The structural basis for 2′−5′/3′−5′-cGAMP synthesis by cGAS

Wu, Shuai; Gabelli, Sandra B.; Sohn, Jungsan

doi:10.1038/s41467-024-48365-3

Download PDF

Article
Open access
Published: 13 May 2024

The structural basis for 2′−5′/3′−5′-cGAMP synthesis by cGAS

Nature Communications volume 15, Article number: 4012 (2024) Cite this article

1 Altmetric
Metrics details

Subjects

Abstract

cGAS activates innate immune responses against cytosolic double-stranded DNA. Here, by determining crystal structures of cGAS at various reaction stages, we report a unifying catalytic mechanism. apo-cGAS assumes an array of inactive conformations and binds NTPs nonproductively. Dimerization-coupled double-stranded DNA-binding then affixes the active site into a rigid lock for productive metal•substrate binding. A web-like network of protein•NTP, intra-NTP, and inter-NTP interactions ensures the stepwise synthesis of 2′−5′/3′−5′-linked cGAMP while discriminating against noncognate NTPs and off-pathway intermediates. One divalent metal is sufficient for productive substrate binding, and capturing the second divalent metal is tightly coupled to nucleotide and linkage specificities, a process which manganese is preferred over magnesium by 100-fold. Additionally, we elucidate how mouse cGAS achieves more stringent NTP and linkage specificities than human cGAS. Together, our results reveal that an adaptable, yet precise lock-and-key-like mechanism underpins cGAS catalysis.

Development of multiplexed orthogonal base editor (MOBE) systems

Article 21 May 2024

Plasmid targeting and destruction by the DdmDE bacterial defence system

Article 13 May 2024

The β-subunit of tryptophan synthase is a latent tyrosine synthase

Article 14 May 2024

Introduction

Cyclic GMP-AMP (cGAMP) synthase (cGAS) is essential for the host defense against cytosolic double-stranded (ds)DNA arising from various maladies (e.g., pathogen invasion, ionizing irradiation, and genotoxic chemicals)^1,2,3. Upon directly binding to and dimerizing on dsDNA, cGAS cyclizes ATP and GTP into 2′−5′/3′−5′-linked cGAMP, a unique metazoan second messenger for initiating type-I interferon (IFN-I)-mediated inflammatory responses (Fig. 1A)^1,4,5,6,7. cGAS is central to antitumor immunity, host defense against an array of pathogens, and regulating autoimmunity^2,3,8.

**Fig. 1: The activation of cGAS entails a disorder-to-order transition.**

It is increasingly appreciated that cGAS-like enzymes have an ancestral origin and cGAMPs are widely employed as second messengers for both prokaryotic and eukaryotic innate immune pathways⁹. However, only metazoan enzymes generate the 2′−5′/3′−5′-linkage combination^4,5,9,10, and currently, the mechanisms by which cGAS specifically generates this uniquely linked cyclic dinucleotide remain poorly understood. Persisting questions include: how cGAS specifically recognizes ATP and GTP at each substrate binding pocket and coordinates its signature 2′−5′ linkage formation; how the same substrate binding sites then switch their nucleotide specificity and precisely position the GTP-AMP (pppGpA) intermediate for the second 3′−5′-linkage formation while barring the cyclization of other off-pathway dinucleotides; how dsDNA and dimerization regulate these processes, how it utilizes different divalent metals; and finally, how active site reactivity and promiscuity are regulated across different mammalian species. Here, we resolve these numerous fundamental mechanistic questions in innate immunology and present a unifying catalytic mechanism of cGAS.

We find that apo-cGAS assumes an array of inactive conformations and binds ATP/GTP nonproductively without involving divalent metals, which is then affixed into the catalytically competent conformation by dimerization-coupled dsDNA binding. We also delineate how web-like network of protein•NTP, intra-NTP, and inter-NTP interactions underpin the substrate-dependent linkage specificity. One Mg²⁺ is sufficient for productive substrate binding and Mn²⁺ is preferred as the second catalytic metal without involving an inverted intermediate. In the cyclization step, the adenosine of the pppGpA intermediate binds ~30° rotated compared to the guanine of the GTP substrate to precisely position the 3′-OH for the second linkage formation. cGAS employs multiple proof-reading mechanisms including an intermediate-dependent nucleoside trap to prevent the cyclization of noncognate dinucleotides. Finally, we reveal how mouse cGAS achieves more stringent substrate selectivity and higher catalytic efficiency than the human enzyme. Together, we set forth a unifying catalytic mechanism of cGAS in which dsDNA-binding pre-organizes the active site into a rigid yet adaptable lock for two sets of keys (substrates and intermediate) necessary to specifically synthesize 2′−5/3′−5′-cGAMP.

Results

Activation of cGAS entails a disorder-to-order transition

Previous structural studies suggested that the active site of apo-cGAS is occluded from binding ATP/GTP (we denote a 1:1 mixture of NTPs and NTP/NTP hereafter). dsDNA binding reconfigures the “activation loop” and the three β-sheets that harbor catalytic acidic residues for coordinating NTPs and two Mg²⁺ ions from the inactive to active conformation while allowing substrate binding^4,7. We had intended to decipher why nucleic acids other than dsDNA bind cGAS but fail to activate the enzyme (e.g., refs. ^11,12) by crystallizing the catalytic domain of mouse (m)cGAS^(cat) with double-stranded (ds)RNA. However, mcGAS^cat crystallized without the ligand, which instead provided us with an independent view of apo-mcGAS^cat at 1.7 Å resolution. Here, the three β-sheets at the active site did not align particularly well with either dsDNA-free or dsDNA-bound mcGAS^cat⁴. (Fig. 1B and Supplementary Fig. 1A), while the activation loop aligned better with mcGAS^cat•dsDNA (Ser¹⁹⁹; Fig. 1B/C). Moreover, the active site appeared entirely solvent accessible, as the cavity volume was ~1000 Å³ greater than that of mcGAS^cat•dsDNA (Supplementary Fig. 1B). To further examine these discrepancies, we crystallized and solved the structure of the catalytic domain of human cGAS (hcGAS^cat). We noted that, unlike the mouse enzyme, the three β-sheets do not move upon dsDNA binding, while the catalytic residues assume different rotamer conformations (Fig. 1D and Supplementary Fig. 1A)^13,14. However, the activation loop in our structure was mostly disordered (thus not modeled in Fig. 1D and Supplementary Fig. 1A). Comparing the B-factors of hcGAS^cat structures with or without bound dsDNA revealed that, while the C-lobe remains steady (likely via dimerization/crystal contact), dsDNA drastically stabilizes the N-lobe containing the activation loop and the catalytic residues (Supplementary Fig. 1C, D). Together, our observations suggest that, instead of being fixed into a specific autoinhibited conformation^4,7, the active site of resting cGAS assumes an array of inactive states and its activation entails a dsDNA-dependent disorder-to-order transition.

