Review

cGAS in action: Expanding roles in immunity and inflammation

See allHide authors and affiliations

Science  08 Mar 2019:
Vol. 363, Issue 6431, eaat8657
DOI: 10.1126/science.aat8657

cGAS sensing of DNA: A comprehensive look

The immune system uses the cGAS-STING (cyclic GMP-AMP synthase–stimulator of interferon genes) signaling pathway to detect the presence of intracellular DNA. This is beneficial because cytosolic DNA is often a sign of host damage or invasion by pathogens. However, inappropriate responses to self-DNA must be suppressed to prevent detrimental effects on the host. Ablasser and Chen review the latest advances uncovering how cGAS and STING control inflammatory responses and are themselves regulated. Special attention is paid to the role of cGAS in sterile inflammatory conditions as well as to the therapeutic potential of modulating this pathway.

Science, this issue p. eaat8657

Structured Abstract

BACKGROUND

The life of any organism depends on the ability of cells to accurately recognize and eliminate harmful microbes. To detect the immense repertoire of pathogenic entities, the mammalian innate immune system has evolved distinct sensing strategies, including a central one based on the recognition of DNA—the basic building block of “life” itself. Integral to this process is the intracellular enzyme cGAS, which upon binding to double-stranded DNA (dsDNA), initiates a tightly regulated signaling cascade involving the adapter STING to trigger a variety of inflammatory effector responses. Although this process was originally discovered as a crucial component of innate immune defense against pathogens, recent work has elucidated a role for cytosolic DNA recognition pathways beyond the “classical” realm of innate immunity. The realization of an important involvement of cGAS and STING in various biological contexts has broadened its implications for human health and disease—much more than initially anticipated.

ADVANCES

Early structural and functional studies on cyclic guanosine monophosphate–adenosine monophosphate (GMP–AMP) synthase (cGAS) have established its capability to interact with dsDNA in a sequence-independent manner. Although this indiscriminate sensing strategy ensures the detection of almost all pathogenic entities, it also enables immune responses to be elicited upon encountering self-DNA. Mechanistically, self-DNA–sensing phenomena can be provoked by diverse alterations of both the extracellular and intracellular milieu, such as perturbations of DNA compartmentalization or disturbances in endogenous DNA metabolism. Initial studies on the relevance of cGAS-dependent recognition of self-DNA have largely focused on bona fide immunological consequences, such as inherited autoimmune and autoinflammatory disorders. Indeed, mutations in genes affecting intracellular DNA turnover can cause rare monogenic autoinflammatory syndromes in which the aberrant stimulation of innate DNA sensing is unequivocally central in driving associated pathologies. In addition to these rather traditional immunological disease entities, the aberrant activation of innate DNA sensing has recently emerged as an underlying cause for a number of distinct biological phenomena. Studies have documented the benefits of innate self-DNA sensing through cGAS by facilitating the recognition of cellular damage and indirectly, the presence of pathogens. Likewise, cGAS and STING have proven to be a central element in both iatrogenic and naturally occurring antitumor immunity and in promoting cellular senescence. However, the inflammatory consequences of the cGAS-STING pathway can also become maladaptive through the potentiation of tissue destruction or through the initiation of more subtle forms of chronic inflammatory diseases.

OUTLOOK

The broad biological roles of intracellular DNA sensing create new opportunities for the exploration and therapeutic manipulation for the prevention and treatment of multiple human diseases. Initial successes with therapies targeting the immunostimulatory effects of the cGAS-STING pathway suggest a major clinical impact in areas of cancer immunotherapy and vaccine development. Furthermore, pharmacological interventions aimed at antagonizing cGAS or STING functions hold similar promise, not only in the context of classical autoinflammatory conditions, but also in the treatment of more complex diseases. We are optimistic that an improved understanding of the molecular basis of innate DNA sensing and signaling via cGAS and STING will aid the design of new therapeutic strategies to manipulate its outcomes in a safe and specific manner. At the same time, we envision that this evolutionarily conserved DNA-sensing system may participate in diverse biological processes that are just beginning to be explored.

Infectious and noninfectious functions of cGAS.

cGAS is a universal sensor of dsDNA. A well-established function of cGAS in infections is the recognition of foreign DNA linked to the orchestration of host defense programs. Through sensing of self-DNA, cGAS also participates in various noninfectious contexts, including antitumor immunity, cellular senescence, and inflammatory diseases.

Abstract

DNA is highly immunogenic. It represents a key pathogen-associated molecular pattern (PAMP) during infection. Host DNA can, however, also act as a danger-associated molecular pattern (DAMP) and elicit strong inflammatory responses. The cGAS-STING pathway has emerged as a major pathway that detects intracellular DNA. Here, we highlight recent advances on how cGAS and STING mediate inflammatory responses and how these are regulated, allowing cells to readily respond to infections and noxious agents while avoiding the inappropriate sensing of self-DNA. A particular focus is placed on the role of cGAS in the context of sterile inflammatory conditions. Manipulating cGAS or STING may open the door for new therapeutic strategies for the treatment of acute and chronic inflammation relevant to many human diseases.

