Structural insight into the cGAS active site explains differences between therapeutically relevant species

Abstract

Cyclic GMP-AMP synthase (cGAS) is an intracellular sensor of double-stranded DNA that triggers a pro-inflammatory response upon binding. The interest in cGAS as a drug discovery target has increased substantially over the past decade due to growing evidence linking its activation to numerous peripheral and neurological diseases. Here, we report the binding mode of previously described cGAS inhibitors while also uncovering the structural basis for the interspecies potency shifts within this chemotype. A single threonine to isoleucine substitution between human and mouse cGAS drives compound activity, as demonstrated by biochemical, cellular, and in vivo studies. Finally, we utilize a structurally enabled design approach to engineer a novel chemical inhibitor with excellent potency for both human and mouse enzymes by targeting key interactions within the enzyme active site. Overall, this work provides the framework for rational optimization of cGAS inhibitors and potential preclinical translational strategies.

Cyclic GMP-AMP synthase (cGAS), an intracellular sensor of dsDNA, is a central driver of inflammatory diseases and an emerging therapeutic target, although optimization of cGAS inhibitors for the clinic is challenging because of the incomplete understanding of their mode of action. Here, the authors study the binding mode of G-chemotype human cGAS inhibitors, revealing key structural differences driving cGAS inhibitor cross-species potency shifts, and discover a novel cGAS inhibitor with exceptional mouse and human potency.

Introduction

The innate immune system coordinates the early host response to pathogens and the recognition of host danger signals. Cyclic GMP-AMP synthase (cGAS) is a cytosolic pattern recognition receptor that detects foreign or mislocalized endogenous double-stranded DNA (dsDNA)¹. Binding of dsDNA leads to the dimerization of cGAS and initiates its nucleotidyl transferase enzymatic activity to catalyze the cyclization of ATP and GTP into the cyclic dinucleotide 2′,3′-cyclic GMP-AMP (cGAMP). cGAMP is a second messenger that binds to its cognate receptor, Stimulator of interferon genes (STING), resulting in the activation of TANK-binding kinase 1 (TBK1), which phosphorylates the cytosolic transcription factor interferon regulatory factor 3 (IRF3) and thereby causes its nuclear translocation and the induction of IFNB1 transcription and downstream interferon stimulated gene (ISG) expression. Activation of the cGAS-STING pathway is also coupled to the NF-κB pathway, and triggers interferon independent pro-inflammatory responses^2–5.

Under homeostatic conditions, endogenous cellular dsDNA is restricted to the nucleus and mitochondria, but stress or the malfunction of key regulatory pathways can result in its escape into the cytosol to activate cGAS. Thus, while nucleic acid recognition by cGAS contributes to the immune response against pathogenic microbes^6–8, its irregular activation can lead to inflammatory disorders and neurodegenerative diseases¹. Deficiencies in TREX1, an exonuclease that regulates cytoplasmic dsDNA levels and consequently cGAS activity, has been genetically linked with the type I interferonopathy Aicardi-Goutières syndrome (AGS), systemic lupus erythematosus (SLE), and familial chilblain lupus^9–12. Alternatively, mutations to the histone complex proteins LSM11 and RNU7-1 that sequester inactive cGAS in the nucleus are also associated with AGS¹³. Recent studies have further revealed a far broader role for cGAS beyond type I interferon (IFN) driven pathologies, extending to neurodegenerative diseases, including Alzheimer’s disease, Parkinson’s disease, amyotrophic lateral sclerosis, and the process of ‘inflammaging’ at advanced age^14–18.

The contribution of cGAS activation to human disease and its potential as a therapeutic target has spurred numerous drug discovery campaigns to identify small molecule inhibitors^19–23. Current efforts have identified multiple chemotypes that most commonly bind within the enzyme’s active site, demonstrating in vitro and cellular pathway inhibition. Despite significant progress, the optimization of these active site binding inhibitors has been hindered by an incomplete understanding of their binding mode and their mechanism of action.

A recent high throughput screen campaign to identify small molecule inhibitors of human cGAS (h-cGAS) yielded a novel G chemotype chemical series with low μM inhibitory potency in primary human cells¹⁹. However, reported x-ray co-crystal structures of these G-chemotype inhibitors bound to apo h-cGAS could not reconcile multiple facets of their behaviour. Curiously, the most active h-cGAS inhibitors were dramatically less potent towards mouse cGAS (m-cGAS), a result that could not be rationalized structurally. Overall, these discrepancies suggested our understanding of this chemotype needed refinement to permit further efficient optimization.

Here, we applied biochemical, biophysical, and new x-ray crystallography studies to redefine the binding mode and mechanism of action of these G-chemotype cGAS inhibitors. This work revealed an ATP-uncompetitive and GTP-competitive mechanism of action, and a drug binding site localized to the enzyme’s GTP pocket. The generation of novel co-crystal structures provided a molecular understanding of inhibitor potency across multiple mammalian species. Specifically, a single amino acid difference, corresponding to a threonine to isoleucine substitution at m-cGAS position 309, was responsible for the cross-species potency shift observed for mouse, as well as other species used for pharmacology and toxicology purposes. The importance of this residue was further confirmed through in vitro biochemical experiments and extended to an in vivo pharmacodynamic (PD) model using a novel humanized cGAS^I309T mouse strain. Finally, we utilized structurally enabled design principles to discover compound 2, a molecule with high potency against both h- and m-cGAS²⁴. Overall, these results provide a structural basis for the optimization of this promising series of cGAS inhibitors and inform our understanding of potential translational challenges across preclinical mammalian animal species that bear the mouse GTP pocket isoleucine.

Results

G chemotype cGAS inhibitors are ATP-uncompetitive and GTP-competitive

To better characterize previously described G chemotype cGAS inhibitors¹⁹, we determined the biochemical mechanism of action of a representative compound G150 (Fig. 1A), one of the most potent h-cGAS inhibitors from this series. Purified full-length h-cGAS enzyme was stimulated with 100 base-pair (bp) dsDNA while the substrates ATP or GTP were titrated, at a fixed, sub-saturating concentration of the other NTP (50 μM), close to the calculated K_{m app} of ATP/GTP (80.2 ± 8.6 μM)²² (Fig. S1A). cGAMP concentrations were then quantified by LC-MS to assess enzymatic activity. As ATP concentrations decreased, G150 biochemical potency was reduced (Fig. 1B). In contrast, declining concentrations of GTP enhanced its inhibitory activity (Fig. 1C). Together, these results signified an ATP-uncompetitive and GTP-competitive mechanism of action²⁵ (Fig. 1D). This profile was differentiated from that of the reference cGAS inhibitor PF-06928215²⁰ (Fig. 1A), that behaved in an ATP- and GTP-competitive manner (Fig. S1B, D), aligning with its reported binding to the ATP pocket of apo-cGAS.

Fig. 1. G-chemotype cGAS inhibitors have an ATP-uncompetitive and GTP-competitive mechanism of action.

A Chemical structures of G150, G140, and PF-06928215. B, C G150 h-cGAS biochemical potency as determined by the inhibition of cGAMP production measured by LC-MS. GTP concentrations were varied while ATP concentrations were fixed at 50 μM (B) or ATP concentrations were varied while GTP concentrations were fixed at 50 μM (C) cGAS was activated by stimulation with 25 nM of 100 bp dsDNA. Graphs depict one of three independent experiments. Error bars represent standard deviations. D IC₅₀ values from (B) and (C) representing the mean ± SD of three independent experiments. E, F Thermal shift experiment of h-cGAS melting temperature in complex with 25 bp dsDNA and cGAS inhibitors (E) or in complex with dsDNA, ATP, and cGAS inhibitors (F). The graphs depict the mean ± SD of one of two independent experiments. G SPR relative response measurements depicting G140 binding to h-cGAS, h-cGAS + 45 bp dsDNA, h-cGAS + ATP, and h-cGAS + ATP + dsDNA. The K_D value represents the mean ± SD of two independent experiments.

Given the observed ATP-uncompetitive mechanism of action of G150, we next employed two orthogonal biophysical assays to confirm that the presence of ATP is required for the binding of the G chemotype compounds. A thermal shift assay demonstrated a significant increase in the melting temperature of the h-cGAS:dsDNA complex in the presence of PF-06928215 (ΔT° = 8.9 °C), whereas G150 or its related analog G140 (Fig. 1A), had no effect (Fig. 1E). However, the inclusion of ATP led to a substantial increase in h-cGAS melting temperature with G150 (ΔT° = 21.9 °C) or G140 (ΔT° = 23.2 °C) (Fig. 1F). As expected, the shift in melting temperature induced by PF-06928215 was reduced upon ATP addition due to competition with the substrate. Surface plasmon resonance (SPR) studies performed with G140, further substantiated these observations. G140 did not bind to apo h-cGAS, h-cGAS complexed to dsDNA, or h-cGAS with ATP added to the buffer (Fig. 1G). However, the presence of all three components (h-cGAS, dsDNA, ATP) led to G140 binding to the complex with a measured K_D of 141 ± 80 nM. Corresponding to published results, PF-06928215 bound to apo h-cGAS with a K_D of 135 ± 37 nM²⁰, but did not bind in the presence of ATP. Together, these results established that ATP is required for the binding and activity of G chemotype cGAS inhibitors.

