New Frontiers in the cGAS-STING Intracellular DNA Sensing Pathway

Abstract

The cGAS-STING intracellular DNA sensing pathway has emerged as a key element of innate antiviral immunity and a promising therapeutic target. The existence of an innate immune sensor that can be activated by any double-stranded DNA of any origin raises fundamental questions about how cGAS is regulated and how it responds to “foreign” DNA while maintaining tolerance to ubiquitous self-DNA. In this review, we summarize recent evidence implicating important roles for cGAS in detection of foreign and self-DNA. We describe two recent and surprising insights into cGAS-STING biology: that cGAS is tightly tethered to the nucleosome, and that the cGAMP product of cGAS is an immunotransmitter acting at a distance to control innate immunity. We consider how these advances influence our understanding of the emerging roles of cGAS in the DNA damage response, senescence, aging, and cancer biology. Finally, we describe emerging approaches to harness cGAS-STING biology for therapeutic benefit.

Etoc blurb:

The cGAS-STING pathway detects intracellular DNA and activates a potent antiviral response. Dvorkin and colleagues summarize recent insights into cGAS-STING biology and its role in host defense, immunopathologies, and anti-tumor immunity.

Introduction

Our current understanding of innate antiviral immunity is rooted in elegant experiments conducted during the 1950s in which Isaacs and Lindenmann found that cells treated with inactivated influenza virus produced a soluble factor that could protect fresh cells from subsequent infection with live virus. They isolated this factor and termed it “interferon” owing to its ability to interfere with viral infection^1,2. We now know that the type I interferons (IFNs) are a family of many closely related cytokines that are produced in response to virus infection³.

The induction of IFNs is, in most cases, under the control of innate immune sensors of nucleic acids. Because viruses can have a genome composed of either RNA or DNA, there are specific sensors that can detect either RNA or DNA to activate the antiviral IFN response. These include several Toll-like receptors (TLRs) expressed by sentinel immune cells that sample endosomal compartments for the presence of unusual nucleic acids⁴. In addition to the TLRs, there are broadly expressed sensors that detect intracellular nucleic acids and signal the presence of viral infection from within the infected cell itself⁵. The RIG-I-like receptors detect structural features of viral RNAs that are scarce within our own cells and signal through the adapter protein MAVS to activate the IFN-mediated antiviral response⁶. The principal sensor of intracellular DNA in vertebrates is cyclic GMP-AMP synthase (cGAS)⁷, a member of the nucleotidyltransferase family of enzymes that includes the oligoadenylate synthetases (OAS). cGAS binds the sugar-phosphate backbone of double-stranded DNA (dsDNA) in a sequence-independent fashion, resulting in structural rearrangements that activate the production of cyclic GMP-AMP (cGAMP) from GTP and ATP. cGAMP rapidly diffuses throughout the cell and binds to STING, a transmembrane protein that resides on the endoplasmic reticulum (ER). cGAMP binding to STING results in a conformational change in STING that exposes binding sites for the kinase TBK1 and the transcription factor IRF3^8,9. IRF3 undergoes TBK1-mediated phosphorylation, dimerization, and nuclear translocation to activate the IFN genes. In addition to activating IFNs, STING also engages the NF-kB transcription factor to induce the expression of proinflammatory cytokines and chemokines¹⁰.

Our understanding of the mechanisms and functions of the cGAS-STING DNA sensing pathway has grown exponentially since the first descriptions of intracellular DNA sensing in 2006^11,12, the discovery of STING in 2008^13,14, and the identification of cGAS and cGAMP in 2013^7,15. Among the many exciting developments in the field, we have learned that cGAS, cGAMP, and STING have deep evolutionary roots in prokaryote anti-phage immunity. Hundreds of bacterial cGAS-like enzymes have been identified that produce an array of cyclic dinucleotides and trinucleotides¹⁶. Even the ligand specificities and of eukaryotic cGAS enzymes have diverged over evolutionary time: cGAS-like receptors in Drosophila fruit flies can detect dsRNA, not dsDNA^17–19.

In vertebrates, the activation of cGAS by dsDNA of any potential origin has created a conundrum of self/non-self discrimination that has intrigued the field since its inception: how does a sequence-independent sensor of dsDNA avoid constant activation by the billions of base pairs of self-DNA in all nucleated cells? Indeed, we have learned of important instances in which cGAS can be activated by self-DNA. An extreme example of this is the rare human “interferonopathy” Aicardi-Goutieres Syndrome, some cases of which are caused by mutations in enzymes that metabolize, modify, or package self-DNA, resulting in inappropriate cGAS activation and severe pathology^20–22. Perhaps less extreme but more pervasive, accumulated damage of nuclear and mitochondrial self-DNA during cellular aging appears to activate cGAS and contribute to the processes of senescence and organismal “inflammaging,” both of which are typically viewed as pathological events that reduce healthy lifespan. In contrast, a growing body of evidence implicates these same mechanisms of self-DNA reactivity in protective immunity to cancer stimulated by radiation and chemotherapy. Thus, the high potential for activation of cGAS by self-DNA may have both pathological and protective functions.