The recognition of “foreign” DNA is one of the most fundamental mechanisms of host defense. Spanning across diverse forms of life, host cells have evolved various strategies to detect and to respond to incoming DNA from pathogens to maintain host–microbial homeostasis. In bacteria, restriction endonucleases and CRISPR protect the host from invasive plasmid DNA and bacteriophages. In mammals, innate DNA sensing involves intracellular signaling receptors that elicit powerful antimicrobial immune responses. The clinical potential of the immunostimulatory capacity of DNA was recognized and even medically exploited over a century ago. In the early 20th century, the surgeon J. von Mikulicz-Radecki injected nucleic acids to prevent “acute, fatal peritonitis” in humans. However, a molecular understanding of the key steps underlying innate DNA recognition and signaling has only very recently emerged. Today, we know that one of the major pathways that mediate the immune response to DNA is governed by the DNA-sensing enzyme cyclic guanosine monophosphate–adenosine monophosphate (cyclic GMP–AMP) synthase (cGAS) (1, 2). cGAS is activated upon binding to double-stranded DNA (dsDNA). Activated cGAS converts adenosine 5′-triphosphate (ATP) and guanosine 5′-triphosphate (GTP) into cyclic GMP–AMP (cGAMP), which functions as a secondary messenger that binds and activates STING (stimulator of interferon genes), an adapter protein on the endoplasmic reticulum (ER) membrane (3, 4). STING then triggers a signaling cascade, which leads to production of a battery of immune and inflammatory mediators, including type I and type III interferons. Both cGAS and STING are pivotal to withstanding infection by numerous pathogens. The identification of homologs of cGAS and/or STING in various, evolutionarily distant organisms suggests that this system represents an ancient recognition mode deployed by many species for host defense. Over the past few years, accumulating evidence indicates that the physiological and pathological relevance of cGAS and STING extends far beyond “traditional” antimicrobial immunity by affecting various noninfectious cellular stress response and stress adaptation programs. This broad biological role largely stems from the unique potential of cGAS to interact with dsDNA in a sequence-independent manner. Such a “universal” sensing mechanism violates one of the most fundamental rules of the classical pattern recognition dogma, which is based on pathogen-specific structural patterns instructing self versus nonself discrimination. Accordingly, to function as a well-balanced innate recognition system, the cGAS-STING pathway requires a number of additional layers of control to guard against erroneous activation of innate immune responses. In this review, we will discuss current knowledge and emerging principles that regulate the (self-) DNA-sensing capacity through the cGAS-STING pathway. We also highlight how insights into the molecular mechanisms and biological effects may be exploited in new therapeutic concepts. Owing to space limitations, we will not cover innate DNA-sensing mechanisms through the Toll-like receptor (TLR) pathway or through other intracellular DNA-sensing mechanisms, including HIN-domain binding proteins or RNA polymerase III, and refer the reader to other excellent recent reviews on these topics (57).

Molecular mechanisms of cGAS activation by dsDNA

The sequence-independent interaction with dsDNA constitutes the front end of the cGAS-governed signaling cascade. Insight into how binding of dsDNA elicits the activation of cGAS has been established by crystallographic and biochemical studies. A key functional element of cGAS is its C-terminal nucleotidyl transferase (NTase) domain, which forms a bilobate architecture that comprises a central catalytic domain and two separate positively charged surfaces (Fig. 1A). It is through these platforms that cGAS interacts with the sugar phosphate backbone of the DNA duplex, an interaction that is further fostered by a conserved zinc ribbon (812). Upon binding to dsDNA, cGAS assembles into a 2:2 cGAS-dsDNA complex, in which two cGAS molecules embrace two molecules of dsDNA (Fig. 1B) (11, 12). cGAS dimers can form “ladder-like” networks between two separate stretches of dsDNA or on one long curved dsDNA helix (13). These have been shown to cooperatively stabilize each individual cGAS-dsDNA complex. This latter observation provides one possible explanation as to why long dsDNA is a more potent activator of cGAS in cells. A recent study showed that human-specific substitutions specifically decrease the affinity of human cGAS toward short dsDNA, providing an alternative but not mutually exclusive explanation for the preference of human cGAS for long DNA (14). Once an interaction between cGAS and dsDNA has occurred, a structural switch rearranges the catalytic pocket such that cGAS carries out the cyclization of ATP and GTP to form cGAMP through a two-step process with an intermediate linear dinucleotide (9, 1517). The cGAMP molecule produced by cGAS contains one phosphodiester bond between the 2′-hydroxyl of GMP and 5′-phosphate of AMP and another between the 3′-hydroxyl of AMP and 5′-phosphate of GMP. This unique isomer, termed 2′3′-cGAMP, binds to STING with a dissociation constant (Kd) of ~4 nM. Notably, although dsRNA or single-stranded DNA (ssDNA) can also bind to cGAS, neither molecule can rearrange the catalytic pocket, which explains the specific activation of cGAS by dsDNA.

Fig. 1 Molecular mechanism of DNA sensing by cGAS.

(A) Schematic representation of the distinct steps of cGAS activation by dsDNA in two perpendicular views. Double-stranded DNA (dsDNA) binding proceeds through two separate DNA-binding sites, one of which is depicted in the cartoon. A ligand-induced conformational change alters the activation loop and, thereby, rearranges the active site. This step enables the binding of the cGAS substrates ATP and GTP, which is followed by a two-step catalysis mechanism leading to the production of cGAMP. One cGAS molecule is omitted for clarity. (B) Crystal structure of the cGAS-dsDNA complex composed of two protein molecules and two DNA molecules. Two perpendicular views are shown. [Adapted from (8, 12)]

In addition to its C-terminal catalytic domain, cGAS contains an unstructured and highly positively charged N-terminal domain (1). Together, both domains bind to dsDNA through multivalent interactions, which drive the liquid–liquid phase separation of cGAS and DNA (18). cGAS-DNA complexes are apparently concentrated within the phase-separated liquid droplets. This likely facilitates cGAS dimerization and, consequently activation. Long DNA is more efficient than short DNA in driving cGAS phase separation, explaining why long DNA is more potent than short DNA in activating cGAS. In addition, zinc ions readily promote cGAS liquid droplet formation in an intracellular environment that contains physiological salt concentrations, which would otherwise inhibit the ionic interactions between cGAS and DNA. Inside cells, cGAS and DNA form punctate structures, which exhibit classical liquid-like behavior, including dynamic exchange with cellular contents outside the droplets and fusion into larger spherical units. The cGAS-DNA droplets isolated from cells exhibit high specific activity in catalyzing cGAMP synthesis. Thus, DNA-induced phase separation of cGAS leads to the assembly of membrane-less compartments that function as “microreactors” to facilitate cGAMP synthesis. This phase transition of cGAS is highly sensitive to the concentrations of cGAS and DNA, suggesting a threshold response in which cGAS is activated only when cytosolic DNA reaches a certain level, such as in the case of pathogen infection (see below).

