Structures and mechanism of E2‐CBASS anti‐phage system

Abstract

Bacteria deploy diverse innate immune systems to combat bacteriophage infections. The cyclic‐oligonucleotide‐based anti‐phage signaling system (CBASS) is a type of innate prokaryotic immune system. CBASS synthesizes cyclic‐oligonucleotide through cGAS/DncV‐like nucleotidyltransferases (CD‐NTases) to activate downstream effectors, which kill bacteriophage‐infected bacteria, thereby stopping phage spread. One major class of CBASS contains a homolog of eukaryotic ubiquitin‐conjugating enzymes, either as an E1‐E2 fusion or a single E2 enzyme. Both enzymes function by regulating CD‐NTase activity. Currently, many structures of CD‐NTases have been reported, but there are only a few reports of structures where CD‐NTases form complexes with the associated E2. In this study, we analyzed the length and classification of the CD‐NTase in two types of type II CBASS—E1E2/JAB‐CBASS and E2‐CBASS. We found that the CD‐NTase in E2‐CBASS is longer and predominantly belongs to clade G. We also present the structure of the SmCdnG‐SmE2 complex with the bound GTP substrate, which indicates the conservation of the donor binding pattern. Interestingly, we discovered that SmCdnG contains a conserved C‐terminal α‐helix and β‐sheet structure, which is uniquely involved in forming a complex with SmE2. We also found that the structure of the E2 protein in the E2‐CBASS system is highly conserved. Altogether, we provide mechanistic insights into the E2‐CBASS system.

Impact statement

Cyclic oligonucleotide‐based anti‐phage signaling system (CBASS) is an important component of prokaryotic innate immunity. This study reveals the structural basis and regulatory mechanism of the E2‐CBASS system, a unique pathway that operates without E1 homologs. Through cryo‐EM analysis and functional validation, we identify critical interactions between CD‐NTase and E2 and uncover a distinct mode of complex assembly. These findings provide valuable insights into CBASS defense systems and expand our understanding of the diversity of prokaryotic immune mechanisms.

INTRODUCTION

Bacteria have evolved multiple innate immune systems to defend against bacteriophage infections ¹ , ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ . Current research has found that many bacterial innate immune systems share a high degree of structural and functional conservation with components of innate immune systems in eukaryotes ⁹ , ¹⁰ . Cyclic GMP–AMP synthase (cGAS) has a key role in mammalian cGAS–STING innate immunity ¹¹ , ¹² . Interestingly, the Vibrio cholerae enzyme dinucleotide cyclase (DncV) has a highly conserved catalytic domain similar to cGAS ¹³ , ¹⁴ , ¹⁵ , ¹⁶ ; therefore, this family of enzymes has been classified as cGAS/DncV‐like nucleotidyltransferases (CD‐NTases) ¹⁷ , ¹⁸ . In bacteria, CD‐NTases are involved in cyclic‐oligonucleotide‐based anti‐phage signaling system (CBASS), which stop phage spread ¹⁴ . In CBASS, CD‐NTase synthesizes cyclic oligonucleotides that activate downstream effectors, killing phage‐infected bacteria. This prevents phage proliferation and protects uninfected bacteria.

Besides the cyclase and effector genes, some CBASS operons encode CD‐NTase‐associated proteins ¹⁶ , ¹⁹ , ²⁰ , ²¹ , ²² that are used for classifying CBASS into four major types, I–IV ²³ . Among more than 5000 predicted CBASS systems, 2199 encode homologs of ubiquitin‐related enzymes, which are key components of eukaryotic posttranslational modification. These are collectively classified as type II CBASS ²³ . Type II CBASS systems are categorized into two main classes. The first is E1E2/JAB‐CBASS, which is present in 1583 prokaryotic genomes and contains an E1‐E2 fusion protein and a JAB deubiquitinating peptidase, with no E3 ligase ¹⁹ , ²⁰ , ²¹ . The second is E2‐CBASS that is present in only 616 prokaryotic genomes and contains only an E2 protein ⁹ , ²² , ²³ . We previously showed that the bacterial E2, functioning as a protease, regulates CdnG (CD‐NTase clade G) by mimicking the ubiquitination cascade, involving E1, E2, E3, and target proteins ²² .