G150 binds to the GTP pocket of h-cGAS in complex with ATP

To provide molecular insights into the binding mode of G chemotype cGAS inhibitors, we revisited earlier attempts to generate x-ray co-crystal structures of G150 complexed to h-cGAS¹⁹. A co-crystal structure of G150 with the h-cGAS catalytic domain (AA157-522) in the presence of 17 bp dsDNA, the non-hydrolysable ATP analog AMPPNP, and MgCl₂ was solved at a resolution of 1.5 Å (PDB: 9C8T) (Fig. 2A). Although the dsDNA density was not resolved, h-cGAS adopted a DNA-bound active conformation indicated by two major conformational changes: (1) the Gly207-Val218 residues transitioned from a disordered state to a defined secondary structure (β-strand between Gly207-Asn210 and α-helix between Gly212-Val218); and (2), a 1 Å shift of the β-sheets containing the catalytic acids (Glu225, Asp227, and Asp319) towards the active site^26,27.

Fig. 2. G150 binds within the GTP pocket of human cGAS in the presence of ATP.

A Cartoon representation of the h-cGAS catalytic domain (cyan) in the dsDNA bound active conformation in complex with AMP-PNP (blue) and G150 (green). 2F_O-F_C electron map of ligands contoured at 1.0 sigma. Ligands are represented as sticks and Mg metals are represented as spheres. B Key residues are identified while the interactions between AMP-PNP, Mg²⁺, G150, and h-cGAS are depicted by dotted lines. C Superimposed m-cGAS structure (PDB: 4K97, magenta) with G150 bound h-cGAS structure (cyan). Key residues involved in AMP-PNP and G150 binding are presented as sticks. D T321 is not conserved between human and mouse. Mouse I309 spatially clashes with the pyridine-ring of G150 bound to the GTP pocket.

Within the enzyme active site, AMPPNP resided in the ATP pocket while the GTP pocket was occupied by G150. The orientation of the inhibitor could be traced with a high degree of confidence (2Fo-Fc map in Fig. 2A). A view of the bound G150-AMPPNP-Mg molecules looking into the catalytic pocket of the ternary complex showed a strong π–π stacking interaction between the pyridoindole core of the inhibitor, the adenine ring of AMPPNP, and Tyr436 (Fig. 2B). Two bound metal ions (Mg²⁺) can be clearly identified in this structure, which bridge the interactions between the α- and β-phosphate of the triphosphate and side chains of catalytic acidic residues (Glu225, Asp227, and Asp319), which further contribute to lock the inhibitor in the GTP pocket. Moreover, the hydroxyl-ethanone side chain of G150 also directly interacts with one of the metal ions. The pyridine-ring of bound G150 was anchored to the deep pocket, where it forms additional hydrogen bond interactions with residues Thr321, Met229, and Thr221, which may contribute to the improved potency of G150 relative to analogs lacking this subgroup, such as G022 (Fig. 3A).

Fig. 3. Identification of A-ring methyl compound 2 with high potency for both human and mouse cGAS.

A Reported h-cGAS and m-cGAS biochemical potencies for G150 and G015 show that removal of the D-ring improves mouse potency but erodes human activity. Previously published cGAMP LC-MS IC₅₀ values from Ref. ¹⁹ are shown. B Design strategy to expand from the A-ring leading to the identification of compound 2 and mean IC₅₀ values from at least 2 independent experiments in indicated human and mouse assays. n.d. not determined. C Co-crystal structure of compound 2 (pink) bound to h-cGAS and AMP-PNP superimposed with the G150 (green) co-crystal structure.

Enabled by this new co-crystal structure, we next sought to rationalize the dramatic loss of potency of G150 to m-cGAS relative to the human enzyme, an approximately 2000-fold reduction in biochemical inhibitory activity¹⁹. To visualize the potential impact of structural differences between the two orthologs, the m-cGAS enzymatic pocket (PDB: 4K97) was superimposed onto our newly solved h-cGAS:G150:AMPPNP co-crystal structure (PDB: 9C8T) (Fig. 2C). While the two enzymatic pockets are highly conserved, a key difference in the GTP deep pocket was apparent where Thr321 is replaced by an isoleucine at the corresponding 309 position of m-cGAS. Strikingly, the large bulky side chain of the mouse isoleucine spatially hindered the G150 pyridine-ring while also lacking the productive hydrogen bond interactions of the human threonine residue (Fig. 2D). Together, this structural analysis strongly predicted that this single amino acid change may be responsible for the interspecies potency shift of G150 and structurally similar compounds.

Structure-based design of compound 2, a cGAS inhibitor with high potency for both human and mouse enzymes

This refined understanding of the G chemotype cGAS inhibitor binding mode and the molecular interactions within the enzyme active site uncovered new opportunities for the rational optimization of this chemical series. To validate this approach, we aimed to identify alternate designs that could avoid steric clashing of the inhibitor with the m-cGAS Ile309 residue but maintain the potent interactions with h-cGAS Thr321, resulting in a molecule highly active against both orthologs. Consistent with this notion, the D-ring-deleted compound G022, which would not sterically clash with the m-cGAS Ile309 residue, showed 1000-fold improved mouse potency vs. G150 but also displayed an 11-fold loss of human activity (Fig. 3A). The methylated derivative of G022, G015, was similarly highly active towards m-cGAS, but suffered a 35-fold loss of potency towards h-cGAS relative to G150. To improve the h-cGAS potency of D-ring deleted compounds while maintaining their mouse potency, derivatives of G015 were generated that expanded into the space occupied by the G150 D-ring with groups extending from the 1-position of the A-ring (Fig. 3B). To avoid clashing with the m-cGAS Ile309, efforts focused on small alkyl substitutions on the A ring to increase the probability of maintaining mouse activity. Inhibitor potencies were assessed in a cGAS biochemical assay measuring the enzymatic consumption of the substrate ATP by a luminescence-based readout (kinase-Glo)¹⁹. The addition of a simple methyl group in the A-ring 1-position vector yielded compound 2 with a 13-fold improvement in h-cGAS biochemical activity while maintaining equivalent m-cGAS potency to G015. Importantly, there was a critical stereochemical dependence of the A-ring 1-methyl group on human enzyme potency, as the (S)-isomer 2 was 157-fold more active than the (R)-isomer 1, while the gem-dimethyl 3 was inactive up to 10 μM (Fig. 3B). To assess the cellular activity of these compounds, human monocytic THP1-Dual cells stably expressing an interferon-stimulated response element coupled to a secreted luciferase gene were utilized, thereby allowing a simple downstream readout of cGAS pathway activation following transfection with dsDNA¹⁹. The cellular potency of 2 was also enhanced relative to the parent compound G015, with an IC₅₀ value of approximately 3 μM in THP1-Dual cells, corresponding to a fourfold improvement. Mouse cellular potency was determined using a murine RAW-Lucia macrophage cell line stably expressing an interferon-inducible reporter construct and transfected with 100-bp dsDNA¹⁹, which demonstrated similar potency between compound 2 and G015.

To validate our understanding of the structure activity relationship and methyl group A-ring stereochemistry of these compounds, the co-crystal structure of 2 with h-cGAS, dsDNA, and AMPPNP was solved at a resolution of 1.5 Å (PDB: 9C8N) (Fig. 3C). As compound 2 shares the same pyridoindole core and hydroxyl-ethanone side chain of G150, the key π–π stacking interactions and metal bridged interactions were similar. The structure revealed the (S)-isomer methyl of 2 pointing in a pseudo-axial orientation ‘upward’ into the back pocket space, forming hydrophobic interactions with Pro306 and Ala307. In contrast, the (R)-methyl of the less active isomer 1 was rationalized to likely point ‘down,’ potentially disrupting ATP binding, which could explain its significantly reduced activity. Together, these results represent an example of structurally enabled design to optimize cGAS inhibitor potency, in this case towards a compound with high potency for both mouse and human enzymes.

