cGAS-like receptors sense RNA and control 3'2'-cGAMP signalling in Drosophila

Abstract

Cyclic GMP-AMP synthase (cGAS) is a cytosolic DNA sensor that produces the second messenger cG[2'-5']pA[3'-5']p (2'3'-cGAMP) and controls activation of innate immunity in mammalian cells^1-5. Animal genomes typically encode multiple proteins with predicted homology to cGAS^6-10, but the function of these uncharacterized enzymes is unknown. Here we show that cGAS-like receptors (cGLRs) are innate immune sensors that are capable of recognizing divergent molecular patterns and catalysing synthesis of distinct nucleotide second messenger signals. Crystal structures of human and insect cGLRs reveal a nucleotidyltransferase signalling core shared with cGAS and a diversified primary ligand-binding surface modified with notable insertions and deletions. We demonstrate that surface remodelling of cGLRs enables altered ligand specificity and used a forward biochemical screen to identify cGLR1 as a double-stranded RNA sensor in the model organism Drosophila melanogaster. We show that RNA recognition activates Drosophila cGLR1 to synthesize the novel product cG[3'-5']pA[2'-5']p (3'2'-cGAMP). A crystal structure of Drosophila stimulator of interferon genes (dSTING) in complex with 3'2'-cGAMP explains selective isomer recognition, and 3'2'-cGAMP induces an enhanced antiviral state in vivo that protects from viral infection. Similar to radiation of Toll-like receptors in pathogen immunity, our results establish cGLRs as a diverse family of metazoan pattern recognition receptors.

Fig. 1. Structural remodelling in animal cGLRs enables divergent pattern recognition.

a, Crystal structures and surface electrostatics of hMB21D2 and Tc-cGLR. Structural comparison with the human cGAS (hcGAS)–DNA complex (Protein Data Bank (PDB): 6CTA) reveals that cGLRs have a conserved architecture with a nucleotidyltransferase signalling core and a shared primary ligand-binding surface (dashed lines). The purple and green boxes indicate cutaways in b. b, Zoomed-in cutaways highlighting structural insertions and deletions unique to each cGLR. hMB21D2 and Tc-cGLR lack the Zn-ribbon motif present in cGAS (left) and hMB21D2 contains a C-terminal α-helix extension that contacts the central ‘spine’ helix (right). Alterations in the predicted ligand-binding surfaces suggest individual cGLRs are remodelled to recognize different molecular patterns. c, Thin-layer chromatography analysis and quantification of Tc-cGLR reactions in the presence of nucleic acid ligands. Tc-cGLR is specifically activated by dsRNA recognition to synthesize a nucleotide (nt) product. Data are relative to maximum activity and represent the mean ± s.e.m. for n = 3 independent experiments. Ori, origin; P_i, inorganic phosphate; ss, single-stranded. Source data

Fig. 2. Drosophila cGLR1 senses long dsRNA.

a, Phylogeny representing 153 Diptera cGLR genes clustered into clades 1–5 (less than 30% sequence identity between clades). Forty-one of forty-two analysed Diptera species encode enzymes in clade 5, including D. melanogaster CG12970 (cGLR1) and CG30424. Enzymes analysed by forward biochemical screen (red dot) and identified as dsRNA-sensing cGLRs (blue circle) are denoted. b, Diptera cGLRs identified in the screen are activated to form a nucleotide product by the dsRNA mimic poly I:C. c, Mutation to the Dm-cGLR1 active site ablates enzymatic activity. Data in b and c are representative of n = 3 independent experiments. WT, wild type. d, Dm-cGLR1 in vitro activity was monitored in the presence of a panel of dsRNAs and quantified relative to 40 bp dsRNA. Data are the mean ± s.e.m. of n = 3 independent experiments. e, Analysis of Dm-cGLR1 activity in human cells using mammalian STING and IFNβ reporter induction, quantified relative to WT activity. Dm-cGLR1 signalling in cells is dependent on stimulation of dsRNA and mutation of the catalytic site, or the predicted ligand-binding residues ablates activity. Data are mean ± s.e.m. of n = 3 technical replicates and representative of n = 3 independent experiments. The inset shows a model of the Tc-cGLR–dsRNA complex based on the hcGAS–dsDNA structure (PDB: 6CTA) used to predict the Dm-cGLR1 ligand-binding residues R23, K42, K52, R241 and K251. Source data