In this review, we consider these pathological and protective functions of cGAS. We summarize two recent developments that offer unexpected insights into the regulation of the cGAS-STING pathway. The first involves a foundational concept in the field known as “cytosolic DNA sensing,” which was proposed in 2006 as a logical (but not evidence-based) explanation for how an intracellular DNA sensor could avoid reactivity to genomic DNA¹². This proposal, which was made seven years before the discovery of cGAS as the key DNA sensor, suggested that sequestration of the sensor in the cytosol would avoid activation by nuclear DNA while enabling responses to DNA that appeared inappropriately in the cytosol. As described below, recent functional and structural studies have revealed that the majority of cGAS in most cell lines is tightly tethered to nuclear chromatin in a way that prevents activation by DNA. Thus, “cytosolic DNA sensing” may be an incomplete framework for considering cGAS biology. The second recent insight is the discovery that cGAMP is an “immunotransmitter” that is exported from and imported into cells, thus allowing for a mode of cell-to-cell transmission of cGAS-STING signaling²³. Considered in the context of the emerging understanding of self-DNA detection by cGAS, these insights have important implications for therapeutic approaches to block cGAS signaling in the context of immunopathology, as well as for harnessing cGAS activation to improve anti-cancer immunity.

cGAS is tightly tethered to the nucleosome

Since its discovery⁷, cGAS has been considered a cytosolic PRR. Given cGAS has no sequence specificity, its localization in the cytosol was thought to be important to sequester it away from genomic DNA to prevent aberrant activation of the cGAS-STING pathway. However, several reports indicated a role for cGAS localized in the nucleus in several different functions including detection of nuclear-replicating viruses, regulation of DNA repair, and promotion of senescence^24–27. These studies indicated that cGAS was only localized in the nucleus under certain conditions, such as viral infection or DNA damage, or during mitosis. In 2019, Manel and colleagues showed that cGAS accumulates on chromatin during mitosis and persists within the nucleus in daughter cells through the next interphase²⁸. The N-terminus of cGAS was found to be important for its nuclear localization and enrichment on centromeric DNA and LINE elements²⁸. Gekara and colleagues also demonstrated that cGAS was constitutively present in the nucleus, and found that this localization required cGAS binding to DNA²⁹. In contrast, Funabiki and colleagues found that purified cGAS lacking the N-terminus bound to mononucleosomes with higher affinity than to naked dsDNA³⁰. This interaction was not mediated by cGAS binding to nucleosomal DNA, as DNA-binding mutants of cGAS bound nucleosomes with similar affinity. Instead, direct interactions between cGAS and the H2A/H2B acidic patch of the nucleosome were required, as mutations of the acidic patch reduced binding of cGAS to the nucleosome³⁰.

Volkman and colleagues demonstrated that in many cell human and mouse cell lines, most cGAS remains nuclear regardless of cell cycle³¹. In many cases, such as mouse macrophages and HeLa cells, the portion of cGAS that remains nuclear can be as high as 85–95%, indicating that nuclear cGAS represents a significant pool of cGAS within the cell. Nuclear cGAS is bound tightly to chromatin and is only liberated at very high salt concentrations. This binding is independent of the DNA-binding ability of cGAS and does not require the cGAS N-terminus. Instead, a set of evolutionarily conserved positively charged residues on the cGAS surface are required to tether cGAS to chromatin. Chief among these are two arginine residues, R236 and R255 in human cGAS. Mutation of each of these residues caused cGAS to become untethered from chromatin and highly autoreactive against self-DNA³¹, indicating that tethering of cGAS to chromatin inhibits cGAS activity, which explains how cGAS can remain nuclear at steady-state without inducing aberrant production of cGAMP.

Six independent groups^32–37 published cryo-electron microscopy structures of both mouse and human cGAS bound to the nucleosome core particle (NCP), elucidating many key features of this interaction, and discovering how cGAS can be kept inactive while being held mere angstroms from its activating ligand (Figure 1). Confirming prior reports³⁰, cGAS was found to bind to the nucleosome acidic patch, a charged pocket formed by a group of 8 residues from H2A and H2B that acts as a binding site for numerous chromatin-binding proteins^38,39. The interaction between cGAS and the acidic patch was very strong, with a measured binding affinity (K_d) of 8.6nM for reconstituted nucleosomes and 6.3nM for cell-derived nucleosomes³⁷. This makes the cGAS:nucleosome interaction over a hundred times stronger than the cGAS:DNA interaction. R236 and R255 of human cGAS were found to be the critical residues mediating binding of cGAS to the nucleosome acidic patch, acting as a canonical “arginine anchor” common to other acidic patch binding proteins, including regulator of chromatin condensation 1 (RCC1), high mobility group protein N2 (HMGN2), interleukin-33, and the latency-associated nuclear antigen (LANA) from Kaposi’s sarcoma herpesvirus (KSHV)^38,39. In addition, residues K254 (which stabilizes the cGAS loop within the acidic patch pocket) and R353 (which binds to D51 of H2B), were also shown to strongly regulate cGAS tethering^33,35–37. All of these studies used cGAS mutants lacking the N-terminus, confirming that it is completely dispensable for cGAS-chromatin tethering. Whether the N-terminus plays any role in further stabilizing cGAS on the nucleosome or participates in interactions that could untether cGAS from the nucleosome requires further investigation.

Figure 1. The cGAS-STING DNA sensing pathway.

Multiple sources of self and foreign DNA converge on cGAS activation, which leads to STING-dependent signaling.

cGAS has 3 DNA binding sites, termed A, B, and C. Canonical activation of cGAS requires dimerization, forming a complex of two cGAS molecules and two molecules of double stranded DNA in which both site A and site B are occupied on both cGAS molecules^40–42. Site C promotes further oligomerization and liquid phase separation to enhance cGAS signaling^43,44. In the cGAS:NCP complex, the critical cGAS residues responsible for binding to the acidic patch originate in or are proximal to site B, such that nucleosome binding directly occludes site B. This includes key DNA-binding residues including R236 and K254, which are repurposed in the cGAS:NCP complex for protein-protein interactions. Site A remains solvent exposed and participates only minimally in nuclear tethering but cannot bind DNA due to steric hindrance from the NCP. Taken together, these data reveal that nucleosome binding inhibits cGAS in three ways: 1) key DNA-binding residues on site B are repurposed for acidic patch interactions, preventing DNA-binding to site B; 2) steric hindrance from the proximal NCP prevent dimerization of cGAS which is required for activation; 3) steric hindrance from the NCP prevents binding of DNA to site A.