The threonine to isoleucine amino acid substitution at mouse position 309 is responsible for the potency shifts of GTP-pocket-binding cGAS inhibitors

While our co-crystal structures revealed that the threonine to isoleucine substitution is likely responsible for the observed interspecies potency shifts, we aimed to validate these findings through site directed mutagenesis. ‘Humanized’ mouse I309T and human T321I mutant cGAS enzymes were purified for use in our biochemical enzymatic assays. Both mutant enzymes demonstrated reduced activity relative to their WT counterparts but were capable of cGAMP production (Fig. S2A, B). To determine the impact of these mutations on compound potency, G chemotype inhibitors with or without the D-ring expected to sterically clash with the isoleucine residue were compared (Fig. 4A). Notably, the biochemical potency of G108 and G150 were highly shifted in the mutant enzymes relative to their WT counterparts (Fig. 4A). G150 had similar IC₅₀ values in the threonine bearing WT h- cGAS and the ‘humanized’ mouse I309T cGAS (0.027 μM and 0.076 μM, respectively) but had greatly reduced inhibitor activity in both the WT mouse and human T321I enzymes (33.5 μM and 8.9 μM, respectively) (Fig. 4B). In contrast, the potency of G022 and 2, compounds lacking the D ring, had similar potencies in all four cGAS enzymes with a far lower fold shift between WT and mutant proteins (≤5) (Fig. 4A, B). Consistent with our hypothesis, these results point towards a steric clash between the D ring and mouse isoleucine residue greatly reducing inhibitor potency.

Fig. 4. GTP pocket threonine to isoleucine substitution is responsible for the potency shifts of G-chemotype inhibitors in human and mouse cGAS.

A Structures and potencies of representative G-chemotype cGAS inhibitors in the cGAMP LC-MS biochemical assay. IC₅₀ values represent the mean of two independent experiments. The fold shift between WT and mutant enzymes is in green. B Representative dose response curves of G150 and compound 2 in WT and mutant enzyme cGAMP LC-MS biochemical assays. Error bars represent standard deviations. C–E cGAS inhibitor potency in (C) WT h-cGAS vs WT m-cGAS and (D) WT h-cGAS vs I309T m-cGAS. IC₅₀ values were determined using cGAS Kinase-glo biochemical assays and represent the mean of at least 2 independent experiments. Grey circles represent GTP pocket binding inhibitors while specific compounds or chemotypes are identified by color. Indicated G-chemotype (green), Hall et al.²⁰, (purple), RU.521 (pink), WO 2021/233854 (blue), WO 2024/035622 (orange). E Structures and potencies of representative inhibitors from differentiated chemotypes in cGAS Kinase-Glo biochemical assay. The fold shift between WT h-cGAS vs WT m-cGAS or WT- h-cGAS vs I309T m-cGAS is depicted in green. IC₅₀ values represent the mean of at least 2 independent experiments.

To broadly test the impact of the isoleucine residue on cGAS inhibitor activity, we next screened a larger panel of diverse compounds spanning multiple chemical series. Chemotypes structurally validated to reside within the GTP pocket and divergent in human vs. mouse activity, represented by the grey, cyan, and blue circles (Fig. 4C), were normalized by the mouse I309T substitution (Fig. 4D). Inhibitors of a malononitrile chemotype reported by Roche (blue), as exemplified by compound 111²⁸, was confirmed by us to occupy the GTP pocket and was similarly demonstrated by us to be influenced by the isoleucine residue. For example, the h-cGAS biochemical IC₅₀ of compound 111 was reduced sixfold relative to m-cGAS, but this was equalized by the threonine substitution to the mouse enzyme (Fig. 4E). The modest effect of the amino acid change on this series is consistent with x-ray crystallography studies demonstrating limited interaction with the isoleucine residue in comparison to D-ring containing G-chemotype molecules, e.g., (PDB: 7FTU).

Chemical series with alternate binding modes, such as PF-06928215 (purple) and RU.521 (pink)^21,29 that reside within the ATP pocket, were not affected by the m-cGAS isoleucine to threonine substitution (Fig. 4E). Similarly, a tricycle chemotype reported by Bellbrook Labs, exemplified by compound 17 (orange)³⁰, was confirmed by us to reside deeper within the ATP pocket³⁰, and was found by us to be inactive to both WT and I309T m-cGAS, but inhibited h-cGAS activity. These results suggest that structural differences within the ATP pocket are likely responsible for driving species selectivity for these molecules. Overall, these results are aligned with the known binding modes of the profiled cGAS inhibitors and reveal how a threonine or isoleucine residue at this position can impact the activity of GTP pocket occupying molecules.

Multiple preclinical species share the mouse isoleucine mutation that impacts cGAS inhibitor potency

As the mouse Ile309 residue was confirmed to be critical to inhibitor potency of GTP-pocket residing cGAS inhibitors, we next assessed the amino acid identity at this position for additional mammalian preclinical species. Alignment of the cGAS protein sequence revealed that rat and dog shared an isoleucine at the equivalent amino acid position to mouse, while human, monkey, and cat cGAS possessed a threonine residue (Fig. 5A). To assess if the impact of the amino acid at this position on inhibitor activity extended to these additional orthologs, we purified recombinant cGAS from these four additional species for biochemical testing. All enzymes were capable of cGAMP production, but the catalytic activity varied across species, with mouse and rat enzymes the most active and human the least active (Fig. S2C, D). Agreeing with our earlier results, the presence of an isoleucine or threonine residue was highly predictive of G chemotype inhibitor potency (Fig. 5B, C). G150 and G140 were highly active in the threonine containing enzymes, namely monkey, cat, and human, whereas potency was dramatically reduced in the isoleucine bearing dog, rat, and mouse cGAS. The opposite trend was observed for G022, which was most active in the cGAS enzymes with the isoleucine residue. Finally, the h-cGAS IC₅₀ of compound 2 was within twofold of that obtained in the isoleucine possessing mouse, rat, and dog enzymes, but approximately tenfold less potent towards cat cGAS. Therefore, other structural differences between cGAS orthologs may contribute to shifts in inhibitor potency, but the threonine to isoleucine substitution within the GTP pocket consistently appears to be the dominant determining factor.

Fig. 5. Threonine to isoleucine substitution correlates with cGAS inhibitor potency in multiple preclinical species.

A Sequence alignment of cGAS enzymatic pocket amino acids from indicated mammalian species. Threonine or isoleucine residues at human position 321 are highlighted. B Chemical structures and potencies of representative G-chemotype cGAS inhibitors in the cGAS Kinase-Glo biochemical assay of indicated mammalian species stimulated with 100 bp dsDNA. Values represent the mean of at least 3 independent experiments. C Representative dose response curves of indicated inhibitors in cGAS Kinase-Glo biochemical assay. Error bars represent standard deviations.

Human selective cGAS inhibitors are active in cGAS^I309T knock-in mice

The poor mouse potency of G140 and G150 has prevented their usefulness as tool compounds to study cGAS inhibition in vivo. Given the value of translational animal models to correlate drug exposure and pharmacodynamic effects, we generated a knock-in (KI) mouse bearing the humanized I309T mutation through CRISPR/Cas9-mediated gene editing, hereafter referred to as cGAS^I309T mice (Fig. S3A–C). The cGAS^I309T mice were confirmed to have equivalent protein expression of cGAS and the downstream pathway signaling components STING, IRF3, and TBK1 in the spleen, heart, liver, and lung relative to WT mice (Fig. S3D). The immune cell compositions of the spleen, mesenteric lymph node, and thymus were also comparable between the two strains, suggesting a limited impact of the mutation on the immune system (Fig. S4A).

To determine if the isoleucine- to threonine-dependent potency shift observed in the biochemical assay similarly translated to cellular systems, we first assessed compound potency in WT and cGAS^I309T bone-marrow derived macrophages (BMDMs). BMDMs were preincubated with the inhibitors G140 or compound 2 followed by transfection with 100-bp dsDNA to activate cGAS and stimulate downstream expression of type I IFN and ISGs. The concentration of the secreted chemokine CXCL10, also known as IP-10, in the supernatant was quantified as a surrogate ISG and readout of pathway activation¹⁹. Induced CXCL10 levels were similar between cells from both mouse strains, suggesting equivalent downstream activation of the cGAS pathway (Fig. S5A). Strikingly, while G140 was inactive in WT BMDMs up to a concentration of 30 μM, it was highly potent in the cGAS^I309T cells with an IC₅₀ of 0.82 μM (Fig. 6A). In contrast, compound 2 had comparable activity in both genotypes (Fig. 6B), aligning with our biochemical results (Fig. 4A). We next assessed inhibitor potency in mouse whole blood (WB) to estimate the drug exposures required for systemic pathway inhibition. Blood was collected from WT and cGAS^I309T mice, transfected with dsDNA, and plasma CXCL10 levels were measured to assess pathway activation. Absolute levels of CXCL10 were similar between both strains (Fig. S5B). In agreement with the BMDM results, G140 was more potent in the humanized cGAS mouse, with an IC₅₀ value of 16.2 μM, but inactive in the blood of WT mice (Fig. 6C). The activity of compound 2 was similar between both strains, though three to fivefold less potent than G140, with IC₅₀ values of 75.1 μM and 48.7 μM in the cGAS^I309T and WT mice, respectively (Fig. 6D). The potency of both compounds was highly shifted relative to cellular values, though this can be attributed to increased protein binding in the blood.