In this study, we revealed that the E2‐associated CdnG contains an extra motif (referred to as E2‐binding motif hereafter) at the C‐terminal domain, which mediates its interaction with the E2. We presented the structure of a recombinant SmCdnG‐SmE2 complex with GTP by single‐particle cryo‐EM at 3.36 Å resolution. We identified the interaction between GTP and the SmCdnG active site that is highly conserved in the enzyme. In addition, we showed that the SmCdnG and SmE2 proteins were conserved with eukaryotic or prokaryotic homologs. Our data demonstrate that the binding between CdnG and E2 is conserved and provide structural and mechanistic insights into the regulation of the E2‐CBASS system.

RESULTS

CD‐NTase in E2‐CBASS systems exhibit extended protein length

To characterize the CD‐NTase in type II CBASS, we conducted a statistical analysis of the sequence lengths of two main categories of type II CBASS CD‐NTases (Figure 1A), termed E1E2/JAB‐CBASS and E2‐CBASS. Based on two studies ²³ , ²⁴ , there are 1345 CD‐NTase sequences in the E1E2/JAB‐CBASS category and 406 CD‐NTase sequences in the E2‐CBASS category (Figure 1B). The lengths of E1E2/JAB‐CBASS CD‐NTase sequences are distributed with enrichments on four specific lengths; however, E2‐CBASS CD‐NTase sequences are longer than those in E1E2/JAB‐CBASS (Figure 1A). In previous studies, prokaryote CD‐NTases were divided into eight major clades, labeled A‐H, based on sequence homology ²³ , ²⁴ , ²⁵ . To characterize the types of CD‐NTases associated with E1‐E2 or E2, we performed bioinformatics classification of CD‐NTases in the eight clades. Our data show that the sequences of E1E2/JAB‐CBASS CD‐NTases are found in clade A (42.16%), clade B (22.08%), and clade G (19.18%), while sequences of E2‐CBASS CD‐NTase are predominantly found in clade G (73.65%) (Figure 1B). The similarity in CD‐NTases associated with E2 suggests that these CD‐NTases have conserved catalytic and regulatory mechanisms.

Figure 1.

Statistics of CD‐NTase in two type II CBASS systems and the covalent linkage between SmCdnG and SmE2. (A) Length of CD‐NTase in E1E2/JAB‐CBASS and E2‐CBASS systems. The results show that CD‐NTases in E2‐CBASS are longer than those in E1E2/JAB‐CBASS. (B) Classification of CD‐NTase in two type II CBASS systems. In the E2‐CBASS system, CD‐NTases are mainly concentrated in clade G. (C) SDS‐PAGE analysis showing the in vitro formation of the SmCdnG–SmE2 complex. (D) The C‐terminal sequence of the SmCdnG protein modified by inserting a TEV site and Strep‐Tag II between G397 and G398. (E, F) SDS‐PAGE (E) and Western blot (F) analyses showing that the SmCdnG‐TEV‐Strep‐SmE2 was cut by TEV.

CD‐NTases and E2 are covalently linked in E2‐CBASS

To investigate the interaction between CdnG (CD‐NTase clade G) and E2 in the E2‐CBASS operon, we purified the SmCdnG (CD‐NTase clade G from Serratia marcescens) and SmE2 proteins from the type II CBASS operon from S. marcescens. This system provided protection against T4 phage infection in Escherichia coli B (Figure S1). In vitro incubation of purified CD‐NTase and E2 showed that these two proteins could form a stable covalent protein complex that was not disrupted by SDS (Figure 1C). When GFP was fused to the C‐terminus of SmCdnG, it was observed that SmE2 cleaved the GFP while forming a covalent complex with SmCdnG (Figure 1C). By adding a TEV cleavage site after G397 near the C‐terminus of SmCdnG and cleaving the SmCdnG‐SmE2 complex using TEV after the complex formation, the complex was separated (Figure 1D,E). Western blot results revealed the formation of a band with the molecular weight of the C‐terminus of SmCdnG linked to SmE2, suggesting a covalent linkage between the C‐terminal region of SmCdnG and SmE2 (Figure 1F). In summary, the E2‐CBASS operon SmE2 forms a covalent linkage to the C‐terminus of SmCdnG. In our previous study, we showed that the covalent bond formed is an isopeptide bond and can be converted to a thioester bond ²² .