In most cell types cGAS exists as both a cytosolic and a nuclear pool; however, the factors that regulate this distribution remain uncharacterized. Several studies have reported that cGAS distribution is affected by cell cycle^29,45, whereas others have found that cGAS distribution is unaffected by cell cycle^31,44. This may be due to cell type-specific differences, including the rate of cell division. Moreover, one study suggested that the N terminus of cGAS recruits it to the plasma membrane in the cytosol⁴⁶.

Another mechanism that regulates protein distribution within the cell is the expression of nuclear localization sequences (NLS) and nuclear export sequences (NES). Two putative nuclear localization sequences have been described for cGAS, NLS1 and NLS2. NLS1 resides within the N-terminus and is therefore not required for nuclear localization of cGAS. Furthermore, when expressed alone, the N-terminus localizes primarily to the cytosol^31,47, suggesting that NLS1 is non-functional. In contrast, deletion of NLS2 (residues ²⁹⁵DVIMKRKRGGS³⁰⁵) from human cGAS (cGAS^ΔNLS2) has been shown to reduce the levels of cGAS within the nucleus^27,48,49. NLS sequences target cells for nuclear import through importin. cGAS has been shown to interact with importin proteins, and treatment with the importin-α inhibitor importazole also reduces the level of nuclear cGAS^27,49. Additionally, it should be noted that when expressed in cells, cGAS^ΔNLS2 failed to activate an IFNβ luciferase reporter⁴⁸. However, given that NLS2 resides proximally to the cGAS catalytic site, deletion of NLS2 may disrupt cGAS catalytic activity.

A putative NES sequence has also been described for cGAS (¹⁶⁹LEKLKL¹⁷⁴). Several studies have shown that mutation of the cGAS NES only modestly increases the level of nuclear cGAS^47,50, suggesting that the NES plays a minor role in maintaining the cytosolic pool of cGAS at baseline. Proteins expressing an NES are exported from the nucleus by CRM1, which can be inhibited by the antifungal leptomycin B (LMB). One study observed that in HeLa cells after DNA transfection, a portion of cGAS translocated from the nucleus to the cytosol in a CRM1-dependent manner, as LMB treatment blocked this translocation⁵⁰. This suggests that the cGAS NES is involved in cGAS translocation following activation in the nucleus. However, it should be noted that other studies have failed to see a shift in cGAS localization after DNA transfection in HeLa cells³¹, so the importance of the NES remains controversial.

Function of Nuclear cGAS

The function of nuclear cGAS is still an open question. Given that nuclear cGAS is held in an inactive state on the chromatin, nuclear tethering may merely be a method of sequestering excess cGAS. In this scenario, nuclear cGAS never becomes untethered or activated, and is held on the chromatin until otherwise degraded. We believe this is unlikely for several reasons. First, other acidic-patch binding proteins that bind with similar affinities, such as RCC1 and HMGN2, are known to be released from chromatin^51,52. Therefore, there is precedent for regulated release of acidic patch-binding proteins that may also apply to cGAS. Second, in many cell lines, the majority of cGAS is tethered to chromatin³¹. It is difficult to imagine that a cell would maintain such a large pool of inactive protein as a “dead end”, rather than target it for degradation when in excess. In fact, given that mutations that untether cGAS from chromatin render it massively and constitutively active³¹, it seems dangerous for the cell to keep cGAS tethered on chromatin solely as a sequestration mechanism.

It is tempting then to speculate that cGAS can become untethered under certain conditions and participate in immune responses. Given that most DNA viruses, such as herpesviruses and adenoviruses, replicate within the nucleus, nuclear cGAS is uniquely positioned to detect incoming viral DNA. Perhaps during infection with nuclear replicating viruses, viral or host factors can cause cGAS to become released from the chromatin to detect replicating viral DNA. Indeed, one study found that during infection with herpes simplex virus 1 (HSV-1) or adenovirus (AdV), a portion of cGAS accumulates in the nuclear soluble fraction rather than the chromatin-bound fraction, suggesting that some portion of chromatin-bound cGAS can become untethered during DNA virus infection⁴⁷. An alternative possibility is that cytosolic cGAS is recruited to the nucleus to detect replicating DNA virus.

Because of the high affinity between cGAS and the nucleosome acidic patch, any mechanism to activate nuclear cGAS must overcome this high affinity. In addition to direct displacement from the nucleosome, cGAS tethering may be regulated by post-translational modifications (PTMs) on either cGAS or the histones, which could induce conformational changes or charge differences that could weaken the interaction between cGAS and the nucleosome. Other proteins that bind to the acidic patch are known to be regulated by PTMs. For example, binding of RCC1 is regulated by methylation, while binding of HMGN2 is regulated by phosphorylation^51,52. cGAS, which is highly post-translationally modified⁵³, may be regulated in a similar fashion. Several studies have suggested a role for cGAS PTMs in regulating cGAS distribution and chromatin-binding. Phosphorylation of tyrosine 215 by B-lymphoid tyrosine kinase (BLK) was associated with cytosolic retention of cGAS, and nuclear translocation of cGAS following DNA damage or VEGF-A signaling was accompanied by dephosphorylation of cGAS at that residue^27,49. Overexpression of a Y215E mutant of cGAS impaired nuclear translocation following DNA damage²⁷. Conversely, methylation of cGAS at K362 by the methyltransferase SUV39H1 was shown to promote nuclear localization of cGAS and tighter binding to the nucleosome⁵⁴. Although no clear role has yet been ascribed to these PTMs during antiviral responses or autoimmunity, these examples suggest that post-translational modification of cGAS may be an important regulatory mechanism to control cGAS distribution and mobilization from the nucleosome. Nevertheless, many questions still remain about the role of PTMs in regulating cGAS localization. One such unexplored question is the role of histone PTMs in regulating cGAS tethering. Histones, especially histone tails, are also highly modified, and such modification may regulate the interaction between cGAS and the acidic patch. The H4 tail is known to bind to the acidic patch in a manner regulated by acetylation³⁸, with acetylation of key lysines on H4 inhibiting the acidic patch interaction. Deacetylation of the H4 tail would promote acidic patch binding, which could compete with cGAS or decrease the number of available acidic patches for cGAS to bind. Further research should examine the role of the H4 tail acetylation, and other post-translational modifications of both cGAS and histones, in regulating the cGAS/acidic patch interaction. Other factors, such as histone turnover, variant histones, and chromatin condensation, could play a role in modulating cGAS nuclear tethering as well.

