Molecular basis of Gabija anti-phage supramolecular assemblies

As one of the most prevalent anti-phage defense systems in prokaryotes, Gabija consists of a Gabija protein A (GajA) and a Gabija protein B (GajB). The assembly and function of the Gabija system remain unclear. Here we present cryo-EM structures of GajA and the GajAB complex, revealing tetrameric and octameric assemblies, respectively. In the center of the complex, GajA assembles into a symmetric tetramer, which recruits two sets of GajB dimer at opposite sides of the complex, resulting in a 4:4 GajAB supramolecular complex for anti-phage defense. Further biochemical analysis showed that GajA alone is sufficient to cut double-stranded DNA and plasmid DNA, which can be inhibited by ATP. Unexpectedly, the GajAB displays enhanced activity for plasmid DNA, suggesting a role of substrate selection by GajB. Together, our study defines a framework for understanding anti-phage immune defense by the GajAB complex.

As one of the most prevalent anti-phage defense systems in prokaryotes, Gabija consists of a Gabija protein A (GajA) and a Gabija protein B (GajB).The assembly and function of the Gabija system remain unclear.Here we present cryo-EM structures of Bacillus cereus GajA and GajAB complex, revealing tetrameric and octameric assemblies, respectively.In the center of the complex, GajA assembles into a tetramer, which recruits two sets of GajB dimer at opposite sides of the complex, resulting in a 4:4 GajAB supramolecular complex for anti-phage defense.Further biochemical analysis showed that GajA alone is sufficient to cut double-stranded DNA and plasmid DNA, which can be inhibited by ATP.Unexpectedly, the GajAB displays enhanced activity for plasmid DNA, suggesting a role of substrate selection by GajB.Together, our study defines a framework for understanding anti-phage immune defense by the GajAB complex.
To mitigate phage infections, bacteria have evolved highly diverse anti-phage immune systems 1 .Though some bacterial immune systems like CRISPR-Cas and cyclic-oligonucleotide-based anti-phage signaling system (CBASS) have been extensively studied 2,3 , many newly identified systems remain unexplored [4][5][6] .The study of bacterial immune system not only offers evolutionary perspectives on immune systems but also provides invaluable tools for biomedical research and disease treatment.
As a newly identified bacterial immune system, the bacterial Gabija defense system exists in at least 8.5% of sequenced genomes with two components, Gabija protein A (GajA) and Gabija protein B (GajB) 4,7 .GajA was shown to be an endonuclease that can recognize specific DNA sequence 8 .GajB was predicted to be a UvrD-like helicase 9 .However, whether and how GajA and GajB assemble into a complex for anti-phage defense remains unclear.
In this article, we present the cryo-electron microscopy (cryo-EM) structures of GajA alone and the GajAB complex.GajA assembles into a tetramer via interactions mediated by both the ATPase domain and the nuclease domain.We also reveal that the GajA and GajB assemble into a heteromeric octamer with four molecules of GajA and four molecules of GajB, which is critical for the anti-phage defense.Given many other supramolecular assemblies identified in bacterial immunity, we propose that supramolecular assemblies may represent a unified mechanism in bacterial immune defense.