dsDNA modulates productive vs. nonproductive substrate binding

To understand why dsDNA-free cGAS is only basally active despite the open active site¹⁵, we attempted to capture substrate-bound cGAS without dsDNA. After conducting co-crystallization and soaking trials, we obtained the following structures with either human or mouse cGAS^cat. We co-crystallized hcGAS^cat with Mg²⁺ + ATP/GTP without dsDNA and found a new density corresponding to cGAMP at the active site; the catalytic residues remained in the inactive state (Fig. 1E; 5 mM Mg²⁺ is present in all our soaking/biochemical experiments unless noted otherwise). Co-crystallizing hcGAS^cat with GTP resulted in the pppGpG dinucleotide (GTP-GMP) bound in a conformation unamenable to further reaction: the densities for the triphosphate and Mg²⁺ were missing and the catalytic residues remained in the inactive state (Fig. 1F). Also of note, the active site B-factors remain high despite bound NTPs in both structures (Supplementary Fig. 1D). We next solved the crystal structure of ATP-bound mcGAS^cat without dsDNA via soaking. Here, one ATP was clearly bound at its designated binding pocket (Site-1), but the second NTP binding site (Site-2) remained empty (i.e., the GTP binding site; Fig. 1G). Strikingly, electron densities were missing for both the ribose of ATP and Mg²⁺; the catalytic residues stayed in the inactive state and the triphosphate appeared suboptimal for coordinating Mg²⁺ (Fig. 1G and Supplementary Fig. 1E). These results are consistent with the mechanism in which cGAS binds NTPs without dsDNA¹⁵, but it is only basally active because it can rarely fix NTPs and metals into the catalytically competent conformation (i.e., nonproductive binding).

dsDNA binding provides enthalpy for productive substrate binding

Our observation that dsDNA-free cGAS binds NTPs nonproductively prompted us to reinvestigate the role of dsDNA in catalysis. We largely employ mcGAS^cat hereafter as its dsDNA-bound form was conducive to crystallization and binding assays using isothermal titration calorimetry (ITC). To generate a cGAS construct that binds NTPs but is deficient in catalysis, we neutralized two acidic residues that coordinate Mg²⁺•ATP¹⁶ (E211Q-D213N in mcGAS, denoted as QN hereafter), which essentially abrogated the enzymatic activity (Supplementary Table 10). Soaking QN-mcGAS^cat•dsDNA crystals with ATP/GTP resulted in one Mg²⁺ and each NTP bound in position for the first 2′−5′-linkage formation (Fig. 2A and Supplementary Fig. 2A; the loss of the second Mg²⁺ was likely caused by the mutation, thus precluding catalysis). Soaking QN-mcGAS^cat•dsDNA crystals with GTP showed two GTPs in the catalytically competent state (Supplementary Fig. 2B). Moreover, soaking ATP to QN-mcGAS^cat crystallized with and without dsDNA showed the productive (one Mg²⁺) and nonproductive (no Mg²⁺) binding states seen from WT ± dsDNA (two Mg²⁺ and no Mg²⁺), respectively (Fig. 1G and Supplementary Fig. 2C–E). These observations corroborate that the QN mutant is well-suited for delineating how dsDNA regulates substrate binding.

**Fig. 2: dsDNA regulates productive vs. nonproductive substrate binding.**

Consistent with our crystal structures (Fig. 1E–G), QN-mcGAS^cat bound NTPs in µM affinities (K_Ds) in the absence of dsDNA (Fig. 2B; values are listed in Supplementary Table 11 and representative ITC traces are shown in Supplementary Fig. 7). When QN-mcGAS^cat was pre-complexed with dsDNA, not only did the K_D improve (Fig. 2B), but the overall thermodynamic profile also changed drastically (Fig. 2B–D). For example, when bound to dsDNA, the ∆H of ATP binding decreased by ~10 kcal/mol and the ∆S decreased by 30 cal/mol/deg. The favorable ∆∆H caused by dsDNA binding corroborates the formation of new interactions between the repositioned catalytic residues and Mg²⁺•ATP, and the unfavorable ∆∆S supports the disorder-to-order transition observed in crystal structures (e.g., Supplementary Fig. 2C vs. 2D; e.g., ref. ¹⁷). Also of note, not only was cGAMP binding barely affected by dsDNA, but it also bound more weakly than ATP/GTP, suggesting that the product would not interfere with catalysis (Fig. 2B–D). Combined with our crystal structures, these results suggest an activation mechanism in which dsDNA-free cGAS binds NTPs nonproductively; DNA binding affixes Mg²⁺•NTPs into the catalytically competent conformation (favorable ∆∆H for bond formation and unfavorable ∆∆S for the decreased degree of freedom).

Dimerization-coupled dsDNA-binding directs site-specific substrate binding

For the first linkage formation, Site-1 and Site-2 specifically bind ATP and GTP, respectively (Fig. 2A). Currently, mechanisms that direct the site-specific substrate binding remain unknown. Of note, ATP has always been found in Site-1 but not in Site-2 (Supplementary Fig. 2C–E and refs. ^4,14), GTP appeared to be capable of binding at both Sites (Supplementary Fig. 2B), and each NTP was bound at its designated site upon soaking ATP/GTP (Fig. 2A and Supplementary Fig. 2A). Combined, these observations suggest that Site-1 intrinsically prefers ATP and Site-2 favors GTP while discriminating strongly against ATP. Indeed, WT-mcGAS^cat showed the highest nucleotidyl transferase (NTase) activity toward ATP/GTP, yet it favored GTP alone (GTP/GTP) over ATP/ATP by 10-fold (Fig. 2E and Supplementary Table 10; i.e., the overall catalytic activity against any NTPs^15,18,19).

dsDNA-dependent dimerization is tightly coupled to the activation of cGAS^6,15 (Fig. 1A). To determine how cGAS links dsDNA binding and dimerization to NTP binding, we titrated GTP to mcGAS^cat pre-saturated with ATP and vice versa in the presence and absence of dsDNA in ITC. Sans dsDNA, GTP-bound QN-mcGAS^cat failed to bind ATP, while pre-saturating the enzyme with ATP decreased GTP binding affinity by 4-fold (Fig. 2B–D). By contrast, each NTP bound to QN-mcGAS^cat•dsDNA pre-saturated with the other NTP even more tightly compared to the unliganded complex (Fig. 2B–D; the favorable ∆∆H and ∆∆S are consistent with NTP binding at correct positions and eventually displacing (partially) pre-bound NTPs), suggesting that dsDNA directs site-specific NTP binding. Additionally, without dsDNA, QN-mcGAS^cat and a dimerization deficient mutant^6,15 (K382E/QN; see Supplementary Fig. 2F) bound ATP and GTP with similar affinities, indicating that dimerization per se is not required to bind NTPs (Fig. 2B–D). However, dsDNA failed to enhance the NTP binding of K382E/QN-cGAS^cat•dsDNA (Fig. 2B–D; e.g., the lack of favorable ∆∆H and now favorable ∆S). Moreover, neither NTP bound to K382E/QN-cGAS^cat•dsDNA pre-saturated with the other (Fig. 2B). We thus surmised that dsDNA binding and dimerization are tightly coupled to direct site-specific (productive) ATP/GTP binding.

Productive NTP binding underpins the distinction between ATP and deoxy-ATP at Site-1

When ATP is bound to mcGAS^cat•dsDNA at Site-1, its 2′-OH makes H-bonds with Glu³⁷¹ and Ser³⁶⁸, while the 3′-OH interacts with Lys⁴²⁴ and its own β-phosphate, apparently stabilizing the substrate (Fig. 3A and Supplementary Fig. 3A). Considering that the ribose of bound ATP was unresolved without dsDNA (Fig. 1G), we hypothesized that these interactions underpin productive substrate binding, which would also allow cGAS to discriminate against deoxy(d)-ATP. Indeed, mcGAS^cat•dsDNA displayed severely impaired NTase activities against 2′-dATP/GTP and 3′-dATP/GTP (Fig. 3B), revealing that the H-bonds from the 2′/ 3′-OH are important even for the first linkage formation. Although both dATPs bound to cGAS without dsDNA, the presence of the activator only marginally enhanced the K_D and ∆H (Fig. 3B). Moreover, compared to when it was pre-bound with ATP, QN-mcGAS^cat•dsDNA pre-saturated with either dATP bound GTP at least 5-fold more weakly (Fig. 3B vs. Fig. 2B). To determine the mechanism, we soaked 2′- or 3′-dATP to WT-mcGAS^cat•dsDNA crystals. Here, the catalytic residues retained the active conformation; however, the ribose of bound 2′-dATP was flipped out of the active site, apparently resulting in distorted α-phosphate and the loss of the second Mg²⁺ (Fig. 3C). Furthermore, similar to nonproductively bound ATP in dsDNA-free cGAS (Fig. 1G), the densities for the ribose of 3′-dATP and the second Mg²⁺ were missing (Supplementary Fig. 3B). These results demonstrate that cGAS utilizes intricate protein•NTP and intra-NTP interactions for productive substrate binding, concomitantly permitting the distinction between ATP and dATPs.