Fig. 6. G140 inhibits in vivo cGAS activity in cGAS^I309T, but not WT mice.

A, B Dose response curves of (A) G140 and (B) compound 2 inhibition of 100 bp dsDNA transfected WT or cGAS^I309T BMDM secreted CXCL10. The IC₅₀ value represents the mean ± SD of three independent experiments. C, D Dose response curves of (C) G140 and (D) compound 2 inhibition of 100 bp dsDNA transfected WT or cGAS^I309T mouse whole blood plasma CXCL10. The IC₅₀ value represents the mean ± SD of two independent experiments. E Blood concentration of G140 after 100 mg/kg PO dosing in WT and cGAS^I309T mice at indicated timepoints. Mouse whole blood IC₅₀ values from (C) are indicated. F Splenic cGAMP levels 16 h post ConA treatment of WT and cGAS^I309T mice pretreated with G140 dosed 100 mg/kg PO. The results from one of two independent experiments is depicted. G Terminal blood concentrations of G140 in ConA treated mice. Statistical analysis represents 1-way ANOVA with Dunnett’s multiple comparisons test. *** p < 0.0001.

To assess the potential of G140 and compound 2 to be used as in vivo tool compounds, drug pharmacokinetics were analyzed in both WT and cGAS^I309T strains (Fig. 6E). In cGAS^I309T mice, an oral dose of 100 mg/kg G140 was predicted to cover the mouse WB CXCL10 IC₅₀ for approximately 9 h, a suitable profile to test in vivo modulation of the cGAS pathway. However, due to poor bioavailability and overall pharmacokinetic properties, compound 2, as well as G150, were deemed unsuitable for pharmacodynamic studies (Table S2).

We next employed a concanavalin A (ConA)-induced autoimmune hepatitis mouse model to explore in vivo modulation of cGAS activity. ConA administration activates cGAS in mice, potentially through the release of dsDNA containing neutrophil extracellular traps³¹. Despite the resulting inflammatory hepatitis phenotype, a 10 mg/kg i.v. dose of ConA resulted in only a slight elevation of cGAMP above the limit of quantification in the liver (Fig. S5C). However, a robust production of cGAMP in the spleen was observed at 12 h and peaked at 16 h (Fig. S5D). The substantially greater production of cGAMP in the spleen can potentially be attributed to the high density of immune cells present. As longer durations led to reduced viability of the mice, a timepoint of 16 h was chosen for PD studies. To assess the pharmacodynamic effect of G140 in this model, the compound was dosed orally at 100 mg/kg in either WT or cGAS^I309T mice, followed by ConA injection. In line with the predicted coverage of the WB IC₅₀, G140 was able to suppress cGAMP levels by approximately 50% in the cGAS^I309T mice but was inactive in WT mice (Fig. 6F). The mean terminal concentrations of G140 in the spleen were similar between both strains, 11.9 μM in the cGAS^I309T mice and 7.3 μM in the WT mice. Overall, these results demonstrate that the cGAS^I309T humanized mouse strain is a suitable tool to assess cGAS inhibition with molecules that occupy the GTP pocket and may clash with the WT m-cGAS Ile309 residue.

Discussion

There is now a wealth of evidence linking cGAS-STING pathway activation to numerous inflammatory and neurodegenerative diseases, including many with few therapeutic options and high unmet need^1,32. Despite the discovery of multiple small molecule cGAS inhibitor chemotypes, a well understood molecular characterization of their binding mode, has not always been available, limiting the efficient optimization and improvement of these molecules towards potential drug candidates. In this work, we aimed to build on the structural understanding of a previously discovered series of inhibitors¹⁹. We characterized their binding mode through biochemical, biophysical, and co-crystal structural studies, utilizing G150 and G140 as representative compounds. This identified an ATP-uncompetitive and GTP-competitive mechanism of action and the inhibitor binding site within the enzyme’s GTP pocket (Fig. 2). Importantly, these results demonstrate that the previously reported co-crystal structure of G150 bound to apo-cGAS (PDB: 4O68) represented an unnatural binding state not relevant for compound activity. Our new co-crystal structure (PDB: 9C8T) has now identified the basis for the species selectivity observed between compounds of this chemical series, which was confirmed through biochemical, cellular, and in vivo site directed mutagenesis studies. This structure is aligned with a recently published co-crystal structure of G150 with h-cGAS and ATP, demonstrating a similar occupancy of the GTP pocket by the inhibitor (PDB: 7FUR).

The human and mouse cGAS proteins share only approximately 60% amino acid identity, however, the enzyme active site is largely conserved. Remarkably, the potencies of various GTP-competitive inhibitors tested in this study were dependent on a single amino acid substitution, a threonine in place of isoleucine at residue 321 of the human enzyme (mouse residue 309). The impact of this residue was confirmed by multiple approaches and extended to additional mammalian species commonly used in preclinical safety studies, including rat and dog. While our work was focused towards understanding compound potency, we also determined that the isoleucine substitution led to a reduction in h-cGAS cGAMP production (Fig. S3B). A recent publication reported a similar effect for h-cGAS^T321I, which was attributed to reduced AMP‐2′‐GTP linear intermediate cyclization relative to the WT enzyme³³. Interestingly, the isoleucine residue at this position also impacted substrate promiscuity, as h-cGAS^T321I demonstrated increased enzymatic activity towards ATP/GTP or GTP/GTP substrate pairs relative to ATP/ATP. Therefore, the presence of an isoleucine residue at this site can impact both inhibitor binding and cGAS activity.

Drug discovery programs require a clear understanding of the target pharmacology to enable translational success. The correlation between drug exposure levels, in vivo pharmacodynamic effects, and efficacy require experimental validation, which may initially present challenges for novel therapeutic targets. Preclinical animal models provide an important tool to better understand these relationships. However, the lack of high-quality in vivo tool compounds has hindered the exploration of these questions regarding cGAS, with the majority of literature utilizing genetic approaches, e.g., cGAS-deficient mice, to implicate cGAS in disease^34,35. The cGAS^I309T mice developed in this work could provide an opportunity for the study of GTP-pocket-residing compounds, such as G140, or other molecules that were likely optimized for human potency at the expense of mouse. However, for compounds with alternative binding modes, such as to the ATP pocket, other mutations or a fully humanized cGAS transgenic mouse may be required.

We utilized a mouse model of ConA-induced cGAS activation to study target modulation. It was found that in vivo coverage of the mouse WB CXCL10 IC₅₀ by G140 translated well to inhibition of splenic cGAMP production. However, as CXCL10 is a downstream readout of cGAS activation, the in vivo correlation between cGAMP and ISG inhibition remains to be defined. The testing of additional compounds is required to solidify this correlation and more accurately determine the relationship between exposure levels and PD effects. The conditions utilized in our ConA model require sufficient target coverage over a period of 16 h, which may prohibit the testing of less optimized compounds with poor pharmacokinetic properties. However, variations to the dosing paradigm or alternate acute triggers of cGAS activation, such as HSV viral infection⁷, could be utilized as substitute models to study in vivo cGAS inhibition.

Once a sufficient understanding of PK/PD relationships has been built, the requirements for efficacy can be properly assessed. A cGAS-driven disease model, such as Trex1^−/− mice, that develop a type-I-interferonopathy-like disease with features resembling that of AGS or chilblain lupus could be utilized for this purpose^34,35. While we have confirmed in vivo inhibition of cGAMP with a small molecule, additional exploration is required to determine the relationship between cGAMP levels and downstream IRF3 or NF-κB dependent proinflammatory gene expression. In addition to understanding the PK/PD/efficacy relationship, this work could potentially identify downstream biomarkers for translational purposes.

Clinically, the blockade of type-I-IFN-signaling has been validated as a treatment for SLE with the approval of the anti-IFNAR1 antibody anifrolumab³⁶, while other pathway modulators, such as JAK inhibitors, are being investigated^37–39. The presence of cGAMP in SLE patient PBMCs suggests that cGAS activation may be the upstream source of type-I-IFN induction⁴⁰. Given the concomitant induction of the NF-κB pathway, inhibition of cGAS could potentially lead to enhanced efficacy relative to downstream targeting of the type I IFN pathway. A clear understanding of the PD/efficacy relationships of these drugs and their signaling pathways will provide guidance for clinical treatment strategies and build confidence for translation to the clinic.

Given the challenges associated with designing highly active molecules with an acceptable balance of properties, structure-based rational design is a vital tool for compound optimization. Here, we demonstrated an example of this approach through the discovery of compound 2, that utilizes an alternate vector to boost human potency, while minimizing steric interference with the isoleucine residue of mouse, rat, and dog enzymes. While compound 2 suffered from poor PK properties, further optimization, potentially aided by more advanced computational approaches, could build off this initial effort. Studies resulting in the identification of a compound 2 derivative with improved mouse PK have been described in a separate manuscript⁴¹.