To summarize, cGAS is both a cytosolic and nuclear protein, and in many cells the nuclear pool of cGAS represents the majority of total cGAS within the cell. Structural analyses show elegant negative regulation of cGAS activation through tethering to the nucleosomal acidic patch, in which the DNA-binding residues of cGAS are repurposed for binding to the acidic patch and cGAS dimerization is impossible. Mutations within the tethering surface render cGAS massively autoreactive, highlighting the importance of this negative regulatory mechanism. Future research must consider this pool of cGAS and work to dissect the different contributions of the cytosolic and nuclear pool of cGAS under different conditions and in response to different sources of immunostimulatory DNA.

cGAMP is an immunotransmitter that can act at a distance

The generation of cGAMP by activated cGAS results in its rapid diffusion throughout the cell and engagement of STING on the ER membrane. Because cGAMP is such a potent immunostimulatory molecule, the question arose of whether and how cGAMP is metabolized. Cyclic AMP (cAMP) and cyclic GMP (cGMP), two cyclic mononucleotides that regulate cell signaling, are both metabolized by intracellular phosphodiesterases that cleave the intramolecular bond to yield AMP and GMP respectively⁵⁵. It would make sense, therefore, if an analogous phosphodiesterase could cleave and inactivate cGAMP.

In 2014, Lingyin Li, Tim Mitchison, and colleagues identified the first enzyme that degrades cGAMP: ectonucleotide pyrophosphatase/phosphodiesterase 1 (ENPP1)⁵⁶. ENPP1 was already known to metabolize ATP into AMP and pyrophosphate, a process that is important for regulation of skeletal and tissue mineralization⁵⁷. Elegant biochemical and structural studies revealed that ENPP1 metabolizes cGAMP by cleaving both phosphodiester bonds, resulting in the release of GMP and AMP^58,59. However, the identification of ENPP1 as the principal cGAMP degrading enzyme raised a conundrum that baffled the field for years and slowed acceptance of ENPP1 as a relevant regulator of cGAMP signaling: ENPP1 is an extracellular enzyme, whereas cGAMP is produced and signals inside of cells. Thus, ENPP1 was not truly analogous to the intracellular cAMP and cGMP phosphodiesterases. cGAMP transfer between certain cell types was known to occur via movement through gap junctions, and the incorporation of cGAMP into viral particles was later described as a means to transfer cGAMP from cell to cell. However, neither of these mechanisms results in exposure of cGAMP to extracellular space. The simple topological question remained: how could an extracellular phosphodiesterase metabolize an intracellular cyclic dinucleotide?

Early studies of cGAMP activity pointed to a potential solution that was not fully appreciated at the time: injection of “naked” cGAMP into mice caused potent STING activation, and the simple addition of cGAMP to cell culture media was sufficient to activate STING signaling⁶⁰. Because nucleotides cannot passively cross the lipid bilayer of the cell membrane, these findings implied a mechanism of cGAMP import into cells. In 2019, two labs employed CRISPR screens to identify SLC19A1 as the first cGAMP importer^61,62. Shortly thereafter, two additional cGAMP importers were identified, SLC46A2 and LRRC8, that mediate cell type specific cGAMP import^63–65. Finally, another study found that the antimicrobial peptide LL-37 can mediate cGAMP transfer into cells⁶⁶.

The question of how cGAMP exits cells remained. Two studies demonstrated that tumor cells produce and release cGAMP that activates STING signaling in neighboring cells to stimulate anti-tumor immunity^67,68. In 2022, the ABCC1 transporter was identified as an ATP-dependent cGAMP exporter⁶⁹. ABCC1, also known as multidrug resistance protein 1 (MRP1), was already well known for its ability to export antibiotics, chemotherapeutic drugs like doxorubicin, and glutathione-conjugated macromolecules⁷⁰. Loss of ABCC1 potentiated STING signaling within cells and exacerbated disease in the Trex1 knockout mouse model of AGS, a disease that is caused by cGAS-STING signaling. Interestingly, ABCC1-deficient cells retained significant residual cGAMP export that was inhibited by the drug MK-571, which can block additional members of the ABCC exporter family⁶⁹. This demonstrates the existence of multiple cGAMP exporters that are perhaps structurally related, the identification of which will broaden our understanding of how cGAMP exits cells (Figure 2).

Figure 2. Localization of cGAS.

At homeostasis, two pools of cGAS reside within the cell. The majority of cGAS is localized within the nucleus where it is sequestered and held inactive on the nucleosome acidic patch. A second, smaller pool of cGAS is located within the cytosol and is competent to signal. cGAS can be activated by DNA from pathogenic sources (e.g. viral DNA) or by self-DNA released after mitochondrial or genomic DNA damage. Activation of cGAS stimulates production of the secondary messenger molecule cGAMP. Many open questions remain in the regulation of cGAS distribution, including whether cGAS can be mobilized from the nucleosome, whether cGAS is dynamically trafficked between the nucleus and the cytosol, and how histone modifications impact cGAS tethering on the nucleosome.

