Activation of cyclic GMP-AMP synthase by self-DNAcauses autoimmune diseasesDaxing Gaoa,1, Tuo Lia,1, Xiao-Dong Lia, Xiang Chena,b, Quan-Zhen Lic, Mary Wight-Carterd, and Zhijian J. Chena,b,2

aDepartment of Molecular Biology, University of Texas Southwestern Medical Center, Dallas, TX 75390-9148; bHoward Hughes Medical Institute, Universityof Texas Southwestern Medical Center, Dallas, TX 75390-9148; cDepartment of Immunology, University of Texas Southwestern Medical Center, Dallas, TX75390-9148; and dAnimal Resource Center, University of Texas Southwestern Medical Center, Dallas, TX 75390-9148

Contributed by Zhijian J. Chen, August 19, 2015 (sent for review August 12, 2015; reviewed by Jean-Laurent Casanova, Jae U. Jung, and Jenny P.-Y. Ting)

TREX1 is an exonuclease that digests DNA in the cytoplasm. Loss-of-function mutations of TREX1 are linked to Aicardi–GoutieresSyndrome (AGS) and systemic lupus erythematosus (SLE) in humans.Trex1−/− mice exhibit autoimmune and inflammatory phenotypesthat are associated with elevated expression of interferon (IFN)-in-duced genes (ISGs). Cyclic GMP-AMP (cGAMP) synthase (cGAS)is a cytosolic DNA sensor that activates the IFN pathway. Uponbinding to DNA, cGAS is activated to catalyze the synthesis of cGAMP,which functions as a second messenger that binds and activates theadaptor protein STING to induce IFNs and other cytokines. Here weshow that genetic ablation of cGas in Trex1−/− mice eliminated alldetectable pathological and molecular phenotypes, including ISG in-duction, autoantibody production, aberrant T-cell activation, and le-thality. Even deletion of just one allele of cGas largely rescued thephenotypes of Trex1−/− mice. Similarly, deletion of cGas in mice lack-ing DNaseII, a lysosomal enzyme that digests DNA, rescued the lethalautoimmune phenotypes of the DNaseII−/− mice. Through quantita-tive mass spectrometry, we found that cGAMP accumulated in mousetissues deficient in Trex1 or DNaseII and that this accumulation wasdependent on cGAS. These results demonstrate that cGAS activationcauses the autoimmune diseases in Trex1−/− and DNaseII−/− mice andsuggest that inhibition of cGASmay lead to prevention and treatmentof some human autoimmune diseases caused by self-DNA.

cGAS | cGAMP | autoimmune disease | Trex1 | DNaseII

Recognition and elimination of invading genetic materials is afundamental mechanism of host defense. In vertebrate ani-

mals, the immune system deploys sensors of DNA and RNA todetect microbial infections (1–3). In addition to a subset of Toll-like receptors that detect microbial nucleic acids in the lumen ofendosomes, cytosolic nucleic acid sensors also play crucial rolesin detecting pathogens, especially those that have successfullybreached the membrane barriers and replicated in the interior ofa cell. The cytosolic nucleic acid sensors include cyclic GMP-AMP(cGAMP) synthase (cGAS) and retinoic acid inducible gene I(RIG-I)-like receptors, which detect DNA and RNA, respectively,to induce type-I interferons (IFNs) and other inflammatory cyto-kines (4–6). cGAS binds to double-stranded DNA (dsDNA) ina sequence-independent manner (2, 7, 8). This binding causes aconformational change in the active site of cGAS, which then usesATP and GTP as the substrates to synthesize cGAMP that containsmixed 2′–5′ and 3′–5′ phosphodiester bonds (9–15). cGAMP thenbinds to and activates the endoplasmic reticulum membrane proteinSTING (14, 16–18). STING in turn activates the protein kinasesIKK and TBK1, which activate the transcription factors NF-κB andIRF3, respectively. NF-κB and IRF3 enter the nucleus and functiontogether to induce IFNs and cytokines. RIG-I and its homologMDA5 detect viral RNA in the cytoplasm and induce IFNs througha similar pathway, except that the essential adaptor protein func-tioning downstream of RIG-I and its homolog MDA5 is the mito-chondrial membrane protein MAVS, not STING (2, 19).Although detection of microbial nucleic acids provides a ver-

satile and highly effective mechanism for the immune system todetect infections, inadvertent reactions to self nucleic acids pose

a risk of triggering autoimmune and autoinflammatory diseases(20). In the case of RIG-I, the problem of avoiding activationby cytoplasmic self-RNA is solved by the ability of RIG-I to detectspecifically viral RNA that contains 5′-triphosphate and -diphosphate(21–23). Cellular RNA contains modifications such as the 5′ capin mRNA, which usually has 2′-O-methylation at the N1 positionthat prevents its recognition by RIG-I (24). For cGAS, which canbe activated by dsDNA irrespective of its sequence or origin,avoidance of triggering autoimmunity is largely achieved by confiningDNA in the nucleus, mitochondria, lysosome, and other membra-nous compartments, and by degrading DNA in the cytoplasm.TREX1 is an exonuclease that degrades DNA in the cytoplasm