Overall, this work provides new insight into the mechanism of action, design, and translation of cGAS inhibitors towards clinical development. Rational design approaches building off results presented herein may aid in developing more potent inhibitors with improved properties leading to potential therapies for numerous diseases.

Methods

Protein expression and purification

For h-cGAS, h-cGAS^T321I, m-cGAS, and m-cGAS^I309T proteins used in the LC-MS cGAMP assay (Figs. 4A, B, and S2A, B), full-length (residues 1–522 for h-cGAS and 1–507 for m-cGAS) protein expression and purification was performed as described previously^19,21. Mutant h-cGAS^T321I and m-cGAS^I309T plasmids were generated by site directed mutagenesis using a QuikChange II XL kit (Agilent) following the manufacturer’s protocol. Mutations were confirmed by sequencing.

For the cGAS protein used in all other experiments, the DNA sequences corresponding to the full-length (residues 1–522) and catalytic domain (residues 157–522) of h-cGAS, full-length m-cGAS (residues 1–507) and I309T mutant, full-length rat cGAS, full-length cat cGAS, full-length dog cGAS and full-length monkey cGAS were synthesized and cloned into a custom pET vector for expression of an N-terminal His-SUMO fusion protein. The fusion cGAS proteins were expressed in BL21-CodonPlus (DE3)-RIPL Competent Cells (Agilent #230280). The cells were grown at 37 °C until OD₆₀₀ reached approximately 1. The temperature was then shifted to 18 °C and recombinant protein expression was induced by supplementation with 0.3 mM IPTG. Cultures were incubated at 18 °C with shaking for 16 h before harvest.

Bacterial pellets were resuspended in lysis buffer (20 mM Tris-HCl pH 7.5, 500 mM NaCl, 30 mM imidazole, 10% glycerol, 1 mM TCEP) chilled on ice and lysed by sonication. The fusion protein was purified over a Ni-NTA affinity column. The His6-SUMO tag was removed by ULP1 cleavage during dialysis against buffer containing 20 mM Tris-HCl pH 7.5, 0.3 M NaCl, 1 mM TCEP. After dialysis, the protein sample was flowed through a second Ni-NTA affinity column and was further purified over a Heparin HP ion-exchange column, followed by gel filtration on a 16/60 G200 Superdex column (GE Healthcare) equilibrated with storage buffer (20 mM Tris-HCl pH 7.5, 0.3 M NaCl, 1 mM TCEP). The final sample of cGAS was concentrated to approximately 10 mg/ml, flash-frozen in liquid nitrogen, and stored at −80 °C for crystallography and biochemical experiments.

Thermal shift assay

Purified 2.5 μM h-cGAS protein and 10 μM 25-bp dsDNA were mixed with 200 μM ATP plus cGAS inhibitors in buffer containing 20 mM Tris-HCl pH 7.5, 50 mM NaCl, 5 mM MgCl₂, 1 μM ZnCl₂ and Protein Thermal Shift Dye (Thermo Fisher Scientific #44-611-46) in a total reaction volume of 20 μl. Thermal scanning (25–75 °C at 1.5 °C/min) was performed, and melting curves were recorded on a QuantStudio RT–PCR machine (Thermo Fisher Scientific). Data analysis was performed using Protein Thermal Shift software (Thermo Fisher Scientific).

Protein crystallization

For crystallization studies, purified h-cGAS catalytic domain at 5 mg/ml was mixed with 17-bp DNA in a molar ratio of 1:1.25 protein:DNA in a buffer containing 20 mM Tris pH 7.5, 150 mM NaCl, 5 mM MgCl₂, 1 mM TCEP 5 mM AMPPNP, and 1 mM of cGAS inhibitor, incubated on ice for 30 min, and centrifuged to pellet precipitates prior to crystallization trials. Crystals were obtained by hanging drop vapor diffusion in drops mixed 1:1 over a reservoir of 0.1 M Tris-HCl pH 8.0, 0.1 M lithium sulfate, 20% PEG 4000 at 18 °C. Crystals were cryo-protected using reservoir solution supplemented with 25% glycerol and flash-frozen in liquid N₂.

The diffraction data sets for cGAS in complex with AMPPNP and G150 were collected at ALBL Synchrotron. The diffraction data sets for cGAS in complex with AMPPNP and 2 were collected at National Synchrotron Light Source II. All the datasets were processed with XDS and AIMLESS⁴². Phases were determined with molecular replacement using Phaser in PHENIX and the PDB of 6CTA used as search model⁴³. The model building was conducted using the program COOT and structural refinement was conducted using the program PHENIX⁴⁴.

Surface plasmon resonance

Purified biotinylated full-length human cGAS apo protein or in complex with biotinylated 45-bp dsDNA was immobilized onto a neutravidin surface. Protein:dsDNA complexes were made by incubating 45-bp dsDNA in a 1:1 molar ratio with human cGAS at 1 µM in 25 mM Tris pH 7.5, 50 mM NaCl, and 0.5 mM TCEP. The assay was run in 25 mM Tris-HCl pH 7.5, 100 mM NaCl, 5 mM MgCl₂, 0.5 mM TCEP, ±0.5 mM ATP, 0.05% Tween-20, 1% DMSO. Compounds were run with either a multi-cycle or single-cycle method and injected over surface with a contact time of 120 s on and 600 s off. A 5-point solvent correction curve was run pre and post run to correct for the refractive index of DMSO. Data was then analyzed using the Biacore evaluation software with either multi-cycle or single-cycle kinetics.

cGAS cGAMP inhibition biochemical assay

The biochemical inhibitory potency of cGAS inhibitors was determined by measuring the in vitro inhibition of cGAMP production. For experiments comparing h-cGAS, h-cGAS^T321I, m-cGAS, and m-cGAS^I309T(Figs. 4A, B, and S2A, B) the protocol was used with slight modifications from that described previously^19,21. Final m-cGAS and h-cGAS protein concentrations of 60 nM and 100 nM were used, respectively. The final reaction buffer for h-cGAS proteins was composed of 20 mM Tris-HCl pH 7.4, 50 mM NaCl, 10 mM MgCl₂, 1 µM ZnCl₂, and 0.01% Tween-20. A reaction incubation period of 30 min and 120 min was selected for m-cGAS and m-cGAS^I309T, respectively and 2 h and 6 h was for h-cGAS and h-cGAS^T321I, respectively based on the formation of equivalent amounts of cGAMP product (Fig. S2). Samples were analyzed using Rapid-Fire mass spectrometry as described previously^19,21.

For in vitro inhibition of cGAMP mechanism of action experiments, the reaction was run in an assay buffer consisting of 20 mM Tris-HCl pH 7.4, 25 mM NaCl, 10 mM MgCl₂, 1 µM ZnCl₂, 0.01% Tween-20, and 1 µM DTT. Compounds were preincubated with 0.5 nM h-cGAS for 2 min before initiating the reaction by the addition of 25 nM 100-bp dsDNA, 100 µM ATP, and 100 µM GTP in a final volume of 40 µl. One of ATP or GTP were used at a fixed concentration of 50 µM, while the second substrate was titrated. After a 4 h incubation at room temperature, the reaction was stopped by the addition of 120 µl of 70% acetonitrile/30% H₂O containing 150 nM of a ¹³C₁₀,¹⁵N₅-cGAMP internal standard⁴⁵. Samples were then centrifuged at 3700 × g for 10 min and 120 μl supernatant was collected and added to 80 μl of acetonitrile. cGAMP quantification was performed by liquid chromatography with tandem mass spectroscopy (LC-MS) using a XBridge BEH Amide Column (Waters #186004863). A gradient was performed starting with 98% mobile phase B (acetonitrile) and 2% mobile phase A (95% 10 mM ammonium acetate/5% acetonitrile) at a flow rate of 1 ml/min and held for 0.5 min. The gradient was then ramped to reach 40% B/60% A at 1 min, then to 5% B/95% A at 1.1 min, and resumed the initial conditions at 4 min.

cGAS Kinase-Glo inhibition biochemical assay

The biochemical inhibitory potency of cGAS inhibitors was also assessed using a luminescence-based readout of ATP consumption, determined by Kinase-Glo Max (Promega # V6071). Enzyme concentrations of 40 nM h-cGAS, 0.5 nM m-cGAS, 10 nM m-cGAS (I309T), 1.5 nM rat cGAS, 15 nM dog cGAS, 15 nM cat cGAS, 15 nM monkey cGAS were selected to yield equivalent ATP depletion. Compounds were first preincubated with recombinant cGAS for 5 min following initiation of the reaction by the addition of 25 nM 100-bp dsDNA, 100 µM ATP, and 100 µM GTP in a final volume of 20 µl in 384-well plates. After a 3 h incubation at room temperature, 20 μl of Kinase-Glo Max reagent was added to each well and luminescence was determined on a Enspire multimode plate reader (PerkinElmer).