Together, the definition of mechanisms of cGAMP export, import, and extracellular degradation have led to the description of cGAMP as an “immunotransmitter” that can signal to distal cells (Figure 2). Conceptually, the biology of the cGAS-cGAMP-STING pathway can now be expanded to include a three-cell model: one cell can produce and export cGAMP, another cell can import cGAMP and signal through STING, and a third cell can contribute the ENPP1 that degrades extracellular cGAMP. Very recently, another ENPP1 paralog, ENPP3, was identified that can also degrade extracellular cGAMP. Interestingly, ENPP1 and ENPP3 have distinct tissue dependent expression patterns, suggesting they are not redundant enzymes⁷¹. Whereas ENPP1 and ENPP3 are expressed at steady state, a third inducible cGAMP degrading enzyme, sphingomyelin phosphodiesterase acid-like 3A (SMPDL3A), was recently identified⁷². SMPDL3A expression is undetectable under steady state conditions, but it is upregulated through activation of Liver X receptor (LXR), which has a well-known role in lipid metabolism⁷². Thus, the regulation of cGAMP as an immunotransmitter will undoubtedly be further refined in the near future.

cGAS detection of self-DNA

In the initial description of intracellular DNA sensing¹², it was postulated that the DNA sensor (identified later as cGAS) might aberrantly sense self-DNA, leading to autoimmunity. Potential sources of this self-DNA include multiple types of DNA damage, both endogenous and exogenous in origin⁷³. These include chromosomal instability, which results from defects in chromosome cohesion, mitotic checkpoint function, centrosome copy number, kinetochoremicrotubule attachment dynamics, and cell-cycle regulation⁷⁴. Another form of chromosome abnormality is micronuclei, which arise from chromosomes or chromosome fragments that fail to incorporate into daughter nuclei during telophase and recruit their own defective nuclear envelope⁷⁵. An abundant source of extranuclear self-DNA is contained within mitochondria; this mitochondrial DNA can escape into the cytosol in the context of mitochondrial dysfunction and damage^76,77. Additionally, neutrophil extracellular traps (NETs) are an abundant source of extracellular genomic DNA and have been show to activate cGAS after uptake by phagocytes⁷⁸.

In 2011, damage of genomic DNA was shown to induce IFN production in HeLa cells and primary human monocytes treated with etoposide, thus linking DNA damage and the innate immune response⁷⁹. It has since been shown that mitotic progression in cells following genotoxic cancer therapy leads to the formation of micronuclei⁸⁰ and chromatin bridges⁸¹, with an ISG signature dependent on the cGAS-STING pathway. Moreover, the cGAS-STING pathway is activated by mitochondrial dysfunction due to loss of apoptotic caspases^82,83, defective mitochondrial DNA condensation⁸⁴, or defects in mitophagy⁸⁵. Mitochondrial DNA damage and release have been implicated in cGAS responses in several infectious and hereditary diseases, including COVID19⁸⁶ and ALS⁸⁷. Z-DNA Binding Protein (ZBP1) was recently identified as a cooperative partner for cGAS sensing of mitochondrial DNA, sustaining and enhancing IFN production⁸⁸. Together, these and many other studies highlight the ability for cGAS to detect both genomic and mitochondrial DNA.

Perhaps the most compelling demonstrations of the pathological consequences of cGAS detection of self-DNA are the human type I interferonopathies, of which Aicardi-Goutieres Syndrome (AGS) is the most well studied^20,89. AGS is a monogenic disorder caused by mutations in any of nine genes that have been defined to date, seven of which have been demonstrated in mouse models to result in activation of the cGAS-STING pathway. First, the DNA exonuclease Trex1 degrades single stranded DNA and is particularly adept at resecting DNA from a nick or discontinuity in dsDNA⁹⁰. Trex1 was found to exhibit potent anti-retroviral activity against both endogenous retroelements and infectious HIV^91,92. Trex1 has also been shown to limit cGAS activity in micronuclei (MN) by resection of micronuclear DNA⁴⁵. The dNTP phosphohydrolase SAMHD1 limits cellular dNTP pools and is a potent restriction factor for HIV that is antagonized by the virus-encoded Vpx factor^93,94. The RNaseH2 enzyme, which is composed of three subunits, each of which can be mutated in AGS, is essential for the removal of misincorporated ribonucleotides that appear with a surprisingly high frequency during genomic DNA replication^95,96. The two most recently identified AGS genes, RNU7-1 and LSM11, form part of a ribonucleoprotein complex that is essential for the processing of mRNAs encoding replication-dependent histones⁹⁷. Fibroblasts from AGS patients harboring these mutations had normal levels of core histones, but the abundance of linker histone H1.4 was significantly reduced, which led to IFN production in a cGAS-STING-dependent manner. Intriguingly, nuclear cGAS in patient fibroblasts was enriched in the nuclear periphery and patient fibroblasts also had more misshapen nuclei, including the presence of nuclear herniations enriched for cGAS⁹⁷. These data suggest linker histones may regulate cGAS by either “shielding” genomic DNA from cGAS or by maintaining nuclear integrity.

