Control of innate immunity by the cGAS-STING pathway

Abstract

Within the cytoplasm of mammalian cells is a protein called cyclic GMP-AMP synthase (cGAS), which acts to defend against infection and other threats to the host. cGAS operates in this manner through its ability to detect a molecular occurrence that should not exist in healthy cells—the existence of DNA in the cytosol. Upon DNA binding, cGAS synthesizes cyclic GMP-AMP (cGAMP), a cyclic dinucleotide that activates the endoplasmic reticulum-localized protein stimulator of interferon genes (STING). STING-mediated signaling culminates in host defensive responses typified by inflammatory cytokine and interferon expression, and the induction of autophagy. Studies over the last several years have established a consensus in the field of the enzymatic activities of cGAS in vitro, as it relates to DNA-induced production of cGAMP. However, much additional work is needed to understand the regulation of cGAS functions within cells, where multiple sources of DNA can create a problem of self and non-self discrimination. In this review, we provide an overview of how the cGAS-STING pathway mediates innate immune responses during infection and other cellular stresses. We then highlight the recent progress in the understanding of the increasingly diverse ways that this DNA-sensing machinery is regulated inside cells, including how cGAS remains inactive to host-derived DNA under conditions of homeostasis.

Graphical Abstract

The cGAS-STING pathway plays a central role in health and disease. In this review, we describe the current understanding of the molecular mechanisms that control activity of this pathway.

Introduction

Cells are equipped with protein-based machineries that link the detection of microorganisms to innate immune responses that prevent pathogen infections¹. Central to the function of these host-defensive molecular machines are members of the pattern recognition receptor (PRR) superfamily. PRRs recognize conserved microbial molecules that are known as pathogen-associated molecular patterns (PAMPs), with the most common being nucleic acids and microbial cell-wall components¹. In the case of DNA sensing, several proteins have been identified as PRRs².

The endosome-localized Toll-like receptor 9 (TLR9) is responsible for the detection of unmethylated CpG-containing DNA that entered cells from the extracellular space³. Among intracellular DNA sensors are absent in melanoma 2 (AIM2), which induces inflammasome activation in response to cytosolic double-stranded (ds) DNA⁴, and the enzyme cyclic GMP-AMP synthase (cGAS), which recognizes B-form DNA independent of its sequence through contacts with the sugar-phosphate backbone^{5, 6}. Upon dsDNA binding, cGAS produces 2’3’-cyclic-GMP-AMP (2’3’-cGAMP, hereafter cGAMP), a second messenger molecule that binds to the protein stimulator of interferon genes (STING, also known as MITA, MPYS, and ERIS) to induce innate immune responses^7–9. These cGAS-STING-mediated responses include the transcriptional induction of antiviral cytokines such as type I interferon (IFN) and proinflammatory cytokines, as well as the induction of autophagy and cell death. Due to its lack of sequence-specificity¹⁰, cGAS has the capacity to recognize dsDNA from any source, including that from viruses, bacteria and endogenous host-derived DNA. cGAS deficiencies in cells and mice are associated with alterations in a variety of infectious and non-infectious diseases, with the latter including cancer and autoinflammatory disorders¹¹. As such, there is an increasing interest in understanding the regulators of cGAS-STING signaling events in natural and therapeutic settings.

In this review, we will first overview the cGAS-STING signaling pathway and the consequent cellular responses, with a focus on the molecular mechanisms by which cGAS is activated by DNA. We will further discuss how cGAS activation is regulated within the cells by subcellular compartmentalization and post-translational modifications, which helps us understand how the endogenous DNA escapes cGAS detection.

The Overview of cGAS-STING Signaling

cGAS is an enzyme that contains latent nucleotidyltransferase (NTase) activity. In vitro studies have demonstrated that binding to dsDNA or RNA-DNA hybrids stimulates cGAS enzymatic activity, resulting in the synthesis of cGAMP from ATP and GTP⁵ (Figure 1). Within cells, cGAMP serves as a second messenger that binds to the endoplasmic reticulum (ER)-resident protein STING. STING also has the ability to detect bacteria-derived cyclic dinucleotides (e.g. cyclic di-AMP, cyclic di-GMP, and 3’3’-cGAMP)^12–15. Binding to cGAMP or bacterial cyclic dinucleotides triggers conformational changes in STING, followed by its trafficking from the ER to the ER-Golgi intermediate compartment (ERGIC) and the Golgi apparatus^16–18. As in the case of many other ER proteins, STING trafficking is mediated by vesicles formed by the GTPase SAR1 and the COPII complex¹⁸. Upon reaching at the ERGIC and Golgi, STING is palmitoylated and oligomerizes into a signaling platform that recruits downstream kinases and other molecular complexes to initiate the cytokine transcription and other innate immune responses^{19, 20}, which are described in detail below.