Structure of GajA
To biochemically characterize GajA, we expressed and purified GajA in Escherichia coli BL21(DE3) (Extended Data Fig. 1a,b).The elution volume of GajA on gel filtration indicated that GajA formed an oligomer (Extended Data Fig. 1a).To reveal the assembly of GajA, we employed cryo-EM single particle analysis to determine the structure of GajA.However, GajA had a severe orientation preference problem on grids, leading to a reconstruction with a nominal resolution of 2.9 Å but poor densities (Extended Data Fig. 1c-g).To resolve this issue, we optimized conditions for grid preparation and eventually obtained a cryo-EM https://doi.org/10.1038/s41594-024-01283-wEach protomer of GajA is composed of an N-terminal ATPase domain that is divided into two halves by an inserted dimerization domain, and a C-terminal nuclease domain (Fig. 1c-e).The N-terminal ATPase domain is composed of an 11-stranded mixed-paralleled β-sheet, sandwiching α1 and surrounded by α2-α8 (Fig. 1d and Extended Data Fig. 3a).Structural comparison revealed that the GajA ATPase domain resembles the canonical ATP-binding cassette (ABC) ATPase proteins with a highly conserved ATP binding site 10,11 (Extended Data Fig. 3b).structure of GajA with a resolution of 3.2 Å by collecting a dataset using a grid with relatively thicker ice (Table 1 and Extended Data Figs.1a and 2a-e).Although the nominal resolution is lower from reconstruction of images with thicker ice, the EM densities have been substantially improved in comparison to the 2.9 Å structure (Extended Data Figs.1g and 2f).The cryo-EM structure of GajA revealed a symmetric tetrameric assembly with dimensions of 175 Å × 115 Å × 50 Å (Fig. 1a,b).The GajA C-terminal domain folds as a topoisomerase-primase (Toprim) domain with a central four-stranded parallel β-sheet surrounded by α-helices 12,13 (Fig. 1d and Extended Data Fig. 3c).Both the N-terminal ATPase domain and the C-terminal Toprim domain were clearly resolved in our cryo-EM structure (Fig. 1d).In contrast, the dimerization domain, predicted to consist of three α-helices by Alpha-Fold 14 , was invisible, indicating the flexibility of this domain (Fig. 1d,e).A recent cryo-EM structure of Bacillus cereus GajAB in complex with a phage protein Gad1 revealed that the dimerization domain of GajA not only contributed to the oligomerization of GajA but was also involved in recruiting Gad1 (ref.15) (Extended Data Fig. 3d).

Assembly of GajA tetramer
The tetramerized GajA is arranged as a dimer of dimer with three types of interface, denoted as type I, type II and type III interfaces (Fig. 2a).The type I interface, located at the center of the GajA tetramer, is mediated by the first half of the ATPase domain with a buried surface area of 609 Å 2 (Fig. 2b).Detailed analysis revealed that both hydrophobic and hydrophilic interactions are critical for the formation of type I interface (Extended Data Fig. 4a).In contrast, both hydrophobic and hydrophilic residues in the second half of the ATPase domain contribute to the formation of type II interface that is characterized by a buried surface area of 325 Å 2 (Fig. 2c and Extended Data Fig. 4b).Additionally, Toprim domains are arranged along a twofold axis to form the type III interface with an extensively buried surface area of 1,471 Å 2 (Fig. 2d and Extended Data Fig. 4c).Together, interactions of these three interfaces govern the assembly of GajA tetramer.

Active site of GajA
The Toprim domains in GajA tetramer, which belong to a class of DNA endonucleases known as OLD (overcoming lysogenization defect) 12 , dimerize at opposite sides of the elongated tetramer.Analysis of the two proximal Toprim domains reveals that the two active sites are about 20 Å away from each other, suggesting that the two active sites work independently (Fig. 2e).Similar to other OLD nucleases 12,13 , the active site of GajA is composed of a conserved DxD motif between β3 and β4, an invariant glutamate following β1, and two glutamate residues on α2 and α7 (Fig. 2f).Collectively, these residues form a very acidic active site in GajA.Studies on BpOLD suggested a two-metal catalysis mechanism 13 , which may be shared by GajA due to the structural similarity of the active sites between GajA and BpOLD (Extended Data Fig. 4d).
The conserved DxD motif and E379 may coordinate a magnesium ion while the other magnesium ion may be coordinated by E513 and E383 (refs.13,16) (Extended Data Fig. 4d).Consistently, the E379A mutation abolished the nuclease activity of GajA, highlighting the notable role of E379 in catalysis 8 .
To further understand how the Toprim domain specifically recognizes and cleaves double-stranded DNA (dsDNA), we generated a structural model of the Toprim domain with dsDNA using RoseTTAFold2NA 17 (Extended Data Fig. 4e).In the predicted model, the cutting site on dsDNA was precisely nested in the catalytic center of the Toprim domain (Extended Data Fig. 4f,g), poised for cleavage.Positively charged residues and hydrophilic residues surrounding the catalytic center established intensive interactions with the dsDNA substrate (Extended Data Fig. 4f,g), which may partially explain the substrate specificity of the Toprim domain.Moreover, the narrow slit between two neighboring Toprim domains in the tetrameric GajA appears to be insufficient to accommodate the dsDNA substrate (Fig. 2d).As such, we speculated that the Toprim domains would undergo large conformational changes upon DNA binding.

