Structural and functional insights into the adenosine deaminase of the type III-B CRISPR–Cas system
Zhaoxing Li, Jianping Kong, Wanqian Wu, Yan Duan, Ziyi Zhu, Chenyang Hua, Purui Yan, Chen Cao, Xu Cao, Yibei Xiao, Meiling Lu, Meirong Chen
TL;DR
This study reveals how two CRISPR effectors, CAAD and Csx1, work together in a bacterial immune system to fight viral infections using a signaling molecule called cOA.
Contribution
The paper introduces a dual-targeting mechanism involving CAAD and Csx1 for coordinated antiviral defense in type III-B CRISPR systems.
Findings
CAAD and Csx1 are co-activated by cOA to degrade viral RNA and disrupt nucleotide metabolism.
cA4/cA6 binding stabilizes CAAD hexamers and triggers ATP deamination.
Both CAAD and Csx1 can degrade cA4, enabling cross-regulation and immune response termination.
Abstract
Type III CRISPR–Cas (Clustered Regularly Interspaced Short Palindromic Repeats and CRISPR-associated proteins) systems confer antiviral immunity via cyclic oligoadenylate (cOA) signaling. Here, we elucidate a cooperative bacterial defense strategy involving two cOA-activated CRISPR-associated Rossmann fold (CARF)-containing effectors, adenosine deaminase CAAD and ribonuclease Csx1, in Thermoanaerobaculum aquaticum. Genomic analyses indicate widespread co-occurrence of CRISPR-associated adenosine deaminase (CAAD) with ancillary CARF-containing effectors in type III CRISPR systems, suggesting that multiple CARF-containing proteins may contribute to a coordinated cOA-dependent defense. Biochemical and structural studies reveal the intrinsic dynamics of CAAD hexamer, and demonstrate that cA4/cA6 binding stabilizes CAAD hexamers, triggering metal-ion-dependent conversion of ATP into inosine…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5- —National Key Research and Development Program of China10.13039/501100012166
- —National Natural Science Foundation of China10.13039/501100001809
- —Basic Research Program of Jiangsu
- —State Key Laboratory of Natural Medicines10.13039/501100011360
- —China Pharmaceutical University10.13039/501100002857
- —Fundamental Research Funds for the Central Universities10.13039/501100012226
- —National Key Research and Development Program of China10.13039/501100012166
- —National Natural Science Foundation of China10.13039/501100001809
- —Basic Research Program of Jiangsu
- —Project Program of State Key Laboratory of Natural Medicines
- —China Pharmaceutical University10.13039/501100002857
- —Fundamental Research Funds for the Central Universities10.13039/501100012226
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsCRISPR and Genetic Engineering · interferon and immune responses · RNA regulation and disease
Introduction
CRISPR–Cas (Clustered Regularly Interspaced Short Palindromic Repeats and CRISPR-associated proteins) systems equip prokaryotes with an adaptive immunity against mobile genetic elements such as invading plasmids and viruses, via an interference mechanism [1–3]. Based on the divergence of the Cas nucleases, CRISPR–Cas systems are classified into types I, II, III, IV, etc [4, 5]. In recent years, type III CRISPR–Cas systems have gained considerable attention. Recent studies have updated the classification of these systems, grouping them into subtypes III-A to III-I based on the evolutionary analysis of Cas10 phylogeny, gene loss, and fusion events [6–8]. In recent years, studies have extensively explored the Type III-B CRISPR–Cas system, unraveling its intricate machinery and RNA-guided DNA cleavage mechanism. Type III CRISPR–Cas systems in prokaryotes provide immunity against invading nucleic acids by coordinated degradation of transcriptionally active DNA and its transcripts by the Csm or Cmr effector complex [9–15]. The Cas10 subunit (named Csm1 and Cmr2 in the III-A and III-B systems, respectively) in addition to utilizing the traditional N-terminal HD nuclease domain for non-specific ssDNA cleavage when present, more commonly employs the Palm polymerase domain to generate a range of nucleotide second messengers, exemplified by cyclic oligoadenylates (cOAn, n = 2–6), which activate a diverse family of effector proteins to defend against mobile genetic elements, including plasmids, archaeal viruses, and phages [7, 15–19].
The CARF (CRISPR-associated Rossmann fold) domain can bind cOAs synthesized by the CSM or CMR complexes and is widely present in type III CRISPR systems [20, 21]. CARF domain-containing proteins encompass diverse functional types, including nucleases (e.g. Csx1/Csm6) [22–25], membrane proteins (e.g. Csx23/CorA) [26–29], NADases (e.g. Cat1) [30], and other effectors [31–34]. Beyond their role in effector activation, accumulating evidence indicates that CARF domain–containing proteins also participate in the regulation of cOA signaling. Several Csx1 and Csm6 proteins have been shown to combine cOA sensing with intrinsic ring nuclease activity, thereby coupling immune activation to signal termination [22, 35, 36]. This regulatory capacity is thought to be important for preventing prolonged immune activation and collateral damage to the host. The CRISPR-associated adenosine deaminase (CAAD) protein, a key component in the type III-B CRISPR–Cas system, has been reported to exhibit deaminase activity [33, 34]. However, its mechanistic role in CRISPR interference and functional interplay with other effectors remains poorly understood, motivating further investigation in this study. Understanding the molecular basis of CAAD’s activity and its relationship with the RNA-guided complex is crucial for harnessing a full comprehension of the mechanisms of immunoregulation mediated by the type III-B CRISPR–Cas system.