**Fig. 3: Molecular bases for ATP recognition at Site-1.**

Molecular basis for distinguishing ATP from GTP at Site-1

Two GTPs bind essentially in the same manner as ATP/GTP in QN-mcGAS^cat•dsDNA (Supplementary Fig. 2A, B). Nevertheless, although Tyr⁴²¹ stacks on both ATP and GTP, unlike ATP, the -NH₂ of guanine was in proximity to His⁴⁶⁷ near Site-1 (Fig. 3D; ≤3 Å). We reasoned that, in our buffer conditions (pH 7.4 for reaction (i.e., physiological) and 6.5 for crystallization), the interaction between the -NH₂ (H-bond donor) of guanine and the partially protonated imidazole ring of His⁴⁶⁷ could be unfavorable due to a steric clash. Moreover, in the ATP/GTP-bound structure, the -NH₂ of ATP at Site-1 appeared to donate a H-bond to the α-phosphate oxygen of GTP at Site-2; however, such an inter-NTP interaction beyond base-stacking is unlikely when GTP binds at Site-1 (Fig. 3D; the carboxyl oxygen of GTP is a H-bond acceptor). To test functional implications, we soaked QN-mcGAS^cat•dsDNA with inosine triphosphate (ITP)/GTP or ATP/ITP. In the ITP/GTP-soaked structure, judged by the electron density for the -NH₂ of GTP, it appeared that ITP and GTP bind at Site-1 and Site-2, respectively (Fig. 3E). Moreover, the ATP/ITP-soaked structure suggested that ATP and ITP are bound at Site-1 and Site-2, respectively (Fig. 3F); the carboxyl oxygen of ITP appeared to accept a H-bond from Ser³⁶⁶ as seen from GTP, and ATP fails to bind stably at Site-2 in all known structures (see also below for the pppGpA-bound structure). These results support the idea that the -NH₂ of GTP is disfavored at Site-1 while the -NH₂ of ATP at Site-1 provides a favorable interaction for site-specific ATP/GTP binding. Consistent with the idea that GTP is disfavored over ITP at Site-1, mcGAS^cat was more active toward ITP/ITP vs. GTP/GTP (Fig. 3G). Moreover, supporting the idea that His⁴⁶⁷ interferes with GTP binding at Site-1, although H467A-mcGAS^cat was less active toward ATP/GTP than WT, it showed a 2-fold higher activity toward GTP/GTP (Fig. 3H). Resolving reaction products via HPLC revealed that, unlike WT, H467A-mcGAS^cat produced off-pathway pppGpG even with ATP/GTP (Fig. 3I); the H467A mutant was indeed more active toward GTP/GTP (Fig. 3J, K). Additionally, supporting that the -NH₂ of ATP contributes positively to the recognition of the ATP/GTP pair by donating the H-bond to the α-phosphate of GTP, mcGAS^cat showed a significantly higher NTase activity toward ATP/GTP vs. purine triphosphate (PuTP)/GTP or ATP/PuTP (Fig. 3G; PuTP lacks the NH₂ of ATP).

The GTP binding at Site-2 dictates the signature 2′-5′-linkage specificity

It is thought that Thr¹⁹⁷ at Site-2 recognizes the -NH₂ of GTP, thereby prescribing its specificity⁴ (Fig. 4A). However, not only did mcGAS^cat show a robust NTase activity toward ATP/ITP, but ITP/ITP was also more favored over GTP/GTP (Fig. 3G), indicating that the interaction between the -NH₂ of GTP and Thr¹⁹⁷ is not critical. We noted that Arg³⁶⁴ appears to stabilize GTP at Site-2 by donating H-bonds at the Hoogsteen edge (Fig. 4A). However, the guanidium of Arg³⁶⁴ would repel the positive dipole of adenine -NH₂, thereby suppressing ATP binding at Site-2. Supporting this notion, ATP/PuTP showed 3-fold higher NTase activity vs. ATP/ATP (Fig. 3G). Moreover, compared to WT, R364A-mcGAS^cat•dsDNA showed 33-fold and 10-fold lower NTase activities against ATP/GTP and GTP/GTP, respectively; however, it was 4-fold more active toward ATP/ATP (Fig. 4B). Additionally, unlike WT-mcGAS^cat which generated 2′-5′-linked pppApA, R364A-mcGAS^cat predominantly produced 3′-5′-linked pppApA (Fig. 4C; S1 nuclease resistant vs. susceptible).

**Fig. 4: Mechanisms that define the GTP specificity at Site-2 and Mn²⁺ utilization.**

Although the second Mg²⁺ is absent in our ATP/GTP-bound QN-mcGAS^cat•dsDNA structure, not only is the 2′-OH of GTP already inline to attack the α-phosphate of ATP at Site-1, but it also makes a H-bond with Asp³⁰⁷, the residue that deprotonates the nucleophile (Fig. 4A). These interactions then place the 3′-OH of GTP too far to interact with Asp³⁰⁷ (4 Å), barring it from participating in the first linkage formation. Indeed, mcGAS^cat was essentially inactive toward ATP/2′-dGTP in our NTase assay (Supplementary Table 10). Moreover, analogous to 2′/3′-dATP, dsDNA failed to enhance the affinity or ∆H of 2′-dGTP binding; QN-mcGAS^cat•dsDNA pre-saturated with ATP also failed to enhance 2′-dGTP binding (Fig. 4D vs. Fig. 2B). These results suggest that the H-bond between Asp³⁰⁷ and the 2′-OH of GTP is integral to dsDNA-dependent productive substrate binding. To further delineate the mechanism, we soaked WT-mcGAS^cat•dsDNA crystals with ATP/2′-dGTP. Here, although 2′-dGTP at Site-2 was bound in the same manner as GTP (and 1Mg²⁺•ATP at Site 1; Fig. 4E), the densities for the second Mg²⁺ and the bottom half of the deoxyribose were missing (Fig. 4E). These results consistently suggest that pre-establishing the H-bond between Asp³⁰⁷ and the 2′-OH of GTP is essential for productive substrate binding, prescribing the signature 2′-5′ linkage specificity, and capturing/retaining the second divalent metal.