Structure of GajAB complex
To understand the assembly of GajAB, we further reconstituted the GajAB complex and determined a 3.0 Å cryo-EM structure of the complex (Fig. 3a,b, Table 1 and Extended Data Figs.5a-c and 6a-c).The cryo-EM structure of GajAB reveals a 4:4 assembly of GajA and GajB with a dimension of 175 Å × 145 Å × 95 Å, contrasting with a previous assumption that GajA and GajB form a complex with variable stoichiometries 9 (Fig. 3a,b).Consistently, gel filtration analysis and native mass spectrometry analysis further confirmed that GajA and GajB assemble into a complex with a stoichiometry of 4:4 while GajB alone exists as a monomer (Extended Data Fig. 5a-c).In the GajAB complex, the tetrameric GajA is decorated by a pair of GajB dimers at either end (Fig. 3a,b).https://doi.org/10.1038/s41594-024-01283-wGajB is composed of four structural domains: 1A, 1B, 2A and 2B, resembling superfamily 1A helicase proteins such as UvrD, PcrA and Rep that function to unwind and translocate DNA [18][19][20] (Fig. 3c,d and Extended Data Fig. 7a).In our cryo-EM structure, domains 1A, 1B and part of 2A are visualized, while domain 2B is completely absent (Fig. 3e).
Sequence alignment revealed that GajB contains the eight sequence motifs in domains 1A and 2A of UvrD, which have been identified to be involved in ATP binding 20 (Extended Data Fig. 7b).These features indicated that GajB is capable of hydrolyzing ATP.Consistently, ATPase activities are detected in GajB in the presence of DNA 9 , indicating that https://doi.org/10.1038/s41594-024-01283-wDNA substrates are required to stimulate the ATPase activity of GajB.Structural comparison to UvrD revealed that domain 2A in GajB is not well positioned to coordinate ATP (Extended Data Fig. 7c).As such, conformational changes are required for GajB to bind and hydrolyze ATP upon DNA binding.The structure of UvrD in complex with ds-ss DNA junction revealed that the single-stranded DNA (ssDNA) binds to domains 1A and 2A at their interfaces with 1B and 2B while the DNA duplex is coordinated by domains 1B and 2B 20 .Comparisons with UvrD demonstrate that GajB contains all the key residues for coordinating ssDNA, suggesting that GajB may use a similar strategy for binding ssDNA (Extended Data Fig. 7d).In contrast, the domain 2B in GajB is much smaller than that of UvrD and lacks key residues for coordinating dsDNA, raising questions whether GajB can efficiently bind to dsDNA (Extended Data Fig. 7a,e).

Mechanism of GajAB assembly
The structure of the GajAB complex revealed that the GajA recruits a pair of GajB molecules via its ATPase domain (Fig. 4a-c).Each GajB molecule interacts with two GajA molecules (cis and trans) that form a head-to-head dimer via their ATPase domain (Fig. 4b,c).The interaction between GajB and GajA in cis is quite extensive with a buried surface area of 755 Å 2 , which is dominated by interactions between the GajB 1B domain and the GajA ATPase domain (Fig. 4b and Extended Data Fig. 8a,b).Both hydrophobic and hydrophilic residues form an extensive network to dock the GajB 1B domain onto the ATPase domain of GajA, positioning the GajB 1A domain adjacent to the GajA ATPase domain with relatively weaker interactions (Extended Data Fig. 8a,b).In contrast, the interactions between GajB and GajA in trans are much weaker with a buried surface area of 140 Å 2 , which is mediated by the GajB 1A domain and the GajA ATPase domain (Fig. 4c and Extended Data Fig. 8c).Additionally, the two paired GajB molecules form relatively weak interactions with each other with a buried surface area of 380 Å 2 , which is dominated by hydrophilic residues in the 1B domain of GajB (Fig. 4d and Extended Data Fig. 8d).Collectively, these three interfaces together with the extensive interactions among tetrameric GajA form the basis of GajAB supramolecular complex assembly.