(25–28). Trex1 deficiency in humans has been linked to severalautoimmune and inflammatory diseases, including Aicardi–GoutieresSyndrome (AGS), systemic lupus erythematosus (SLE), familialchilblain lupus, and retinal vasculopathy with cerebral leukodystrophy(29). A common feature of these diseases is the elevated expressionof IFN-stimulated genes (ISGs), suggesting that a defect in clearingcytosolic DNA leads to the activation of the IFN pathway. Trex1-deficient mice exhibit inflammatory diseases and prematuredeath accompanied by elevated ISG expression. Despite somecommon autoimmune phenotypes, Trex1−/− mice manifest myo-carditis, whereas human AGS patients suffer severe encephalop-athy. The autoimmune and myocarditis phenotypes in Trex1−/−

Significance

The immune system detects microbial DNA in the cytosol ofinfected cells and mounts effective antimicrobial responses,including the production of type-I interferons. However, whenself-DNA enters or accumulates in the cytosol, it can causeautoimmune diseases. Mutations of the exonuclease Trex1 inhumans have been linked to autoimmune diseases includingAicardi–Goutieres Syndrome (AGS) and systemic lupus ery-thematosus (SLE). In mice, genetic deletion of Trex1 or the ly-sosomal nuclease DNaseII leads to lethal autoimmune diseases.Here we show that cyclic GMP-AMP synthase (cGAS) activationby self-DNA is responsible for the lethal autoimmune diseasesin these models. These results provide the proof-of-conceptthat inhibition of cGAS may be an effective therapy for someautoimmune diseases such as AGS and SLE.

Author contributions: D.G., T.L., and Z.J.C. designed research; D.G., T.L., X.-D.L., andX.C. performed research; D.G., T.L., X.-D.L., X.C., Q.-Z.L., M.W.-C., and Z.J.C. analyzeddata; D.G. and Z.J.C. wrote the paper; Q.-Z.L. performed autoantigen array analysis;and M.W.-C. performed histology and pathology analyses of mouse tissues.

Reviewers: J.-L.C., The Rockefeller University; J.U.J., University of Southern California; andJ.P.-Y.T., University of North Carolina at Chapel Hill.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.

See Commentary on page 12903.1D.G. and T.L. contributed equally to this work.2To whom correspondence should be addressed. Email: Zhijian.Chen@UTSouthwestern.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1516465112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1516465112 PNAS | Published online September 14, 2015 | E5699–E5705

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Dec

embe

r 14

, 202

0

mice are rescued by deleting Sting, supporting an importantrole of the cytosolic DNA-sensing pathway in causing the in-flammatory diseases in the absence of Trex1 (26, 27, 30). In-terestingly, patients with gain-of-function mutations in STINGalso develop lupus-like syndromes; however, these patients alsoexhibit unique features of vascular and pulmonary inflamma-tion (31, 32).

Another major cellular nuclease is DNaseII, which is localizedin the lysosome and is largely responsible for the clearance ofDNA of dead cells and expelled nuclei that are engulfed bymacrophages (33–35). So far, no known DNaseII deficiency inhumans has been reported. DNaseII−/− mice die in utero, and theembryos become severely anemic by embryonic day 17.5 (E17.5),owing to the inability of macrophages to digest nuclear DNAexpelled from erythroid precursor cells. The undigested DNAinduces excessive production of IFNs, which kill the embryos.Deletion of the IFN receptor IFNAR1 rescued the embryoniclethality of DNaseII−/− mice, but the double-knockout mice de-veloped polyarthritis because of increased production of in-flammatory cytokines such as TNF-α and IL6 (36). The increasedproduction of these cytokines as well as IFNs is dependent onSTING, because deletion of Sting in DNaseII−/− mice not onlyrescued the embryonic lethality but also prevented the devel-opment of polyarthritis (37).A key question that remains is the identity of the DNA sensor

that triggers the autoimmune diseases caused by defectiveclearance of self-DNA. Although cGAS is the dominant DNAsensor responsible for activating the IFN pathways in response toinfections by a variety of pathogens, including DNA viruses,retroviruses, and some bacteria such as Mycobacterium tubercu-losis (13, 38–43), the role of cGAS in autoimmune diseases hasnot been extensively investigated. Here we show that deletion ofcGas in Trex1−/− and DNaseII−/− mice rescued the lethal auto-immune phenotypes of these mice and prevented the accumu-lation of cGAMP and the expression of ISGs in the mutantanimals. These results demonstrate that cGAS activation causesthe autoimmune diseases in Trex1−/− and DNaseII−/− mice.

Fig. 1. cGAS deletion rescues the lethality of Trex1−/− mice. Survival curves ofTrex1−/−, Trex1−/−cGas+/−, and Trex1−/−cGas−/−mice are shown. All mice were on aC57BL/6 background. Statistical analysis was performed with a Mantel–Cox test.