Insights into the cGAS-STING signal transduction cascade

Upon binding DNA, cGAS is activated to produce cGAMP, which binds to and activates STING on the ER membrane (3). STING exists as a preformed dimer with two STING C-terminal domains (CTDs) forming a V-shaped binding pocket facing the cytosol. Upon cGAMP binding, the CTD undergoes extensive conformational alterations, including the adoption of a more closed conformation by the V-shaped dimer and the formation of a “lid” covering the cGAMP binding site (19). Through an unknown mechanism, this conformational change in STING ultimately triggers its translocation from the ER to the Golgi via the ER-Golgi intermediate compartment (ERGIC) (2022). The translocation of STING to the Golgi is necessary to engage downstream signaling components and to regulate the induction of type I interferon (IFN) transcription, a hallmark output signal of cGAS-STING activity (Fig. 2). This functional property emerges from the exquisite ability of STING to create a signaling platform allowing for the recruitment of TBK1 and IRF3. After trafficking to the Golgi, STING is palmitoylated at two cysteine residues (Cys88 and Cys91). This process occurs at the Golgi and may explain the requirement to traffic to this specific organelle (23). This palmitoylation may further promote STING oligomerization and subsequent activation of TBK1. However, TBK1 recruitment per se is insufficient to activate IRF3. Rather, the phosphorylation of a conserved consensus motif (pLxIS; p is a hydrophilic residue and x is any residue), present in the C-terminal portion of STING (S366), must occur for STING to be licensed for the interaction with IRF3. This key step is mediated by TBK1 (24). Through this mechanism, TBK1 phosphorylates IRF3, which subsequently dimerizes and translocates to the nucleus to drive type I IFN expression. After initiation of downstream signaling, STING is degraded in endolysosomes (21, 25).

Fig. 2 The cGAS–cGAMP–STING signaling pathway.

The presence of cytosolic dsDNA—for example, through DNA virus infection—is recognized by cGAS. Binding of cGAS leads to the formation of cGAS-DNA complexes, which condense into liquid-like droplets through phase separation. This molecular process allows proper activation of the enzymatic function of cGAS leading to the synthesis of 2′3′-cGAMP, a potent ligand of STING. cGAMP-activated STING translocates to the Golgi, where palmitoylation on STING occurs. STING is phosphorylated by TBK1, which licenses STING to recruit the TBK1 substrate IRF3, leading to IRF3 phosphorylation. IRF3 then dimerizes and translocates to the nucleus to drive the expression of type I IFNs and ISGs. The cGAScGAMPSTING signaling pathway also promotes the activation of NF-κB transcription factors, which induce proinflammatory genes. cGAMP can also activate innate immune defense in neighboring cells through gap-junction transfer of cGAMP.

Although this review focuses on self-DNA sensing and the effects of the endogenous cyclic dinucleotide cGAMP, STING can also directly become engaged by prokaryotic cyclic dinucleotides, containing two 3′-5′-phosphodiester linkages (3′3′-CDNs) (26). These 3′3′-CDNs, including cyclic di-GMP, cyclic di-AMP, and 3′3′-cGAMP, are specific to bacterial microbes and control a variety of bacterial cellular processes, including essential activities such as those regulating growth and metabolism. Thus, this feature defines STING as a direct pathogen recognition receptor (PRR) that may influence innate immune responses independently of cGAS. Compared to other 3′3′-CDNs, however, the 2′3′-cGAMP interaction with STING is unique. Relative to 3′3′-cGAMP and other CDNs, 2′3′-cGAMP shows an exquisite high affinity for STING and elicits stronger type I IFN responses (1517). Notably, 2′3′-cGAMP adopts a more ordered conformation as a free ligand compared to other 3′3′ CDNs (27). This favorable intrinsic thermodynamic property of 2′3′-cGAMP itself may explain the high-affinity binding of the 2′3′-linked molecule to STING. In addition, on the basis of structural studies, it appears that 2′3′-cGAMP can form more-extensive interactions inside the STING ligand-binding pocket, thereby stabilizing a specific intermediate state with a readily covered lid. It is currently unknown why this distinctive state would allow for a more potent stimulatory capacity or stronger cellular responses.

cGAS-STING–initiated cellular effector mechanisms

The hallmark signaling output of the cGAS-STING pathway is the transcriptional up-regulation of type I IFNs and other IRF3 target genes, including several IFN-stimulated genes (ISGs). In addition, cGAS-STING signaling has been shown to activate canonical and—in cancer cells—noncanonical nuclear factor κB (NF-κB), mitogen-activated protein (MAP) kinases, and signal transducer and activator of transcription (STAT) transcription factors (28, 29). The precise molecular mechanism regulating the engagement of these latter transcription factors is incompletely understood. What is known, however, is that the C-terminal tail (CTT) of STING and, in particular, the TBK1-controlled phosphorylation of S366, are not essential to activate NF-κB (30). This observation is interesting from an evolutionary perspective because the CTT is a rather recent functional element of STING, which emerged simultaneously with the IFN antiviral system (31). CTT-independent responses may thus represent “primitive” forms of STING-mediated pathogen defense that may even rely on a distinct repertoire of antimicrobial effector functions not controlled by de novo gene expression (31). In support of this concept, STING has been reported to mobilize autophagy to resist Mycobacterium tuberculosis infection (32). STING activation by cyclic di-AMP has also been positioned upstream of an integrated ER stress response that enforces innate defense selectively against Gram-positive bacteria (33). Another type I IFN-independent antiviral effector function of STING, comprising the suppression of RNA translation, has recently been proposed in the context of RNA virus infection (34).