Fig. 3. Discovery of 3′2′-cGAMP as a metazoan nucleotide second messenger.

a, High- performance liquid chromatography (HPLC) analysis of the Dm-cGLR1 reaction (orange) and comparison with synthetic standards (black or dashed lines) demonstrates that Dm-cGLR1 synthesizes 3′2′-cGAMP as the major product. A minor Dm-cGLR1 reaction product is 2′3′-c-di-AMP (see also Extended Data Fig. 6a). b, Thin-layer chromatography analysis of mouse cGAS and Dm-cGLR1 reactions labelled with either α-³²P-ATP or α-³²P-GTP (indicated as [α-³²P]NTP) and treated as indicated. Pairwise labelling and nuclease P1 digestion verify that cGAS and Dm-cGLR1 synthesize distinct cGAMP isomers with opposite phosphodiester linkage specificities. Representative of n = 3 independent experiments. High-resolution mass spectrometry confirms the major Diptera cGLR product as 3′2′-cGAMP (see also Extended Data Fig. 6b). c, HPLC quantification of insect cGLR nucleotide products. 3′2′-cGAMP is the dominant product of each identified Diptera cGLR (denoted by a black line), and 2′3′-cGAMP is the dominant product of cGAS and Tc-cGLR. Data are the mean quantified product of n = 3 independent experiments. d, Thermal denaturation assay showing that dSTING selectively recognizes 3′2′-cGAMP (see also Extended Data Fig. 8b, c). Representative of n = 3 independent experiments. e, Crystal structure of the dSTING–3′2′-cGAMP complex reveals a tightly closed homodimer and an ordered β-strand lid, indicating high-affinity engagement of the endogenous Drosophila second messenger 3′2′-cGAMP. f, Alignment and conservation of the stem helix and β-strand lid in human and insect STING proteins. Critical ligand-binding residues (blue dot) and adaptations specific to Diptera (red outline) are denoted. g, Superposition of the dSTING–3′2′-cGAMP (blue–orange) complex and the human STING–2′3′-cGAMP (grey–pink) (PDB: 4KSY) complex reveals that human STING readout of the 2′–5′ phosphodiester bond by R232 is absent in dSTING (left). Human STING S162 (grey) contacts the free 3′ OH of the guanosine base in 2′3′-cGAMP (pink). dSTING N159 (blue) extends across the ligand-binding pocket to contact the free 3′ OH of the adenosine base in 3′2′-cGAMP (orange) (right).

Fig. 4. 3′2′-cGAMP activates STING-dependent antiviral immunity in Drosophila.

a, Synthetic 3′2′-cGAMP or 3′3′-c-di-GMP was injected into the body cavity of flies and gene expression was measured after 24 h. Sting-regulated gene 1 (Srg1) RNA levels are shown as fold induction compared with buffer control in WT. dSTING^Mut is the RXN mutant, as previously characterized^,. Data in a and b are mean ± s.e.m. of RNA levels measured relative to the control gene Rpl32 from n = 3 independent experiments of n = 6 flies. The P values, calculated by unpaired t-test, were not significant (NS) unless otherwise noted; NS P > 0.05. ^****P < 0.001, ^*P = 0.0119. b, Viral RNA loads 3 days after infection with Drosophila C virus (DCV) demonstrate significantly diminished viral replication in WT flies injected with 3′2′-cGAMP. ^**P = 0.051. c, Survival analysis of animals infected with DCV demonstrates that injection of 3′2′-cGAMP results in a Sting-dependent response that significantly delays mortality. Data are mean ± s.e.m. ^****P < 0.001. Data in c and d are each from n = 3 independent experiments of n = 30 flies. dpi, days post-infection. d, Survival analysis directly comparing the effects of cGAMP isomers 7 days after DCV infection. 3′2′-cGAMP injection increases animal survival in a dose-dependent manner and confers greater protection than 2′3′-cGAMP (see also Extended Data Fig. 10d). e, Proposed model for cGLR–STING signalling. Upon recognition of distinct molecular patterns, animal cGLRs synthesize a nucleotide second messenger that activates antiviral immunity through STING. Source data