Fig. 2. cGAS mediates multiorgan inflammation in Trex1−/− mice. (A and C) Representative H&E-stained heart (A) and skeletal muscle (C) sections from12-wk-old Trex1−/−, Trex1−/−cGas+/−, and Trex1−/−cGas−/− mice. Blue-stained cells indicate leukocytes that infiltrate the heart. (B and D) Blinded analysis of theindicated tissues of 12-wk-old Trex1−/−, Trex1−/−cGas+/−, and Trex1−/−cGas−/− mice as well as Trex1+/− mice. Histological scores were calculated as described inMaterials and Methods. Statistical analysis was performed with a two-tailed, unpaired Student’s t test. ***P < 0.001.

E5700 | www.pnas.org/cgi/doi/10.1073/pnas.1516465112 Gao et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 14

, 202

0

ResultscGAS Is Responsible for Inflammatory Diseases and Death of Trex1−/−

Mice. Trex1−/− deficient mice die within a few months after birthdue to severe inflammation in multiple organs, especially in theheart (Fig. 1; refs. 26 and 27). To determine whether cGAS acti-vation—presumably by self-DNA accumulated in the cytosol ofTrex1−/− cells—is responsible for the inflammatory diseases andreduced survival of Trex1−/− mice, we generated Trex1−/− cGas−/−

and Trex1−/−cGas+/− mice. Strikingly, all Trex1−/−cGas−/− mice thatwere monitored (89/89) survived >200 d after birth, whereas ∼39%(65/167) of Trex1−/−mice survived within the same period (Fig. 1A).Even removal of just one allele of cGas in Trex1−/− mice allowed∼93% (91/98) of the mice to survive (Fig. 1). Hematoxylin and eosin(H&E) staining of the heart, skeletal muscle, skin, and kidney bi-opsies showed that the inflammatory pathology observed in thesetissues of Trex1−/−mice was eliminated in Trex1−/−cGas−/− mice andmarkedly reduced in Trex1−/−cGas+/−mice (Fig. 2 and Fig. S1 A–E).Moreover, splenomegaly found in Trex1−/− mice was not ob-served in Trex1−/−cGas−/− mice (Fig. S1 F–H). These results

demonstrate that cGAS is responsible for the inflammation andlethality of Trex1−/− mice.

cGAS Deletion Abolishes the Expression of ISGs and InflammatoryCytokines in Trex1−/− Mice. Gene-expression analysis by quantita-tive RT-PCR (qRT-PCR) of RNA from the hearts of 12-wk-oldTrex1−/− and Trex1+/− mice showed that Trex1−/− mice had elevatedlevels of ISGs, including IFIT3, CXCL10, and IRF7, as well as theinflammatory cytokines TNF-α and IL-12 (Fig. S2A and Table S1).The expression of these ISGs and cytokines was largely abolished byremoving one or both alleles of cGas in Trex1−/− mice (Fig. 3 A–Eand Fig. S2A). Similar results were obtained by using kidneys andperipheral blood monocytes (PBMCs; Fig. S2 B and C) or by mea-surements of IL-12 protein in the mouse sera (Fig. S2D). Elevatedlevels of ISGs in bone marrow dendritic cells (BMDCs) (Fig. S3A),macrophages (Fig. S3B), and embryonic fibroblasts (Fig. S3C) fromTrex1−/− mice were largely blunted by the deletion of cGas.

Trex1−/− Mice Overproduce cGAMP in a cGAS-Dependent Manner.Quantitative mass spectrometry (MS) was used to measure the

Fig. 3. cGAS is essential for the expression of ISGs and inflammatory cytokines as well as cGAMP overproduction in Trex1−/−mice. (A–E) qRT-PCR analysis of indicated ISGs(A–C), TNF-α (D), and IL12p40 (E) in hearts from 12-wk-old mice of indicated genotypes. Fold changes are relative to Trex1−/−cGas−/−mice. Error bars represent SD. *P < 0.05;**P< 0.01; ***P< 0.001. (F andG) cGAMP levels inmouse hearts of indicated genotypeswere quantified by LC-MS.Upper shows zoomed chromatograms displaying relativeintensities of endogenous cGAMP (red) and internal standard (green). Lower shows the calculated amounts of cGAMP (fmol) in each sample. Error bars represent SEM.

Gao et al. PNAS | Published online September 14, 2015 | E5701

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Dec

embe

r 14

, 202

0

absolute amounts of cGAMP in mouse hearts. This measurementwas made possible by spiking an internal standard of 13C10

15N5-labeled cGAMP in the heart extracts (Materials and Methods). Thismeasurement revealed elevated levels of cGAMP in Trex1−/− mice,which were markedly reduced in Trex1−/−cGas+/− mice and abol-ished in Trex1−/−cGas−/− mice (Fig. 3F). Interestingly, the levels ofcGAMP in Trex1−/−Sting−/− mice were approximately eightfoldhigher than those in Trex1−/−mice (Fig. 3G), suggesting that STINGfacilitates the clearance of cGAMP through an unknown mecha-nism. Because Trex1−/−Sting−/− mice do not display any apparentinflammation or other abnormal phenotypes (30), these resultsimply that high levels of cGAMP do not appear to cause adverseeffects in mice in the absence of STING.