RAW Lucia luciferase reporter assay

RAW-Lucia ISG cells (Invivogen #rawl-isg, confirmed mycoplasma negative) were cultured in DMEM media (Gibco #11995-065) supplemented with 10% FBS HI, and 1× Pen/Strep. For compound screening, 35,000 cells were plated per well of a 96-well plate in complete media and incubated overnight. The following day, media was replaced with growth media containing 1.8% FBS in a final volume of 100 μl. Cells were pre-incubated with a serial dilution of compounds in OPTI-MEM for 60–90 min at 37 °C, 5% CO₂ in a humidified incubator. The cGAS pathway was activated by transfecting the cells with a final concentration of 0.5 µg/ml 100-bp dsDNA and 0.5 µl/ml Lipofectamine 2000 for 24 h at 37 °C, 5% CO₂. After the incubation, the supernatant was collected and luciferase expression was quantified to assess ISG reporter induction using QUANTI-Luc (Invivogen #rep-qlcg5) following the manufacturer’s protocol. Luminescence was measured using an Enspire multimode plate reader (PerkinElmer).

THP-1 Dual luciferase reporter assay

THP1-Dual cells (Invivogen, #thpd-nfis, confirmed mycoplasma negative) were cultured in RPMI 1640 media (Gibco #61870-036) supplemented with 25 mM HEPES (Gibco #15630080), 10% FBS (Gibco #12484028), 1× Pen/Strep (Gibco #15140122). For compound screening, the cells were resuspended in fresh assay media (2% FBS) and 51,000 cells were plated per well of a 96-well plate in a final volume of 100 µl. Cells were pre-incubated with a serial dilution of compounds in OPTI-MEM for 90 min at 37 °C, 5% CO₂ in a humidified incubator. The cGAS pathway was activated by transfecting the cells with a final concentration of 0.5 µg/ml 100-bp dsDNA and 0.5 µl/ml Lipofectamine 2000 for 20–24 h at 37 °C, 5% CO₂. After the incubation, the supernatant was collected and luciferase expression was quantified to assess ISG reporter induction using QUANTI-Luc (Invivogen #rep-qlcg5) following the manufacturer’s protocol. Luminescence was measured using an Enspire multimode plate reader (PerkinElmer).

Cellular viability assay

The effect of compounds on THP-1 Dual and RAW Lucia cellular viability was determined using CellTiter-Glo luminescent cell viability assay (Promega), following the manufacturers instructions.

THP-1 Dual cGAMP assay

THP1-Dual cells (Invivogen, #thpd-nfis) were cultured in RPMI 1640 media (Gibco #61870-036) supplemented with 25 mM HEPES (Gibco #15630080), 10% FBS (Gibco #12484028), 1× Pen/Strep (Gibco #15140122). For compound screening, the cells were resuspended in fresh assay media (2% FBS) and 270,000 cells were plated per well of a 96-well plate. Cells were pre-incubated with a serial dilution of compounds in OPTI-MEM for 90 min at 37 °C, 5% CO₂ in a humidified incubator. The cGAS pathway was activated by transfecting the cells with a final concentration of 1.0 µg/ml 100-bp dsDNA and 2.63 µl/ml Lipofectamine 2000 for 24 h at 37 °C, 5% CO₂. After the incubation, 200 µl of supernatant was transferred into a new 96-well plate. The plates were centrifuged (3700 × g, 10 min) and 180 µl of supernatant was removed. The pellets were then frozen at −80 °C for cGAMP quantification.

Mouse BMDM dsDNA assay

The femur and tibia from age and sex matched C57BL/6 WT and cGAS^I309T mice were harvested and bone marrow collected. Following red blood cell lysis (Biolegend #420301), the remaining cells were plated in differentiation media DMEM (Gibco #11995-065) supplemented with 10% FBS (Gibco #12484028), 1× Pen/Strep (Gibco #15140122), and 50 ng/ml M-CSF (Proteintech #315-02) and incubated at 37 °C, 5% CO₂ overnight. The following day, cells were transferred to a new flask to remove fibroblasts. After 7 days of differentiation, the BMDMs were collected and plated at 80,000 cells/well of a 96-well plate in media containing 10 ng/ml M-CSF in a final volume of 100 μl. Cells were pre-incubated with a serial dilution of compounds in OPTI-MEM for 90 min at 37 °C, 5% CO₂ in a humidified incubator. The cGAS pathway was activated by transfecting the cells with a final concentration of 0.5 µg/ml 100-bp dsDNA and 0.5 µl/ml Lipofectamine 2000 for 24 h at 37 °C, 5% CO₂. After incubation, the supernatant was collected and CXCL10 levels were quantified by mesoscale discovery assay (MSD #K152UFK) following the manufacturer’s instructions.

Mouse WB dsDNA assay

Blood from male C57BL/6 WT and cGAS^I309T mice aged 9–12 weeks was collected in hirudin-coated tubes (Sarstedt #04.1959.001) and 135 µl was added to wells of a 96-well plate. cGAS inhibitor treatment was performed by adding 7.5 µl of a 20× compound solution or vehicle to each well and the plate was incubated on a shaker (400 rpm) in a humidified incubator at 37 °C, 5% CO₂ for 5 min, then incubated for 1 h without shaking. The cGAS pathway was activated by transfecting the cells with a final concentration of 1.0 µg/ml 100-bp dsDNA and 2.0 µl/ml Lipofectamine 2000 for 6 h at 37 °C, 5% CO₂. After incubation, the plates were centrifuged (1000 × g, 10 min) and the plasma was collected and stored at −80 °C for future analysis. CXCL10 levels in the plasma were quantified by mesoscale discovery assay (MSD #K152UFK) following the manufacturers’ instructions.

¹³C₁₀,¹⁵N₅-cGAMP synthesis

A ¹³C₁₀,¹⁵N₅-cGAMP internal standard was used for LC-MS quantification of cGAMP⁴⁵ and produced by the incubation of 1 mM ATP, 1 mM ¹³C₁₀,¹⁵N₅-GTP (Sigma #645680), 12.5 nM m-cGAS and 25 nM 100-bp dsDNA for 24 h at 37 °C. The identity and purity of the enzymatically synthesized ¹³C₁₀,¹⁵N₅-cGAMP product was confirmed by LC-MS and molecular weight calculated to be 390.125.

cGAMP quantification of cell and mouse tissue lysates

Cell and homogenized tissue lysates were incubated with 300 µl of 0.5 nM ¹³C₁₀,¹⁵N₅-cGAMP internal standard solution prepared in 70% acetonitrile. The plate was vortexed followed by centrifugation (3200 × g, 10 min, room temperature). The supernatant was then transferred to a 1.2 ml deep-well plate to evaporate overnight under nitrogen gas. To purify cGAMP and remove any potential substances that would interfere with its quantification, a magnetic pull-down using His-tagged recombinant STING was performed⁴⁶. The sample pellets were resuspended in 100 µl of solution containing 100 nM of His-tagged STING protein in disodium phosphate buffer and 2 µl of magnetic nickel beads. The plates were incubated for 2 h at room temperature with shaking at 450 RPM and then placed on a magnetic stand and the supernatant was removed. The samples were washed twice with 200 µl of water and 120 µl of 90% MeOH/10% (10 mM ammonium acetate pH 8, 0.1% NH₃) was added to each well and incubated for 20 min at room temperature with shaking at 450 RPM. On a magnetic stand, the supernatant was then transferred to an injection plate. A volume of 20 µl was injected on a Q-Exactive LC-MS using a Halo Hilic PRM method to quantify cGAMP. A gradient was performed with 70% mobile phase B (acetonitrile) and 30% mobile phase A (95% 10 mM ammonium acetate/5% acetonitrile) with a flow rate of 0.75 ml/min. At 1.8 min the gradient was ramped to reach 5% B/95% A and held for 3.3 min, before resuming the initial conditions at 3.4 min.

Mouse experiments

Mice were maintained under pathogen-free conditions and were used in compliance with protocols approved by the Institutional Animal Care and Use Committees of Admare Institute, which conform to the Canadian Council of Animal Care regulatory standards on experimental animal usage. Mice were single housed in Tecniplast ventilated cages and acclimated for at least 6 days. Mice were fed Teklad Global Soy Protein-Free Extruded Rodent 2920X Diet. Irradiated and contaminant-free water was provided in prefilled bottles ad libitum. All experiments were non-randomized and unblinded. Male 7- to 10-week-old C57Bl/6 mice were purchased from Charles River Laboratory (Saint-Constant, QC, Canada).