cGAS and STING also control cellular proliferation and the initiation of cell death programs in certain cell types. The activation of cGAS or STING within T and B lymphocytes can slow down proliferation or trigger apoptosis (30, 35, 36). This is reflected by gain-of-function mutations within STING, which leads to T cell apoptosis in humans and to severe T cell lymphopenia in mice (37, 38). Similarly, human monocytes undergo cell death upon stimulation of STING (30, 39). The molecular determinants of these cell-specific outcomes are not entirely resolved but may depend on differences in the steady-state expression of STING and/or in a distinct ability for negative-feedback inhibition (36).

The cGAS-STING signaling cascade, featuring a transmittable second messenger, also functions across neighboring cells via gap-junction communication and may perhaps even connect more distant cell types or tissues through the viral transfer of cGAMP (4042). This intercellular cGAMP-mediated effector response has been proposed to support growth of brain metastatic cells by promoting a tumor-conducive cytokine environment (43).

It is important to note that studies of cGAS- or STING-mediated effector responses are often performed using a single well-defined ligand. However, within the context of genuine stress responses, infectious and noninfectious alike, cells usually receive multiple inputs or experience many environmental alterations simultaneously. Accordingly, the cellular repercussions of cGAS-STING activation may differ in such more complex settings. For example, in the context of HIV infection of dendritic cells, subtle cGAS-dependent type I IFN responses are markedly potentiated by secondary inflammatory cues (44).

Although cGAS and STING generally function in the same pathway, they may also function independently of one another in some cases. Best studied in this context is the function carried out by STING as a direct sensor of CDNs during bacterial infection, which appears to be conserved across several species, including Drosophila melanogaster (26, 45). cGAS-independent roles of STING during RNA virus infection have also been reported. However, it is unclear what activates STING in this setting (34). Likewise, cGAS has been shown to trigger autophagy independently of STING (46). cGAS-deficient cells also show a higher tendency to immortalize compared to STING-deficient cells (47, 48). Whether this phenomenon relies on the role of cGAS in autophagy, or is perhaps due to its impact on suppressing DNA repair pathways, remains to be clarified. Finally, it is worth noting that 3′3′-linked CDNs have recently been shown to confer STING-independent cellular responses, including the negative regulation of NF-κB (49).

Regulatory mechanisms restricting abnormal responses to self-DNA

The evolutionary conservation of DNA-sensing strategies in pathogen recognition and defense illustrates the fundamental advantage of a sensitive, universal, and efficient recognition strategy, which is highly advantageous when considering the necessity to generate a rapid response to microbial invasion. However, the lack of “pathogen-specific” attributes requires additional mechanisms that allow cGAS to be part of a cell-autonomous inflammatory defense program. We propose below conceptually distinct regulatory mechanisms that not only facilitate the “preferred” immune recognition of pathogens but, simultaneously, constrain activation by self-DNA species so to avoid untoward consequences for the host (Fig. 3).

Fig. 3 Regulatory mechanisms governing self-DNA sensing.

(A) Conceptually distinct mechanisms are illustrated encompassing regulated DNA turnover, subcellular DNA compartmentalization, and the existence of a dynamic activation threshold of cGAS and STING, respectively. (B) Fluorescent images of cGAS (green) colocalizing with DNA in micronuclei (left), entering the nucleus (middle), and attaching to chromatin herniations (right) are shown. DAPI (4′,6-diamidino-2-phenylindole) is shown in blue and the nuclear lamina is stained in red. Scale bars, 5 μm.

Nucleases as natural antagonists of innate DNA sensing

The critical role of host deoxyribonucleases (DNases) in counteracting aberrant innate immune responses was first revealed by studying mouse models and human genetic disorders with defects in specific nucleases (see below). It has become clear that a critical function of intracellular nucleases is the preservation of cellular “DNA” homeostasis under physiological conditions. DNases continuously operate at baseline levels to ensure the disposal of potentially harmful DNA that accumulates as a consequence of normal cellular or organismal function. The 3′-5′ exonuclease TREX1, for example, is tasked with the degradation of cytosolic DNA substrates originating from active retroelements or damaged DNA (50, 51). The lysosomal endonuclease DNase II acts to degrade phagocytosed DNA material stemming from apoptotic cells, but may also influence cell-intrinsic degradation of DNA damage products from the nucleus (52, 53). Loss of TREX1 or DNase II has been shown to trigger lethal autoinflammatory diseases that can be rescued by deletion of cGAS or STING in mouse models (5457).

Regulation of the activity of DNA, cGAS, and cGAMP

At the receptor level, regulatory mechanisms act on cGAS activity, which collectively may facilitate its primary function in antimicrobial defense signaling. First, cGAS is an interferon-stimulated gene (ISG) (58). Thus, during an infection, cGAS levels build up and thereby lower the threshold to permit DNA sensing and cellular activation. Mechanistically, this threshold response can be explained by the cGAS phase separation boundary that depends on the concentrations of both cGAS and DNA in the cytosol. Additionally, it has been reported that posttranslational modifications can both negatively (e.g., phosphorylation) and positively affect cGAS activity, as does the ionic environment (e.g., Zn2+ or Mn2+ ions) (18, 59).