Mn²⁺ is preferentially incorporated as the catalytic metal

Although cGAS employs two divalent metals in catalysis⁴, their roles remain poorly defined. Our structures consistently suggest that only one Mg²⁺ is necessary for productive substrate binding (e.g., Fig. 4A), while the second Mg²⁺ drives chemistry (i.e., neutralizing negative charges upon deprotonating 2′- or 3′-OH by Asp³⁰⁷). Of note, cytosolic Mn²⁺ released from damaged organelles accentuates the dsDNA sensing activity of cGAS in cells and in vivo²⁰. Furthermore, it was reported that Mn²⁺ activates cGAS in the absence of dsDNA without involving the canonical catalytic residues by invoking an inverted intermediate, albeit necessary [Mn²⁺] is unphysiologically high²¹ (>1 mM vs. ≤ 50 µM in vivo). Nevertheless, although we also found that high [Mn²⁺] can activate cGAS without dsDNA¹⁸, we showed that excess Mg²⁺ and dsDNA binding synergistically allow cGAS to incorporate physiological concentrations of Mn²⁺ to boost its catalytic activity¹⁸. To elucidate the mechanism by which dsDNA-bound cGAS incorporates Mg²⁺ and Mn²⁺ in catalysis, we first soaked WT-mcGAS^cat•dsDNA crystals with Mn²⁺ + ATP and found 2Mn²⁺•ATP as seen from the 2Mg²⁺•ATP-bound structure (Fig. 4F and Supplementary Fig. 2E). Next, to determine whether Mn²⁺ is preferred at either metal binding site, we soaked WT-mcGAS^cat•dsDNA crystals with ATP pre-complexed with increasing concentrations of Mn²⁺ (0.015-1 mM) in the presence of excess Mg²⁺ (10 mM); the presence of Mn²⁺ was tracked via its anomalous signal at 1.8961 Å (6539 eV) during data collection. Here, we found that cGAS specifically favors Mn²⁺ over Mg²⁺ by ≥100-fold at the catalytic metal binding site (Fig. 4G); the binding here also cooperatively increased the affinity for Mn²⁺ at the first site (Fig. 4G and Supplementary Fig. 3C). Together, we concluded that Mn²⁺ accelerates the signaling activity of cGAS•dsDNA by preferentially incorporated as the catalytic metal over Mg²⁺ without requiring any other reaction pathways.

Mechanisms that govern the second 3′-5′ linkage specificity and product fidelity

cGAS can generate linear dinucleotides from various NTPs, but it remains unknown why only ATP/GTP can be readily cyclized^4,5,18. Indeed, cGAS failed to produce cyclic dinucleotides from GTP/GTP, ATP/ATP, ATP/ITP, and ITP/GTP (Figs. 3J and 4C and Supplementary Fig. 4). To decipher the cyclization mechanism, we sought to capture the cognate pppGpA intermediate in crystallo. We noted that Mn²⁺ can activate QN-mcGAS^cat•dsDNA to some extent (the catalytic Asp³⁰⁷ that deprotonates the OH for nucleophilic attack is still intact). Soaking these crystals with ATP/GTP precomplexed with 5 mM Mg²⁺/ 1 mM Mn²⁺ resulted in 1Mn²⁺•pppGpA trapped in the active site (Fig. 5A and Supplementary Fig. 5A; the second metal was missing likely due to mutation). Strikingly, compared to the guanine of GTP at Site-2, the adenine of pppGpA was rotated ~30°, exposing the N1 imine as the H-bond acceptor from Ser³⁶⁶ instead of the carboxyl of guanine (Fig. 5A, B; the N1 amide in guanosine is a H-bond donor). Consequently, the 3′-OH of adenosine was aligned to be deprotonated by Asp³⁰⁷ for the subsequent nucleophilic attack, while the 2′-OH was completely occluded (Fig. 5A, B). Moreover, Arg³⁶⁴ was pushed away by the repulsive positive dipole of adenine (Fig. 5A, B). Similar to the β-phosphate•3′-OH interaction seen from productive ATP binding, the -NH₂ of guanine was then no longer close to His⁴⁶⁷ but appeared to donate an intramolecular H-bond to the α-phosphate to stabilize the intermediate for cyclization (Fig. 5A).

**Fig. 5: Molecular determinants that drive the selective cyclization of pppGpA.**

To test whether such a precise NTP-specific coordination is critical for cyclization, we soaked Mg²⁺+ITP/GTP, ATP/ITP, or GTP alone into WT-mcGAS^cat•dsDNA, all of which produced respective linear dinucleotides bound at the active site. First, pppGpI (Fig. 5C) was bound analogous to the previously reported pppGpG⁴ (Supplementary Fig. 5B, C). Here, the catalytic Mg²⁺ was absent, and unlike pppGpA where the 2′-OH was occluded, both 2′-OH and 3′-OH were near Asp³⁰⁷ (Fig. 5C bottom and Supplementary Fig. 5C), which likely have sterically hindered the second metal binding (also possibly obscuring the deprotonation target). Next, we identified pppIpA based on how the nucleoside at Site-2 resembles the adenine of pppGpA (Fig. 5D and Supplementary Fig. 5D); Arg³⁶⁴ was also pushed away as seen from pppGpA and the 3′-OH was in position for catalysis. Nevertheless, the second Mg²⁺ was again missing, and the α-phosphate of pppIpA appeared misaligned compared to pppGpA (Fig. 5D, bottom). These observations suggest that the H-bond between the α-phosphate and the -NH₂ of guanine is important for positioning the metal•intermediate complex for cyclization. Even more strikingly, when we soaked GTP alone, we found a different conformation of pppGpG in which the guanosine was no longer bound at Site-2 but trapped via intra-NTP H-bonds despite both Mg²⁺ ions being present for catalysis (Fig. 5E and Supplementary Fig. 5E). Soaking Mn²⁺•GTP yielded the same “flipped-out” pppGpG seen from soaking Mg²⁺•GTP (Supplementary Fig. 5F), indicating that this apparent proof-reading mechanism is not metal-dependent. Together, we concluded that cGAS precisely positions the cognate pppGpA intermediate not only for the 3′-5′ linkage specificity, but also for the likelihood of cyclization.

Human and mouse cGAS enzymes display different active site reactivity and fidelity

Considering the primary sequence similarity at the active site (only three residue differences between mouse and human: Ile³⁰⁹/Thr³²¹, Cys⁴¹⁹/Ser⁴³⁴, and His⁴⁶⁷/Asn⁴⁸²), we propose that our findings are broadly relevant across vertebrates. Of note, it was recently reported that recombinant mcGAS^cat is more active than hcGAS^cat because the former relies less on dsDNA length for binding/activation¹⁴. However, others have reported that mcGAS and hcGAS display similar dsDNA length-dependence²². Indeed, we found that recombinant mcGAS still binds dsDNA in a length-dependent manner (Fig. 6A); however, the mouse enzyme was at least 5-fold more active than hcGAS (Fig. 6B). Interestingly, compared to hcGAS^cat, mcGAS^cat was more active toward GTP/GTP, but less active toward ATP/ATP (Fig. 6C); the pppApA linkage was predominantly 3′-5′ for hcGAS^cat unlike mcGAS^cat (Supplementary Fig. 6). These results suggested the two cGAS enzymes have different active site reactivity and promiscuity. We then noted that Ile³⁰⁹ near GTP at Site-2 is Thr³²¹ in hcGAS (Fig. 6D), the former of which could stack better against the base than the latter. Indeed, T321I-hcGAS^cat showed markedly higher NTase activities against ATP/GTP and GTP than WT, while that for ATP was reduced (Fig. 6C). Moreover, pppApA resulting from T321I-hcGAS^cat was 2′-5′ linked (mcGAS-like), while I309T-mcGAS^cat produced an increased amount of the 3′-5′-linked dinucleotide (hcGAS-like; Supplementary Fig. 6). These observations indicate that Ile³⁰⁹ of mcGAS suppresses ATP binding while stabilizing GTP at Site-2.