Together, the AGS genes suggest potential sources of endogenous DNA that may contribute to cGAS activation. Two AGS genes encode potent antiretroviral enzymes (TREX1 and SAMHD1), implicating reverse-transcribed endogenous retroelements as a potential source of immunostimulatory DNA. Interestingly, treatment of Trex1−/− mice with reverse transcriptase inhibitors (RTIs) ameliorates disease⁹⁸, and treatment of TREX1-mutant human AGS patients with RTIs reduces the interferon signature in peripheral blood⁹⁹. Four of the AGS genes encode enzymes that act to prevent the formation of micronuclei (RNaseH2)¹⁰⁰ or metabolize damaged DNA within micronuclei (Trex1)⁴⁵. Finally, two AGS genes (LSM11, RNU7) contribute to the proper expression of histones, most notably linker histones⁹⁷. It is worth considering these processes in light of a predominant pool of nucleosome-tethered cGAS. Whereas the cytosolic DNA detection model would require the leakage or ejection of DNA fragments from the nucleus to engage cytosolic cGAS, the existence of a substantial pool of nuclear cGAS raises the strong possibility that each of these DNA abnormalities could be detected in situ, in the nucleus (or micronucleus) itself.

In addition to AGS, other diseases similarly provide insight into cGAS-STING biology. Bloom Syndrome is a rare genetic disease characterized by marked genomic instability resulting from a deficiency in the BLM RecQ-like helicase, a DNA helicase essential for maintaining genome integrity. Bloom Syndrome patient fibroblasts show an increase in micronuclei and an ISG signature dependent on the cGAS-STING pathway¹⁰¹. cGAS activation has also been found in cells from Ataxia-Telangectasia (A-T) patients that have mutations in the ATM kinase¹⁰², and cGAS activation has been implicated in neurodegeneration in a rat model of A-T¹⁰³.Together, the study of human diseases that result in inappropriate cGAS-STING activation, along with new insights into how cGAS can respond to genotoxic stress, reveal an emerging paradigm that places cGAS as a self-referential innate immune sensor positioned to detect perturbations in nuclear and mitochondrial DNA.

In addition to being activated downstream of DNA damage, the cGAS-STING pathway has also recently been implicated in regulating and maintaining DNA damage response (DDR) pathways. However, both the outcome of cGAS-STING signaling on DDR pathways and the mechanism by which these outcomes are produced remain controversial. Several studies have shown that signaling through the cGAS-STING pathway is required to prevent chromosomal instability^104–106. Conversely, several studies have pointed to a role for the cGAS-STING pathway in promoting, rather than suppressing, genomic instability through inhibition of homology-directed repair (HDR)^27,29,107. Further work is needed in order to elucidate the multifaceted role of cGAS in modulating DDR and future studies should examine more rigorously which pool of cGAS contributes to regulation of DDR. One possibility is that the nuclear pool and the cytosolic pool of cGAS differ in their regulation of DDR.

cGAS-STING in Senescence

Many recent studies have demonstrated a critical role for cGAS in promoting senescence and regulating inflammation in senescent and aging cells^25,108–118. Loss of cGAS led to enhanced proliferation and spontaneous immortalization in MEFs^25,108, as well as loss of senescence markers such as production of β-galactosidase and cytokines representative of the secretory-associated-senescence phenotype (SASP), suggesting cGAS is important for senescence induced by serial passage. cGAS was also required for senescence induced by DNA damage and oxidative stress, as well as oncogene-induced senescence^25,108,109. Initial reports suggested that STING was dispensable for senescence²⁵, however most studies have found that STING is required for cGAS-mediated senescence by activating IRF3 and NF-kB to upregulate the SASP^{108–114,116}.

Senescence is a hallmark of aging, and the cGAS-STING pathway plays a key role in regulating age-induced senescence and inflammation through several different mechanisms. Senescence is accompanied by a loss of Lamin B1 and a corresponding loss of nuclear integrity^119,120, potentially leading to activation of cGAS^25,108,109. Indeed, cells from patients with aging-related diseases including ataxia telangiectasia, Hutchinson-Gilford progeria (HGPS), and Werner syndrome have evidence for elevated cGAS-STING activation^{25,108,109,121,122}. Senescent cells also downregulate TREX1 and DNase2, leading to enhanced accumulation of cytosolic DNA in senescent cells¹²¹. In stromal cells and professional contractile cells, aging was also associated with a loss of YAP/TAZ activity and nuclear localization¹¹³. YAP and TAZ are transcriptional coactivators that regulate gene expression downstream of mechanical signaling¹²³. Loss of YAP/TAZ function led to a loss in nuclear integrity that promoted senescence and inflammation primarily through activation of the cGAS-STING pathway, with cGAS shown to accumulate at the nuclear border in complex with DNA¹¹³.

Aging is also associated with de-repression of endogenous retroviral elements. Two of these endogenous retroviral elements, Line elements and human endogenous retrovirus K (HERVK) have been shown to induce senescence and inflammaging in both human cells and mice in part by activation of the cGAS-STING pathway and SASP induction^{110–112,114,116}. In human cells harboring mutations that induce progeroid diseases including HGPS and Werner syndrome, upregulation of HERVK activated cGAS-STING signaling to drive senescence and inflammation¹¹⁶.

Mitochondrial dysfunction has also been linked to aging, with two recent studies showing that mitochondrial DNA (mtDNA) released into the cytosol in aged cells activates the cGAS-STING pathway to promote senescence and neurodegeneration^117,118. Mitochondrial outer membrane permeabilization (MOMP), a key feature of apoptosis, was found to occur only in a minority of mitochondria in senescent cells, leading to inflammation but not cell death¹¹⁸. This process, termed “minority MOMP” (miMOMP), required BAX and BAK, and drove expression of the SASP in senescent cells through activation of the cGAS-STING pathway. Treatment of senescent fibroblasts with the BAX inhibitor BAI1 reduced mtDNA release and IL-6 secretion, and treatment of aged mice with BAI1 reduced brain inflammation, ameliorated the age-related decline in neuromuscular coordination, improved bone density, and reduced SASP expression¹¹⁸. Similarly, aged mice treated with the STING inhibitor H-151 performed better in tests of strength and memory¹¹⁷. This was accompanied by reduced microgliosis and increased neuronal density in the hippocampus. cGAS-STING activation was suggested to be driven by mtDNA, as aged microglia had misshapen mitochondria and increased amounts of cytosolic mtDNA.