Nuclease activities and anti-phage defense of GajAB complex
Consistent with previous studies 8,9 , GajA has nuclease activities in the presence of Mg 2+ and is capable of cleaving dsDNA while GajB has no nuclease activities (Fig. 5a,b).Unexpectedly, the complex of GajAB displayed similar nuclease activities toward dsDNA as GajA (Fig. 5c).The nuclease activities for both GajA and GajAB can be effectively inhibited by 0.5 mM ATP (Fig. 5a,c), indicating that ATP may serve as a critical factor in regulating the Gabija system.Contrasting a previous study 8 , we found that both GajA and GajAB are capable of cleaving plasmid pUC19 (Fig. 5d,e).Moreover, GajAB displayed higher nuclease activities toward pUC19 than GajA alone, highlighting the importance of GajAB complex assembly in effectively cleaving plasmid DNA (Fig. 5d,e).In addition, we found that GajB alone or the GajAB complex can effectively hydrolyze ATP in the presence of ssDNA, consistent with a recent study 9 .Therefore, ssDNA may activate the ATPase activity of GajB to lower the ATP level in cells for triggering the activation of GajA.Consistently, phage resistance assay showed that the GajAB complex is more effective in anti-phage defense compared to GajA or GajB alone (Fig. 5f).Moreover, we also showed that the nuclease activity of GajAB is essential for anti-phage defense (Fig. 5f).As such, the supramolecular complex assembly of GajAB is critical for anti-phage defense.

Discussion
We find that GajA and GajB assemble into a supramolecular complex for anti-phage defense (Fig. 6).Our structural analysis revealed that GajA alone forms a tetramer and further assembles into a heteromeric octamer with GajB (Fig. 6).Moreover, GajA alone has nuclease activity while GajB adopts a similar fold as UvrD that can bind DNA substrates.No obvious conformational changes have been observed in GajA upon binding to GajB.As such, we propose that GajB may function to assist GajA to better recognize its substrates and thereby promotes the activity of GajA.Consistent with this assumption, we found that the complex of GajAB displays higher nuclease activities toward plasmids than GajA alone (Fig. 5d,e).Given that the nuclease activity of GajA can be inhibited by ATP and GajB has ATPase activity in the presence of ssDNA, it is also possible that GajB may promote the activity of GajA through lowering cellular ATP level under certain conditions.Additionally, the nuclease activity of GajA is inhibited by ATP through a mechanism yet to be revealed.ATP binds to the ATPase domain and stabilizes the oligomerization of GajA.As large conformational changes are coupled with DNA binding, perhaps the presence of ATP may hinder the conformational changes of GajA and thereby prevents the DNA substrate from effectively binding to GajA.
Supramolecular assembly appears to be an emerging theme in anti-phage immune defense.More and more studies have revealed that bacterial immune systems tend to form large complexes for anti-phage defense [21][22][23][24] .For example, RdrA and RdrB in the restriction by an adenosine deaminase acting on RNA (RADAR) system assemble into a giant assembly with a molecular weight of up to 10 MDa (refs.21,22).Here, we present another example to show the supramolecular assembly by the Gabija system.As both the RADAR system and the Gabija system contain and oligomerize via ATPase domains, we believe that other ATPase-containing bacterial immune systems may also assemble into supramolecular complexes for anti-phage defense.https://doi.org/10.1038/s41594-024-01283-w
For the assembly of the GajAB complex, we incubated GajA and GajB with molar ratio of 1:1 on ice for 1 h followed by further purification via gel filtration.

Cryo-EM data collection
Three microliters of sample at 1.8 mg ml −1 was applied to a glow-discharged Quantifoil R1.2/1.3 400 mesh gold grid (Electron Microscopy Sciences), blotted for 4 s in 100% humidity at 4 °C and plunged into liquid ethane using an FEI Vitrobot Mark IV (Thermo Fisher).All grids were screened using a Thermo Fisher Glacios microscope (OSU Center for Electron Microscopy and Analysis).
For GajA tetramer (4 GajA) in thicker ice, 1,364 micrographs were collected using a 300 kV Titan Krios microscope equipped with a K3 direct electron detector (Thermo Fisher) in counting mode with a nominal magnification of 81,000×, and a physical pixel size of 0.899 Å with defocus values ranging from −1.0 to −2.0 μm.For GajA in thin ice, 6,370 images were collected using similar parameters.