Extended Data Fig. 1. Sequence and structural analysis of hMB21D2 and Tc-cGLR.

a, Structure guided sequence alignment of the catalytic domain of hcGAS (PDB: 4KM5), hMB21D2 and Tc-cGLR. Strict secondary structure conservation further supports conserved structural homology despite primary sequence divergence. The [D/E]hD[X_50–90]D catalytic triad is highlighted with a red outline and the human Zn-ribbon insertion that is absent in other cGLRs is denoted with a red dashed outline. hMB21D2 contains an additional 61 residues that are not resolved in the crystal structure and are absent from the alignment. b, c, Zoomed-in cutaways of the hMB21D2 (b) and Tc-cGLR (c) crystal structures highlighting positioning of conserved catalytic residues in the nucleotidyltransferase active site. In hcGAS, the analogous residues coordinate two Mg²⁺ metal ions to control synthesis of 2′3′-cGAMP (inset, middle; PDB: 6CTA). The hMB21D2 structure is in an inactive state distinguished by misaligned catalytic residues and occlusion of the active site by an extended Gly-Gly activation loop, indicating that catalytic activation is probably controlled by a conformational rearrangement. d, e, TLC analysis of in vitro tests for potential activating ligands of hMB21D2. No nucleotide products were identified upon stimulation with 40-nt or 40-bp nucleic acid ligands (d) or ligands known to activate mammalian Toll-like receptors (e). Data shown are representative of n = 3 independent experiments. f, Z-score structural similarity plot showing homology between hMB21D2 and Tc-cGLR with representative structures in the PDB (PDB90). Increasing Z-score indicates greater homology, confirming the close relationship between animal cGLR enzymes and mammalian cGAS, and more distant similarity to cGAS/DncV-like nucleotidyltransferases (CD-NTases) in bacterial antiphage defence systems and human oligoadenylate synthase 1 (refs. ^,,,–). Z-score cut-offs are 13 and 15 for hMB21D2 and Tc-cGLR, respectively.

Extended Data Fig. 2. Forward biochemical screen of predicted cGLRs in Diptera.

a, Violin plot showing the number of predicted cGLRs in Diptera genomes. Drosophila genomes (n = 31 species) have a median of four predicted cGLRs in contrast to a median of two predicted cGLRs in other dipteran insects (n = 11 species). b, Schematic of the in vitro screen of predicted cGLRs in the order Diptera. Fifty-three sequences were selected representing each clade in the phylogeny in Fig. 2a. Following recombinant protein expression in E. coli, lysates were split into two samples for parallel TLC analysis of in vitro enzymatic activity and HPLC-MS analysis of lysate nucleotide metabolites.c, d, Purified cGLR proteins were incubated overnight at 37 °C with α³²P-radiolabelled nucleotides, a mixture of Mn²⁺ and Mg²⁺, and the 45-bp immunostimulatory DNA ISD45 or the synthetic dsRNA analogue poly I:C as potential nucleic acid ligands, and reactions were visualized by PEI-cellulose TLC. Wild-type (WT) and catalytically inactive mouse cGAS enzymes were used as controls for each sample set. Note that mouse cGAS exhibits dsDNA-independent activity in the presence of Mn²⁺ (ref. ). Predicted Diptera cGLRs are grouped by clade (DC01–05) and numbered within each clade. Ligand-dependent activity was identified for DC02_01, DC05_03, DC05_19 and DC05_21; species listed below. We observed ligand-independent activity for two enzymes in clade 3. Data represent n = 2 independent experiments. e, SDS–PAGE and Coomassie stain analysis of NiNTA-purified cGLR protein fractions used for the biochemical screen. f, SDS–PAGE and Coomassie stain analysis of final cGLR proteins used for biochemical studies, which were purified by NiNTA-affinity, ion-exchange chromatography and size-exclusion chromatography.