**Fig. 6: Mechanisms that prescribe the intrinsically higher catalytic activity of mcGAS vs. hcGAS.**

We noted that T321I-hcGAS^cat is defective in cyclization although it efficiently generates the pppGpA intermediate (Fig. 6E). Further examining pppGpA-bound mcGAS^cat indicated that Cys⁴¹⁹, which forms a H-bond with the ribose moiety of guanosine, is Ser in human (Fig. 6D). T321I/S431C-hcGAS^cat was not only even more active (Fig. 6C), but also readily cyclized pppGpA into cGAMP (Fig. 6E). Finally, when transfected into HEK293T cells, mcGAS^FL and hcGAS^FL showed comparable human STING-dependent IFN-Luc (luciferase) activities (Fig. 6F), indicating that in cellulo signaling outputs of cGAS from different species are likely similar despite varying catalytic potentials at the active site. Nevertheless, consistent with our biochemical assays, the signaling activity of T321I- hcGAS^FL was compromised, but T321I/S431C-hcGAS^FL displayed a significantly higher IFN-Luc activity (Fig. 6F). Together, we concluded that human and mouse cGAS enzymes have different active site reactivity and promiscuity. Moreover, we identify additional key interactions that regulate the site-specific substrate binding (Ile³⁰⁹ for GTP binding at Site-2 and ensuring the 2′-5′ linkage) and the cyclization efficiency (Cys⁴¹⁹ for stabilizing pppGpA).

Discussion

Not only is cGAS essential to the innate immune system^1,2,3, but cGAS-like receptors are also central to the host defense across a wide range of organisms⁹. We present here a unifying catalytic mechanism of mammalian cGAS in which the dsDNA-dependent disorder-to-order transition generates a rigid yet adaptable lock to specifically recognize its cognate substrates and intermediate (Fig. 7).

**Fig. 7: The catalytic mechanism of cGAS.**

The role of dsDNA and dimerization in cGAS activation

Resting cGAS would assume an array of inactive conformations without being fixed into a specific autoinhibited state (Figs. 1 and 7). NTPs can bind cGAS without dsDNA; however, the enzyme remains only basally active as it can rarely achieve the site-specific metal•NTP binding necessary for catalysis (Figs. 1 and 2). Dimerization-coupled dsDNA binding then establishes a web-like network of highly sensitive protein•NTP, intra-NTP, and inter-NTP interactions leading to productive nucleotide transactions (Figs. 2 and 4).

Mechanisms that underpin NTP and linkage specificities

We propose that the productive 1Mg²⁺•triphosphate binding at Site-1 provides the most significant interaction energy (∆H; Fig. 2). The intramolecular H-bond between the β-phosphate and the 3′-OH of ATP stabilizes its α-phosphate electrophile and the second divalent metal for catalysis, while precluding any reaction with dATP (Fig. 3). Moreover, His⁴⁶⁷ sterically suppresses GTP binding at Site-1, while ATP binding here allows its differentially positioned -NH₂ (positive dipole) to interact favorably with the triphosphates of GTP at Site-2 (Fig. 3). Considering that ATP outnumbers GTP by ~6-fold in vivo²³, the additional 18-fold preference for ATP created by these interactions (3-fold from ATP/GTP vs. PuTP/GTP for WT and ~6-fold from WT vs. H476A against ATP/GTP; Fig. 3G) would synergistically increase the likelihood of ATP binding at Site-1 by ~110-fold over GTP. At Site-2, Ser³⁶⁶ acts as the most energetically important H-bond donor to GTP, and Ile³⁰⁹ and Arg³⁶⁴ fix GTP in place for the 2′-5′ bond formation while discriminating against ATP (Figs. 4, 6, and 7). Also of note, pre-establishing the H-bond between the 2′-OH of GTP and Asp³⁰⁷ is critical not only for aligning the obligate nucleophile for the first linkage formation, but also for retaining the catalytic divalent metal (Figs. 4 and 7).

For cyclization, pppGpA would also be first recognized by the favorable 1Mg²⁺•triphosphate•enzyme interactions at Site-1. This anchoring interaction then allows the adenine of the intermediate to accept a H-bond from Ser³⁶⁶ at Site-2 while overcoming the Ile³⁰⁹/Arg³⁶⁴ blockade (Figs. 5–7). This binding mechanism then precisely positions the 3′-OH of adenosine for cyclization while excluding its 2′-OH (Fig. 5). Additionally, similar to the 3′-OH and β-phosphate of ATP, the -NH₂ of guanine (pppGpA) is crucial for correctly positioning the target α-phosphate to retain the catalytic metal for cyclization; Cys⁴¹⁹ also helps stabilizing the intermediate (Figs. 5–6). Importantly, our results consistently indicate that any deviations from such a precise coordination would prevent cyclization (Figs. 5 and 7): any misalignments of the 3′-OH nucleophile (pppGpI) or the α-phosphate electrophile (pppIpA) would disrupt catalytic metal binding, and the enzyme-mediated intra-dinucleotide interactions would trap pppGpG in a nonproductive state.

Metal utilization

We show that capturing/retaining the second divalent metal is crucial for distinguishing NTPs from d-NTPs, defining the signature 2′-5′-linkage, and selectively cyclizing the cognate intermediate (Figs. 3–5 and 7). Moreover, Mn²⁺ is specifically preferred as the second catalytic metal over Mg²⁺ (Fig. 4F-G), explaining how the emergence of cytosolic Mn²⁺ from damaged organelles can potentiate the signaling activity of cGAS²⁰. Importantly, when cGAS is bound to dsDNA, Mn²⁺-mediated catalysis does not invoke alternate mechanisms (Fig. 4F-G). We reason that Mn²⁺ being a transitional metal with more flexible polarizable options than alkaline Mg²⁺ likely allows the former to bind more tightly as the catalytic metal.

Of human and of mice

In recent years, species-dependent specificities of innate immune pathways have garnered much attention. For instance, small molecule modulators arising from targeting mcGAS and mouse STING have failed to be effective toward the human counterparts^24,25,26,27. We show that, compared to hcGAS, not only does mcGAS enforce a stricter GTP specificity at Site-2 (Ile³⁰⁹ vs. Thr³²¹; Fig. 6), but it also cyclizes pppGpA more efficiently (Cys⁴¹⁹ vs. Ser⁴³⁴; Fig. 6). Nonetheless, our cellular assays suggest that the species-dependent active site reactivity per se does not dictate the overall signaling output of cGAS (Fig. 6F). It is still possible that there might be small but functional differences that our transfection system failed to capture. Nonetheless, we speculate that other factors such as higher dimerization propensity¹⁵ and the ability to form phase-separated condensates¹¹ compensate hcGAS for its lower catalytic efficiency in vivo.

Future studies focusing on detailed kinetic analyses coupled with molecular dynamics simulations will provide a deeper understanding of cGAS catalytic mechanisms, which include exactly what the rate-limiting step is and how the intermediate is released and recaptured for cyclization.

Methods

Protein expression and purification

Human cGAS (full-length (hcGAS^FL) and the catalytic domain (hcGAS^cat, residue 157-522)) and mouse cGAS (mcGAS^FL and mcGAS^cat (residue 147-507)) were cloned into the pET28b vector (Novagen) with an N-terminal 6×His-MBP-tag containing a TEV protease cleavage site. Plasmids encoding wild-type and mutant cGAS constructs were expressed in E. coli BL21-(DE3) cells at 16 °C using 0.3 mM IPTG for 18 hours. Cells were lysed by sonication in the lysis buffer (20 mM Na-phosphate at pH 7.5, 500 mM NaCl, 40 mM imidazole, and 0.5 mM TCEP) and purified by the Ni-NTA chromatography. The fusion protein was cleaved by the TEV protease at 4 °C overnight and the MBP tag was removed by binding them to the Ni-NTA column and amylose column. The untagged protein (flow-through) was further purified by size-exclusion chromatography (Superdex 200 for full-length, Superdex 75 for catalytic domain, Cytiva). Human cGAS protein variants were concentrated and stored in 20 mM Tris-HCl at pH 7.5, 300 mM KCl, and 0.5 mM TCEP. Mouse cGAS protein variants were concentrated in 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 0.5 mM TCEP. All purified cGAS proteins were flash-frozen in liquid nitrogen and stored at −80 °C.

