quantum energy filter (slit width of 20 eV). Defocus values ranged from ‒0.8 µm to ‒1.8 µm. Images were recorded automatically in super-resolution counting mode using EPU software. PsCaspase in the apo form was recorded at a binned pixel size of 0.844 Å, PsCaspase in complex with N7-cADPR, and PsCaspase in complex with N7-cADPR and inhibitor at a binned pixel size of 0.660 Å. All three data sets were collected with a total dose of about 50 e−/Å2 over 32 frames. Frame alignment and dose-weighted summation were performed using the patch motion correction in CryoSPARC.
Cryo-EM data processing
The cryo-EM data processing workflow is outlined in Supplementary Fig. S1. All cryo-EM data were processed using CryoSPARC, with patch-CTF employed to estimate the contrast transfer function (CTF) parameters and the blob picker used for particle picking. A total of 1,734,613/ 2,303,669/1,043,446 particles were extracted from 1,203/5,250/2,078 micrographs for PsCaspase in the apo form, PsCaspase in complex with N7-cADPR, and PsCaspase in complex with N7-cADPR and inhibitor, respectively (unless otherwise indicated, all subsequent occurrences of the three sets of numbers refer to them in the order listed above). After multi-2D classification, 521,572/240,890/247,307 particles were selected for further processing.
Subsequent multi-class ab initio reconstruction and heterogeneous refinement yielded high-quality models and their corresponding particle sets. These models were further refined using non-uniform refinement, resulting in final cryo-EM maps of PsCaspase in the three respective states at resolutions of 2.58 Å (from 229,477 particles), 2.60 Å (from 112,822 particles), and 2.21 Å (from 162,289 particles). The final reconstructions exhibited C1 symmetry, and resolution was estimated using the gold-standard Fourier shell correlation (FSC) with a 0.143 criterion. Data collection, refinement, and validation statistics are provided in Supplementary Table S3.
Cryo-EM data processing for PsCaspase in the apo paired form follows a similar protocol, yielding 12,360 final particles selected from 5,965 micrographs and a 6.72 Å resolution map reconstructed with C2 symmetry imposed.
Cryo-EM model building and refinement
The initial model of
PsCaspase in
apo form was derived using AlphaFold3 and served as the starting model for building the other two states. Predicted models were fitted into the cryo-EM density maps using UCSF ChimeraX
33. Model building was further improved in Coot
34. Final models for all datasets were refined against their respective maps in real space using PHENIX, with secondary structure and geometry restraints applied. Model validation was carried out by assessing clash scores, MolProbity scores, and Ramachandran plot statistics, as reported by PHENIX
35. All the figures were created in UCSF ChimeraX.
Negative-Stain electron microscopy sample preparation
BW25113 cells were transformed with either pET28a-Tac-PsCaspase or pET28a-Tac-PsTIR-PsCaspase plasmids. For each construct, 50 mL cultures were prepared, with two parallel cultures set up for the pET28a-Tac-PsTIR-PsCaspase group. When the cultures reached an OD600 of 0.6, IPTG was added to a final concentration of 0.2 mM, followed by induction at 37 °C for 2 h. Subsequently, one of the pET28a-Tac-PsTIR-PsCaspase cultures was infected with T6 phage at an MOI of 2 and further incubated at 37 °C for 30 min. The remaining two cultures were incubated under the same conditions without phage infection. After incubation, all three cultures were adjusted to an OD600 of 0.6, and 50 mL of each culture was harvested by centrifugation. Cell pellets were resuspended in 1 mL buffer containing 10 mM Tris-HCl (pH 8.0), 500 mM NaCl, and 5 mM DTT. Cells were lysed by sonication and centrifuged at 13,000 rpm for 30 min. The supernatants were collected for subsequent negative-staining electron microscopy sample preparation.
The pET28a-His6-SUMO-PsCaspase plasmid was transformed into BL21(DE3) cells for large-scale protein expression. Cell pellets obtained from 2 L bacterial culture were resuspended in 200 mL lysis buffer containing 50 mM Tris-HCl (pH 8.0), 500 mM NaCl, and 30 mM imidazole, followed by cell disruption. The lysate was clarified by centrifugation, and the supernatant was collected. Aliquots of clarified lysate (2, 4, 8, 16, and 32 mL) were individually loaded onto Ni-NTA affinity columns. After extensive washing to remove nonspecifically bound proteins, target proteins were eluted with 2 mL elution buffer containing 50 mM Tris-HCl (pH 8.0), 500 mM NaCl, and 300 mM imidazole. ULP1 protease was added to the eluted fractions for tag cleavage, with the 32 mL loading group receiving a two-fold amount of ULP1. After digestion, 1 mL of each sample was directly subjected to size-exclusion chromatography (SEC) using a Superose6 column. Based on the SEC profile of the 32 mL loading group, fractions corresponding to the PsCaspase-containing peak were collected. For the remaining samples, fractions within the same elution volume range were collected regardless of whether the target protein was detected.
Negative-stain electron microscopy
Negative-stain specimens were prepared by applying 5 μL of sample to glow-discharged 200-mesh copper grids coated with an ultrathin carbon support film. After incubation for 30 s, excess solution was removed with filter paper, and the grids were immediately stained with 3 μL of 1% (w/v) uranyl acetate for 30 s. The staining step was repeated once, after which excess stain was removed and the grids were air-dried at room temperature.
Negative-stain electron microscopy data were collected on an FEI Tecnai Spirit equipped with an iCorr and operated at 120 kV, with a defocus range of ‒2 to ‒3 μm. Low-magnification micrographs were recorded at 1,100×, and high-magnification micrographs were collected at 67,000×.
To assess the relationship between protein concentration and right-handed helical filament formation, protein samples at a series of concentrations (0.08 μM, 0.54 μM, 1.35 μM, 2.43 μM, and 6.21 μM) were obtained by direct purification and were subjected to negative-stain grid preparation and imaging at high magnification.
To monitor the time-dependent formation of N7-cADPR-induced left-handed double-helical filaments, N7-cADPR was added at a fivefold molar excess relative to protein (0.68 μM final concentration). Samples were collected at 0 min (without N7-cADPR addition) and 2, 10, 20, and 40 min after ligand addition, followed by negative-stain grid preparation and imaging at high magnification.
For negative-stain characterization of cell lysates, lysates from each condition were prepared for negative staining and imaged at both low and high magnifications.
Quantification and statistical analysis
The number of replicates is specified in the associated figure legends. Each replicate represents a biological replicate of the specified experiment. Two-tailed Student t-test was performed for statistical analyses. P values above 0.05 were considered non-significant.
DATA AVAILABILITY
Cryo-EM density maps have been deposited in the Electron Microscopy Data Bank (EMDB), and the corresponding atomic coordinates in the Protein Data Bank (PDB), with the following accession codes: EMD-67943 and PDB 21RJ (PsCaspase in the apo form); EMD-67942 and PDB 21RH (PsCaspase in complex with N7-cADPR); EMD-67944 and PDB 21RK (PsCaspase in complex with N7-cADPR and inhibitor). The cryo-EM density map of PsCaspase in the apo paired form has been deposited in the EMDB under accession code EMD-68137. This paper does not report original code.
The Author(s) 2026. Published by Higher Education Press. This is an Open Access article distributed under the terms of the CC BY license (https://creativecommons.org/licenses/by/4.0/).