cGAS Activation Causes Autoantibody Production and T-Cell Activation inTrex1−/− Mice. Trex1−/− mice, both 12-wk- and 5-mo-old, developedautoantibodies against DNA, nuclear antigens (ANAs), and otherself-antigens, including histone H3, M2, and the U1 subunit of smallnuclear ribonucleoprotein (U1-snRNP; Fig. 4 A–C and Fig. S4). Allof these autoantibodies were greatly reduced by deleting one orboth alleles of cGas. The CD4 and CD8 T cells from Trex1−/− micehad higher expression of T-bet than T cells from Trex1−/−cGas−/−

mice and produced more IFN-γ in response to stimulation withphorbol myristate acetate (PMA) and ionomycin (Fig. 4 D andE; Fig. S5). Hearts from Trex1−/− mice, but not Trex1−/−cGas−/−

mice, had elevated expression of IFN-γ and T-bet (also knownas Tbx21) RNA, indicating spontaneous infiltration of acti-vated immune cells to the heart, which was blocked by de-leting cGas (Fig. S6). Autoreactive T cells in Trex1−/− mice wereevident by enhanced surface expression of CD69 for CD4 and CD8T cells and Ly6c for CD8 T cells (Fig. S7 C–E). Deletion of cGasin Trex1−/− mice markedly reduced the numbers of these activated

T cells (Fig. S8 C–E). Elevated levels of memory CD4 and CD8T cells (CD44hi/CD62Llo) in Trex1−/− mice were also reduced inTrex1−/−cGas−/−mice (Fig. 4 F andG; Fig. S7 A and B; Fig. S8 A andB). These results further demonstrate that cGAS activation causesthe autoimmune phenotypes in Trex1−/− mice.

cGAS Activation Causes Lethal Autoimmunity in DNaseII−/− Mice. To in-vestigate the role of cGAS in another autoimmune disease model, wedeleted cGas in DNaseII−/− mice. DNaseII−/−cGas+/− mice were borndead, similar to DNaseII−/− mice (34). In contrast, DNaseII−/−cGas−/−

mice appeared to be developmentally normal, similar to WT andDNaseII+/−cGas+/−mice.DNasII−/−cGas+/− embryos were also smallerand pale compared with DNaseII−/−cGas−/− and DNaseII+/−cGas+/−

embryos (Fig. S9A). Thus, deletion of both cGas alleles is nec-essary to rescue the lethality of DNaseII−/− mice. As shownpreviously, DNaseII−/−Ifnar1−/− mice developed polyarthritis andproduced antibodies against dsDNA (Fig. 5; ref. 36). In contrast,DNaseII−/−cGas−/− mice did not have apparent signs of arthritis, andthe anti-DNA antibody was greatly reduced (Fig. 5 A–C). These re-sults, together with the previous data showing that deletion of Stingrescued the phenotypes of DNaseII−/− mice, strongly suggest thatcGAS is the dominant DNA sensor that activates STING to causelethal autoimmunity in DNaseII−/− mice.qRT-PCR analyses of RNA isolated from the fetal liver,

embryonic limb, and fibroblasts at E15.5 showed that the ex-pression levels of IFIT3, CXCL10, and ISG15 were muchhigher in DNaseII−/−cGas+/− mice than in DNaseII+/−cGas+/− orDNaseII−/−cGas−/− mice (Fig. 6A; Fig. S9 B and C). Measurementof cGAMP in fetal livers showed that DNaseII−/−cGas+/− miceproduced higher amounts of cGAMP than in DNaseII+/−cGas+/−

mice. cGAMP was not detected in DNaseII−/−cGas−/− mice, but wasstrongly elevated in DNaseII−/−Sting−/− mice (Fig. 6B), consistent

Fig. 4. cGAS deletion mitigates autoimmunity in Trex1−/− mice. (A–C) Detection of autoantibodies, including ssDNA IgG (A), ssDNA IgG2b subtype (B), andANA IgG (C) in the sera of 12-wk-old mice of indicated genotypes. (D and E) Flow-cytometric analysis of intracellular IFN-γ in response to PMA plus ionomycinin splenic CD4+ T cells (D) and CD8+ T cells (E) from 12-wk-old Trex1−/− and Trex1−/−cGas−/− mice. (F and G) Flow-cytometric analysis of CD44hiCD62lo cells insplenic CD4+ T cells (F) and CD8+ T cells (G) from 12-wk-old Trex1−/− and Trex1−/−cGas−/− mice. Error bars represent SD. **P < 0.01; ***P < 0.001.

E5702 | www.pnas.org/cgi/doi/10.1073/pnas.1516465112 Gao et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 14

, 202

0

with higher levels of cGAMP in Trex1−/−Sting−/− mice (Fig. 3G).These results further suggest that cGAS is indispensable for theproduction of cGAMP in DNase-deficient cells, whereas STINGappears to play a role in facilitating the clearance of cGAMP.