Among CARF domain-containing proteins, the most extensively studied are the nucleases Csx1 and Csm6 [22–25, 35, 36]. Both contain CARF and HEPN domains, and function as ribonucleases in type III CRISPR–Cas systems through their non-specific ssRNase activity. Previous studies have demonstrated that these enzymes play crucial regulatory roles in antiviral defense by processing invading nucleic acids.
In this study, we investigated the mechanism of CARF-containing deaminase CAAD in type III CRISPR systems and the collaborative effects of multiple CARF-containing effectors. While previous studies established that type III CRISPR systems often associate a single CARF-containing effector (e.g. CAAD with a Nudix hydrolase) to mediate abortive infection [33, 34], our work reveals an expanded strategy wherein two distinct CARF proteins—a deaminase (CAAD) and a nuclease (Csx1)—are convergently recruited. This dual-effector architecture enables simultaneous targeting of nucleotide metabolism and viral nucleic acids, thereby enforcing a more robust antiviral state. Overall, our study elucidates the molecular mechanism of CAAD and proposes novel insights into the functional modes of multiple effectors.
Materials and methods
Plasmid and strain construction
pPS22 (pET23a-Cas10/Csm/csm2H6N) was purchased from Addgene (https://www.addgene.org/70038/) for the expression of Csm ribonucleoprotein complexes [9]. Full-length CAAD and Csx1 of *Thermoanaerobaculum aquaticum (*NCBI reference sequence: JMFG01000000) were synthesized and codon-optimized for expression in Escherichia coli by Genscript. The genes encoding TaqCAAD and TaqCsx1 were cloned into the pET28a vector with a C-terminal His6 tag. All point mutations were introduced by PCR-based mutagenesis using the wild-type constructs as templates. The target or nontarget of the Csm complexes, was inserted into the cloning site of pCDFDuet vector, designated as pCDFDuet-Target and pCDFDuet-Non-Target. For the toxicity assay in E. coli, the plasmids pRSFDuet-CAAD-Csx1, pRSFDuet-CAADH^461A−^Csx1, pRSFDuet-CAAD-Csx1H^409A^, and pRSFDuet-CAADH^461A−^Csx1H^409A^ were constructed for coexpression of CAAD and Csx1 or their variants. All obtained plasmids were verified by DNA sequencing.
Plasmid challenge assays
pPS22 (Cas10^H14A, D15A^), pCDFDuet-Target (or pCDFDuet-NonTarget), and pRSFDuet-CAAD-Csx1 (or pRSFDuet-CAAD^H461A^-Csx1, pRSFDuet-CAAD-Csx1^H409A^, and pRSFDuet-CAAD^H461A^-Csx1^H409A^) were co-transformed into E. coli BL21 (DE3) cells. Single colonies were picked into 1 ml of LB medium supplemented with 100 μg/ml ampicillin, 50 μg/ml kanamycin, and 50 μg/ml streptomycin and grown at 220 rpm at 37°C for 2 h, then induced with 0.2 mM IPTG at 37°C for 4 h. Next, 2.5 μl of a 10-fold serial dilution was applied in duplicate to LB agar plates supplemented with 100 μg/ml ampicillin, 50 μg/ml kanamycin, and 50 μg/ml streptomycin. Plates were incubated at 37°C overnight. The experiment was performed as two independent trials, each with three biological replicates and at least two technical replicates.
Protein expression and purification
Sequence-verified plasmids were transformed into E. coli BL21 (DE3) cells. The cell cultures were grown at 37°C in Lysogeny broth (LB) medium until the optical density at 600 nm reached 0.6–0.8, then induced with 0.5 mM Isopropyl β-d-1-Thiogalactopyranoside (IPTG) overnight at 25°C. Cells were collected by centrifugation at 4000 rpm for 10 min and lysed by high-pressure homogenizer in lysis buffer containing 20 mM HEPES pH 7.5, 20 mM imidazole and 500 mM NaCl. The lysate was centrifuged at 18 000 rpm for 30 min twice at 4°C, and the supernatant was applied onto the pre-equilibrated Ni-NTA column (GE Healthcare). After washing the column with 200 ml buffer containing 20 mM HEPES pH 7.5, 500 mM NaCl, the bound protein was eluted with buffer containing 20 mM HEPES pH 7.5, 500 mM NaCl, and 300 mM imidazole and further purified using a HiLoad 16/600 Superdex 200 column (GE Healthcare) equilibrated in 20 mM HEPES pH 7.5 and 500 mM NaCl. The corresponding peak fractions containing pure proteins were pooled and concentrated to >10 mg/ml, snap-frozen in liquid nitrogen, and stored at −80°C for later use. All mutants’ purification methods were the same as the wild type.
SEC analysis
The oligomeric states of the full-length TaqCAAD protein and its mutants were determined by SEC analysis. The pure protein was loaded onto the Superdex 200 Increase 10/300 GL pre-equilibrated with buffer (20 mM HEPES pH 7.5, 500 mM NaCl). Furthermore, the same amount of apo-TaqCAAD protein was mixed with cA_3_, cA_4_, or cA_6_ in 1:1.2 (TaqCAAD–cOAs) molar ratio and incubated at 4°C for 10 min. Absorption at 280 and 260 nm was monitored. The data were analyzed using GraphPad Prism 10.4.