Generation of cGAS^I309T mice

Mice harboring the I309T point mutation in exon 3 of the cGAS gene (official name: C57BL/6NTac-Cgasem7978(I309T)Tac) were generated by CRISPR/Cas9 genome editing, in accordance with the strategy shown in Fig. S3A. The single-guide RNA (sgRNA; 5′-AATCTCTGTGGATATAATTC-3′) was designed using the CRISPOR gRNA design tool⁴⁷. An 84-nt single-stranded oligodeoxynucleotide (IDT, Coralville, IA, USA) containing the I309T point mutation and the recognition site of the restriction enzyme EcoRV was synthesized. The Cas9 protein, (4 µM; NEB, Ipswich, MA, USA), gRNA (4 µM, IDT) and oligonucleotide (20 µM; IDT) were electroporated into C57BL/6N mouse embryos to generate knock-in mice. Genotyping was performed by PCR amplification using primers flanking the insertion site. Restriction digest analysis of the PCR product with EcoRV and sequence analysis were used to validate the knock-in allele.

Pharmacokinetic analysis

Drug pharmacokinetics of 100 mg/kg oral dosing of G140 were determined in age and sex matched C57BL/6 WT and cGAS^I309T mice. For oral dosing a vehicle composed of 0.5% methyl cellulose + 0.1% Tween-80 was used. For i.v. dosing, a vehicle of 40–60% PEG 400 and 5–10% DMSO was used. Blood was collected at multiple timepoints from the jugular vein post-dosing in 0.1 M sodium citrate and compound levels were quantified by LC-MS. PK parameters were calculated using Phoenix WinNonlin (Certara) 8.3.5.340.

Mouse Concanavalin A model

Male C57BL/6 WT and cGAS^I309T mice aged 9–11 weeks were intravenously dosed at 10 mg/kg Concanavalin A (ConA) (Sigma # C5275) solution prepared at a concentration of 2 mg/ml in D-PBS. For cGAS inhibitor studies, G140 was resuspended in 0.5% Methocel + 0.1% Tween 80 in water and administered orally to mice at a dose of 100 mg/kg, 1 h preceding ConA administration. At 16 h post-ConA dosing, mice were sacrificed under isoflurane and blood and tissues were collected for processing. Blood compound levels were by determined LC-MS. The liver and spleen were homogenized in PBS using a Bead Ruptor 24 homogenizer instrument (Omni international) in ceramic bead containing tubes (Omni international #19-627) for cGAMP quantification.

SDS-PAGE and immunoblotting

Mouse tissues from WT and cGAS^I309T mice were homogenized in CytoBuster Protein Extraction Reagent (Sigma, #71009-M) supplemented with 1× protease inhibitor cocktail (Thermo Fisher Scientific #1861279) using a Bead Ruptor 24 homogenizer (OMNI International). Samples were centrifuged to remove debris and the supernatant was collected for protein quantification by BCA protein assay (Thermo Fisher Scientific #23227). Equivalent amounts of protein were mixed with 4× sample buffer (Bio-Rad, #1610747) and 2-mercaptoethanol (Bio-Rad, #1610710) and boiled for 5 min at 95 °C. SDS-PAGE was performed on precast criterion gradient 4–15% gels (Bio-Rad, #5678085). Proteins were then transferred to a nitrocellulose membrane and total protein was determined by Ponceau staining. Samples were blocked in 5% skim milk in TBS 0.1% Tween 20 for 1 h, then incubated overnight at 4 °C with primary antibodies in 5% skim milk TBS-T. The following day, the membranes were washed, incubated with secondary anti-rabbit-HRP antibodies for 1 h at room temperature, and then imaged using ECL SuperSignal™ West Dura Extended Duration Substrate reagent (Thermo Fisher Scientific #34075) and a ChemiDoc™ Touch Imaging System (Bio-Rad).

Flow cytometry

Tissues from WT and cGAS^I309T mice were crushed using glass microscope slides, resuspended in HBSS, filtered through a 70 μm cell strainer to remove debris, followed by red blood cell lysis (Biolegend #420301). Single cell suspensions were counted, stained with a viability dye (Biolegend, #423101) in PBS to identify dead cells, blocked with anti-CD16/anti-CD32 (Biolegend, #101320) in 2% FBS in PBS, and then stained with surface marker antibodies on ice for 20 min. Cells were then washed and fixed in 2% PFA for analysis. Samples were acquired using a LSR II flow cytometer (BD Biosciences) and data was analyzed using FlowJo software.

Statistical analyses

Data were expressed as mean + SD or mean ± SD from two to three replicates. Statistical analysis and the calculation of IC₅₀ values were performed using Prism v10 (GraphPad Software) or Dotmatics (Insightful Science). For comparing multiple groups, statistical significance was determined by one-way ANOVA with Dunnett’s multiple comparisons test. P values are denoted in figures.

General synthetic considerations

Unless otherwise noted, all reagents and solvents were purchased from vendors such as Aldrich, Enamine, Combi-Blocks, or VWR and used as received. NMR solvent (d6-DMSO) was purchased from Cambridge Isotopes Lab Inc. and used as received.

¹H and ¹³C NMR were recorded on a Bruker 400 MHz spectrometer. Chemical shifts are expressed in ppm values. ¹H NMR spectra are referenced to the solvent residual peak of 2.50 for d6-DMSO, and 13 C NMR spectra are referenced to the solvent residual peak of 39.52 for d6-DMSO. Peak multiplicities are designated by the following abbreviations: s, singlet; br.s, broad singlet; d, doublet; t, triplet; q, quartet; m, multiplet; r, rotomers; J, coupling constant in Hz. If a coupling pattern can be assigned as a combination of multiplicities, then the listed abbreviations are combined to provide an appropriate descriptor for the observed patterns (e.g., dt - doublet of triplets).

Analytical purity of all final compounds was greater than 95% as determined by LCMS using UV 254 nM detection unless otherwise stated.

Synthesis of compounds 1 – 2

Preparation of 6,7-dichloro-1-methyl-2,3,4,5-tetrahydro-1H-pyrido[4,3-b]indole (Compound S1)

To a solution of (2,3-dichlorophenyl)hydrazine hydrochloride (100 g, 468 mmol, 1.00 equiv) and (2S)-2-methylpiperidin-4-one hydrochloride (70.1 g, 468 mmol, 1.00 equiv) in EtOH (500 ml) was added concentrated H₂SO₄ (459 g, 4.68 mol, 10.0 equiv) at room temperature. The mixture was stirred overnight at 80 °C then cooled to room temperature and poured dropwise into ice water. The mixture was extracted with EtOAc (5 × 1000 ml). The EtOAc extracts were combined, washed with water (3 × 1000 ml) and dried over anhydrous Na₂SO₄, filtered, and concentrated under reduced pressure. The crude residue was triturated in EtOAc (40 ml), and the resulting precipitated solids were collected by filtration and washed with ethyl acetate (3 × 5 ml) to provide solids which were ~1:3 ratio of S1:S2 regioisomers by LCMS and as an unequal mixture of enantiomers of each regioisomer by chiral HPLC indicating epimerization during the reaction. The solids were then purified by C-18 reverse phase flash chromatography (mobile phase: H₂O (0.5% TFA)/MeOH (55:45) in three batches to provide 6,7-dichloro-1-methyl-2,3,4,5-tetrahydro-1H-pyrido[4,3-b]indole trifluoroacetate (Compound S1, 10.1 g, 6.1% yield) as a light brown solid and as an undetermined ratio of enantiomers. LCMS (ES) m/z: 255.05 [M + H]⁺, ¹H NMR (400 MHz, DMSO-d₆) δ 7.54 (d, J = 8.5 Hz, 1H), 7.22 (d, J = 8.4 Hz, 1H), 4.77 (q, J = 6.7 Hz, 1H), 3.62–3.53 (m, 1H), 3.46–3.40 (m, 1H), 3.04 (q, J = 5.7 Hz, 2H), 1.64 (d, J = 6.7 Hz, 3H).