Figure 1. Overview of cGAS activation and cGAS-STING signaling.

cGAS binds to both pathogen-derived and host organelles-derived DNA in the cytosol and micronuclei, which then forms cGAS-DNA 2:2 complexes. These complexes undergo phase separation to further assemble into larger oligomers with liquid-like biophysical properties. These cGAS-DNA condensates allow efficient catalytic action of cGAS, which turns ATP and GTP into cGAMP. cGAMP binds to STING on the ER membrane, which in turn traffics to the ERGIC/Golgi and oligomerizes. Activated STING recruits TBK1 that phosphorylates STING at its C-terminal tail (CTT). IRF3 is recruited to the pLxIS motif among the phosphorylated residues at the CTT of STING, then gets phosphorylated by TBK1. Phosphorylated IRF3 dimers and translocates to the nucleus, where it induces the transcription of type I IFN and proinflammatory cytokines in collaboration with NF-κB, which is also activated by TBK1 and IKKε downstream of STING.

Transcriptional and non-transcriptional host defensive responses induced by cGAS-STING

Within the Golgi apparatus, palmitoylated STING oligomers interact with the kinase TANK-binding kinase 1 (TBK1)¹⁹. This interaction between STING and TBK1 dimer is mediated by the PLPLRT/SD domain present in the cytoplasmic C-terminal tail (CTT) of STING, leading to TBK1 auto-phosphorylation and activation^{21, 22}. Activated TBK1 phosphorylates STING at Serine 366 (S366, S365 in mice) within a sequence also present in its C-terminus called a pLxIS motif (p, hydrophilic residue; x, any residue; S, phosphorylation site)^{22, 23}. This TBK1-mediated phosphorylation of STING is dependent on the oligomerization, as the TBK1 protein cannot phosphorylate the STING dimer it is bound to due to the large space between the TBK1 active site and S366 of directly bound STING molecule²². The transcription factor interferon regulatory factor 3 (IRF3) is then recruited to the phosphorylated pLxIS motif, where is in turn phosphorylated by TBK1^{22, 23}. Activated IRF3 then dimerizes and translocates to the nucleus, where it induces the transcription of type I IFN genes in collaboration with NF-κB²⁴. Type I IFN then elicits an antiviral state of cells in an autocrine or paracrine manner, through binding to its receptor IFN-α/β receptor (IFNAR)²⁵. The IFNAR1-IFNAR2 complex activates the Janus Kinase (JAK)-signal transducer and activator of transcription (STAT) pathway, leading to the transcriptional induction of interferon-stimulated genes (ISGs) that target multiple stages in the viral life cycle²⁵.

As mentioned above, NF-κB is another transcription factor that is activated downstream of STING. Although the activation of NF-κB is dependent on TBK1 and IKKε, it does not require IRF3 activation²⁶. In addition to synergizing with IRF3 to induce type I IFN, NF-κB induces the transcription of inflammatory cytokines and chemokines. Mouse STING mutants that contain a defective pLxIS motif (STING-S365A), which lack the IRF3 binding site and do not activate IRF3, revealed that STING induces IRF3- and IFN-independent antiviral responses^27–29. These IRF3-independent factors that are upregulated during these responses include CXCL1, CXCL2, 4-BBL1, and COX2, whose expression is dependent on NF-κB²⁷. Moreover, STING-mediated NF-κB activation in dendritic cells upregulates the surface expression of major histocompatibility complex II (MHC-II) and costimulatory molecules such as CD80 and CD86²⁷. These events are considered hallmark activities needed to stimulate T cell-mediated inflammatory responses³⁰. The IRF3-independent transcriptional responses are important for host defense, as the STING-S365A mice are resistant to herpes simplex virus 1 (HSV-1) infection^27–29 or develop polyarthritis in DNA clearance-defective DNase II^−/− mice even in the absence of IRF3 activation³¹.