Cryo-EM data processing
The detailed flowcharts for data processing of all the datasets are illustrated (Extended Data Figs.1e, 2c and 5c).The datasets were imported into cryoSPARC (v4.1.1)implementation of patch motion correction and patch contrast transfer function estimation 25 .Initial particle picking was done by blob picking to generate initial two-dimensional (2D) classes.Representative 2D classes were then selected as templates to pick all the particles for reconstruction.
For GajA tetramer, 849,640 particles were picked and extracted.After two rounds of 2D classification, 583,860 particles were selected and merged for further three-dimensional (3D) reconstruction and heterogeneous refinement.The best class of 96,633 particles was selected for further nonuniform refinement with D2 symmetry, resulting in a 3.23 Å map.
For GajAB complex, 6,928,153 particles were picked and extracted.After two rounds of 2D classification, 2,682,203 particles were selected for further ab initio reconstruction to generate three initial models for further refinement.The best class was selected for further 3D classification and heterogeneous refinement.The final best class of 942,091 particles was selected for nonuniform refinement with C1 symmetry and D2 symmetry, resulting in a 2.98 Å map and a 2.79 Å map, respectively.
All reported resolutions were estimated on the basis of the gold-standard Fourier shell correlation (FSC) 0.143 criterion 26 .

Model building and refinement
Two initial models of GajA and GajB were predicted by AlphaFold 14 and fit into the cryo-EM maps of GajA tetramer or GajAB complex using Chimera 27 .Manual adjustments were done using Coot to yield the final atomic model 28 .Real-space refinement was performed to refine the model against cryo-EM density map with secondary structure and geometry restraints in PHENIX 29 .The all-atom contacts and geometry for the final models were validated by Molprobity 30 .All the structural figures were generated using PyMOL 31 , Chimera 27 and ChimeraX 32 .

Nuclease assays
For pUC19 plasmid, a final concentration of 400 nM protein and 1,000 ng plamid substrate were incubated in reaction buffer (20 mM Tris-HCl pH 8.0) at 37 °C for 10 min.Samples were separated in 1% (w/v) agarose gel in 1× tris acetate EDTA (TAE) buffer for 30 min.
For dsDNA substrate, we annealed 5′-TTTTTTTTTTTTTTTT TAATAACCCGGTTATTTTTTTTTTTTTTTTTTTTTTTTTT-3′ and 5′-AAAAAAAAAAAAAAAAAAAAAAAAAATAACCGGGTTAT TAAAAAAAAAAAAAAAAA-3′ to generate a dsDNA substrate.Then, 400 nM protein and 800 nM nucleic acid substrate were incubated in reaction buffer (20 mM Tris-HCl pH 8.0) at 37 °C for 10 min.Samples were separated in 12% polyacrylamide gel electrophoresis gel in 1× tris borate EDTA (TBE) buffer for 1 h.In each reaction, a final concentration of 5 mM MgCl 2 or 0.5 mM ATP was added as indicated.All gels were stained by running buffer containing 0.5 ng μl −1 ethidium bromide and visualized by Imaging System (Bio-Rad).

ATPase assays
The ATPase activity was measured by an ATPase/GTPase Assay Kit (Sigma-Aldrich, #113CB04A30).GajB or GajAB in the absence or presence of ssDNA (5′-TTTTTTTTT TTTTTTTTAATAACCCGGTTATTTTTT TTTTTTTTTTTTTTTTTTTT-3′) was incubated with ATP in the buffer of 20 mM Tris pH 8.0, 75 mM KCl and 2 mM MgCl 2 at 37 °C.The reaction products were monitored by BioTek Synergy HT microplate reader at 600 nm.ATPase activities were calculated on the basis of the released phosphate.

Structural prediction
We utilized RoseTTAFold2NA 17 , a deep learning technique, to predict the complex between the GajA protein and dsDNA.This method relies solely on protein and DNA sequences for predicting complex structures.We used the GajA Toprim domain (residues 372-574) and a dsDNA segment (5′-GAATAACCCCGGATATT-3′ and 5′-AATATCCGGGGT TATTC-3′) as the input of the method.To predict the protein's structure, 3,361 homologous sequences of the GajA top domain were identified.RoseTTAFoldNA also offers a confidence measure for its predictions, known as predicted aligned error (PAE).PAE estimates the positional discrepancy in angstroms between predicted and actual structures, calculated for each pair of residues.In the predicted model, the average PAE between residues of the protein and residues of DNA was 16.8, indicating high confidence in the predicted complex's structure.