DiscussionA number of DNA sensors have been proposed to induce type-IIFNs. Besides cGAS, several other proteins, including DAI,IFI16, DDX41, and DNA-PK, were reported to mediate IFN

Fig. 5. cGAS activation causes polyarthritis in DNaseII−/−Ifnar1−/− mice. (A and B) Representative foot pad pictures (A) and arthritis scores (B) of 7-mo-old WT,DNaseII−/−Ifnar1−/−, and DNaseII−/−cGAS−/− mice. Arthritis scores were calculated as described in Materials and Methods. (C) Detection of serum autoanti-bodies against dsDNA. Statistical analysis was performed with a two-tailed, unpaired Student’s t test. Error bars represent SD. *P < 0.05; ***P < 0.001.

Fig. 6. cGAS is essential for ISG up-regulation and cGAMP production in fetal livers of DNaseII−/− mice. (A) qRT-PCR analysis of indicated ISGs in fetal liverfrom E15.5 mouse embryos of indicated genotypes (n = 3). (B) cGAMP levels in mouse fetal livers of indicated genotypes were quantified by LC-MS. Uppershows zoomed chromatograms displaying relative intensities of endogenous cGAMP (red) and internal standard (green). Lower shows the calculated amountsof cGAMP in each sample. Error bars represent SD (A) or SEM (B). *P < 0.05; **P < 0.01.

Gao et al. PNAS | Published online September 14, 2015 | E5703

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Dec

embe

r 14

, 202

0

induction through STING (44). However, so far, only cGAS hasbeen demonstrated by genetic experiments to function as anonredundant and essential cytosolic DNA sensor that activatesthe type-I IFN pathway in vivo (13, 41). In particular, cGas−/−

mice and cells derived from these mice failed to produce IFNs inresponse to infection by several DNA viruses. TLR9 is a membraneprotein localized on the endosomal membrane, with its ligand-binding domain facing the lumen of endosome. TLR9 is activatedby certain single-stranded DNA oligos that contain CpG-rich se-quences. The binding of CpG DNA to TLR9 in plasmacytoiddendritic cells leads to production of copious amounts of IFN-α.However, induction of IFNs by mammalian DNA in myeloid den-dritic cells was shown to be independent of TLR9 (45), and deletionof TLR9 in DNaseII−/−mice failed to rescue the lethal autoimmunediseases of these mice (46). Thus, another DNA sensor is likelyresponsible for triggering the inflammatory cascades when cellularDNA clearance is impeded.Initial evidence that cGAS is responsible for activating the type-I

IFN pathway in Trex1−/− cells was provided by findings that retro-viruses such as HIV-1 robustly induced IFNs and ISGs in Trex1-deficient cells and that such induction was abolished by TALEN-mediated knockout of cGas (39). Subsequent studies furthershowed that deletion of cGas by the CRISPR/Cas9 technology incell lines lacking Trex1 or DNaseII abrogated the induction ofseveral ISGs (47, 48). In this study, we demonstrated that Trex1−/−

cGas−/− mice were viable and free of any observable signs of dis-eases that would otherwise cause mortality in Trex1−/− mice. Wefurther showed that cGas deletion rescued the embryonic lethalityof DNaseII−/− mice and that the DNaseII−/−cGas−/− mice did notexhibit polyarthritis or other autoimmune phenotypes found inDNaseII−/−Ifnar1−/− mice. Taken together, these results providethe in vivo proof that cGAS activation is responsible for auto-immune diseases caused by defective clearance of self DNA.Although the rescue of DNaseII−/− mice requires the deletion of

both alleles of cGas, deletion of just one allele of cGas is sufficientto largely eliminate the inflammatory disease and rescue the le-thality of Trex1−/− mice. It is not clear what causes the difference inthese two models, but one possibility is that DNaseII−/− cells maycontain more DNA in the cytoplasm than Trex1−/− cells, therebycausing strong inflammatory responses with just one allele of cGas.Alternatively, the cell types responsible for triggering the in-flammatory responses may be different between DNaseII−/− andTrex1−/− mice; hence, the threshold of cGAS expression required tocause autoimmune diseases may vary in different cells. Future re-search should be directed at determining the cell types responsiblefor causing the autoimmune diseases in DNaseII−/− and Trex1−/−

mice. Another important direction is to investigate the origin andnature of self-DNA that accumulates in these DNase-deficient mice.Through quantitative MS, we have provided direct evidence that

tissues from Trex1−/− andDNaseII−/−mice contain elevated levels ofcGAMP, which explains why the IFN pathway is activated in thesemice. Deletion of cGas ablated cGAMP production in these mice,formally demonstrating that cGAS is solely responsible for synthe-sizing cGAMP in response to accumulation of self-DNA in thecytosol. Interestingly, deletion of Sting further elevated the cGAMPlevels in Trex1−/− and DNaseII−/− mice, suggesting that STING fa-cilitates the clearance of cGAMP. The mechanism by which STINGfacilitates cGAMP clearance requires further investigation.The demonstration that cGAS activation is responsible for

the severe autoimmune diseases and lethality in Trex1−/− andDNaseII−/− mice strongly suggests that inhibition of cGAS couldprovide therapeutic benefits to patients with AGS and otherautoimmune diseases such as SLE. cGAS is an enzyme that islikely amenable to inhibition by small-molecule drugs, and therecent determination of the high-resolution structures of cGASin its apo- and DNA-bound forms should facilitate the devel-opment of these inhibitors (9–12, 14, 15, 49, 50). The patientswho are most likely to benefit from cGAS inhibitors are those

with elevated cGAMP production. Our finding that cGAMPlevels are elevated in Trex1−/− and DNaseII−/− mice provides aproof of concept for extending such measurements to humanpatients with a variety of autoimmune diseases, especially thosewith interferonopathies (exhibiting IFN signatures).