Analysis of deamination reactions
For cOAs screening, 50 nM TaqCAAD and 500 nM cA_4_/cA_6_ were incubated in an enzymatic reaction buffer containing 150 mM NaCl, 20 mM HEPES pH 7.5, and 5 mM MgCl_2_. For substrate screening, the reactions contained 50 nM TaqCAAD, 500 nM cA_4_ or cA_6_, and 200 μM final concentration of each substrate (ATP, dATP, AMP, adenosine, UTP, CTP, GTP). All reactions were incubated at 37°C for 60 min and quenched with 100 μl acetonitrile. After centrifugation at 15 000 rpm for 10 min, the reaction products were passed through a 0.22-μm filter for high-performance liquid chromatography (HPLC) analysis.
cOAs cleavage assay
For the cOAs cleavage assay, 50 μM TaqCAAD or TaqCsx1 and 50 μM cA_4_/cA_6_ were incubated in a buffer containing 150 mM NaCl, 20 mM HEPES pH 7.5, and 5 mM MgCl_2_ at 37°C for 1 h. The reactions were then stopped by the addition of acetonitrile.
HPLC analysis
The samples were analyzed using an Agilent 1260 Infinity HPLC system (Agilent Technologies, USA) or Shimadzu LC-40 Nexera HPLC system (Shimadzu Corporation, Japan) equipped with an Elite C_18_ column (4.6 mm × 250 mm, 5 μm) or WELCH Ultimate Polar RP C18 column (4.6 mm × 250 mm, 5 μm). The mobile phases were freshly prepared, where eluent A was acetonitrile and eluent B was 0.1% ammonium acetate aqueous solution. The separation was carried out at a flow rate of 1.0 ml/min, employing a gradient program starting with 1.5% eluent A and linearly increasing to 10% eluent A over 30 min. The column temperature was set at 25°C, the detection wavelength was 254 nm, and the injection volume was 10 μl per sample.
Mass spectrometry analysis
LC-MS analysis was performed using a Q-TOF mass spectrometer (Agilent Technologies, USA) equipped with an electrospray ionization source and a UPLC system (Agilent 1290 UPLC). Samples were analyzed on an Agilent ZORBAX Eclipse Plus C_18_ column (3.0 mm × 100 mm, 1.8 μm). The mobile phases consisted of eluent A (acetonitrile) and eluent B (0.1% formic acid aqueous solution). The flow rate was set at 0.4 ml/min with a linear gradient program starting at 1.5% eluent A and increasing to 10% eluent A over 10 min. The column temperature was maintained at 25°C, and the injection volume was 1.0 μl per sample. Both positive and negative ion modes were employed, with calibration of accurate mass performed prior to sample analysis. The mass scan range was set at m/z 100–1500. N_2_ was used as the desolvation gas, with a flow rate of 6.8 l/min and a voltage of 4.0 kV.
RNA cleavage assays
5′ FAM-labeled single-stranded RNA (ssRNA) oligonucleotides were purchased from GenScript (AGAUAGAUGUAAUUCCGGUGUAGA and CCCAAAAAGCUAAUACAGUAAACC). 5′ FAM-labeled ssRNA (100 nM) was incubated with 50 nM Csx1 supplemented with 200 nM cA_4_ in the buffer (20 mM HEPES pH 6.5, 150 mM NaCl). The reactions were performed at 37°C for 30 min and quenched by adding 2× RNA loading dye under 95°C for 5 min. The conditions are the same with cA_6_ and cA_3_ except that Csx1 was 500 nM and cA_6_ or cA_3_ was 5 µM. All the samples were analyzed by 20% Urea-PAGE (20% acrylamide, 6 M urea, and 1× TBE) and scanned using Tanon MINI SPACE 3000. To illustrate the cleavage site, the marker sample was formed by using nuclease P1 (NEB) to digest the ssRNA at 37°C for 10 s, followed by adding 0.1 M ethylenediaminetetraacetic acid to stop digestion.
For the pre-incubation assay, 200 nM cA_4_ were pre-incubated with TaqCsx1 or TaqCAAD, followed by addition of Csx1 (in the CAAD pre-incubation assay) and RNA substrate to initiate the reaction.
Microscale thermophoresis
The binding affinity between wild-type CAAD or Csx1 with cOAs was determined by microscale thermophoresis (MST). CAAD or Csx1 (10 μM) was fluorescently labeled in 2× NHS binding buffer using 30 μM RED-NHS labeling dye (Nanotemper Technologies) and incubated in the dark at room temperature for 30 min. cOAs were serially diluted in a 1:1 ratio with PBS. The labeled proteins were then mixed with cOA in a 1:1 ratio. The supernatants were then loaded into Monolith NT.115 Series K022 Premium capillaries in triplicate, and the thermophoresis was detected with 100% excitation power and 60% IR-laser power for an on-time of 20 s at 25°C. The binding affinities of cOAs to CAAD or Csx1 were analyzed according to the law of mass action in a standard fitting mode of MO.Affinity analysis software (version 2.2.4).
Electron microscopy sample preparation and data collection
An aliquot of 3.5 μl of purified CAAD at a concentration of 2.5 mg/ml, with 2 mM cA_6_, 2 mM ATP, and 2 mM MgCl_2_ was applied to glow-discharged Quantifoil holey gold grids (R 1.2/1.3, Au 300 mesh), followed by a 15 s wait time and 4.5 s blot time with the blot force set to 2. The grids were then plunge frozen in liquid ethane using a Virtrobot Mark IV (Thermo Fisher Scientific). The Vitrobot chamber was maintained at ∼100% humidity and 16°C. Cryo-EM images were collected manually on an FEI Titan Krios G3i (Thermo Fisher Scientific) operated at 300 kV and equipped with a Falcon 4i detector. All cryo-EM movies were recorded in counting mode using EPU. The detailed parameters, including electron dose, pixel size, and magnifications of electron microscopy data collection, are listed in Supplementary Table S2.