Plaque assays
Plaque assays were performed as previously described 4,33 .Briefly, reconstructed plasmid was transformed into E. coli DE3 competent cell.A single bacterial colony was picked from a fresh LB agar plate and grown in LB broth containing antibiotic at 37 °C to an OD 600 of ∼0.4.Protein expression was induced by the addition of 0.2 mM IPTG.After further growth for ∼3 h, 500 μl of the bacterial cultures was mixed with https://doi.org/10.1038/s41594-024-01283-w14.5 ml of 0.5% LB top agar, and the entire samples were poured onto LB plates containing antibiotic and IPTG (0.1 mM).Plates were spotted with 4 μl of the T4 phage diluted in LB at eight tenfold dilutions, namely 10 0 to 10 −7 .Lysate titer was determined using the small drop plaque assay method as previously described 33,34 .Plates were incubated at 37 °C overnight and then imaged.

Native mass spectrometry analysis
The sample underwent online buffer exchange before mass spectrometry analysis.A mobile phase of 200 mM ammonium acetate at pH 7.0 was maintained at a flow rate of 50 μl min −1 .A Vanquish Duo Ultra-High-Performance LC system (Thermo Scientific) equipped with a dual pump and autosampler was used to load protein samples without further purification into a MAbPac size exclusion column (MAbPac SEC1, 2.1 × 300 mm, 300 Å, 5 μm; Thermo Scientific, 008789).One microgram of the sample was injected onto the column and was further tracked by ultraviolet absorbance.From 0 min to 16.8 min, the flow was diverted from the column to the mass spectrometer, while after this time (16.8 min to 30 min), flow was diverted to the waste.The timing of this switch was optimized to allow for buffer-exchanged protein samples to make it to the mass spectrometer, while salts from the original buffer were passed to the waste.Buffer-exchanged samples were then analyzed using a Q Exactive Ultra-High Mass Range Hybrid Quadrupole-Orbitrap mass spectrometer (Thermo Scientific).The following settings were used for mass spectrometry analysis: the scan range was set to 1,000-20,000.The resolution was set to 12,500.The number of microscans was set to 5. The maximum injection time was set to 200 ms.The sheath gas was set to 60 psi.The auxiliary and sweep gas were set to 0 psi.Spray voltage was set to 3.85 kV in positive ion mode.The capillary temperature was set to 275 °C.S-lens RF level was set to 200 V.The in-source trapping was set to −120 V.The in-source dissociation was set to 5 V.The source d.c.offset was set to 21 V.The injection flatapole d.c. was set to 5 V.The inter-flatapole lens was set to 4 V.The bent flatapole d.c. was set to 2 V.The tapping gas pressure was set to 7. The total run time for the method was 30 min.The mass spectrum was deconvolved using Unidec 35 .

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Fig. 1 |
Fig. 1 | Cryo-EM structure of GajA.a,b, A cryo-EM density map (a) and ribbon diagrams (b) of GajA tetramer on two different planes: top view (left) and vertical rotation (right).c, Domain architecture of GajA.The ABC ATPase domain is indicated in green, dimerization domain in orange and Toprim in blue.d,e, A ribbon diagram of a GajA protomer determined by cryo-EM reconstruction (d) or AlphaFold prediction (e) with domains colored as in c.

Fig. 2 |
Fig. 2 | Assembly of GajA.a, The assembly of tetrameric GajA with three key interfaces indicated, which are denoted as interfaces I, II and III.b, Details of interface I mediated by the first halves of ABC ATPase domains with secondary structures indicated.c, Details of interface II mediated by the second halves of ABC ATPase domains with secondary structure indicated.d, Details of interface

3 |
Structure of the GajAB complex.a,b, A cryo-EM density map (a) and ribbon diagrams (b) of the GajAB complex with GajA in cold colors and GajB in warm colors on two different planes: top view (left) and vertical rotation (right).c, Domain architecture of GajB.The 1A domain is indicated in pink, 1B domain in magenta, 2A in yellow and 2B in orange.d,e, Ribbon diagram of a GajB protomer predicted by AlphaFold (d) or determined by cryo-EM reconstruction (e) with domains colored as in c.