Together, these studies indicate a critical role for cGAS in senescence and inflammaging and point to the cGAS-STING pathway as a potential therapeutic target for progeroid and aging-related diseases. Future studies should investigate in greater depth whether cGAS^−/− and STING^−/− mice are protected from aging-related conditions and assess the translational potential of cGAS and STING inhibitors in treating progeroid and aging-related disorders. Furthermore, it remains unclear whether this function is unique to cGAS, or whether other PRRs may also contribute to senescence and inflammaging. One study found that senescent cells upregulated RIG-I, and that knockout of RIG-I or MAVS reduced expression and secretion of IL-6 and IL-8, two key SASP genes, in senescent cells¹²⁴. The intracellular form of Klotho, a protein identified as an anti-aging factor¹²⁵, was shown to block RIG-I signaling and suppress IL-6 and IL-8 secretion¹²⁴. Still, the contribution of RIG-I or MDA5 in promoting senescence in vivo remains unclear, and future studies using MAVS^−/− and MDA5^−/− mice will be needed to establish their relative importance. Nevertheless, it is intriguing to speculate that detection of a variety of mislocalized or aberrantly expressed nucleic acids by diverse PRRs may be a general feature of aging-related senescence and inflammation.

cGAS in cancer biology

Recent advances in immuno-oncology highlight the importance of the immune system in the prevention, control, and elimination of cancer. Many immunotherapies function by helping the body’s own immune system attack and eliminate tumor cells. Immune checkpoint blockade of PD-1/PD-L1 and CTLA4 has revolutionized cancer therapy by unleashing T cell responses to tumors. However, since activation of most adaptive immunity relies on innate immune responses¹²⁶, it is not surprising that a growing body of evidence highlights the critical role of the innate immune system in these anti-tumor responses.

Innate immune cells can directly detect and eliminate tumor cells, as well as induce and amplify adaptive immune cell responses. Activation and/or manipulation of the innate immune system has unlocked potential new avenues for anti-tumor immunotherapies. The cGAS-STING pathway is particularly well positioned to contribute to anti-tumor immunity: as described above, cGAS can be activated during key processes of oncogenesis by endogenous DNA damage, chromosomal instability, and micronuclei formation; and by DNA damage resulting from therapeutics including chemotherapy and radiation. Gajewski and colleagues demonstrated in 2014 that STING-deficient mice support accelerated tumor growth in numerous transplantable tumor models; this defective anti-tumor immunity was accompanied by weaker tumor-specific cytotoxic T cell responses¹²⁷. In other work, dendritic cells (DCs) were shown to present tumor antigens and generate anti-tumor CD8 + T cell responses in both humans and mice¹²⁸. It was further demonstrated that STING signaling in DCs themselves was required for effective tumor antigen cross-presentation and subsequent CD8+ T cell activation¹²⁹. Together, these findings highlight how the activation of the STING pathway in antigen presenting cells (APCs) contributes to activation of anti-tumor CD8+ T cells.

Natural Killer (NK) cells are another type of innate immune cell that can detect and kill tumor cells directly. Recently it was shown that their activation and effector function also depend on the activation of the STING pathway. Intratumoral injection of STING agonists in the form of cyclic-di-nucleotides (CDNs) causes NK cell-mediated killing of tumor cells that is independent of CD8+ T cell responses. DCs also contributed to NK anti-tumoral activity in vivo through their production of type I IFNs¹³⁰. Mice depleted of NK cells before tumor implantation were not able to control tumor growth, highlighting the importance of these cells in tumor rejection. Furthermore, in human melanoma patients there is a positive correlation between cGAS expression, NK cell receptor expression, and survival⁶⁷.

Interestingly, it has been shown that cGAMP produced by the tumor cells themselves is required to activate STING in non-tumor cells and to drive anti-tumor responses. cGAS-deficient mice control growth of implanted tumors as well as wild type mice. However, when tumor cells were deleted for cGAS before implantation into WT mice, tumor growth was no longer restricted⁶⁷. Further investigation found that many human and mouse cancer cells produce and export cGAMP at steady state, but this exported cGAMP could only be detected when ENPP1 was either genetically deleted or chemically inhibited^56,68. Mice deficient for ENPP1 or treated with a chemical ENPP1 inhibitor, showed greater immune activation and control of tumor burdens⁶⁸. Furthermore, ENPP3 activity can be tumor promoting in some settings, and mice expressing a hydrolase-deficient ENPP3 have fewer lung metastases in a model of melanoma. However, because both tumor cells and the surrounding TME can express different levels of ENPP1 and ENPP3, the contribution of either enzyme in anti-tumor immunity may depend on the tumor type⁷¹. cGAMP therefore serves as a potent immunotransmitter from tumor cells to the surrounding non-tumor cells. The recent identification of specific mechanisms of cGAMP export and import raises new questions about how these transporters contribute to anti-tumor immunity. It should also be noted that increased chromosomal instability is also associated with greater metastatic potential, and in some settings, chronic activation of the cGAS-STING pathway in tumors is associated with metastasis¹³¹. This suggests that outcomes of STING activation are more nuanced, potentially leading to both tumor restrictive and tumor promoting outcomes depending on context.

Innate immune activation through the cGAS-STING pathway is not only important for cancer therapies that cause DNA damage, but also plays a role in the anti-tumor effects of checkpoint blockade. In mice it was shown that cGAS and STING were required for the antitumor effects of anti-PD-L1 therapy, both in restricting tumor growth and recruitment of tumor specific CD8+ T cells¹²⁹. This suggests a direct interplay between these two pathways that could be leveraged to improve existing anti-tumor therapies.