CdnG and E2 undergo spontaneous covalent ligation in vivo

A previous study has identified several CBASS operons that encode fusion proteins containing both a CD‐NTase domain and an E2‐like domain ⁹ , suggesting a covalent linkage between SmCdnG (46 kD) and SmE2 (18 kD) ²² . In this study, we cloned the gene encoding SmCdnG with an N‐terminal 6× His tag into the pET28a vector, and the gene encoding SmE2 without a tag into the pQE‐82L vector, and then both plasmids were co‐transformed into E. coli BL21(DE3) for co‐expression of SmCdnG‐SmE2 complex (Figure S2A). Sodium dodecyl sulfate‐polyacrylamide gel electrophoresis (SDS‐PAGE) analysis confirmed the presence of SmCdnG‐SmE2 complex at a 1:1 molar ratio (62 kD) (Figure S2B,C). The SmCdnG‐SmE2 fusion could not be separated by SDS or dithiothreitol (DTT), confirming a covalent linkage between SmCdnG and SmE2.

CdnG and E2 form a stable heterodimer that binds GTP

In the CD‐NTase cGAMP synthesis reaction, the substrate GTP acts as a donor ²² , ²⁵ . To elucidate the structural basis of the SmCdnG‐SmE2 complex with GTP, we purified the protein complex, followed by incubation with GTP for 4 h. We then performed single‐particle cryo‐EM to visualize the structure of the SmCdnG‐SmE2 complex with GTP. The 2D class averages of the particle images demonstrated good convergence (Figure S2D). After multiple rounds of 3D classification and 3D refinement, the final 3D reconstruction has a resolution of 3.36 Å. The cryo‐EM map displayed densities of two proteins with clearly visible side chains for most residues (Figures 2A and S3). Our structure shows that SmCdnG and SmE2 form a heterodimeric complex in a 1:1 ratio, with the complex containing one GTP molecule at the active site (Figure 2A). SmCdnG forms a cage‐like structure and has an additional β‐sheet at its C‐terminus (Figure 2A). A GTP molecule is present in the nucleotide‐binding pocket of the SmCdnG.

Figure 2.

Cryo‐EM structure of the SmCdnG‐SmE2 complex with GTP. (A) The cryo‐EM map of the SmCdnG–SmE2 complex with GTP, overlaid with a cartoon representation of the model created with ChimeraX. SmCdnG‐SmE2 is a heterodimer, with SmCdnG binding one molecule of GTP. (B) Overview of the SmCdnG‐SmE2 interfaces, with the covalent linkage between SmCdnG and SmE2 highlighted by a red dashed line. The SmCdnG‐SmE2 complex also has three additional interaction interfaces. (C–E) Detailed view of interacting residues at interface 1 (C), interface 2 (D), and interface 3 (E) between SmCdnG and SmE2. Black dashed lines represent hydrogen bonds. (F) SDS‐PAGE analysis of SmCdnG–SmE2 complex formation in wild‐type and interface‐mutant variants.

CdnG and E2 engage in extensive interactions

SmCdnG interacts with SmE2 through the C‐terminal region of SmCdnG to form a complex. In contrast to the E1E2/JAB‐CBASS system, where cGAS weakly interacts with dimerized E1‐E2 fusion ¹⁹ , ²⁰ , SmCdnG of CBASS forms extensive interactions with E2 at three interfaces in addition to the covalently‐linked C‐terminus (Figure 2B). At interface 1, K380 of SmCdnG interacts with S123 of SmE2 through hydrogen bond interactions (Figure 2C). At interface 2, E340 of SmCdnG interacts with S49 of SmE2 through hydrogen bonding, while L385 of SmCdnG interacts with P48 of SmE2 through hydrophobic interactions (Figure 2D). At interface 3, R360 of SmCdnG interacts with E147 of SmE2 through a salt bridge, while Q368 of SmCdnG interacts with Y136 of SmE2 through hydrogen bonding (Figure 2E). To check whether all CD‐NTases and associated E2s form the same interactions, we used AlphaFold 3 to predict the structure of the bdCdnG‐bdE2 complex, and found that it also has these three binding interfaces (Figure S6), indicating that the interactions at these interfaces are conserved in all CdnG‐E2 complexes.