Materials and MethodsMice. Trex1+/− mice were from Deborah Barnes (Cancer Research U.K., London)and Nan Yan (University of Texas Southwestern Medical Center), and DNaseII+/− mice were from Shigekazu Nagata (Kyoto University, Kyoto). cGas−/− micewere generated in our laboratory as described (13). All mice used in this studywere on C57BL/6 background. The mice were bred and maintained underspecific pathogen-free conditions in the animal facility of the University ofTexas Southwestern Medical Center at Dallas according to experimental pro-tocols approved by the Institutional Animal Care and Use Committee.

Pathology. Tissues were fixed in 4% (wt/vol) paraformaldehyde, paraffin-embedded, cut into 5-μm sections, and stained with H&E. Heart tissues werestained with Picro-Sirius red, and kidney samples were stained with periodicacid–Schiff. Inflammation and fibrosis were evaluated based on degree ofseverity. The sum of individual scores was used to obtain a total tissue his-tological score. Specific information about the scoring criteria for each tissuewas described (30). Clinical assessment of foot pads for arthritis was per-formed as described (36).

cGAMP Extraction and Quantification. Fresh organs were snap-frozen in liquidnitrogen, minced immediately with dissecting scissors in cold 80% (vol/vol)methanol with 2% (vol/vol) acetic acid (HAc), and stored at −80 °C beforefurther processing. On the day of analysis, the frozen samples were thawedon ice, and 13C10

15N5-labeled cGAMP (+15 atomic mass units) internal stan-dard was supplemented. Samples were then homogenized with a tissuetearor (Biospec Products) for 30 s and cleared by centrifugation (10,000 × g)for 10 min. The pellets were further extracted in 20% (vol/vol) methanol and2% HAc for two more rounds, and all of the cleared extracts were com-bined. From these extracts, cGAMP was enriched by solid-phase extraction(SPE) using HyperSep Aminopropyl SPE Columns (Thermo Scientific). Briefly,the columns were first activated by methanol and washed twice with 2%HAc; after drawing through the extracts, columns were washed twice with2% HAc and once with 80% methanol and, finally, eluted with 4% (vol/vol)ammonium hydroxide in 80% methanol. The eluents were spin-vacuumed todryness, reconstituted in liquid chromatography (LC)/MS-grade water, clearedby centrifugation, and transferred to autosampler vials for MS analyses.

MS analyses were performed as described (38). Briefly, the SPE eluentswere separated on an Xbridge Amide column (3.5 μm, 3.5 mm ID × 100 mm L;Waters) on a Dionex Ultimate 3000 Rapid Separation Liquid Chromatographysystem (Thermo Scientific). Mobile phase A was 20 mM ammonium bi-carbonate with 20 mM ammonium hydroxide in water, and mobile phase Bwas acetonitrile. The separation ran at a flow rate of 400 μL/min for the first14.5 min and 800 μL/min for the remaining 8.5 min, through the followinggradient: 0 min, 85% B; 3 min, 85% B; 10 min, 2% B; 14 min, 2% B; 14.5 min,85% B; and 23 min, 85% B.

The LC eluent was ionized by an Ion Max NG heated electrospray source,with a spray voltage of +3,750 V; an ion transfer tube temperature of 342 °C;a vaporizing temperature of 292 °C; and the sheath, auxiliary, and sweep gasat 45, 17, and 1 arbitrary units, respectively. The spray was analyzed onlineon a TSQ Quantiva triple quadruple mass spectrometer (Thermo Scientific),which performed continuous multiple reaction monitoring scans with adwell time of 50 ms, Q1 and Q3 resolutions of 0.7 FWHM, and the collision-induced dissociation gas of 1.5 units. cGAMP and the internal standard weremonitored in the positive mode with four transitions, respectively (cGAMP:675–136, 675–152, 675–476, and 675–524; and the internal standard: 691–146, 691–152, 691–491, and 691–539). Raw MS Data were converted to themzXML format with ReAdW and read into MATLAB for noise reductionand data processing. Absolute quantities of endogenous cGAMP were cal-culated with the light:heavy ratios and the molar quantity of supplementedinternal standard.

Statistics. Statistical analysis of mouse survival was performed by using theMantel–Cox test. Other statistical analyses were performed with a two-tailed, unpaired Student’s t test.

Note. While our manuscript was in submission, Gray et al. independentlyshowed that cGas ablation rescued the autoimmune phenotypes of Trex1−/−

mice (51).