Crystallization

All human cGAS crystals were grown by vapor diffusion in hanging drops at 25 °C in 24-well plates within 3 days. apo-WT-hcGAS^cat crystals were obtained by mixing 1 μl of protein (8 mg/ml) with 1 μl of reservoir solution containing 0.1 M Tris pH 8.5, and 8 % (w/v) PEG8000. For co-crystallization of WT-hcGAS^cat with nucleotides ATP/GTP or GTP, 8 mg/ml protein was firstly preincubated with 2 mM nucleotides, and then 1 μl of protein -nucleotides mixture was mixed with 1 μl of reservoir solution containing 5 mM magnesium chloride, 0.1 M Tris-HCl pH 8.5, and 8% (w/v) PEG8000.

All mouse cGAS crystals were obtained using the hanging-drop vapor-diffusion method at 4 °C in a 24-well plate within one week. For crystals of the apo-mcGAS^cat (WT or QN), 10 mg/ml of protein was mixed with 1.2 mM of 16-bp dsRNA (5′- UCU GUA CAU GUA CAG A-3′) at a ratio of 1:1.2 and incubated on ice for 30 min. 1 μl of the mixture was mixed with 1 μl of reservoir solution containing 0.2 M magnesium acetate tetrahydrate, 0.1 M sodium cacodylate trihydrate (pH 6.5), and 20% (w/v) PEG8000. mcGAS^cat was crystallized without dsRNA. For crystallization of mcGAS^cat•dsDNA, 10 mg/ml of protein (WT or QN) was mixed with 1.2 mM of palindromic 18-bp dsDNA (5′-ATC TGT ACA TGT ACA GAT-3′) at the 1:1.2 molar ratio and incubated on ice for 30 min. 1 μl of the protein•DNA solution was then mixed with 1 μl of reservoir solution containing 0.2 M ammonium acetate, 32% (v/v) MPD, and 0.1 M Bis-Tris pH 6.5.

Crystal preparation (soaking)

The mouse cGAS^cat and human cGAS^cat crystals were harvested and cryoprotected using the reservoir buffer supplemented with 20% (v/v) ethylene glycol before flash-frozen in liquid nitrogen. For substrate binding or reaction in crystals, mcGAS^cat or mcGAS^cat•dsDNA crystals (WT or QN) were first transferred to the reservoir buffer supplemented with 2 mM NTPs (ATP, GTP, ATP/GTP, ATP/ITP, ITP/GTP ATP/2′-dGTP, 2′-dATP, and 3′d-ATP) plus 5 mM MgCl₂ (or MnCl₂ when noted)) then incubated at 4 °C for 20 min before flash-frozen in liquid nitrogen without cryoprotection. For tracking the anomalous signal of Mn²⁺, WT-mcGAS^cat•dsDNA crystals were transferred to the reservoir buffer supplemented with 4 mM ATP, 10 mM MgCl₂, and varying MnCl₂ (1000, 500, 200, 100, 40, 15, and 0 μM) for 20 min at 4 °C then flash-frozen in liquid nitrogen without cryoprotection.

Data collection and structure determination

All diffraction data were collected from the National Synchrotron Light Source II (NSLS-II) at Brookhaven at 100 K on beamlines 17-ID-1 (AMX) or 17-ID-2 (FMX). For X-ray anomalous scattering, the data were collected at the absorption edge of manganese (6539 eV) on 17-ID-2 (FMX). Data were indexed and integrated with XDS²⁸ and scaled and merged with AIMLESS²⁹. All the structures were solved by molecular replacement using the MOLREP module in the CCP4i suite²⁹. We used an apo-mcGAS^cat monomer (from PDB: 4lez) and the 1:1 mcGAS^cat•dsDNA complex (without bound cGAMP and the second cGAS monomer and dsDNA in 4lez) as the initial search models for mcGAS^cat and mcGAS^cat•dsDNA, respectively. For hcGAS^cat, we used PDB ID: 4k8v as the search model. We ensured that all search models have empty active sites. The subsequent model building and structure refinement were conducted in Coot³⁰ and Phenix_Refine³¹. The Mn²⁺ anomalous maps were calculated using REFMAC5 and quantified in Coot using the contour level (σ). The structure figures were prepared using CCP4MG²⁹ and PyMOL (Schrödinger). Data collection and refinement statistics are summarized in Supplementary Tables 1–9. The coordinates and structural factors of all structures have been deposited to the Protein Data Bank with the respective PDB IDs listed in Supplementary Tables 1–9.

Pyrophosphatase-coupled cGAS activity assay

The catalytic activity of cGAS (NTase) was measured by the pyrophosphatase-coupled assay tracking the production of PP_i as reported^15,18,19. Briefly, 200 nM cGAS was incubated with 50 nM of E. coli pyrophosphatase, 200 μM of NTPs (ATP/GTP, ATP/ITP, PuTP/GTP, PuTP/ATP, ATP/ATP, GTP/GTP, ITP/ITP, ATP/2′-dGTP, 3′-ATP/GTP, 2′-dATP/GTP) plus 200 nM 60-bp dsDNA in the reaction buffer (25 mM Tris-acetate (Ac) pH 7.4, 125 mM KAc, 1 mM TCEP, 5 mM MgAc₂ at pH 7.4, and 5% glycerol) at 25 ± 2 °C (RT). 60-bp dsDNA sequence: 5′-ATG GAA GAT CCG CGT AGA AGG ACG ACG GCG CCA CGC GCT AAG AAG CCG TCC GCG AAG CGC-3′. A total of 4 aliquots were taken from the reaction at designated time points and mixed with an equal volume of the quench buffer (the reaction buffer plus 25 mM EDTA) in a 384-well plate (Corning). Quenched reactions were mixed with 10 µl malachite green solution and incubated for 45 min at RT. The absorbance at 620 nm was recorded using a Tecan M1000 plate reader, and data were compared to an internal standard curve of inorganic phosphate to determine the concentration of phosphate released in each well. Phosphate concentrations of control reactions without cGAS were subtracted from reactions containing the enzyme. Apparent catalytic rates were calculated from the slopes of control-subtracted phosphate concentrations over time.

Isothermal titration calorimetry

Experiments were performed using a MicroCal VP-ITC microcalorimeter (Malvern) under the reaction buffer containing 20 mM HEPES at pH 7.5, 125 mM KAc, 5 mM MgCl₂, and 1 mM TCEP. For nucleotide-cGAS interactions, 0.8 mM nucleotide in the injection syringe was titrated into the sample cell loaded with 10 μM cGAS protein with or without 10 μM 18-bp dsDNA (5′-ATC TGT ACA TGT ACA GAT-3′). For interactions of nucleotide with cGAS pre-bound with NTPs, 0.8 mM nucleotide in the injection syringe was titrated into the sample cells loaded with 10 μM cGAS and 120 μM nucleotide with or without 10 μM 18-bp dsDNA. Titrations consisted of 20 injections (size 15 μl, duration 18 sec), with 300 sec equilibration time. The data were analyzed using Origin 7.0.