To validate the functional importance of these interactions, we performed site‐directed mutagenesis of key interface residues. Notably, the R360D mutation (interface 3) abolished complex formation, as shown by SDS‐PAGE. At interface 2, E340K and L385T mutations reduced covalent linkage formation and prevented the accumulation of high‐molecular‐weight products (Figure 2F). These results demonstrate that noncovalent interactions facilitate formation of the covalent linkage between SmCdnG and SmE2, suggesting a recognition mechanism in which specific binding precedes and enables covalent bond formation.

GTP binds to a conserved catalytic site in the CD‐NTases of E2‐CBASS

To understand the impact of the substrate GTP on the structure of SmCdnG, we compared the structures of SmCdnG in apo and GTP‐bound states. The SmCdnG structure exhibits a canonical CD‐NTase fold, featuring a central pocket for nucleotide binding and a lid that covers the active site (Figure 3A). In SmCdnG, there is an elongated groove, dividing the molecule into two distinct domains, named N‐lobe and C‐lobe (Figure 3A,B). Compared to the SmCdnG apo structure, the N‐lobe of GTP‐bound SmCdnG rotates at about 3.4° (Figures 3B and S4). The active site of CD‐NTases contains discrete “donor” and “acceptor” nucleotide binding pockets; a GTP molecule is present within the donor pocket (Figure 3C). Structural analysis reveals that a conserved helix, functioning like a lid, is positioned above the acceptor and donor nucleotide pockets in the active site, with residues S167 and Y172 interacting with GTP. Notably, residues S167 and Y172 are conserved specific amino acids in CdnG ²⁵ . There is a QGSV (66‐69) sequence, belonging to the conserved XGSX motif in the superfamily of nucleotidyl transferases, which forms a small helix interacting with GTP nucleotides (Figure 3C). The aspartic acid residues are the key active site residues of nucleotidyl transferase superfamily (NTS), such as in (2′, 5′)‐ligoadenylate synthetase (OAS1) and poly(A) polymerase ²⁶ . Two conserved amino acids, D84 and D86, on the β2 strand, coordinate the donor substrate GTP. These residues are highly conserved in CdnG and likely contribute to the enzymatic reaction (Figure 3D). In summary, the core structure of the SmCdnG protein is highly conserved, and GTP binding induces a subtle conformational change, likely as a result of nucleotide coordination and priming of the catalytic activity.

Figure 3.

Structural and sequence analysis of SmCdnG with GTP. (A) Structure of SmCdnG with bound GTP. (B) Alignment of the structure of SmCdnG with bound GTP and the apo SmCdnG structure. The N‐lobe of SmCdnG with bound GTP is tilted by 3.4°. (C) Detailed view of the GTP binding site in SmCdnG. Multiple conserved residues interact with GTP. (D) Structure‐guide alignment of different CdnG subclasses, with the catalytic site sequences highlighted, performed using ESPript 3.2.