EM data processing
Images were processed in cryoSPARC (v.4.2) [37]. Movie frames were aligned and summed using the patch motion module. Contrast transfer function (CTF) parameters were estimated for individual particles on each micrograph using the patch CTF module. After patch CTF estimation, micrographs with resolution estimates worse than 6 Å were discarded. The blob picker was used to pick particles with a circular diameter ranging from 120 to 280 Å. Particle picks were inspected, and particles with NCC scores below 0.25 were discarded. 2D classification, 3D classification, and 3D refinement were also performed using cryoSPARC. All refinements followed the gold-standard procedure, in which two half-datasets were refined independently. Overall resolutions were estimated based on the gold-standard criterion of Fourier shell correlation (FSC) = 0.143. Local resolutions were calculated in cryoSPARC and visualized using ChimeraX [38]. Reconstructions were sharpened and filtered to overall resolution by using DeepEMhancer with the default tight mask preset [39].
Model building
Structures of the TaqCAAD were individually predicted using AlphaFold3 [40]. The model was rigid-body fitted into the density map using UCSF ChimeraX and manually rebuilt in Coot [41]. Amino acid residues were mutated to reflect the correct sequence of the construct. The models were subsequently refined with phenix.real_space_refine [42]. The quality of the structural model was assessed using the MolProbity program in Phenix [43].
Results
Adenosine deaminase cooperates with a CARF-containing nuclease to enforce multifaceted CRISPR immunity
Expanding our prior bioinformatic interrogation of 121 type III CAAD systems, we observed that 76% of these systems are genetically linked to one or more downstream genes encoding CARF-containing effector proteins, including Csm6, Card1, and Csx1 (Fig. 1A and Supplementary Table S1). Such genomic architectures imply potential synergistic coordination between CAAD-mediated ATP deamination and ancillary CARF-driven immune responses (e.g. RNA degradation, transcriptional silencing). We focused on the CAAD system coupled with ribonuclease Csx1 from Thermoanaerobaculum aquaticum (Fig. 1B). In this system, downstream of the cas10-cmr genes within a type III-B CRISPR locus lie two proteins fused to CARF effector-binding domains: CAAD, which is fused to an adenosine deaminase (ADA) domain, and Csx1, which is fused to a higher eukaryotes and prokaryotes nucleotide-binding (HEPN) domain.
CAAD and CARF effectors synergize in type III CRISPR immunity. (A) Prevalence of ancillary CARF effectors co-occurring with CAAD in type III CRISPR systems. (B) Gene loci of deaminase-containing type III CRISPR–Cas from Thermoanaerobaculum aquaticum and domain architecture of TaqCAAD and TaqCsx1. (C) Plasmid challenge assay of E. coli BL21 carrying plasmids encoding Csm, CAAD, and Csx1 or its variant. The Csm system utilized in this study was derived from S. epidermidis, distinguishing it in origin from both the TaqCAAD and TaqCsx1.
To investigate the activity and function of CAAD and Csx1 in vivo, we co-transformed E. coli with plasmids encoding the type III-A Csm complex from S. epidermidis for cOA synthesis, and CAAD and Csx1, both derived from T. aquaticum [9]. Co-expressed wild-type CAAD and Csx1 showed potent target-dependent bacterial growth suppression. This inhibitory activity could be progressively abolished through active-site mutations (Fig. 1C). These findings suggest an additive immune response of CAAD and Csx1 in type III CRISPR–Cas-mediated antiviral immunity, potentially allowing hosts to dynamically adjust antiviral responses based on infection severity, as also discussed in previous studies on CARF proteins and cOA signaling [19, 44, 45].
TaqCAAD exhibits cA4/cA6-dependent oligomerization and stringent catalytic specificity
Next, we further characterized the deaminase activity of TaqCAAD. Consistent with previous studies [27, 28], cA_4_ or cA_6_, but not cA_3_, could activate TaqCAAD to efficiently convert ATP into ITP (Fig. 2C and Supplementary Fig. S2A). Quantitative binding analyses via MST confirmed high-affinity interactions between TaqCAAD and cA_4_ or cA_6_. TaqCAAD (500 nM) exhibited high-affinity binding to cA_4_ (K_D_ = 82.4 ± 0.4 nM) and cA_6_ (K_D_ = 370.4 ± 0.2 nM) (Fig. 2B). In contrast, cA_3_ showed no detectable binding ability to CAAD (Supplementary Fig. S1).
cOAs regulation of CAAD assembly and activity. (A) The effects of different cOA ligands on CAAD oligomeric states were assessed by size-exclusion chromatography. (B) Quantitative binding analyses via MST confirmed high-affinity interactions between TaqCAAD and cA4 or cA6. KD are 82.4 ± 0.4 nM and 370.4 ± 0.2 nM, respectively. (C) HPLC profiles of ATP incubated with CAAD in the absence or presence of cA3, cA4, or cA6. (D) HPLC analysis of a reaction incubating cA4/cA6 with TaqCAAD by the TaqCAAD in comparison with cA4/cA6 alone.