Harnessing cGAS biology for therapeutic benefit

Ample human genetic evidence and mouse models have connected the cGAS-STING pathway to numerous human immune diseases, including AGS, familial chilblain lupus (FCL), STING-associated vasculopathy with onset in infancy (SAVI), and some cases of systemic lupus erythematosus (SLE). STING signaling has also been implicated in mouse models of numerous diseases, including Parkinson’s disease, Amyotrophic Lateral Sclerosis (ALS), and Niemann-Pick disease. Conversely, our emerging understand of the anti-cancer functions of cGAS-STING has motivated interest in triggering this axis for cancer therapy. Immense resources in biotech and pharma have been invested in the development of two classes of drugs: cGAS-STING inhibitors, and STING agonists. These efforts have been reviewed recently¹³²; we will summarize them here and consider how our emerging understanding of the biology of this pathway may point to new avenues and applications.

cGAS and STING inhibitors have been reported and are advancing in clinical development¹³². Obvious applications of such inhibitors include AGS, FCL, SLE, and SAVI. Importantly, the disease mechanisms might warrant the selective use of such antagonists. In particular, mouse models of Niemann-Pick disease have demonstrated that STING activation can occur in the absence of cGAS¹³³; such mutants would benefit from STING inhibitors but not cGAS inhibitors.

More broadly, the emerging contributions of cGAS-STING to chronic conditions associated with aging warrant consideration of the possibility that low-dose, long term STING inhibition might contribute to preserving healthy life span. Just as low dose aspirin is widely used to reduce risk of cardiovascular disease, could STING inhibition broadly improve tissue function and ameliorate inflammaging? At this point, this is more of a rhetorical question, and only the development of safe, inexpensive, orally bioavailable STING inhibitors would enable such a possibility.

In immuno-oncology, STING agonists have shown considerable promise in mouse models for their ability to stimulate inflammation in tumors and enhance antigen presentation and cytotoxic T cell responses^134,135. Clinical trials of these STING agonists have been performed in human cancer patients, based on the idea that introduction of a STING agonist might turn immunologically “cold” tumors “hot.” So far, these clinical trials have yielded disappointing results, with a general failure of efficacy either as a monotherapy or in combination with checkpoint blockade¹³⁶. Although the reasons underlying these outcomes remain unclear, three possibilities may contribute. First, dose-limiting toxicity might prevent the administration of a therapeutic dose of STING agonist. Second, because the efficacy of STING agonists in mouse models requires STING signaling in the host and not in the tumor cells, it is possible that treating “cold” tumors might result in failure to activate sufficient immune cells because they are not present in the tumor at the time of treatment. Third, STING signaling causes toxicity in T cells, which may cripple the very cells that are important for immunotherapy.

Based on the emerging paradigms of nuclear cGAS and the immunotransmitter function of cGAMP, an alternative approach to boost cGAS-STING signaling in tumors could be explored in which the endogenous signaling pathway can be optimized and amplified. Such an approach would require four elements. First, optimizing the activation of cGAS in tumor cells to maximize intratumoral cGAMP production. This will require an understanding of how cGAS activation is restrained in the context of chemotherapy and radiation. A simple way to bypass such restraint would be to introduce constitutively active cGAS into tumors¹³⁷. Second, enhancing cGAMP export from tumor cells, in part by identifying those tumors with high ABCC1 expression that closely correlates with efficiency of cGAMP export⁶⁹. Third, promoting cGAMP import into the relevant immune cells that contribute to anti-tumor immunity while perhaps avoiding cGAMP import into T cells that may restrain their function²³. Finally, preventing degradation of extracellular cGAMP by inhibiting ENPP1¹³⁸, ENPP3, and perhaps SMPDL3A. While this approach is currently aspirational, a first step would be to leverage the deep data sets and single cell atlases of gene expression within tumors to identify those with tumor cells that express the most cGAS and ABCC1, those that contain the highest frequency of immune cells that are important for the response to cGAMP, and those that would be most responsive to inhibition of ENPP1, ENPP3, and SMPDL3A. Just as tumor genomics and expression profiles now contribute to the decision of whether to administer checkpoint blockade, we predict that an analogous set of criteria focusing on cGAS-STING biology might inform how to enhance endogenous activation of this pathway in tumors.

Concluding thoughts

The cGAS-STING DNA sensing pathway has emerged as a central mediator of host defense, a cause of numerous human immune diseases, a contributor to fundamental pathological processes of senescence and aging, and a new avenue for triggering inflammation in tumors to improve anti-cancer immunity. All of these myriad roles for cGAS can be attributed to its ability to be activated by any double stranded DNA of any origin. The study of the high potential for self-reactivity of cGAS has revealed elegant new regulatory mechanisms that restrain cGAS and touch on broad swaths cell biology. Thus, a new frontier of cGAS-STING biology will be to leverage our understanding of this cGAS regulation to develop new ways to treat a variety of human diseases, from autoimmunity to inflammaging to cancer.

Figure 3. cGAMP signaling and regulation.

  cGAMP can act in both an autocrine and paracrine manner when activating STING.  cGAMP is exported from cells by ABCC1 and other unknown transporter(s). Export can be chemically inhibited by MK-571, which acts on several ATP-binding cassette (ABC) transporters including ABCC1. cGAMP import then occurs through several known importers, (SLC19A1, SLC46A2, and LRRC8), thus allowing cGAMP to act as an immunotransmitter. Concentrations of extracellular cGAMP can be regulated via degradation by both membrane-bound ENPP1 and ENPP3, or soluble ENPP1. SMPDL3A, another cGAMP degrading enzyme, may act both intra and extracellularly.