Extended Data Fig. 3. Sequence analysis and mutagenesis of insect cGLRs.

a, Alignment of the catalytic domain of hcGAS and active cGLRs identified in T. castaneum, D. eugracilis, L. cuprina, D. erecta, D. simulans and D. melanogaster. The EhD[X_50–90]D catalytic triad is highlighted with a red outline and the human Zn-ribbon insertion that is absent in insect cGLRs is denoted with a red dashed outline. Predicted basic ligand-binding residues selected for mutational analysis denoted by black circles. cGLRs from D. erecta and D. simulans are close homologues of Dm-cGLR1 (76% and 91% sequence identity, respectively) and thus are also referred to as ‘cGLR1’. All biochemical experiments with Ds-cGLR1 were performed with a construct beginning at M19. b, In vitro reactions demonstrating that mutation of the catalytic residues ablates nucleotide product synthesis by Ds-cGLR1 in response to poly I:C. c, d, In vitro reactions analysing dsRNA recognition through the putative ligand-binding surface by Ds-cGLR1 (c) or Tc-cGLR (d). The insets for panels c and d show models of the Tc-cGLR–dsRNA complex based on the hcGAS–dsDNA structure (PDB: 6CTA), indicating predicted dsRNA-interacting residues in Ds-cGLR1 (c) or Tc-cGLR (d). Charge swap mutation to these residues variably disrupted poly I:C-stimulated activity by Ds-cGLR1 and Tc-cGLR, shown by TLC (left) and quantified relative to WT activity (right). Data in b–d are representative of n = 3 independent experiments. e, SDS–PAGE and Coomassie stain analysis of purified WT and mutant proteins, as labelled in the above TLC images. f, IFNβ luciferase assay in which cGLRs are expressed in human cells and CDN synthesis is detected by mammalian STING activation, as in Fig. 2e. IFNβ was quantified relative to the empty vector control. In comparison to hcGAS control, which is activated by expression vector-plasmid DNA, Dm-cGLR1 (left) and Ds-cGLR1 (right) strictly require poly I:C stimulation to activate a downstream STING response. Mutation to catalytic residues or putative ligand-binding residues ablates cGLR1 signalling. See Fig. 2e: Dm-cGLR1 activity quantified relative to WT activity upon poly I:C stimulation. Data are mean ± s.e.m. of n = 3 technical replicates and representative of n = 3 independent experiments. Source data

Extended Data Fig. 4. Analysis of RNA recognition by insect cGLRs.

a–c, In vitro activity assays for each active insect cGLR demonstrating that dsRNA recognition is required for enzyme activation. Reactions were performed with 40-nt or 40-bp synthetic ligands. Weak Deu-cGLR ssRNA-stimulated activity may be explained by transient short duplex formation similar to observations that some ssDNA oligos can stimulate mouse cGAS dsDNA-dependent activity. b, TLC and quantification for enzyme activation in the presence of a panel of 10–40-bp synthetic dsRNA ligands. dsRNA (30 bp) is sufficient to stimulate maximal activity for Tc-cGLRs, Dm-cGLRs and Lc-cGLRs, while Ds-cGLR1 requires 35 bp and Deu-cGLR can be activated by dsRNAs as short as 15 bp. Data are mean ± s.e.m., quantified relative to maximum observed activity. c, Reactions with 146-bp in vitro-transcribed dsRNAs containing either a 5′ triphosphate or 5′ OH termini demonstrate that dsRNA recognition by insect cGLRs does not involve 5′-end discrimination. d, Deconvolution of catalytic metal requirements for enzymatic activity by insect cGLRs. Insect cGLRs require Mn²⁺ for maximal catalytic activity, with weak product formation observed in the presence of Mg²⁺. e, Poly I:C titration demonstrates that dsRNA stimulation of Drosophila cGLR1 activity in cells is dependent on RNA concentration. IFNβ luciferase assay in which cGLRs are expressed in human cells and CDN synthesis is measured by mammalian STING activation, as in Fig. 2e and Extended Data Fig. 3f. IFNβ quantified relative to the empty vector control. Data are mean ± s.e.m. of n = 3 technical replicates. All data in a–e represent n = 3 independent experiments. Source data

Extended Data Fig. 5. Characterization of Ds-cGLR1–dsRNA condensate formation.