E5704 | www.pnas.org/cgi/doi/10.1073/pnas.1516465112 Gao et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 14

, 202

0

ACKNOWLEDGMENTS. We thank Drs. Deborah Barnes and Nan Yan forproviding the Trex1+/− mice; Dr. Shigekazu Nagata for the DNaseII+/− mice; andMr. John Shelton (University of Texas Southwestern) for performing the histologyanalyses of mouse tissues. This work was supported by National Institutes of

Health Grant AI-93967, the Lupus Research Institute Distinguished InnovatorAward, Welch Foundation Grant I-1389, and Cancer Research Prevention Instituteof Texas Grant RP120718. T.L. was supported by a Cancer Research InstitutePostdoctoral Fellowship. Z.J.C is a Howard Hughes Medical Institute Investigator.

1. Pandey S, Kawai T, Akira S (2015) Microbial sensing by Toll-like receptors and in-tracellular nucleic acid sensors. Cold Spring Harb Perspect Biol 7(1):a016246.

2. Wu J, Chen ZJ (2014) Innate immune sensing and signaling of cytosolic nucleic acids.Annu Rev Immunol 32:461–488.

3. Goubau D, Deddouche S, Reis e Sousa C (2013) Cytosolic sensing of viruses. Immunity38(5):855–869.

4. Sun L, Wu J, Du F, Chen X, Chen ZJ (2013) Cyclic GMP-AMP synthase is a cytosolic DNAsensor that activates the type I interferon pathway. Science 339(6121):786–791.

5. Yoneyama M, et al. (2004) The RNA helicase RIG-I has an essential function in double-stranded RNA-induced innate antiviral responses. Nat Immunol 5(7):730–737.

6. Wu J, et al. (2013) Cyclic GMP-AMP is an endogenous second messenger in innateimmune signaling by cytosolic DNA. Science 339(6121):826–830.

7. Cai X, Chiu YH, Chen ZJ (2014) The cGAS-cGAMP-STING pathway of cytosolic DNAsensing and signaling. Mol Cell 54(2):289–296.

8. Xiao TS, Fitzgerald KA (2013) The cGAS-STING pathway for DNA sensing. Mol Cell51(2):135–139.

9. Ablasser A, et al. (2013) cGAS produces a 2′-5′-linked cyclic dinucleotide secondmessenger that activates STING. Nature 498(7454):380–384.

10. Civril F, et al. (2013) Structural mechanism of cytosolic DNA sensing by cGAS. Nature498(7454):332–337.

11. Diner EJ, et al. (2013) The innate immune DNA sensor cGAS produces a noncanonicalcyclic dinucleotide that activates human STING. Cell Reports 3(5):1355–1361.

12. Gao P, et al. (2013) Cyclic [G(2′,5′)pA(3′,5′)p] is the metazoan second messengerproduced by DNA-activated cyclic GMP-AMP synthase. Cell 153(5):1094–1107.

13. Li XD, et al. (2013) Pivotal roles of cGAS-cGAMP signaling in antiviral defense andimmune adjuvant effects. Science 341(6152):1390–1394.

14. Zhang X, et al. (2013) Cyclic GMP-AMP containing mixed phosphodiester linkages isan endogenous high-affinity ligand for STING. Mol Cell 51(2):226–235.

15. Zhang X, et al. (2014) The cytosolic DNA sensor cGAS forms an oligomeric complexwith DNA and undergoes switch-like conformational changes in the activation loop.Cell Reports 6(3):421–430.

16. Ishikawa H, Barber GN (2008) STING is an endoplasmic reticulum adaptor that facil-itates innate immune signalling. Nature 455(7213):674–678.

17. Zhong B, et al. (2008) The adaptor protein MITA links virus-sensing receptors to IRF3transcription factor activation. Immunity 29(4):538–550.

18. Jin L, et al. (2008) MPYS, a novel membrane tetraspanner, is associated with majorhistocompatibility complex class II and mediates transduction of apoptotic signals.Mol Cell Biol 28(16):5014–5026.

19. Yoneyama M, Onomoto K, Jogi M, Akaboshi T, Fujita T (2015) Viral RNA detection byRIG-I-like receptors. Curr Opin Immunol 32:48–53.

20. Crow YJ (2015) Type I interferonopathies: Mendelian type I interferon up-regulation.Curr Opin Immunol 32:7–12.

21. Goubau D, et al. (2014) Antiviral immunity via RIG-I-mediated recognition of RNAbearing 5′-diphosphates. Nature 514(7522):372–375.

22. Hornung V, et al. (2006) 5′-Triphosphate RNA is the ligand for RIG-I. Science314(5801):994–997.

23. Pichlmair A, et al. (2006) RIG-I-mediated antiviral responses to single-stranded RNAbearing 5′-phosphates. Science 314(5801):997–1001.

24. Schuberth-Wagner C, et al. (2015) A conserved histidine in the RNA sensor RIG-Icontrols immune tolerance to N1-2’O-methylated self RNA. Immunity 43(1):41–51.

25. Mazur DJ, Perrino FW (1999) Identification and expression of the TREX1 and TREX2cDNA sequences encoding mammalian 3′–>5′ exonucleases. J Biol Chem 274(28):19655–19660.

26. Yang YG, Lindahl T, Barnes DE (2007) Trex1 exonuclease degrades ssDNA to preventchronic checkpoint activation and autoimmune disease. Cell 131(5):873–886.