We then evaluated the enzymatic properties of TaqCAAD. ITP production by TaqCAAD strictly requires divalent metal ions, such as Mg^2+^, Mn^2+^, and Ca^2+^ (Supplementary Fig. S2B). Although Zn^2+^ is not necessarily supplemented to the reaction, subsequent structural analysis revealed Zn^2+^ binding in the active site of CAAD, which is probably sequestered from the expression culture. Extended substrate profiling revealed that CAAD possesses stringent substrate specificity. Notably, no deamination activity was detectable with adenosine, AMP, dATP, UTP, GTP, or CTP, even in the presence of 20 µM cA_6_ (Supplementary Fig. S2C and D). Moreover, under identical conditions, TaqCAAD selectively cleaved cA_4_ while leaving cA_6_ intact, as demonstrated by HPLC retention time shifts and validated by mass spectrometry (Fig. 2D and Supplementary Fig. S2E). Meanwhile, substitution of K106 with alanine abolished cA_4_ degradation, implicating K106 in the ring nuclease activity of the CARF domain, consistent with observations reported in a recent independent study [46] (Fig. 2D).
Strikingly, the binding of cA_4_/cA_6_ induced stable oligomerization of TaqCAAD. In the apo state, CAAD exhibited a dynamic oligomerization equilibrium between dimeric and hexameric forms on a Superdex 200 Increase 10/300 GL column (Fig. 2A). Whereas supplementation with cA_4_ or cA_6_ yields a stable monodisperse hexameric complex as evidenced by a single symmetric peak at 10 ml retention volume (Fig. 2A). Consistent with the binding affinity as well as the activating ability, this ligand-induced oligomerization was not observed with cA_3_ (Fig. 2A).
Cryo-EM structures reveal dynamic assembly of cA6-bound TaqCAAD hexamer with domain-specific flexibility
Overall structures of TaqCAAD bound with cA6 and ATP
To elucidate the structural basis of cA_6_-dependent activation, we determined cryo-EM structures of wild-type TaqCAAD in complex with cA_6_ and ATP. Initial 2D classification revealed significant structural heterogeneity of TaqCAAD, particularly manifesting as pronounced flexibility in the CARF domains (Supplementary Fig. S3). Iterative 3D classification resolved five structural states of the TaqCAAD hexamer, which predominantly adopts a trimer-of-dimers configuration, with three CARF domain dimers radially associating around a central deaminase core (Fig. 3A and Supplementary Fig. S5A). This overall trimer-of-dimers architecture closely resembles the recently reported CARF-deaminase assemblies of LngCAAD and BbaCAD [33, 34] (Supplementary Fig. S5B and C), and is also reminiscent of the CARF-based nuclease SisCsx1, underscoring a conserved higher-order organization among CARF effector proteins [26] (Supplementary Fig. S5D). Remarkably, the CARF domains exhibit exceptional flexibility, with only one CARF dimer resolved in three reconstructions (States I–III; global resolutions: 2.74, 2.79, and 2.94 Å), whereas two and three CARF dimers were captured in State IV (global resolution: 2.94 Å) and V (global resolution: 2.84 Å), respectively (Supplementary Fig. S4). In contrast, the densities of catalytic deaminase domains can be well-defined across all states. The TaqCAAD monomers are curled around the two-fold axis to form a dimer (Supplementary Fig. S5E). The primary distinction between different states lies in the spatial arrangement of these CARF domains. Comparison of a well-resolved dimer from State I with one from State V revealed that the CARF domain undergoes significant rigid-body rotation relative to the deaminase domain at the dimer interface. Specifically, while the CARF and deaminase domains assume an approximately parallel arrangement in State I, they undergo rigid-body rotation of ~90° to adopt a near-perpendicular orientation (Fig. 3B and Supplementary Fig. S5F and G and Supplementary Video S1). The CARF and deaminase domains within the CAAD dimer exhibit rotational flexibility at the 6H domain, with each dimer containing one cOA-binding pocket and two catalysis pockets (Supplementary Fig. S5E). When aligning the structures of different states by the CARF dimer, State V adopts a more compact conformation, moving closer to the bound cA_6_ (Fig. 3C).
Cryo-EM structure of TaqCAAD. (A) Overall structure of a ligands-bound TaqCAAD hexamer, colored by chain. (B) Superposition of representative dimer from State I and representative dimer from State V. (C) Structural rearrangement of CARF dimers aligned by cA6. State I CARF dimer superimposed with State V CARF dimer. (D) Top view of the TaqCAAD dimer showing the cOAs-binding pocket built by the CARF domains. (E) Amino acid residues of the CAAD hexamer involved in interactions with cA6. The cA6 is enclosed by the density map. Black dashes show hydrogen bond interactions. (F) Sequence alignment of representative CAAD proteins. The residues implicated in the assembly of cA6-binding are marked by cycles, mutants are highlighted in red. (G) HPLC analysis of ATP deamination by CAAD CARF domain representative mutants. (H) Bottom view of the TaqCAAD dimer showing the catalytic pockets built by the CARF domains. (I) Amino acid residues of the CAAD hexamer involved in interactions with ATP. The ATP and metal ions are enclosed by the density map. Black dashes show hydrogen bond interactions. (J) Sequence alignment of representative CAAD proteins. The residues implicated in the assembly of ATP and metal ion binding are marked by cycles, mutants are highlighted in red. (K) HPLC analysis of ATP deamination by CAAD deaminase domain representative mutants. (L) Key interaction sites at the intra-dimer AB interface. (M) Sequence alignment of representative CAAD proteins. Key dimer/hexamer assembly residues are marked by cycles, mutants are highlighted in red. (N) Key interaction sites at the inter-dimer AD interface in hexamer. (O) ATP deamination activity of interface mutants.