The C‐terminal structural motif of CdnG in E2‐CBASS recruits E2

To further characterize the E2‐associated CdnGs, we analyzed the structure of SmCdnG and found that its C‐terminus contains an extra structure motif that is unique in E2‐associated CdnGs (Figures 2A and 4A). The C‐terminal region (residues 352–398) of SmCdnG contains an ⍺‐helix and a three‐stranded β‐sheet that interacts with SmE2 (Figures 2A and 4A). This C‐terminal motif is completely unique compared to human cGAS and CD‐NTase from the E1E2/JAB‐CBASS system ²² . Interestingly, a three‐stranded β‐sheet is also present in the C‐terminal region of ubiquitin that interacts with eukaryotic E2. We used AlphaFold 3 to predict the structures of CD‐NTase proteins in several other E2‐CBASS systems (Figures 4B,C and S5). We found that the C‐terminal helix and β‐sheet are conserved in the CD‐NTase protein structures of these E2‐CBASS systems; bdCdnG (PDB ID: 7LJN) also belongs to the E2‐CBASS system. In the crystal structure of bdCdnG, the C‐terminal sequence was deleted ²⁵ , while using AlphaFold 3 prediction, we found that bdCdnG also contains a conserved C‐terminal structure (Figure 4B). By aligning the C‐terminal amino acid sequences of these CD‐NTase proteins, we found significant sequence variation (Figure 4D), indicating that the C‐terminal structure of the CdnG, rather than its sequence, plays a crucial role in the interaction between CdnG and E2. Notably, the glycine at the C‐terminus is consistent with a previous report and highly conserved ²² . Our results suggest that the C‐terminal region of SmCdnG resembles the C‐terminal region of ubiquitin and mediates the formation of SmCdnG‐SmE2 complex, suggesting a possible evolutionary ancestor of ubiquitin in bacteria. Interestingly, these ubiquitin‐like interactions between CdnG and E2 are not present in the E1E2/JAB‐CBASS systems ²⁰ .

Figure 4.

Structural and sequence analysis of CD‐NTase from different E2‐CBASS systems. (A) The extra α‐helix and β‐sheet in the C‐terminus of the SmCdnG. (B) Superposition of truncated BdCdnG structure (PDB ID: 7LJN) and AlphaFold 3 predicted full‐length BdCdnG structure. The extra α‐helix and β‐sheet in the C‐terminus of the BdCdnG. (C) Protein structure of Clade H protein GjCdnH in E2‐CBASS predicted by AlphaFold 3. (D) Structure guide alignment of CD‐NTase C‐terminal region from different E2‐CBASS systems by ESPript 3.2.

E2 in E2‐CBASS mimics human ubiquitin‐conjugating E2 enzyme

In order to understand the characteristics of E2, we analyzed the structure of SmE2. The structure of the SmE2 protein contains an α‐helix and four β‐strands at the N‐terminus and an α‐helix and some unstructured regions at the C‐terminus (Figure 5A). There is an unstructured loop between the N‐terminal β‐sheets and the C‐terminal α‐helix (Figure 5A). The conserved site C101 is located within this unstructured loop, near the last β‐strand. The amino acid K159, which forms an isopeptide bond with SmCdnG, is found in the unstructured sequence at the C‐terminus ²² . The structure of the SmE2 is very similar to that of human E2 HsUbcH5B (PDB ID: 4AUQ), with a root mean square deviation (RMSD) of 2.5 Å (Figure 5A). To verify the structural conservation of E2 in the E2‐CBASS systems, we used AlphaFold 3 to predict multiple E2 structures located in different types of CD‐NTase operons (Figure 5B,C). Structural alignment revealed remarkable similarity in the E2 structures within the E2‐CBASS systems (Figure 5D). The conserved catalytic site residue, Cys, is located at the same position. The structure of SmE2 is also similar to the E2 structure (PDB ID: 7TO3) in the E1E2/JAB‐CBASS system, with a root mean square deviation (RMSD) of 11.3 Å (Figure S7); however, the E2‐CBASS system lacks the E1 protein ²⁰ . This indicates that the E2 protein in the E2‐CBASS system has a unique role in regulating CD‐NTase activity.

Figure 5.

Structural and sequence analysis of E2 proteins from different E2‐CBASS systems. (A) Superposition of SmE2 structure used in this study and human E2 structure (PDB ID: 4AUQ). (B) Structure of GjE2 of E2‐CBASS predicted by AlphaFold 3. (C) Structure of BdE2 of E2‐CBASS predicted by AlphaFold 3. (D) Structure‐guide alignment of E2 by ESPript 3.2.

DISCUSSION

In this study, we analyzed the length distribution of two types of type II CBASS CD‐NTases. The average length of the CD‐NTases in E2‐CBASS is longer than that in E1E2/JAB‐CBASS. Moreover, the classification is primarily concentrated in clade G, indicating that the E2‐CBASS phage defense system is very unique. We used cryo‐electron microscopy to determine the structure of the SmCdnG‐SmE2 complex with GTP. Our structure resolved the heterodimer structure at 3.36 Å, showing the structures of SmCdnG, the folded C‐terminus of SmCdnG, and SmE2, from the E2‐CBASS system. The core structure of SmCdnG is consistent with other types of CD‐NTase structures. However, the C‐terminus of the SmCdnG protein is found exclusively in the E2‐CBASS system, featuring an α‐helix and a three‐stranded β‐sheet. The protein structure of SmE2 is conserved with human E2 and the E2 of type II CBASS, but the E2‐CBASS system does not contain any E1 proteins, highlighting the unique function of the E2 protein in the E2‐CBASS system.