a, Electrophoretic mobility shift assay (EMSA) showing binding between Ds-cGLR1 or the C-terminal NTase domain of hcGAS (hcGAS-NTase) and a 40-bp dsRNA or 45-bp dsDNA. Ds-cGLR1 preferentially binds to dsRNA and more weakly interacts with dsDNA, consistent with observed binding between hcGAS and dsRNA. b, EMSA comparison of Ds-cGLR1–dsRNA binding and mammalian cGAS–dsDNA binding. Similar to hcGAS, Ds-cGLR1 forms a higher-order protein–nucleic acid complex that does not migrate through the gel, in contrast to the 2:2 binding observed between mouse cGAS and dsDNA. Data in a and b are representative of n = 3 independent experiments. c, Analysis of the effect of AF488 labelling on Ds-cGLR1 enzymatic activity. Similar to previous observations with hcGAS, AF488 labelling negatively impacts enzymatic activity but has minimal effect at the ratio of 90% unlabelled and 10% labelled protein used for all imaging experiments. Data are mean ± s.e.m. of n = 3 independent experiments. d, e, Analysis of hcGAS (d) and Ds-cGLR1 (e) phase separation with AF488-labelled protein. Mammalian cGAS contains a highly disordered N-terminal extension of approximately 150 residues, but this unstructured extension is absent in insect cGLR sequences. In the presence of dsDNA, full-length hcGAS forms highly dynamic liquid droplets^,,, whereas the minimal hcGAS NTase domain forms rigid protein–DNA condensates similar to those formed by Ds-cGLR1–RNA complexes. hcGAS exhibits a preference for condensate formation in the presence of dsDNA (d), whereas Ds-cGLR1 exhibits a preference for dsRNA (e), as observed in panel a. Scale bars, 10 μm. Analysis of Ds-cGLR1 dsRNA length specificity for condensate formation demonstrates clear length dependency (e) and supports that long dsRNA and condensate formation are required for maximal Ds-cGLR1 activation. Source data

Extended Data Fig. 6. Synthesis of 3′2′-cGAMP by Diptera cGLRs.

a, HPLC analysis of the nucleotide products of Tc-cGLR, Dm-cGLR1, Ds-cGLR1, Lc-cGLR and Deu-cGLR reactions compared with relevant synthetic controls. Integration of major and minor product peaks in n = 3 independent experiments was used to calculate relative product ratios shown in Fig. 3c. b, The Drosophila cGLR major reaction product was purified from Deu-cGLR reactions and compared with synthetic 3′2′-cGAMP with tandem mass spectrometry analysis. Parent mass extracted ion trace (left) and tandem mass spectra comparison (right) validate the chemical identity of the Drosophila cGLR product as 3′2′-cGAMP. c, Identification of widespread 3′2′-cGAMP synthesis by Diptera cGLRs. The heat map shows the relative concentrations of cGAMP isomers detected by HPLC-MS in bacterial lysates expressing Diptera cGLRs (as described in Extended Data Fig. 2b) (left). In all cases, 3′2′-cGAMP was present as the dominant product with trace amounts of 3′3′-cGAMP and 2′3′-cGAMP detected in some samples as minor species. Right, inset of clade 5 in the Diptera cGLR phylogeny from Fig. 2a annotated to show all enzymes identified to synthesize 3′2′-cGAMP.

Extended Data Fig. 7. Mechanism of 3′2′-cGAMP bond formation and resistance to degradation by viral poxin enzymes.