Recognition of cOAs by the CARF domain
The CARF domain adopts a canonical Rossmann fold, with a ligand-binding pocket at the dimer interface (Fig. 3D). Structural analysis reveals that cA_6_ binding involves conserved residues (e.g. E17, T35, H126) that engage in hydrogen bonding and electrostatic interactions with the phosphate backbone and adenine bases of cA_6_ (Fig. 3E). Additionally, the T10 forms a hydrogen bond with the phosphate backbone, while Y105 mediates π–π stacking with the adenine base (Fig. 3E). Sequence and structural alignment of the TaqCAAD CARF domain with those of recently reported CAAD homologs revealed a conserved overall fold and binding-pocket architecture, with RMSDs of 0.854 Å over 151 pruned atom pairs for LngCAAD and 1.051 Å over 155 pruned atom pairs for BbaCAD (Supplementary Fig. S5H and I). Superposition of the cA_4_-bound homolog structures onto the TaqCAAD-cA_6_ CARF domain further suggested a largely conserved ligand-binding mode, in which the shared adenines occupy the same core pocket region, whereas the additional two adenines in cA_6_ appear to contribute minimally to specific recognition. Given that the core pocket residues are highly conserved and have been extensively studied, we prioritized residues with comparatively variable conservation in our alignment that are positioned to contact cOAs, including T10 and Y105, for functional interrogation. We generated alanine substitutions at T10 and Y105 and measured their impact on deaminase activity in vitro. HPLC analysis revealed that the T10A mutation minimally affected cA_4_-activated deaminase activity but caused a moderate reduction in cA_6_-activated ATP deamination, whereas the Y105A mutation completely abolished activation by both ligands (Fig. 3F and G). Therefore, Y105 mediates indispensable π–π stacking interactions with the adenine base of cOAs, making this residue essential for ligand recognition and subsequent effector activation.
ATP recognition and catalysis by the deaminase domain
Diffuse density was observed in each deaminase pocket of the cA_6_-CAAD complex, wherein Zn^2+^, Mg^2+^, and ATP were modeled (Fig. 3H and I). Within this deaminase active pocket, Zn²⁺ stabilizes the catalytic site by coordinating H260, H262, and H461. Residues S319, S352, N355, Y356, and T396 engage the phosphate group and ribose moiety of the substrate. Additionally, Mg²⁺ bridges the β- and γ-phosphates of the substrate with G314, forming a metal ion bridge. The 6-amino group of ATP’s adenine is specifically engaged by H485 and D548, which have been documented as critical residues for substrate discrimination and deaminase activity [27, 28]. We introduced alanine substitutions at residues H262, S319, H485, and D548 and assessed their impact on deaminase activity. As expected, all single-point mutants (H262A, S319A, H485A, and D548A) severely impaired CAAD’s deaminase activity, with only trace-level product formation detectable for some variants (Fig. 3J and K).
CAAD hexameric assembly is essential for catalytic activity
Hexameric stability of CAAD involves interactions both within and between dimers. The intradimer interface is extensive, featuring S109–S109 and W167–W167 interactions in the CARF domain (Fig. 3L and M), along with multiple hydrogen bonds within the deaminase domain, including R453–E469, V454–R497, H452–W472, S479–R502, and R481–E501 (Fig. 3L and M). The interdimer interface involves π–π stacking between symmetrically related W364 residues and a hydrogen bond between S362 and H419 (Fig. 3M and N). To dissect the structural determinants of CAAD oligomerization and function, we engineered site-specific mutants at conserved interfacial residues. SEC revealed that the mutant bearing W167A at the dimer interface exhibited heterogeneous oligomeric states, while deaminase activity was reduced but not abolished (Fig. 3O and Supplementary Fig. S6). The S362A/W364A double mutation located at the inter-dimer interface disrupted hexamer formation and exhibited complete loss of activity, establishing that hexameric assembly is strictly required for catalysis (Fig. 3O and Supplementary Fig. S6). These results demonstrate that specific intra- and inter-dimer interactions are essential for maintaining the stable CAAD hexamer. Although the R453A mutant maintained hexameric integrity, it abolished deaminase activity (Fig. 3O and Supplementary Fig. S6), suggesting that hexameric integrity is strictly required but not sufficient for deaminase activity.
cA4 specifically activates TaqCsx1 ribonuclease activity
To determine which cAn could activate the TaqCsx1 RNase activity, we incubated TaqCsx1 and 5′ FAM-labeled 24-mer ssRNA with commercially available chemically synthesized cA_3_, cA_4_, and cA_6_ for 30 min. The ribonuclease activity assay demonstrated that Csx1 requires weakly acidic conditions and exhibits optimal catalytic activity at 50°C, which corresponds to the optimal growth temperature of its host, T. aquaticum (Supplementary Fig. S7A). Only cA_4_ is capable of activating its single-stranded RNA cleavage function (Fig. 4A). Notably, we observed that TaqCsx1 selectively cleaves phosphodiester bonds near uridine residues, exhibiting a strong preference for 5′-U-A-3′ linkages (Fig. 4A and Supplementary Fig. S7B). Cleavage also occurs at 5′-G-U-3′ and 5′-U-C-3′ sites with markedly lower efficiency. Furthermore, MST assay further confirmed that TaqCsx1 is specifically regulated by cA_4_, exhibiting a strong binding affinity exclusively for cA_4_ while showing no detectable interaction with cA_6_ (Fig. 4B and Supplementary Fig. S7C). Interestingly, TaqCsx1 exhibited a strong binding affinity for cA_3_ (224.8 ± 0.5 nM), although cA_3_ failed to activate Csx1 (Supplementary Fig. S7C). This phenomenon has not been previously reported in related studies. We speculate that this may be a result of non-specific binding due to the smaller molecular size of cA_3_, which, however, cannot induce the conformational changes required for enzymatic activation.