In addition to the transcription of type I IFN, cGAS-STING signaling has been implicated in other innate immune responses. Autophagy is one such non-transcriptional consequences of STING activation¹⁸. Evolutionarily more ancient than the IFN genes, autophagy is a cellular process that was first described to maintain cell homeostasis under conditions of starvation³². Under these conditions, intracellular contents are encapsulated into membrane-bound structures called autophagosomes, which are delivered to lysosomes and degraded. The degraded materials can then be recycled for use by the cell as a means of self-sustenance. Autophagy also operates as a host-defense mechanisms to encapsulate cytosol-localized pathogens, leading to their eventual delivery to and destruction within lysosomes³³. During infections with the DNA virus HSV-1 or Mycobacterium tuberculosis (Mtb), DNA from these pathogens can activate cGAS^34–37. The resulting cGAMP production leads to STING-dependent autophagy of the pathogens and lysosomal degradation^{35, 38}. Mechanistically, the STING-containing ERGIC serves as a source of the hallmark processes in autophagosome formation—the lipidation of the protein LC3¹⁸. LC3 lipidation is dependent on autophagy protein 5 (ATG5) and WD repeat domain phosphoinositide-interacting protein 2 (WIPI2) but not on Unc51-like autophagy activating kinase 1 (ULK1), ULK2, or Beclin 1, the components in the conventional autophagy¹⁸. A STING mutant lacking its C-terminal activation domain still triggers autophagy, suggesting that STING induces autophagy in a IRF3-independent manner¹⁸. STING-induced autophagy also functions as negative feedback of cGAS-STING pathway, as TBK1 activates p62/SQSTM1-dependent autophagy targets STING for lysosomal degradation³⁹. Because STING orthologs in Xenopus tropicalis and the sea anemone Nematostella vectensis induce autophagy without IRF3 activation or I IFN transcription¹⁸, autophagy induction is thought to be the primordial function of cGAS-STING signaling.

Overall, cGAS-STING pathway induces IFN and other cytokine responses as well as cytokine-independent autophagy, all of which play significant roles in restricting pathogens.

cGAS-STING activities that impact cell viability and mitosis

cGAS-STING mediated induction of diverse cell death pathways

In addition to the aforementioned activities that mediate host defensive type I IFN and autophagy responses, recent studies have revealed cGAS-STING activities that impact cells in unexpectedly diverse manners. For example, several instances of STING-mediated induction of programed cell death have been reported^40–49 (Table 1). T cells that experience intense STING activation trigger type I IFN-independent apoptosis by inducing the transcription of BH3-only proteins and other proapoptotic genes through the activities of IRF3 and p53⁴⁰. STING also induces transcription-independent T cell apoptosis by disrupting calcium homeostasis and therefore sensitizing cells to ER stress and the unfolded protein responses (UPR), as revealed by a study using an autoinflammatory disease-associated STING mutant⁴¹. A gain-of-function STING mutant N154S (N153S in mouse) develops lung inflammation, myeloid cell expansion, and T cell cytopenia in IRF3-lacking mice⁴², and the T cell death caused by this mutant was abrogated by the ER stress inhibitors⁴¹. More recent data have indicated that STING signaling renders tumor cells more sensitive to apoptosis induced by reactive oxygen species (ROS)⁴³. This process may be mediated by the actions of ROS-metabolizing ISGs⁴³, which impact the extent of cellular damage that can be inflicted by oxidation.

Table 1.

cGAS-STING-dependent cell death. Summary of cGAS-STING-dependent cell death described in the main text.

Cell Death	Cell Types	Stimulation / Model	References
Apoptosis	T cells	Small molecule STING agonist	38
		STING gain-of-function mutant (N154S in human, N153AS in mice)	39, 40
	HNSCC cells	DNA-damaging agent, radiation	41
Necroptosis	L929 fibroblasts	Sendai virus (SeV), murine gammaherpesvirus-68 (MHV68)	42
	HT29 colon cancer cells	SMAC mimetic (LBW-242) + pan-caspase inhibitor (z-VAD-fmk)	43
	Mice (in vivo)	Ischemia reperfusion (I/R) injury	44
Pyroptosis	BMDMs	Chlamydia trachomatis, Francisella novicida	45, 46
	BLaER human monocytes	Horse testis (HT)-DNA lipofection	47

HNSCC, Head and neck squamous cell carcinomas; BMDM, bone marrow-derived macrophage; BLaER, tamoxifen-inducible derivative of the RCH-ACV B-cell leukemia cell line.