As detailed above, there are also important regulatory elements surrounding the DNA ligand. As such, pathogens may deliver optimal cGAS ligands—long double-stranded B-form helices and dsDNA associated with nucleoid proteins—which are rarely found in the cytoplasm of cells under normal conditions. Conversely, self-DNA may be less immunogenic. It is possible that the structural organization of chromatin or alternative intrinsic biophysical properties restrain cGAS activity toward endogenous DNA. In this context, it is notable that cGAS is also present in the nucleus. Nuclear positioning may be necessary for cGAS to execute its antiviral function given that many viruses (e.g., herpesviruses) recognized by cGAS release their genomic content only after having reached the nuclear compartment. Thus, cGAS may be able to distinguish self- from nonself DNA inside the nucleus, although the mechanistic basis for such discriminative behavior remains to be established. Finally, levels of cGAMP are likely subject to cellular self-control. For example, the phosphodiesterase ENPP1 was shown to cleave cGAMP, although this appears to occur extracellularly (60).

Cellular compartmentalization

Compartmentalization has been regarded as one of the most obvious prerequisites for the ordered functioning of cytosolic DNA sensors in innate immunity by physically separating self-DNA from cGAS under steady-state conditions. This conceptual model of compartmentalization of self-DNA has indeed been validated in various contexts, demonstrating that altered self-DNA localization can indeed elicit strong inflammatory responses (see below). However, numerous settings of physiological breakdown of compartment integrity (e.g., mitosis and apoptosis), as well as the nuclear localization of cGAS in some cells, suggest that there are additional mechanisms regulating cGAS activity when it encounters abundant chromatin DNA. We propose here that the cause and kinetics of the loss of organelle integrity dictate whether or not cGAS becomes engaged. During normal biological processes, such as mitosis, tightly orchestrated decompartmentalization may impose transient negative control over cGAS and/or STING and impair activation. Indeed, there is evidence that such cellular self-organization of dsDNA recognition occurs during apoptosis. Despite the release of DNA into the cytosol as a result of inner mitochondrial membrane herniations via BAK or BAX macropores, apoptotic cells restrain innate immune signaling through the action of apoptotic caspases (61, 62). Similar control mechanisms may operate during cell division, yet the nature of the underlying signals remains to be elucidated. In addition, cells may mitigate cGAS activity against transient exposure to (damaged) self-DNA through repair or adaptation processes. For example, acute nuclear membrane damage such as that occurring during cell migration is perceived by cGAS, but rapid sealing of nuclear perforations through the ESCRT machinery possibly prevents the generation of an inflammatory signal (63, 64). By contrast, if unscheduled, protracted, or combined with changes in self-DNA configuration (damaged DNA), loss of organelle integrity provokes cGAS activity. For example, pathogen-mediated destruction of mitochondrial integrity is acute and not programmed (65). Following exposure to acute stressors, such as DNA damage inducers, fragile micronuclei are formed, which are prone to rupture unpredictably and cause cGAS activation (66, 67). After periods of progressive stress, senescent cells remodel their nuclear envelope, and this is accompanied with the repetitive release of cytosolic chromatin fragments, which present potent triggers of inflammation via cGAS (47, 48, 68). Likewise, persistent mitochondrial stress triggers cGAS-dependent cellular responses (69). Chronic alterations in lipid metabolism may similarly interfere with the proper localization of DNA inside mitochondria (70). Future studies are needed to refine our molecular understanding of acute versus chronic, scheduled versus unscheduled, minor versus major decompartmentalization, and normal versus modified DNA in terms of their impact on cGAS activity.

Impact of self-DNA sensing on human health

The homeostasis of innate DNA sensing is predicated in large part on the interplay among distinct regulatory mechanisms (see above). Disruption of any element of the underlying safeguarding network can give rise to excessive or prolonged activation of innate immune effector responses and may eventually contribute to human disease (Table 1). We begin this section by considering how self-DNA–sensing mechanisms participate in the recognition of pathogens. We follow this by reviewing sterile inflammation caused by self-DNA recognition.

Table 1 Overview of diseases linked to cGAS and/or STING.

Abbreviations: KO, knockout; NA, not available; PBMCs, peripheral blood mononuclear cells.

View this table:

The host defense “perspective” of self-DNA sensing

Our prevailing view of innate DNA sensing in the context of pathogen infection is strongly influenced by the traditional PRR paradigm. Although some of the mechanistic details of pathogen-derived DNA sensing warrant further clarification, substantial evidence supports the predominant role of cGAS in sensing viral DNA during infection. However, in addition to exploiting a straightforward PAMP-based recognition strategy, cGAS also facilitates DAMP-based pathogen detection acting against a much broader range of pathogens, including RNA viruses. In the context of viral infection, several reports have exemplified that mounting antiviral responses relies to an appreciable extent on the binding of mitochondrial DNA to cGAS. In herpes simplex virus (HSV)–infected cells, for example, virus-induced mitochondrial DNA (mtDNA) stress and subsequent liberation of mtDNA into the cytosol constitute an important trigger of the ensuing antiviral response (69). Other pathogens, including certain virulent strains of Mycobacterium tuberculosis and the positive-strand RNA virus dengue virus, were shown to cause mitochondrial damage and DNA release that lead to the activation of cGAS-STING signaling (65, 71). The finding that dengue virus encodes proteins that degrade cGAS suggests an important evolutionary pressure for natural antagonists of the self-DNA–sensing machinery for viral propagation (65). In this context, it is interesting to note that cGAS-deficient mice display increased susceptibility against West Nile virus infection, underscoring the biological importance of self-DNA sensing in the context of systemic antiviral immunity (58). In addition to serving at a cellular level, self-DNA triggered inflammation may also contribute to systemic, pathogen-induced host responses at an organismal level. The massive accrual of dying cells is a common feature of sepsis, a detrimental consequence of systemic bacterial infection. STING, presumably activated by self-DNA, has been recognized to precipitate deleterious effects associated with this hyperinflammatory condition (72).