a, Analysis of Dm-cGLR1 reactions with pairwise combinations of α-³²P-labelled nucleotides and non-hydrolyzable nucleotides reveals reaction intermediates and identifies the order of bond formation during 3′2′-cGAMP synthesis. Left: TLC analysis demonstrates that Dm-cGLR1 forms a linear intermediate in the presence of GTP and non-hydrolyzable ATP (Apcpp), indicating that the 2′–5′ phosphodiester bond is synthesized first. Exposed γ-phosphates removed by phosphatase treatment before analysis are indicated in parentheses. Note that while a linear intermediate cannot be formed in the presence of non-hydrolyzable GTP (Gpcpp), Dm-cGLR1 will synthesize the off-product 2′3′-c-di-AMP. Mouse cGAS, which synthesizes 2′3′-cGAMP through the linear intermediate pppG[2′–5′]pA, is shown here for comparison. Right: schematic of the reaction mechanism for each enzyme. b, Poxins are 2′3′-cGAMP-specific viral nucleases that disrupt cGAS–STING signalling. HPLC analysis of synthetic 2′3′-cGAMP or 3′2′-cGAMP treated with poxin from the insect baculovirus Autographa californica nucleopolyhedrovirus (AcNPV) is shown^,. In 1 min, AcNPV poxin cleaves 2′3′-cGAMP into a mixture of intermediate and full-cleavage product; and after 1 h, turnover is complete. No cleavage of 3′2′-cGAMP is observed by AcNPV poxin under these reaction conditions. c, Using TLC as a more sensitive assay, we observed minimal cleavage of 3′2′-cGAMP following overnight incubation with AcNPV poxin. Data in a–c are representative of n = 3 independent experiments. d, Schematic highlighting how an isomeric switch in phosphodiester linkage specificity makes 3′2′-cGAMP remarkably resistant to poxin-mediated cleavage.

Extended Data Fig. 8. Structural and biochemical analysis of dSTING.

a, Alignment of the C-terminal CDN-binding domains of human STING, mouse STING, D. eugracilis STING and D. melanogaster STING. Architecture of the core CDN-binding domain is conserved across metazoans; the disordered C-terminal tail, which controls IRF3–IFNβ signalling, is specific to vertebrates^,. Ligand-interacting residues selected for mutational analysis are denoted with a black circle; Diptera-specific adaptations are highlighted with a red outline. All structural and biochemical experiments were performed with a D. eugracilis STING construct terminating at I340. b, In vitro thermal denaturation assay analysing dSTING interactions with a panel of CDNs. Only 3′2′-cGAMP forms a thermostable complex with dSTING in vitro (see also Fig. 3d). 2′3′-cGAMP is known to be capable of stimulating dSTING-dependent signalling in vivo, supporting that dSTING can engage with 2′3′-cGAMP with lower affinity. This observation is consistent with the weaker recognition of bacteria-derived 3′3′-cGAMP and 3′3′-c-di-GMP by human STING^,. c, In vitro thermal denaturation assay demonstrating concentration-dependent thermal shift induced by 3′2′-cGAMP. d, Dose titration of 2′3′-cGAMP and 3′2′-cGAMP in human cells demonstrating selective response by dSTING to 3′2′-cGAMP. The D. eugracilis CDN-binding domain (CBD) was adapted for downstream signalling in human cells by addition of N-terminal human transmembrane (hTM) domains and the human C-terminal tail (hCTT). e, Comparison of the human STING–2′3′-cGAMP and dSTING–3′2′-cGAMP crystal structures reveals a conserved closed homodimer architecture in which apical ‘wings’ are spread 32–36 Å, demonstrating high-affinity engagement with an endogenous ligand. f, Enlarged cutaways of 3′2′-cGAMP in the dSTING crystal structure. Above: the simulated annealing F_O−F_C omit map (contoured at 3 σ). Below: a top-down view highlighting key dSTING–3′2′-cGAMP contacts. g, Full crystal structure used to determine the structure of D. eugracilis STING in complex with 3′2′-cGAMP. T4 lysozyme is fused to the N terminus of the D. eugracilis STING CBD. h, Thermal denaturation assay as in Fig. 3d demonstrating that N-terminal fusion of T4 lysozyme does not impair dSTING recognition of 3′2′-cGAMP. i, Mutational analysis of key ligand-interacting residues in dSTING; the thermal denaturation assay was used to analyse 3′2′-cGAMP recognition. Mutations that conserve functional contacts with 3′2′-cGAMP (Y164F) maintain ligand recognition; mutations that ablate contacts abrogate ligand binding. N159S exhibits diminished ability to recognize 3′2′-cGAMP. Data in b and i are mean ± s.e.m. of the average T_m calculated from n = 2 technical replicates in n = 3 independent experiments. Data in c are representative of n = 3 independent experiments. Data in d are mean ± s.e.m. of n = 3 technical replicates and representative of n = 3 independent experiments. j, SDS–PAGE and Coomassie stain analysis of purified WT and mutant proteins. Source data