Analytical fractionation of cGAS reactions by HPLC

1 μM cGAS, 1,000 μM NTPs, and 1 μM 60-bp dsDNA were incubated in 50 μl reaction buffer (25 mM Tris-Ac pH 7.4, 125 mM KAc, 1 mM TCEP, 5 mM MgAc₂ at pH 7.4, and 5% glycerol) for the indicated time at 25 ± 2 °C. Reactions were quenched with 25 mM EDTA, diluted to 80 μl with water, and filtered with a 3 kDa cutoff centrifugal filter (VWR). The flow-through was then mixed 1:1 with HPLC buffer A (see below) and fractionated on the Waters 1525 HPLC system using a Poroshell 120 SB-C18 column (2.7 μm; 4.6 × 100 mm) with a 100 μl sample loop. Absorbance at 260 nm was tracked with a Waters 2996 photodiode array detector. To degrade 3′-5′-linked dinucleotides, reaction samples were first passed through in 3 kDa cutoff centrifugal filters, then treated with the S1 nuclease (2 units/µl final; ThermoFisher) for 15 min at 37 °C. All reactions were then quenched with 25 mM EDTA and filtered again. HPLC gradient scheme: 0% B from 0–1 min, linear increase to 50% B from 1–27 min, and linear increase to 100% B from 27–28 min. HPLC gradient scheme to resolve pppGpG and ATP (related to Fig. 3):0% B from 0–1 min, linear increase to 50% B from 1–37 min, and linear increase to 100% B from 37–38 min. Buffer A: 100 mM potassium phosphate monobasic, 5 mM tetrabutylammonium, final pH 6.0; Buffer B: buffer A supplement with 30% acetonitrile. Peak intensities were integrated by the Waters Empower^TM 3 software.

Fluorescence-anisotropy binding assays

An increasing amount of cGAS was added to fluorescein-amidite (FAM)-labeled dsDNA fragments (5 nM final). Fluorescence anisotropy (FA) signal was recorded with a Tecan M1000 plate reader as previously reported^15,18. Changes in FA were plotted as a function of cGAS concentration and fit to the Hill equation. 20-bp FAM-dsDNA: 5′-TAA GAC ACG ATG CGA TAA AA-3′. 39-bp FAM-dsDNA: 5′-TAA GAC ACG ATG CGA TAA AAT CTG TTT GTA AAA TTT ATT-3′. 72-bp FAM-dsDNA: 5′-TAA GAC ACG ATG CGA TAA AAT CTG TTT GTA AAA TTT ATT AAG GGT ACA AAT TGC CCT AGC ACA GGG GTG GGG-3′.

Measuring cGAS signaling activities in HEK293T cells using the luciferase reporter

The Human Embryonic Kidney (HEK) 293T cells (ATCC, CRL-11268) were maintained in the DMEM high glucose medium (ThermoFisher) supplemented with 10% FBS at 37 °C and 5% CO₂. Cells were seeded in a 24-well plate (6 ×10⁴ cells/well) and incubated overnight. 50 ng of pCMV plasmids encoding empty vector or various cGAS^FL variants were then transfected using Lipofectamine 3000 (Invitrogen), along with 5 ng of Renilla Luciferase plasmid, 10 ng of plasmid encoding human STING, and 10 ng of plasmid encoding Firefly Luciferase under an IFN-Iβ promoter. Transfected cells were incubated for 24 hours, washed once with PBS, and lysed with passive lysis buffer (Promega). Lysates were transferred to a white 96-well plate with a reflective flat bottom (Corning) and analyzed for Firefly Luciferase and Renilla Luciferase activities using the Dual-Luciferase Reporter Assay System (Promega) on a Synergy H1 plate reader equipped with the dual-injector system (BioTek). The firefly output was divided by the Renilla output and normalized to the empty vector control; see also ref. ¹⁸.

Statistics and reproducibility

All biochemical experiments were conducted at least three times. No data were excluded from the analyses. p-values were determined using a 2-tailed t-test with equal variances.

Reporting summary

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

Data availability

The structure coordinates generated in this study have been deposited in the Protein Data Bank under the following accession codes: 8SHU (apo mcGAS WT). 8SHK (mcGAS WT + ATP). 8SHY (mcGAS QN + ATP). 8SHZ (apo hcGAS WT). 8SI0 (hcGAS WT + cGAMP). 8SJ8 (hcGAS WT + pppGpG). 7UUX (cGAS QN:dsDNA + ATP). 8SKT (cGAS WT:dsDNA + ATP/Mn²⁺). 7UYQ (cGAS QN:dsDNA + GTP). 7UXW (cGAS QN:dsDNA + ATP/GTP). 8G1J (cGAS QN:dsDNA + ATP/ITP). 8G10 (cGAS QN:dsDNA + GTP/ITP). 7V0W (cGAS QN:dsDNA + pppGpA/Mn²⁺). 7UYZ (cGAS WT:dsDNA + pppGpG/Mg²⁺). 7UZR (cGAS WT:dsDNA + pppGpG/Mn²⁺). 8EAE (cGAS WT:dsDNA + pppGpI). 8G23 (cGAS WT:dsDNA + pppIpA). 8SJ0 (cGAS:dsDNA + 2′-dATP). 8SJ1 (cGAS:dsDNA + 3′-dATP). 8SJ2 (cGAS:dsDNA + 2′-dGTP). 8GIM (cGAS:dsDNA + ATP·10 mM Mg²⁺). 8GIN (cGAS:dsDNA + ATP·10 mM Mg²⁺ + 15 μM Mn²⁺). 8GIP (cGAS:dsDNA + ATP·10 mM Mg²⁺ + 40 μM Mn²⁺). 8GIO (cGAS:dsDNA + ATP·10 mM Mg²⁺ + 100 μM Mn²⁺). 8GIR (cGAS:dsDNA + ATP·10 mM Mg²⁺ + 200 μM Mn²⁺). 8GIS (cGAS:dsDNA + ATP·10 mM Mg²⁺ + 500 μM Mn²⁺). 8GIT (cGAS:dsDNA + ATP·10 mM Mg²⁺ + 1000 μM Mn²⁺). All quantitative data generated in this study are provided in the Supplementary Information file and the Source Data. Source data is provided as a Source Data file. Source data are provided with this paper.