Self-DNA sensing in autoinflammatory and autoimmune diseases

Studies performed in mutant mice harboring gene products that regulate DNA metabolism provided a first causal link between autoinflammation and the cGAS-STING system. For example, Trex1 deficiency has been shown to cause a lethal inflammatory phenotype in mice involving multiple organs, especially the heart, which is entirely rescued by concurrent deletion of cGAS, STING, IRF3, or IFNAR (51, 5557). In humans, TREX1 was the first gene to be linked to the rare genetic disorder Aicardi–Goutières syndrome (AGS), a systemic inflammatory disease with an early onset (73). Mutations in additional genes have been associated with AGS, namely SAMHD1, all components of the RNase H2 endonuclease complex, IFIH1, and ADAR1 (74). Of these, defects arising from SAMHD1 or RnaseH2 mutations can trigger cGAS-STING activity. Thus, the accumulation of endogenous DNA material is predicted to be a common disease-causing event. The precise pathological mechanism of DNA accumulation, however, appears to differ among individual AGS mutant cells, reflecting their distinct biological roles. The heterotrimeric RNase H2 complex removes misincorporated ribonucleotides from genomic DNA, defects of which create genomic instability, DNA damage, and the formation of micronuclei (67). SAMHD1, through its deoxynucleotide triphosphate phosphohydrolase activity, regulates intracellular nucleotide levels, which are important for reverse transcription of retroviral cDNA but also to maintain genome integrity (75, 76). As an additional function, SAMHD1 has recently been shown to regulate DNA replication fork resection and, if inactive or not present, ssDNA accumulates in the cytoplasm upon replication fork stalling (77). Similarly, TREX1 activity is involved in DNA damage responses but has also been linked to suppression of retroelement transposition (50, 51). Collectively, these findings define a common role of AGS genes to suppress the accrual of endogenous DNA metabolites and suggest that malfunctioning of DNA metabolism can cause severe inflammation. Given the shared functioning of the AGS-associated genes in DNA damage responses, additional genetic mutations that activate cGAS may be uncovered. A critical future question is whether in these emerging contexts, cGAS activity has similar detrimental effects for the initiation or aggravation of the disease phenotype, or perhaps even restricts disease phenotypes in some cases. Alternatively, the defects in DNA damage response may cause additional perturbations unrelated to the aberrant action of cGAS. Indeed, despite a similar biochemical pathogenesis, AGS patients show discrete clinical phenotypes. Along these lines, a clinical syndrome referred to as STING-associated vasculopathy with onset in infancy (SAVI) is caused by gain-of-function mutations of STING and characterized by early-onset systemic inflammation with fever, severe vasculopathy that leads to debilitating skin lesions and interstitial lung disease resulting in pulmonary fibrosis (37, 78). Currently, it is unclear why, despite relying on a similar pathological mechanism—aberrant cGAS-STING stimulation—these genetic syndromes exhibit distinct clinical manifestations. This may reflect the differential involvement of cell types and/or participation of distinct inflammatory pathways.

Deleterious systemic inflammatory consequences have also been documented in the context of perturbed extracellular DNA metabolism. Studies in mice revealed that the endonuclease DNase II is critical for the ordered and “immunologically silent” disposal of apoptotic cells by macrophages (efferocytosis). DNase II deficiency in mice causes lethal anemia or debilitating polyarthritis when ablated in adulthood (52). The persistent stimulation of cGAS-STING within DNA-loaded macrophages has been proposed to mediate these adverse effects (52, 54, 56). In humans, a multisystem autoinflammatory syndrome due to hypomorphic mutations in DNASE2 has been described with patients presenting with related symptoms, including neonatal anemia and arthropathy (79).

Systemic lupus erythematosus (SLE) is the most prevalent human disease associated with an increase in type I IFN levels and defects in apoptotic cell clearance alike, raising the possibility that maladaptive self-DNA signaling via cGAS may contribute to the manifestation of SLE and other lupus-like diseases. A strong argument for this hypothesis is a recent study that found increased cGAMP concentrations in the peripheral blood from ~15% of SLE patients (80). Further strengthening a link between SLE and cGAS, a large cohort study identified polymorphisms in TREX1 as markers for predisposition to SLE (81). Moreover, heterozygous mutations of TREX1 are associated with familial chilblain lupus, a monogenic form of cutaneous lupus erythematosus (82). Similarly, RNase H2 mutations have been linked to SLE in addition to AGS (83).