Extended Data Fig. 9. 3′2′-cGAMP induces the expression of dSTING-regulated genes.

a–d, Injection of 3′2′-cGAMP into D. melanogaster has a dose-dependent effect on the expression of Sting-regulated genes (srgs). 2′3′-cGAMP was used as positive control as previously characterized^,. Synthetic nucleotide was injected into the body cavity of WT (w¹¹¹⁸) flies and gene expression was measured after 24 h. RNA levels were measured relative to the control gene RpL32, and nucleotide concentrations are displayed in μg μl⁻¹. Note that for srg2 measurement after injection of 9E−7 μg μl⁻¹ 3′2′-cGAMP, there was one outlier replicate with a value of 0.5977 (data not shown, included in mean analysis). e–k, As in Fig. 4a, RNA expression analysis of Sting-regulated genes (srgs) 24 h after injection with synthetic 3′2′-cGAMP or 3′3′-c-di-GMP. RNA levels are shown as fold induction compared with buffer control in WT flies. dSTING^Mut = RXN mutant; Relish^Mut = Relish^E20 mutant, as previously characterized^,. All data in a–k represent the mean ± s.e.m. of n = 3 independent experiments and each point represents a pool of 6 flies. P value ns (>0.05) unless otherwise noted: ^****P < 0.0001 (e); ^***P = 0.0006, ^*P = 0.0404 (f); ^****P < 0.0001, ^***P = 0.0002 (g); ^****P < 0.0001 (h); ^**P = 0.0015, ^*P = 0.0117 (i); ^***P = 0.0002, ^**P = 0.0076 (j); ^***P = 0.0009 (k). Source data

Extended Data Fig. 10. 3′2′-cGAMP functions as a potent antiviral ligand.

a, Analysis of the effect of 3′2′-cGAMP on Drosophila C virus (DCV) viral RNA load in flies. dSTING WT and mutant flies were co-injected with DCV and 3′2′-cGAMP or buffer control. Viral RNA levels were measured at each time as indicated relative to the control gene RpL32. DCV is a picornavirus-like (+)ssRNA virus in the family Dicistroviridae. ^**P = 0.0051, ^*P = 0.0388. b, Analysis of the effect of 3′2′-cGAMP on vesicular stomatitis virus (VSV) viral RNA load in flies. dSTING WT and mutant flies were co-injected with VSV and 3′2′-cGAMP or buffer control as in a. Viral RNA levels were measured 4 days post-infection (dpi) relative to the control gene RpL32. VSV is a (-)ssRNA virus in the Rhabdoviridae family. ^*P = 0.0185. c, Analysis of DCV viral RNA load in flies injected with increasing doses of 3′2′-cGAMP, 2′3′-cGAMP or buffer control (as in a). Viral RNA levels were measured 2 dpi relative to the control gene RpL32. For 2′3′-cGAMP injection: 9E−1 ^*P = 0.0192. For 3′2′-cGAMP injection: 9E−3 ^*P = 0.0212, 9E−2 ^**P = 0.0075, 9E−1 ^**P = 0.0070. d, Survival curves after DCV infection showing the effect of injection with dose titration of 3′2′-cGAMP or 2′3′-cGAMP compared with buffer control. Both cGAMP isomers significantly delay mortality in a dose-dependent manner; 3′2′-cGAMP provides greater protection in comparison to 2′3′-cGAMP. For 2′3′-cGAMP injection: 9E−3 ^**P = 0.0047, 9E−2 ^**P = 0.0031, 9E−1 ^***P = 0.0002. For 3′2′-cGAMP injection: 9E−4 ^*P = 0.0344, 9E−3 ^***P = 0.0005, 9E−2 ^****P < 0.0001, 9E−1 ^****P < 0.0001. All data in a–d represent the mean ± s.e.m. of n = 3 independent experiments and each point represents a pool of 6 flies (a, b) or 10 flies (c, d). P value is ns unless otherwise noted; ns P > 0.05. Source data