References

Sun, L., Wu, J., Du, F., Chen, X. & Chen, Z. J. Cyclic GMP-AMP synthase is a cytosolic DNA sensor that activates the type I interferon pathway. Science 339, 786–791 (2013).
Article ADS CAS PubMed Google Scholar
Hopfner, K. P. & Hornung, V. Molecular mechanisms and cellular functions of cGAS-STING signalling. Nat. Rev. Mol. Cell Biol. 21, 501–521 (2020).
Article CAS PubMed Google Scholar
Chen, Q., Sun, L. & Chen, Z. J. Regulation and function of the cGAS-STING pathway of cytosolic DNA sensing. Nat. Immunol. 17, 1142–1149 (2016).
Article CAS PubMed Google Scholar
Gao, P. et al. Cyclic [G(2′,5′)pA(3′,5′)p] is the metazoan second messenger produced by DNA-activated cyclic GMP-AMP synthase. Cell 153, 1094–1107 (2013).
Article CAS PubMed PubMed Central Google Scholar
Ablasser, A. et al. cGAS produces a 2′-5′-linked cyclic dinucleotide second messenger that activates STING. Nature 498, 380–384 (2013).
Article ADS CAS PubMed PubMed Central Google Scholar
Li, X. et al. Cyclic GMP-AMP synthase is activated by double-stranded DNA-induced oligomerization. Immunity 39, 1019–1031 (2013).
Article CAS PubMed Google Scholar
Zhang, X. et al. The cytosolic DNA sensor cGAS forms an oligomeric complex with DNA and undergoes switch-like conformational changes in the activation loop. Cell Rep. 6, 421–430 (2014).
Article CAS PubMed PubMed Central Google Scholar
Decout, A., Katz, J. D., Venkatraman, S. & Ablasser, A. The cGAS-STING pathway as a therapeutic target in inflammatory diseases. Nat. Rev. Immunol. 21, 548–569 (2021).
Article CAS PubMed PubMed Central Google Scholar
Kranzusch, P. J. cGAS and CD-NTase enzymes: structure, mechanism, and evolution. Curr. Opin. Struct. Biol. 59, 178–187 (2019).
Article CAS PubMed PubMed Central Google Scholar
Holleufer, A. et al. Two cGAS-like receptors induce antiviral immunity in Drosophila. Nature 597, 114–118 (2021).
Article ADS CAS PubMed Google Scholar
Du, M. & Chen, Z. J. DNA-induced liquid phase condensation of cGAS activates innate immune signaling. Science 361, 704-709 (2018).
Zhou, W., Mohr, L., Maciejowski, J. & Kranzusch, P. J. cGAS phase separation inhibits TREX1-mediated DNA degradation and enhances cytosolic DNA sensing. Mol. Cell 81, 739–755.e7 (2021).
Article CAS PubMed PubMed Central Google Scholar
Kranzusch, P. J., Lee, A. S., Berger, J. M. & Doudna, J. A. Structure of human cGAS reveals a conserved family of second-messenger enzymes in innate immunity. Cell Rep. 3, 1362–1368 (2013).
Article CAS PubMed PubMed Central Google Scholar
Zhou, W. et al. Structure of the human cGAS-DNA complex reveals enhanced control of immune surveillance. Cell 174, 300–311.e11 (2018).
Article CAS PubMed PubMed Central Google Scholar
Hooy, R. M. & Sohn, J. The allosteric activation of cGAS underpins its dynamic signaling landscape. Elife 7, 570 (2018).
Article Google Scholar
Civril, F. et al. Structural mechanism of cytosolic DNA sensing by cGAS. Nature 498, 332–337 (2013).
Article ADS CAS PubMed PubMed Central Google Scholar
Perozzo, R., Folkers, G. & Scapozza, L. Thermodynamics of protein-ligand interactions: history, presence, and future aspects. J. Recept. Signal Transduct. Res. 24, 1–52 (2004).
Article CAS PubMed Google Scholar
Hooy, R. M., Massaccesi, G., Rousseau, K. E., Chattergoon, M. A. & Sohn, J. Allosteric coupling between Mn2+ and dsDNA controls the catalytic efficiency and fidelity of cGAS. Nucleic Acids Res. 48, 4435–4447 (2020).
Article CAS PubMed PubMed Central Google Scholar
Hooy, R. & Sohn, J. A pyrophosphatase-coupled assay to monitor the NTase activity of cGAS. Methods Enzymol. 625, 77–86 (2019).
Article CAS PubMed Google Scholar
Wang, C. et al. Manganese increases the sensitivity of the cGAS-STING pathway for double-stranded DNA and is required for the host defense against DNA viruses. Immunity 48, 675–687.e7 (2018).
Article CAS PubMed Google Scholar
Zhao, Z. et al. Mn(2+) directly activates cGAS and structural analysis suggests Mn(2+) induces a noncanonical catalytic synthesis of 2'3′-cGAMP. Cell Rep. 32, 108053 (2020).
Article CAS PubMed Google Scholar
Andreeva, L. et al. cGAS senses long and HMGB/TFAM-bound U-turn DNA by forming protein-DNA ladders. Nature 549, 394–398 (2017).
Article ADS CAS PubMed Google Scholar
Traut, T. W. Physiological concentrations of purines and pyrimidines. Mol. Cell. Biochem. 140, 1–22 (1994).
Article CAS PubMed Google Scholar
Lama, L. et al. Development of human cGAS-specific small-molecule inhibitors for repression of dsDNA-triggered interferon expression. Nat. Commun. 10, 2261 (2019).
Article ADS PubMed PubMed Central Google Scholar
Vincent, J. et al. Small molecule inhibition of cGAS reduces interferon expression in primary macrophages from autoimmune mice. Nat. Commun. 8, 750 (2017).
Article ADS PubMed PubMed Central Google Scholar
Roberts, Z. J. et al. The chemotherapeutic agent DMXAA potently and specifically activates the TBK1-IRF-3 signaling axis. J. Exp. Med. 204, 1559–1569 (2007).
Article CAS PubMed PubMed Central Google Scholar
Conlon, J. et al. Mouse, but not human STING, binds and signals in response to the vascular disrupting agent 5,6-dimethylxanthenone-4-acetic acid. J. Immunol. 190, 5216–5225 (2013).
Article CAS PubMed Google Scholar
Kabsch, W. Xds. Acta Crystallogr. D Biol. Crystallogr. 66, 125–132 (2010).
Article ADS CAS PubMed PubMed Central Google Scholar
Potterton, L. et al. CCP4i2: the new graphical user interface to the CCP4 program suite. Acta Crystallogr. D Struct. Biol. 74, 68–84 (2018).
Article ADS CAS PubMed PubMed Central Google Scholar
Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot. Acta Crystallogr. D Biol. Crystallogr 66, 486–501 (2010).
Article ADS CAS PubMed PubMed Central Google Scholar
Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213–221 (2010).
Article ADS CAS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

We thank Dr. Mario A. Bianchet and the beamline staff for remote crystallographic data collection at NSLS-II at Brookhaven, NY. This work was supported by NIH grants (R01 GM129342 and R35 GM145363 to J.S.). Work at the AMX (17-ID-1) and FMX (17-ID-2) beamlines was supported by the NIH, the National Institute of General Medical Sciences (P41GM111244), the DOE Office of Biological and Environmental Research (KP1605010), and the National Synchrotron Light Source II at Brookhaven National Laboratory are supported by the DOE Office of Basic Energy Sciences under contract number DE-SC0012704 (KC0401040). In memory of Dr. L. Mario Amzel.

Author information

Sandra B. Gabelli
Present address: Discovery Chemistry, Merck Laboratories, West Point, PA, USA

Authors and Affiliations

Department of Biophysics and Biophysical Chemistry, Johns Hopkins University School of Medicine, Baltimore, MD, USA
Shuai Wu, Sandra B. Gabelli & Jungsan Sohn
Department of Oncology, Johns Hopkins University School of Medicine, Baltimore, MD, USA
Sandra B. Gabelli & Jungsan Sohn
Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, MD, USA
Sandra B. Gabelli
Division of Rheumatology, Johns Hopkins University School of Medicine, Baltimore, MD, USA
Jungsan Sohn

Authors

Shuai Wu
View author publications
You can also search for this author in PubMed Google Scholar
Sandra B. Gabelli
View author publications
You can also search for this author in PubMed Google Scholar
Jungsan Sohn
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

S.W. and J.S. conceptualized the project. S.W. performed all experiments, analyzed data, and wrote the paper. S.B.G. supervised the crystallographic data collection and processing and also edited the paper. J.S. supervised the overall project, analyzed data, and wrote the paper.

Corresponding author

Correspondence to Jungsan Sohn.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Peer Review File

Reporting Summary

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Wu, S., Gabelli, S.B. & Sohn, J. The structural basis for 2′−5′/3′−5′-cGAMP synthesis by cGAS. Nat Commun 15, 4012 (2024). https://doi.org/10.1038/s41467-024-48365-3

Download citation

Received: 31 May 2023
Accepted: 26 April 2024
Published: 13 May 2024
DOI: https://doi.org/10.1038/s41467-024-48365-3

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.