Small molecule inhibition of cGAS reduces interferon expression in primary macrophages from autoimmune mice

Abstract

Cyclic GMP-AMP synthase is essential for innate immunity against infection and cellular damage, serving as a sensor of DNA from pathogens or mislocalized self-DNA. Upon binding double-stranded DNA, cyclic GMP-AMP synthase synthesizes a cyclic dinucleotide that initiates an inflammatory cellular response. Mouse studies that recapitulate causative mutations in the autoimmune disease Aicardi-Goutières syndrome demonstrate that ablating the cyclic GMP-AMP synthase gene abolishes the deleterious phenotype. Here, we report the discovery of a class of cyclic GMP-AMP synthase inhibitors identified by a high-throughput screen. These compounds possess defined structure-activity relationships and we present crystal structures of cyclic GMP-AMP synthase, double-stranded DNA, and inhibitors within the enzymatic active site. We find that a chemically improved member, RU.521, is active and selective in cellular assays of cyclic GMP-AMP synthase-mediated signaling and reduces constitutive expression of interferon in macrophages from a mouse model of Aicardi-Goutières syndrome. RU.521 will be useful toward understanding the biological roles of cyclic GMP-AMP synthase and can serve as a molecular scaffold for development of future autoimmune therapies.Upon DNA binding cyclic GMP-AMP synthase (cGAS) produces a cyclic dinucleotide, which leads to the upregulation of inflammatory genes. Here the authors develop small molecule cGAS inhibitors, functionally characterize them and present the inhibitor and DNA bound cGAS crystal structures, which will facilitate drug development.

Fig. 1

Development of a high-throughput screen for the identification of cGAS inhibitors. a Schematic of immune stimulatory dsDNA-dependent cGAS synthesis of cyclic GMP-AMP (cGAMP). b The enzymatic activity of cGAS is determined by monitoring the consumption of ATP (blue) and GTP (green), and the generation of cGAMP (red) using an RF-MS. The assay was incubated for 120 min using 60 nM cGAS. Normalized extracted ion count values are plotted. c For the high-throughput screen, 300 nM dsDNA was used to stimulate mouse cGAS activity; % cGAMP product formation was measured against 1266 compounds (blue dots) at a final concentration of 12.5 μM using Sigma-Aldrich LOPAC (library of pharmacologically active compound) library plates. The results of two independent days, as technical replicates, are plotted against each other as normalized percent of inhibition (NPI). The negative control DMSO is shown in green and the positive control (no dsDNA) is shown in red. Data were analyzed using Vortex software and the coefficient of correlation was 0.86. d The results using the LOPAC library resulted in the assay having a Z prime factor of 0.76. e Following the high-throughput screen from over 100,000 small-molecule compounds, the following four were selected for additional characterization. See text for details on the triage process

Fig. 2

Ternary crystal structure of cGAS with dsDNA and RU.365. a Overall structure of the cGAS-dDNA-RU.365 complex, with cGAS in violet, dsDNA in light orange, and ligand in yellow. RU.365 is shown in stick representation. b Close-up of the cGAS-binding pocket (shown in surface representation and violet) and bound RU.365 (show in space-filling representation and yellow). The black dashed box indicates the entire binding pocket of cGAS. c The amino acids (shown in dots and stick, in violet) surrounding the bound RU.365 (shown in stick, in yellow). d–g The conserved stacking interaction mediated by Arg364 and Tyr421 with the bound RU.365 (d), ATP (e), 5′-pppGpG (f), and 2′,3′-cGAMP (g). The bound ligands and these two residues are shown in stick representation

Fig. 3

A structural comparison of RU.365 with its derivative RU.521. a Omit Fo-Fc electron density (green mesh) of RU.365 and RU.521 contoured at 3σ. b Superposition of structures of bound RU.365 (yellow) and RU.521 (cyan). c The dichloro substitution target deeper in the cGAS pocket. RU.521 (yellow) and RU.365 (silver) are shown in stick, with two chlorines shown in dots. cGAS (violet) is shown in surface view. d The stacking interaction between RU.521 (yellow)/RU.365 (silver) and R364/Tyr421. e Water-mediated interaction between RU.521 (yellow) and cGAS (violet). Lys350 in RU.365 complex is shown in silver. Superposition of structures of bound RU.365 f–h and RU.521 i–k with those of substrates, intermediate, and product. All the ligands are shown in stick representation. Inhibitors are shown in yellow. Substrates/Intermediate/Product are shown in cyan

Fig. 4

cGAS inhibition by RU.365 and its chemical active analogs. a In vitro concentration response curves for RU.365 and the most potent analog synthesized in-house, RU.521, in the presence of dsDNA, ATP and GTP as measured by RF-MS. The IC50 values for RU.365, RU.521 and less potent analogs are listed in Supplementary Table 4. Values are averages of triplicate determinations with SD indicated. b Isothermal titration calorimetry-binding curves for RU.521 titrated into cGAS/dsDNA complex

Fig. 5

Small-molecule inhibition of cGAS-dependent interferon induction in cellular assays. RU.365, RU.332, and RU.521 were tested in cellular assays for their effectiveness in inhibiting cGAS activity in dsDNA-stimulated RAW macrophages. The inhibition of type I IFN response via dsDNA was monitored via an interferon sensitive promoter coupled to a luciferase gene. The concentration response curves for RU.365 (a), RU.332 (b), and RU.521 (c) in RAW cells are shown. Error bars represent SEM

Fig. 6

Potent and selective inhibition of cGAS activity in RAW macrophage and BMDM cells from an Aicardi-Goutières Syndrome mouse model. RAW luciferase reporter cells were exposed to dsDNA (a), 5′ppp-HP20 RNA (b), Pam3CSK4 (c), poly(I:C) (d), lipopolysaccharide (LPS) (e), or recombinant murine interferon-β (Ifnb) f to promote either a type I interferon (a), (b), (f), or NF-κB (c), (d), (e) response under different immune stimuli and simultaneously treated with indicated small molecule (or vehicle). Type I interferon response was read via luciferase reporter, while NF-κB response was read via qRT-PCR. g RAW KO-cGAS reporter cells were exposed to dsDNA or cGAMP and simultaneously treated with each of the small molecules. h The cytotoxic effects of the lead compounds were tested in RAW macrophages at concentrations spanning the range tested in the dose response curves. Cytotoxicity was measured by the quantitation of ATP (CellTiter-Glo assay) at 72 h and shown as percent viability of cells. i BMDM cells from Trex1 ^−/− mice were treated with the indicated compound for 24 h and then harvested for qRT-PCR analysis. Shown are the results, relative to IFNB1 expression in Trex1 wild-type BMDMs. Error bars represent SEM