Preparation of (R)-1-(6,7-dichloro-1-methyl-1,3,4,5-tetrahydro-2H-pyrido[4,3-b]indol-2-yl)-2-methoxyethan-1-one (Compound 1) and (S)-1-(6,7-dichloro-1-methyl-1,3,4,5-tetrahydro-2H-pyrido[4,3-b]indol-2-yl)-2-methoxyethan-1-one (Compound 2)

To a solution of 6,7-dichloro-1-methyl-2,3,4,5-tetrahydro-1H-pyrido[4,3-b]indole trifluoroacetate (Compound S1, 3.50 g, 9.51 mmol, 1.00 equiv) and methoxyacetic acid (1.71 g, 19.0 mmol, 2.00 equiv) in DMF (35 ml) was added HATU (3.98 g, 10.5 mmol, 1.10 equiv) and DIEA (3.68 g, 28.5 mmol, 3.00 equiv) at room temperature. The resulting mixture was stirred for 1 h at room temperature then poured into water and extracted with EtOAc (3 × 200 ml). The EtOAc extracts were combined, washed with water (5 × 500 ml), dried over anhydrous Na₂SO₄, filtered, and concentrated under reduced pressure. The residue was purified by silica gel flash column chromatography (eluted with 1:4 PE/EA) followed by further purification by prep-HPLC (conditions: Column, XBridge Prep OBD C18 column, 30 × 150 mm, 5 μm; mobile phase A: 10 mM NH₄HCO₃, mobile phase B: ACN; flow rate: 60 ml/min; gradient: 41% B to 43% B in 8 min, 43% B; wavelength: 254 nm; RT1 (min): 7) to afford 1-(6,7-dichloro-1-methyl-1,3,4,5-tetrahydro-2H-pyrido[4,3-b]indol-2-yl)-2-methoxyethan-1-one (Compound S3, 1.07 g, 35% yield) as a 66:34 mixture of enantiomers as an off-white solid.

1-(6,7-dichloro-1-methyl-1,3,4,5-tetrahydro-2H-pyrido[4,3-b]indol-2-yl)-2-methoxyethan-1-one (Compound S3, 1.0 g, 3.06 mmol) as 66:34 mixture of enantiomers was separated into constituent enantiomers by chiral SFC (conditions: system, Mettler Toledo Minigram; column, ChiralPak AD, 10 × 250 mm; gradient, isocratic 25% MeOH / 75% CO₂; flow rate, 10 ml/min) followed by additional chiral SFC separation (conditions: system, Waters Prep 100; column, Lux Amylose 1, 20 × 250 mm; gradient, isocratic 35% MeOH/65% CO₂; flow rate, 50 ml/min) to provide: (R)-1-(6,7-dichloro-1-methyl-1,3,4,5-tetrahydro-2H-pyrido[4,3-b]indol-2-yl)-2-methoxyethan-1-one (Compound 1), 162 mg (16% yield, 5.6% yield from compound S1), peak-2 (second eluting peak), %ee: 99.63% (conditions: ChiralPak AD, 4.6 × 150 mm; 5–60% MeOH/CO₂ over 7 min, rt: 4.148 min), ¹H NMR (400 MHz, DMSO) δ 11.46 (d, J = 7.7 Hz, 1H), 7.55–7.33 (m, 1H), 7.17 (t, J = 8.4 Hz, 1H), 5.61 and 5.14 (q, J = 6.7 Hz, 1H), 4.76–4.09 (m, 2H), 3.98 (dd, J = 13.9, 5.1 Hz, 1H), 3.49–3.35 (m, 1H), 3.31 (s, 3H), 3.12–2.85 (m, 1H), 2.81–2.71 (m, 1H), 1.48 and 1.38 (d, J = 6.5 Hz, 3H) [mixture of two rotomers], ¹³C NMR (101 MHz, DMSO-d₆) δ 167.2, 134.7, 133.6, 125.4, 122.9, 120.3, 117.4, 113.7, 112.2, 71.2, 58.3, 43.9, 37.5, 23.8, 19.2 [mixture of two rotamers, major rotamer peaks listed], HRMS (ES+) m/z: [M + H]⁺ calcd 327.0667, found 327.0663; and (S)-1-(6,7-dichloro-1-methyl-1,3,4,5-tetrahydro-2H-pyrido[4,3-b]indol-2-yl)-2-methoxyethan-1-one (Compound 2), 464 mg (46% yield, 16% yield from compound S1), peak-1 (first eluting peak), %ee: 99.95% (conditions: ChiralPak AD, 4.6 × 150 mm; 5–60% MeOH / CO₂ over 7 min, rt = 3.775 min), LCMS (ES) m/z: 327.05 [M + H]⁺, LCMS % purity by UV @ 254 nm  = 99.1%, ¹H NMR (400 MHz, DMSO) δ 11.46 (d, J = 7.7 Hz, 1H), 7.46 (t, J = 9.5 Hz, 1H), 7.16 (t, J = 8.5 Hz, 1H), 5.61 and 5.14 (q, J = 6.5 Hz, 1H), 4.82–4.08 (m, 2H), 3.98 (dd, J = 14.0, 5.3 Hz, 1H), 3.50–3.36 (m, 1H), 3.31 (s, 3H), 3.14–2.83 (m, 1H), 2.83–2.68 (m, 1H), 1.48 and 1.38 (d, J = 6.6 Hz, 3H) [mixture of two rotomers], ¹³C NMR (101 MHz, DMSO-d₆) δ 167.2, 134.7, 133.6, 125.4, 122.9, 120.3, 117.4, 113.7, 112.2, 71.2, 58.3, 43.9, 37.5, 23.8, 19.2 [mixture of two rotomors, major rotomer peaks listed], HRMS (ES+) m/z: [M + H]⁺ calcd 327.0667, found 327.0664; and Absolute stereochemistry of compounds 1 and 2 were assigned based on the cGAS co-crystal structure of Compound 2.

Synthesis of compound 3

Preparation of 1-(6,7-dichloro-1,1-dimethyl-1,3,4,5-tetrahydro-2H-pyrido[4,3-b]indol-2-yl)-2-methoxyethan-1-one (Compound 3)

A solution of 2-(6,7-dichloro-1H-indol-2-yl)ethanamine (1.00 eq, 54 mg, 0.236 mmol) and acetone (2.00 eq, 0.035 ml, 0.471 mmol) in THF (1 ml) was treated with trifluoroacetic acid (1.50 eq, 0.027 ml, 0.354 mmol) and stirred at rt overnight. The volatiles were then removed under reduced pressure to provide crude compound S5 6,7-dichloro-1,1-dimethyl-2,3,4,5-tetrahydro-1H-pyrido[4,3-b]indole which was used directly in the next step without further purification. The yield was considered quantitative.

Crude compound S5 6,7-dichloro-1,1-dimethyl-2,3,4,5-tetrahydro-1H-pyrido[4,3-b]indole (1.00 eq, 63 mg, 0.234 mmol) and 2-methoxyacetic acid (2.00 eq, 42 mg, 0.468 mmol) were combined DMF (1 ml). To the solution was added DIPEA (3.00 eq, 0.12 ml, 0.702 mmol) and HATU (1.00 eq, 89 mg, 0.234 mmol) and then the mixture was stirred at rt for 20 min. The reaction mixture was then directly purified by C18 reverse phase chromatography eluting with 0-100% MeCN/H₂O to afford 1-(6,7-dichloro-1,1-dimethyl-1,3,4,5-tetrahydro-2H-pyrido[4,3-b]indol-2-yl)-2-methoxyethan-1-one (Compound 3, 23 mg, 29% yield). LCMS (ES) m/z: 341.31 [M + H]⁺, ¹H NMR (400 MHz, DMSO) δ 11.46 (s, 1H), 7.51 (d, J = 8.6 Hz, 1H), 7.16 (d, J = 9.5 Hz, 1H), 4.13 (s, 2H), 3.57 (t, J = 5.4 Hz, 2H), 3.32 (s, 3H), 2.82 (t, J = 5.4 Hz, 2H), 1.80 (s, 6H), ¹³C NMR (101 MHz, DMSO) δ 169.68, 134.56, 133.95, 124.58, 122.68, 120.09, 118.46, 117.36, 113.86, 73.24, 58.09, 57.51, 40.81, 25.93, 23.94, HRMS (ES + ) m/z: [M + H]⁺ calcd 341.0823, found 341.0819.

Author contributions

Conceptualization: K.P., A.M.S., and L.W. Experiments and Methodology: A.M.S., L.W., N.S., R.E.B., S.C., S.D., V.D., N.F., S.G., P.L.G., D.M., R.S., D.S., L.Z., M.O.B., J.B., A.C., L.F., L.L., and W.X. Writing: A.M.S., K.P., L.W., and R.E.B. Review and editing: All Authors. Supervision: K.P., M.A.C., T.T., and D.J.P.

Peer review

Peer review information

Communications Chemistry thanks Junmin Quan, Eileen Parkes, and the other, anonymous, reviewer for their contribution to the peer review of this work.

Data availability

The pdb files for the solved crystal structures of h-cGAS:G150:AMPPNP and h-cGAS:Compound 2:AMPPNP have been deposited in the RCSB Protein Data Bank. All other relevant data that support the findings of this study are available from the corresponding authors upon request.

Competing interests

A.M.S., L.W., N.S., R.E.B., V.D., N.F., S.G., D.M., R.S., L.Z., and K.-A.P. and M.A.C. are employees of Ventus Therapeutics. S.C., S.D., P.L.G., D.V.S., M.-O.B., J.D.B., A.C., and L.D.F. are former employees of Ventus Therapeutics. T.T. is a scientific founder of Ventus Therapeutics. L.L., W.X., and D.J.P. declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42004-025-01481-7.