In the E2‐CBASS system, the C‐terminal structure of CD‐NTase is conserved. However, through sequence alignment, it was found that, except for the terminal residue Gly, the other sequences do not exhibit sequence conservation. In the E1E2/JAB‐CBASS system, Cap3 can specifically recognize and interact with the associated CD‐NTase and cannot act on other CD‐NTases ²⁰ . We anticipate that the E2 protein in the E2‐CBASS system will also specifically recognize the C‐terminal sequence of the CD‐NTase protein and perform cleavage and ligation. Of course, elucidating the precise mechanisms of this cleavage and ligation recognition requires more in‐depth and detailed systematic studies.

MATERIALS AND METHODS

Expression of recombinant proteins

The plasmids expressing SmCdnG‐SmE2 complex were derived from the previously referenced study ²² . Plasmids were co‐transformed into E. coli BL21 (DE3) for protein expression and selection using kanamycin (50 μg/ml) and ampicillin (100 μg/ml). The bacterial culture was initially grown overnight in 10 ml LB medium at 37°C with shaking at 220 rpm. The following day, the culture was scaled up to 1 l of LB medium and incubated for about 3 h until the OD₆₀₀ reached 0.8–1.0. The cultures were then cooled in an ice‐water bath for 10 min, followed by the addition of 0.5 mM IPTG. The cells were then incubated overnight at 18°C to allow expression. Bacterial cells were collected by centrifugation at 7808g for 10 min, and the pellets were frozen and stored at −80°C.

To purify SmCdnG‐SmE2 complex, the frozen bacterial pellet was first thawed and then suspended in binding buffer (20 mM Tris pH 8.0, 300 mM NaCl, 20 mM imidazole) containing 0.2 mM PMSF. The cell suspensions were disrupted using sonication, followed by centrifugation at 30,624g for 30 min at 4°C to clarify the lysates. The clarified supernatant was then loaded onto Ni‐NTA agarose resin, pre‐equilibrated with binding buffer, and incubated at 4°C with gentle agitation for approximately 30 min. The resin was washed three times with 10 volumes of binding buffer. Bound protein was eluted with 20 mM Tris pH 8.0, 300 mM NaCl, and 200 mM imidazole. Proteins were analyzed by SDS–PAGE with Coomassie blue staining (Beyotime). Protein was further concentrated using centrifugation with a 30 kDa cut‐off column (PALL).

Cryo‐EM sample preparation and data collection

The SmCdnG‐SmE2 complex was further purified by size‐exclusion chromatography using a Superdex™ 75 Increase 10/300 GL (Cytiva) column in TBS buffer (20 mM Tris–HCl (pH 8.0), 100 mM NaCl). Proteins were filter‐concentrated to approximately 3 mg ml⁻¹ before use, flash‐frozen in liquid nitrogen, and stored at −80°C. Proteins were analyzed by SDS–PAGE with Coomassie blue staining (Beyotime), and concentration was quantified by UV absorbance at 280 nm.

The purified SmCdnG‐SmE2 complex was supplemented with a final concentration of 5 mM MnCl₂ and 0.5 mM GTP, followed by incubation at 4°C for 4 h. A 3 µl aliquot of the reaction sample was deposited onto glow‐discharged copper UltrAuFoil R1.2/R1.3 grids, which were then blotted for 3 s in 100% humidity at 8°C before being rapidly frozen in liquid ethane using a Vitrobot Mark IV. The grids were subsequently examined using a Glacios microscope, and those exhibiting the optimal ice thickness and particle density were selected for cryo‐EM data acquisition. Data were collected using a 300 keV Titan Krios G4 microscope, equipped with a Gatan K3 direct electron detector and a Gatan Quantum energy filter, at the Cryo‐EM Unit of the Core Facility at Wuhan University.