Self-DNA sensing in cellular senescence

Recent studies have uncovered an important function of cGAS and STING in cellular senescence, a permanent state of cell cycle arrest. In addition to cessation of cell division, senescent cells also secrete inflammatory mediators, growth factors, and proteases, a phenomenon collectively known as the senescence-associated secretary phenotype (SASP) (84). cGAS or STING deficiency markedly compromises the establishment of cellular senescence, including the SASP, during spontaneous immortalization, in response to DNA-damaging agents or upon oncogene activation (47, 48, 68). A common thread that links these diverse senescence-inducing conditions is the appearance of micronuclei or cytoplasmic chromatin fragments (85). Micronuclei exhibit an unstable or ruptured nuclear envelope, thereby allowing cGAS access to the chromatin fragments, which presumably leads to cGAS activation (66, 67). cGAS association with micronuclei and the resultant activation of inflammatory genes have also been found in RNase H2–deficient cells and in cells with chromosome instability. Thus, cGAS activation by self-DNA appears to serve as a mechanism to alert the immune system of the presence of damaged or stressed cells that harbor a compromised genome. Although this may serve as a protective, tumor-suppressive mechanism in young animals, increased numbers of senescent cells in older animals are linked to organ dysfunction and a variety of aging-related diseases (86). Notably, many, if not all, of these disorders and even aging itself are associated with a low-grade inflammatory state, termed “inflammaging.” Remarkably, the genetic ablation of senescent cells in middle-aged mice ameliorates tissue inflammation and enhances the life-span and health of these animals (87). It remains to be seen whether the modulation of cGAS or STING is beneficial in the context of senescence-associated diseases. If so, inhibitors of the cGAS-STING pathway may be used as senomorphic agents.

Self-DNA–mediated activation of cGAS as a pathogenic factor in common diseases

Evidence is emerging that self-DNA–mediated immune activation also gives rise to a diverse spectrum of more prevalent diseases (Table 1). In the following section, we highlight two concrete examples, which provide strong support for the maladaptive roles of cGAS and STING.

Inflammation is tightly linked to both acute and chronic heart failure (88). For example, there is growing recognition that myocardial infarction (MI) is associated with a strong inflammatory response in the inflicted tissue. It has recently been proposed that self-DNA sensing from dying cells leads to MI-associated type I IFN response (89, 90). The inhibition of components of the cGAS–STING–type I IFN axis was shown to hinder pathological myocardial remodeling, preserve cardiac function, and improve post-MI survival in mice. Other conditions associated with severe tissue destruction and the sudden or excessive accumulation of dead cells, such as sepsis or pancreatitis, may fuel a similar self-DNA–initiated inflammatory program and contribute to disease pathology (72, 91).

Recent studies have provided a link between the cGAS-STING pathway and Parkinson’s diseases (92). In familial Parkinson’s disease, loss-of-function mutations have been found in the ubiquitin E3 ligase Parkin (PRKN) and protein kinase PINK1. Both Parkin and Pink1 have been shown to play a critical role in the autophagy of damaged mitochondria, a process known as mitophagy. Mice deficient in Prkn or Pink1 produce increased numbers of inflammatory cytokines and type I interferons when they are under stress (e.g, exhaustive exercise) (92). These inflammatory phenotypes were rescued by deleting STING, suggesting that the STING pathway is activated when damaged mitochondria cannot be removed by Parkin- and Pink1-mediated mitophagy. STING deletion also rescued the loss of dopaminergic neurons and motor defects observed in Parkin-deficient mutator mice, which accumulate damaged mitochondrial DNA. Thus, DNA from damaged mitochondria appears to activate cGAS, which may subsequently drive inflammation in Parkinson’s disease patients. As damaged mitochondria and mitochondrial DNA have been increasingly linked to a variety of chronic diseases, including neurodegeneration or age-dependent macular degeneration (93), it is possible that the activation of cGAS and STING is a mechanism underlying many of these diseases.

cGAS-STING signaling in cancer immunity and immunotherapy

Cancer cells are constantly under stress that results from chromosomal abnormality, genomic DNA damage, and hyperproliferation. Such stress is detected by the body’s immune system, which plays a pivotal role in suppressing cancer. This cancer immune surveillance is known to be mediated by immune cells such as natural killer (NK) and T cells. However, how the immune system seeks out cancer cells from the abundance of normal cells in the body is a fundamental biological question that remains poorly understood. Recent studies suggest that the cGAS-STING pathway plays an important role in cancer immune surveillance in both cell-autonomous and cell-nonautonomous manners. Cancer cells that harbor chromosomal abnormality or genomic DNA damage often form micronuclei or cytoplasmic chromatin fragments that activate cGAS in a cell-autonomous manner. This, in turn, promotes cellular senescence and induces inflammatory and immune-stimulatory molecules (e.g, cytokines and NK ligands) that suppress or eliminate cancerous cells. However, cancer cells are also taken up by antigen-presenting cells (APCs) such as dendritic cells and macrophages in a cell-nonautonomous manner. Through an undefined mechanism, DNA from cancer cells may be delivered to the cytosol of APCs where cGAS is activated, leading to the production of IFNs and immune-stimulatory molecules, which are important to prime and expand tumor-specific T cells and recruit them to tumors (94, 95). Through these antitumor functions, cGAMP and its analogs have been shown to exert potent antitumor effects in mouse tumor models, especially when used in combination with immune-checkpoint blockers such as the antibody against PD-L1 (96, 97).

Like pathogens that have succeeded in coexisting with their hosts, cancer cells have evolved to evade the host immune surveillance. Many cancer cells silence the expression of cGAS or STING, or both, which allows them to escape from senescence and immune surveillance. Epigenetic mechanisms such as DNA methylation have been found to cause the silencing of these genes in cancer cells (98). Alternatively, some cancer cells can even co-opt the cGAS pathway to their own advantage. For example, brain metastatic cancer cells in mice can transfer cGAMP to neighboring astrocytes to activate inflammatory mediators, which in turn promote the metastasis of brain cancer cells (43). Similarly, chromosomal instability was found to promote metastases in a STING-dependent manner (29). Finally, cGAS may also facilitate tumorigenesis by suppressing homologous recombination in the nucleus (99). These findings warrant caution in moving forward clinical programs that aim to boost the cGAS-STING pathway to treat cancers. An additional concern surrounding the use of STING pathway agonists in cancer therapy is the potential for “cytokine storms,” in which the excessive production of cytokines can lead to severe toxicity and even death.