Fig. 4 |Fig. 5 |Fig. 6 |
Fig. 4 | Assembly of GajAB.a, The assembly of GajAB with a dimeric GajA (green and blue) engaged with two GajB protomers (pink and magenta).b, Cis-interactions mediated by the GajA ATPase domain and GajB.c, Transinteractions mediated by the GajA ATPase domain and GajB.d, Interactions between two neighboring GajB protomers, which are mediated by the 1B domain of GajB.

Extended Data Fig. 1 |
Cryo-EM reconstruction of GajA in thin ice.a, b, Gel filtration profile (a) and SDS-PAGE gel (b) of GajA purification.The experiment was replicated at least three times.c, A representative cryo-EM image of GajA in thin ice.Thousands of images were collected.d, Representative 2D class averages of GajA calculated from thin-ice cryo-EM images.e, Data processing workflow for 3D reconstruction of GajA tetramer from thin-ice cryo-EM images.f, The FSC curve of reconstructed GajA tetramer from thin-ice cryo-EM images.g, Representative cryo-EM density of GajA tetramer fitted with α-helices and β-strands.The density map was shown at a contour level of 0.03.Extended Data Fig. 2 | Cryo-EM reconstruction of GajA in thicker ice.a, A representative cryo-EM image of GajA in thicker ice.Thousands of images were collected.b, 2D class averages of GajA calculated from thick-ice cryo-EM images.c, Data processing workflow for 3D reconstruction of GajA tetramer from thick-ice cryo-EM images.d, The FSC curve of reconstructed GajA tetramer from thick-ice cryo-images.e, Local resolution of reconstructed GajA tetramer from thick-ice cryo-images.f, Cryo-EM density of GajA tetramer fit with α-helices and β-strands.The density map is shown at contour levels of 0.03.Extended Data Fig. 4 | Interfaces in GajA tetramer.a-c, Enlarged views of interface I (a), interface II (b), and interface III (c) in GajA tetramer.Key residues on the interfaces were highlighted in sticks.d, Superimposed structures of the active sites from GajA (green) and BpOLD (PDB ID 6NK8, gray).e, Structure of GajA in complex with dsDNA (Yellow) that was predicted by RoseTTAFoldNA.f, Electrostatic surface representation of GajA with dsDNA.The catalytic center of GajA is highlighted by a red circle.Negatively charged residues surrounding the catalytic center of GajA coordinate dsDNA.g, Key residues involved in coordinating dsDNA are highlighted in sticks.Extended Data Fig. 6 | Cryo-EM reconstruction of GajAB.a, A representative cryo-EM image of GajAB complex.Thousands of images were collected.b, 2D class averages of GajAB complex.c, Data processing workflow for 3D reconstruction of GajAB complex.d, e, Local resolution (d) and FSC curve (e) of reconstructed GajAB complex without symmetry setting.f, g, Local resolution (f) and FSC curve (g) of reconstructed GajAB complex with D2 symmetry setting.Extended Data Fig. 7 | Structural comparison of GajB and UvrD.a, Overlaid structures of GajB (magenta, pink, yellow, and orange, AlphaFold predicted structure) and UvrD (PDB ID 2IS2, cyan).b, Sequence alignment of ATP binding motifs between GajB and UvrD.c, Overlaid structures of GajB (magenta) and UvrD (cyan) showed that domain 2A of GajB is not well positioned to coordinate ATP.d, Expanded view of key residues involved in coordinating ssDNA from GajB (magenta, AlphaFold predicted structure) and UvrD (cyan).e, Overlaid structures of GajB (magenta, AlphaFold predicted structure) and UvrD (cyan) revealed that domain 2B in GajB lacks key motifs for coordinating dsDNA.Extended Data Fig. 8 | Interfaces in GajAB.a, Key residues mediating interactions between GajB domain 1B (magenta) and GajA ATPase domain (green).b, Key residues mediating cis-interactions between GajB domain 1A (pink) and GajA ATPase domain (green).c, Key residues mediating transinteractions between GajB 1A (pink) and GajA ATPase (blue).d, Key residues mediating interactions of two neighboring GajB protomers.