Molecular basis of cA4-specific activation in TaqCsx1. (A) cA4-specific activation of TaqCsx1 ribonuclease activity. The reactions were separated using 20% TBE–urea gel. (B) High-affinity cA4 binding to TaqCsx1. KD = 142.6 ± 0.1 nM, measured by MST. (C) HPLC analysis of a reaction incubating cA4/cA6 with TaqCsx1.
Notably, TaqCsx1 exhibited selective cleavage activity toward cA_4._ Consistent with previous reports on Csm6 and Csx1 [18, 19, 22–24, 36], cA_4_ cleavage generated A_4 > p (further processed into A_2 > p), with minor A_4 species also detected (Fig. 4C and Supplementary Fig. S7D). Given the thermophilic origin of TaqCsx1, we examined cA_4 cleavage at elevated temperatures and observed a product distribution similar to that at 37°C, with multiple species detected under both conditions (Supplementary Fig. S7E). Moreover, HPLC analysis of CARF and HEPN variants/truncations showed that these constructs retained cA_4_-processing activity (Supplementary Fig. S7F), suggesting that both the CARF and HEPN modules can contribute to cA_4_ cleavage, likely at distinct scissile sites. The selectivity and cleavage products of TaqCsx1’s cOA processing activity closely resembled those of TaqCAAD, and their genomic co-occurrence suggests potential functional coordination between multiple CARF-containing effectors. To investigate this, we pre-incubated cA_4_ with either Csx1 or CAAD separately and assessed the impact on Csx1’s RNase activity (Supplementary Fig. S7G). The results showed that pre-incubation with either protein reduced Csx1’s RNase activity, though this attenuation was not statistically significant. This attenuation of activity may result from both the cleavage of cA_4_ and competitive binding for cA_4_ by the CARF domains of these two effectors.
Discussion
In this study, we investigated the molecular mechanism of the CARF-containing deaminase CAAD and revealed a cooperative antiviral strategy involving multiple CARF effectors in the type III CRISPR system of Thermoanaerobaculum aquaticum (Fig. 5). Following phage infection, the activated Cmr complex synthesizes cOAs. We demonstrated that both cA_4_ and cA_6_ bind with high affinity to the CARF domain of CAAD, stabilizing its pre-existing hexameric form into a monodisperse, catalytically active complex. This enables CAAD to bind ATP and catalyze its deamination to inosine triphosphate (ITP) with stringent specificity. Catalytic activity strictly requires divalent metal ions and is abolished by mutations in key active-site residues. Cryo-EM structures revealed that cA_6_ binding locks CAAD into a functional hexameric state, wherein conformational flexibility of the CARF domains relative to the catalytic core is constrained. Crucially, hexameric integrity is indispensable for catalysis, as disrupting inter-dimer interfaces dissociated the hexamer and abolished activity. This hexameric trimer-of-dimers architecture shares striking structural similarity with previously reported CAAD/Cad1 family effectors, including LngCAAD [34] and BbaCad1 [33], in which ligand-induced hexamerization and stabilizing inter-dimer interfaces are essential for deaminase activity, as well as with the unrelated CARF-family RNase SisCsx1, where comparable inter-dimer contacts are critical for cooperative cOA-dependent catalysis [26]. Additionally, the allosteric activation of TaqCAAD, in which signal transmission from the CARF domain to the catalytic core is mediated by the 6H domain, closely parallels the mechanism observed in StCsm6, in which cA6 binding drives a ∼60°CARF–HEPN rotation through the 6H solenoid to activate ribonuclease function [22]. These parallels highlight conserved structural and regulatory principles among CARF-family effector nucleases in type III CRISPR systems.
Mechanism of cOA-mediated antiviral defense in Thermoanaerobaculum aquaticum. Following phage infection, the Cmr complex in Thermoanaerobaculum aquaticum synthesizes cOAs, which activate two distinct CARF-containing effectors: CAAD and Csx1. The cOAs, particularly cA4 and cA6, bind to the CARF domain of CAAD, stabilizing its hexameric form and activating its catalytic activity. This allows CAAD to deaminate ATP to ITP, a reaction crucial for metabolic disruption. In parallel, cOAs also activate Csx1, which exhibits specificity for cA4, inducing its RNase activity to cleave ssRNA. Their cooperative action forms a dual-targeting strategy against viral replication, disrupting both nucleotide metabolism and RNA integrity.