27. Morita M, et al. (2004) Gene-targeted mice lacking the Trex1 (DNase III) 3′–>5′ DNAexonuclease develop inflammatory myocarditis. Mol Cell Biol 24(15):6719–6727.

28. Crow YJ, et al. (2006) Mutations in the gene encoding the 3′-5′ DNA exonucleaseTREX1 cause Aicardi-Goutières syndrome at the AGS1 locus. Nat Genet 38(8):917–920.

29. Rice GI, Rodero MP, Crow YJ (2015) Human disease phenotypes associated withmutations in TREX1. J Clin Immunol 35(3):235–243.

30. Gall A, et al. (2012) Autoimmunity initiates in nonhematopoietic cells and progressesvia lymphocytes in an interferon-dependent autoimmune disease. Immunity 36(1):120–131.

31. Jeremiah N, et al. (2014) Inherited STING-activating mutation underlies a famil-ial inflammatory syndrome with lupus-like manifestations. J Clin Invest 124(12):5516–5520.

32. Liu Y, et al. (2014) Activated STING in a vascular and pulmonary syndrome. N EnglJ Med 371(6):507–518.

33. Chang HC, Liao TH (1990) Reassociation of deoxyribonuclease II with the lysosomalmembrane isolated from porcine spleen. Arch Biochem Biophys 280(2):320–324.

34. Kawane K, et al. (2001) Requirement of DNase II for definitive erythropoiesis in themouse fetal liver. Science 292(5521):1546–1549.

35. Lan YY, Londono D, Hacohen N (2012) Lysosomal DNaseII degrades nuclear DNA andprevents self DNA recognition. J Immunol 188(Meeting Abstract Suppl):159.17.

36. Kawane K, et al. (2006) Chronic polyarthritis caused by mammalian DNA that escapesfrom degradation in macrophages. Nature 443(7114):998–1002.

37. Ahn J, Gutman D, Saijo S, Barber GN (2012) STING manifests self DNA-dependentinflammatory disease. Proc Natl Acad Sci USA 109(47):19386–19391.

38. Collins AC, et al. (2015) Cyclic GMP-AMP synthase is an innate immune DNA sensor forMycobacterium tuberculosis. Cell Host Microbe 17(6):820–828.

39. Gao D, et al. (2013) Cyclic GMP-AMP synthase is an innate immune sensor of HIV andother retroviruses. Science 341(6148):903–906.

40. Lam E, Stein S, Falck-Pedersen E (2014) Adenovirus detection by the cGAS/STING/TBK1DNA sensing cascade. J Virol 88(2):974–981.

41. Schoggins JW, et al. (2014) Pan-viral specificity of IFN-induced genes reveals new rolesfor cGAS in innate immunity. Nature 505(7485):691–695.

42. Wassermann R, et al. (2015) Mycobacterium tuberculosis differentially activates cGAS-and inflammasome-dependent intracellular immune responses through ESX-1. CellHost Microbe 17(6):799–810.

43. Watson RO, et al. (2015) The cytosolic sensor cGAS detects mycobacterium tubercu-losis DNA to induce type I interferons and activate autophagy. Cell Host Microbe17(6):811–819.

44. Paludan SR, Bowie AG (2013) Immune sensing of DNA. Immunity 38(5):870–880.45. Martin DA, Elkon KB (2006) Intracellular mammalian DNA stimulates myeloid den-

dritic cells to produce type I interferons predominantly through a toll-like receptor9-independent pathway. Arthritis Rheum 54(3):951–962.

46. Okabe Y, Kawane K, Akira S, Taniguchi T, Nagata S (2005) Toll-like receptor-independent gene induction program activated by mammalian DNA escaped fromapoptotic DNA degradation. J Exp Med 202(10):1333–1339.

47. Ablasser A, et al. (2014) TREX1 deficiency triggers cell-autonomous immunity in acGAS-dependent manner. J Immunol 192(12):5993–5997.

48. Motani K, Ito S, Nagata S (2015) DNA-mediated cyclic GMP-AMP synthase-dependentand -independent regulation of innate immune responses. J Immunol 194(10):4914–4923.

49. Li X, et al. (2013) Cyclic GMP-AMP synthase is activated by double-stranded DNA-induced oligomerization. Immunity 39(6):1019–1031.

50. Kranzusch PJ, Lee AS, Berger JM, Doudna JA (2013) Structure of human cGAS reveals aconserved family of second-messenger enzymes in innate immunity. Cell Reports 3(5):1362–1368.

51. Gray EE, Treuting PM, Woodward JJ, Stetson DB (2015) Cutting edge: cGAS is requiredfor lethal autoimmune disease in the Trex1-deficient mouse model of Aicardi-Goutieres syndrome. J Immunol 195(5):1939–1943.

52. Li QZ, et al. (2005) Identification of autoantibody clusters that best predict lupusdisease activity using glomerular proteome arrays. J Clin Invest 115(12):3428–3439.

Gao et al. PNAS | Published online September 14, 2015 | E5705

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Dec

embe

r 14

, 202

0