Therapeutic targeting of the cGAS-STING pathway

The potent immune-stimulatory potential of the cGAS-STING pathway, as well as its role in a wide range of human diseases, makes it highly attractive for pharmacological interventions. Initial drug discovery efforts have mainly focused on the development of STING agonists for usage as vaccine adjuvants or as immunostimulatory anticancer compounds. Some existing vaccine adjuvants, including alum or chitosan, have been shown to rely—at least to some extent—on self-DNA–triggered activation of cGAS and STING (100, 101). Additional research using natural or synthetic cGAMP derivatives and cGAS- or STING-deficient animals has further established that STING agonists exert highly potent effects in establishing immune memory after prophylactic vaccination (102).

As highlighted above, overactive cGAS and STING can have deleterious consequences. Thus, the pharmacological inhibition of the cGAS-STING pathway is thought to have broad therapeutic potential either by providing an alternative treatment strategy that complements existing therapies or by establishing a unique approach for unmet medical needs. Recently, compounds antagonizing cGAS have been described. However, their further development has been hampered by poor activity in cells or poor selectivity against off-target effects (103, 104). Nevertheless, these proof-of-concept studies, along with the availability of the cGAS structure, offer optimism that more suitable lead candidates may be rationally designed in the near future. Moreover, small-molecule inhibitors of STING have been identified from a phenotypic screening effort and were subsequently found to block STING in a covalent manner (105). Notably, the covalent inhibition of STING is specific, as it does not inhibit other innate immune pathways. Further investigation is required, however, to precisely determine whether targets other than STING may exist. Covalent modifiers of STING displayed high therapeutic efficacy in a mouse model of AGS and provide a compelling argument for targeting STING. One potential problem posed by suppressing cGAS or STING is how to disrupt their activity without increasing the risk of infection. In the Trex1-/- mouse model of AGS, the deletion of only one allele of cGAS or STING is sufficient to completely suppress the disease phenotype (55, 56, 106). Thus, dosing schemes may be elaborated that would not limit the usefulness of cGAS or STING antagonists, even as part of a long-term treatment regimen. Moreover, compared to the blockade of common antiviral components, such as TBK1 or elements of the type I IFN signaling axis, the suppression of cGAS and STING bears less risk of immunosuppression and opportunistic infection, because it would leave intact other PRR systems. In addition, beyond potential safety advantages, cGAS and STING inhibitors may also have higher potency relative to existing therapeutic agents (e.g., JAK inhibitors and antibodies against interferon receptor), because the latter cannot limit the maladaptive effects of other cytokines such as TNF-α and IL-6. Indeed, genetic ablation of IRF3 fails to rescue the deleterious effects of a hyperactive STING allele in mice (38).

Conclusion and future perspective

The innate recognition of cytosolic DNA is a fundamental property of immunity in various species. However, far beyond the original concept of its primary role in antiviral defense, the cGAS-STING pathway constitutes a key cellular stress-sensing pathway that has important implications for many noninfectious states. Research over the past few years has illustrated the benefits and constraints of incorporating a universal dsDNA detection system into the operations of a cell as a mechanism of innate immune defense. The conflict arising from the apparent inability to distinguish “self” from “nonself” on a receptor level requires the implementation of various regulatory strategies, which we have outlined in this review. These programs work synergistically to set up a cellular threshold that is under normal conditions tolerant to self-DNA, but permits pathway activation on encountering noxious challenges. The effects of self-DNA sensing in the context of damage can occasionally provide advantages to the host by facilitating stress adaptation and defense from endogenous derangements (e.g., chromosomal instability), but it can also aggravate maladaptive responses to cellular damage and, thereby, drive deleterious inflammatory diseases.

Much remains to be learned about the rules of engagement that dictate the function and outcome of the self-DNA–sensing capacity of the cGAS-STING pathway in distinct contexts. A critical topic will be to better understand the regulation of cGAS and STING in space and time. In this regard, an important unanswered question is how cells accomplish a “programmed” state of tolerance against self-DNA (e.g., during mitosis or apoptosis). It is unlikely that compartmentalization itself is sufficient to maintain such a tolerogenic state and, thus, the identification of additional factors that act upon cGAS or self-DNA will be of importance. We also note that under some conditions, cGAS and STING can operate independently of each other. Disentangling the individual functions of each pathway component will refine our current understanding of innate DNA sensing as a whole. Along these lines, cGAS and STING are considered to be present in most, if not all, immune and nonimmune cells, but their functions in each individual cell type and tissue remain to be characterized. An in-depth understanding of the cGAS-STING pathway, including the careful consideration of possible species-specific differences, will be instrumental for further development of therapeutics targeting the DNA-sensing pathway. If successful, pharmacological agonists and antagonists of cGAS and STING could become effective therapies for many prevalent and emerging diseases, including infectious, autoimmune, inflammatory, and senescence-associated diseases and cancer.

References and Notes

Acknowledgments: We thank B. Guey for providing fluorescence microscopy images displayed in Fig. 3. Funding: Research in the Chen laboratory has been supported by grants from the NIH (P50AR070594 and U01CA218422), the Cancer Prevention and Research Institute of Texas (RP120718 and RP150498), Lupus Research Alliance Distinguished Innovator Award, the Welch Foundation (I-1389), and ImmuneSensor LLC. Research in the Ablasser laboratory is supported by grants from the SNSF (BSSGI0-155984, 31003A_159836), the Gebert Rüf Foundation (GRS-059_14), the European Research Council (ERC-StG; 804933 ImAgine), and the Fondation Acteria. Competing interests: A.A. is a shareholder in company developing STING- and cGAS-directed therapeutics.
View Abstract

Stay Connected to Science

Navigate This Article