Simultaneously, the same type of second messenger (cOAs) differentially activates a second CARF effector from Thermoanaerobaculum aquaticum, the single-stranded RNA-specific nuclease Csx1. We discovered that Csx1 exhibits exquisite specificity for cA_4_, which activates its RNase activity to cleave ssRNA. Intriguingly, while Csx1 binds cA_3_ with high affinity, it fails to induce RNase activation, suggesting that ligand size and geometry are critical for triggering the productive conformational change. Notably, both CAAD and Csx1 possess intrinsic cOA cleavage activity. Similar ring nuclease activities have been reported previously for other CARF domain–containing effectors, including Csx1 and Csm6 proteins [22, 25, 35], and are now recognized as a widespread feature of type III CRISPR systems. Intrinsic cOA degradation by these effectors is thought to contribute to signal attenuation and immune homeostasis.
Unlike the CAAD–Nudix hydrolase pair reported in prior studies [28], which primarily mitigates metabolic toxicity, the CAAD–Csx1 partnership identified here employs a multipronged attack strategy, simultaneously targeting nucleotide metabolism and nucleic acid cleavage. It is speculated that this dual-targeting mechanism enhances immune robustness by compromising distinct viral processes, thereby reducing the likelihood of phage escape. The observed target-dependent growth suppression in E. coli underscores the physiological relevance of this synergy. The differential cOA specificity of these effectors (CAAD activated by cA_4_/cA_6_ versus Csx1 exclusively by cA_4_) indicates that multiple signaling pathways operate within the type III-B defense system. These effectors act through different complementary mechanisms, with Csx1 driving rapid RNA degradation and CAAD inducing ITP-mediated metabolic disruption, together forming a coordinated interference response against invading nucleic acids. Furthermore, both effectors degrade cA_4_, potentially forming a feedback loop to attenuate immune responses once the infection has been cleared—a hypothesis that warrants future investigation.
Notably, recent work has demonstrated that some viral anti-CRISPR proteins (such as AcrIII-1) specifically degrade cA_4_, thereby undermining cA_4_-dependent CRISPR defense [47]. Because AcrIII-1 is widely distributed among mobile genetic elements and exhibits potent ring nuclease activity against cA_4_, the ability of effectors like CAAD to also respond to cA_6_ may represent an evolutionary strategy to circumvent cA_4_-specific anti-CRISPR activity and sustain an effective immune response when cA_4_ signaling is suppressed by viral antagonists. In this context, responsiveness to multiple cyclic oligoadenylate species could provide functional redundancy within the signaling network, reducing the vulnerability of the defense system to single-point viral countermeasures.
The broader occurrence of multiple CARF-containing effectors in type III CRISPR loci raises the possibility that such proteins do not operate in isolation, but instead cooperate through complementary and potentially self-regulating mechanisms. Their distinct ligand specificities suggest that cOAs may function as layered second messengers, fine-tuning the timing and strength of immune activation. This arrangement could provide both robustness and flexibility, as overlapping activities buffer against viral countermeasures while specialized functions constrain different aspects of viral replication. The intrinsic cOA degradation activities observed for both CAAD and Csx1 further imply that effector engagement contributes not only to defense but also to the controlled termination of signaling. Although these scenarios remain speculative, the frequent genomic linkage of multiple CARF effectors suggests that such cooperative architectures may represent an evolutionarily favored strategy to enhance the versatility and precision of type III CRISPR immunity. Future studies will be required to dissect these interactions in greater detail and to determine whether coordinated activation of multiple CARF effectors constitutes a generalizable principle of type III CRISPR immunity.
Supplementary Material
gkag231_Supplemental_Files
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Barrangou R, Fremaux C, Deveau H et al. CRISPR provides acquired resistance against viruses in prokaryotes. Science. 2007;315:1709–12. 10.1126/science.1138140.17379808 · doi ↗ · pubmed ↗
- 2Marraffini LA, Sontheimer EJ. CRISPR interference: r NA-directed adaptive immunity in bacteria and archaea. Nat Rev Genet. 2010;11:181–90. 10.1038/nrg 2749.20125085 PMC 2928866 · doi ↗ · pubmed ↗
- 3Marraffini LA . CRISPR–Cas immunity in prokaryotes. Nature. 2015;526:55–61. 10.1038/nature 15386.26432244 · doi ↗ · pubmed ↗
- 4Makarova KS, Wolf YI, Alkhnbashi OS et al. An updated evolutionary classification of CRISPR–Cas systems. Nat Rev Microbiol. 2015;13:722–36. 10.1038/nrmicro 3569.26411297 PMC 5426118 · doi ↗ · pubmed ↗
- 5Makarova KS, Wolf YI, Iranzo J et al. Evolutionary classification of CRISPR–Cas systems: a burst of class 2 and derived variants. Nat Rev Microbiol. 2020;18:67–83. 10.1038/s 41579-019-0299-x.31857715 PMC 8905525 · doi ↗ · pubmed ↗
- 6Makarova KS, Haft DH, Barrangou R et al. Evolution and classification of the CRISPR–Cas systems. Nat Rev Microbiol. 2011;9:467–77. 10.1038/nrmicro 2577.21552286 PMC 3380444 · doi ↗ · pubmed ↗
- 7Makarova KS, Aravind L, Wolf YI et al. Unification of Cas protein families and a simple scenario for the origin and evolution of CRISPR–Cas systems. Biol Direct. 2011;6:38. 10.1186/1745-6150-6-38.21756346 PMC 3150331 · doi ↗ · pubmed ↗
- 8Makarova KS, Shmakov SA, Wolf YI et al. An updated evolutionary classification of CRISPR–Cas systems including rare variants. Nat Microbiol. 2025;10:3346–61. 10.1038/s 41564-025-02180-8.41198952 PMC 12669027 · doi ↗ · pubmed ↗