Cryo‐EM data processing and model building

For the dataset, 5064 movies were collected in super‐resolution mode, with 40 frames per video, a 2.16 s exposure time, a 70 e⁻ Å⁻² accumulated dose, and a 0.335 Å pixel size.

The movies were binned by a factor of 2 (pixel size, 0.67 Å), and motion correction was performed using the patch motion correction feature in CryoSPARC ²⁷ . Particles were picked using a general Topaz‐trained early ²² and blob picker. After two rounds of 2D classification, the particles selected by the two pickers were combined, and duplicate particles were discarded. Heterogeneous refinement (3D classification) was performed on 1,692,689 particles from 2D classification, with the map serving as the initial model. One class out of six was selected for extraction from micrographs (extraction box size 512, Fourier crop to box size 360). Particles were used for another round of nonuniform refinement, and the final map resolution of the SmCdnG–SmE2 complex with GTP was 3.36 Å (Figure S8 and Table S1). The initial model of the SmCdnG‐SmE2 complex bound to GTP was obtained by fitting the predicted structure into the cryo‐EM maps using ChimeraX ²⁸ . Subsequent model inspection, building, and manual adjustments were done in Coot ²⁹ , while real‐space refinements were performed using Phenix ³⁰ . The cryo‐EM density maps and structural models were visualized using ChimeraX and PyMOL.

AUTHOR CONTRIBUTIONS

Jun Xiao: Data curation; writing—original draft. Yan Yan: Data curation; resources. Jing Li: Data curation. Greater Kayode Oyejobi: Writing—review and editing. Dongyang Lan: Data curation. Bin Zhu: Conceptualization; resources; writing—review and editing. Zhiming Wang: Data curation; writing—review and editing. Longfei Wang: Funding acquisition; writing—review and editing.

ETHICS STATEMENT

This study did not involve animals or humans.

CONFLICT OF INTERESTS

The authors declare no conflict of interests.

Supporting information

Figure S1. Phage resistance assay of Escherichia coli B expressing the Serratia marcescens CBASS system.

Figure S2. Sample preparation for cryo‐electron microscopy. (A) Schematic diagram showing the constructs for co‐expression of SmCdnG and SmE2. (B) SEC data for the SmCdnG‐SmE2 complex. (C) SDS‐PAGE of SmCdnG‐SmE2 complex. (D) A representative cryo‐EM micrograph and 2D averages of the SmCdnG‐SmE2 complex with GTP.

Figure S3. Chemical properties of the SmCdnG‐SmE2 complex with GTP surfaces. (A) Electrostatic surface analysis of the SmCdnG–SmE2 complex with GTP. (B) Hydrophobic surface analysis of the SmCdnG–SmE2 complex with GTP.

Figure S4. Superposition of the SmCdnG‐SmE2‐bound GTP structure and the apo SmCdnG‐SmE2 structure.

Figure S5. Structure‐guide alignment of cGAS proteins from different type II CBASS E2‐only systems.

Figure S6. Cartoon representation of the bdCdnG–bdE2 complex predicted by AlphaFold 3, highlighting the interface between bdCdnG and bdE2.

Figure S7. Superposition of SmE2 structure and E2 structure from E1E2/JAB‐CBASS system.

Figure S8. (A) Flow chart of cryo‐EM data processing and 3D reconstruction of the SmCdnG‐SmE2 complex. (B) Orientation distribution plot of the 3D reconstruction of the SmCdnG‐SmE2 complex. (C) Directional FSC of the SmCdnG‐SmE2 with GTP cryo‐EM density map.

Cryo‐EM data collection, refinement, and validation statistics.

DATA AVAILABILITY

Data supporting the findings of this study are available in this article and its supplementary files. Genomes of S. marcescens were sequenced previously, and sequences are publicly available. For the SmCdnG‐SmE2 complex with GTP, coordinates are available at the RCSB PDB (http://www.rcsb.org) under accession codes 9KKB, and electron microscopy maps are available at the Electron Microscopy Data Bank (EMDB; https://www.ebi.ac.uk/emdb/) under accession codes EMD‐62384.