In addition to apoptosis, cGAS-STING signaling promotes other forms of cell death. STING activation induces type I IFN and TNFα secretion. These cytokines can act synergistically to stimulate the necroptosis-inducing kinases receptor interacting kinase 1 (RIPK1)- and RIPK3 in neighboring cells⁴⁴. STING-dependent necroptosis is not only induced by DNA viruses such as murine gammaherpesvirus 68 (MHV68), but also by mitochondrial DNA (mtDNA) that has been released into the cytosol^{44, 45}. Indeed, a mtDNA-STING-cell death pathway is required for the ischemia reperfusion (I/R) injury-induced intestinal barrier disruption in vivo⁴⁶.

Pyroptosis, the inflammatory cell death commonly caused by the actions of inflammasomes⁵⁰, is also positively regulated by cGAS-STING activation. cGAS-STING signaling promotes inflammasome assembly and activation in several experimental contexts, resulting in pyroptosis through type I IFN-dependent and -independent mechanisms. Type I IFN upregulates the expression of ISGs such as guanylate-binding proteins (GBPs) and immunity-related GTPases (IRGs)⁵¹. These factors can rupture cytosolic bacterial membranes to expose bacterial DNA to the protein AIM2. DNA-bound AIM2 then seeds the assembly of an inflammasome that stimulates pyroptosis⁵². Type I IFN also has been reported to induce the expression of the bacterial LPS receptor caspase-11, which can stimulate the NOD-, LRR- and pyrin domain-containing protein 3 (NLRP3) inflammasome by Gram-negative bacteria⁵³. Indeed, cGAS and STING are required for cell death in murine macrophages that occurs during infections with Chlamydia trachomatis and Francisella novicida^{47, 48}. A study using BLaER1 human monocytes revealed that cytosolic DNA detection by cGAS causes the translocation of STING to lysosomes, where STING induces cell death via lysosomal rupture⁴⁹. This lysosomal cell death (LCD) elicits pyroptosis by potassium (K⁺) efflux-induced NLRP3 activation.

Altogether, activation of cGAS and STING leads to various forms of cell death in both pathogenic and sterile inflammation, with remaining unclarities as to how the cells determine which forms of cell death to induce in a certain context.

cGAS-STING mediated DNA repair and Cellular Senescence

DNA repair machineries, including homologous recombination (HR) and non-homologous end joining (NHEJ), are crucial for genomic stability and cell viability but also must be tightly regulated, because excessive DNA repair can lead to undesired chromosomal rearrangement⁵⁴. cGAS has been implicated in the regulation of DNA repair, either as a positive and negative regulator^55–57. Negative regulation of HR is mediated by chromatin-bound cGAS in the nucleus⁵⁶. DNA-bound cGAS oligomerizes and therefore compacts DNA into higher-order state, which is resistant to RAD51 recombinase-mediated strand invasion⁵⁶. This STING-independent role of cGAS accelerates the generation of micronuclei and cell death under severe genomic injury caused by irradiation. On the contrary, recently reported positive regulation of DNA repair is mediated by cytosolic DNA-bound cGAS⁵⁷. Upon exposure of cells to ionizing radiation, DNA damage-induced cytosolic DNA leakage stimulates cGAS-STING signaling, leading to TBK1-mediated phosphorylation of phosphoribosyl pyrophosphate synthetases 1 and 2 (PRPS1/2)⁵⁷. Phosphorylated PRPS1 and 2, in turn, promote DNA synthesis and repair. These opposing roles of cGAS in DNA repair indicate that cGAS functions in a context-dependent fashion, including cell types and the magnitude of DNA damages, which warrants future investigations.

Cellular senescence defines the irreversible cell cycle arrest caused by various cellular and environmental stresses, including inflammation and aging⁵⁸. In the case of DNA damage, the cGAS-STING pathway induces a senescence-associated secretory phenotype (SASP), the hallmark phenotype of senescence in which cells secrete a variety of proteins including inflammatory cytokines, chemokines, growth factors, and proteases^59–61. SASP is not only the hallmark of senescence, but is also an amplifier of this process. The requirement for cGAS-STING in cellular senescence has been revealed by both in vitro and in vivo studies using ionizing radiation and other DNA-damaging agents^59–61. In response to these agents, wild type cells arrest their growth rates, whereas cGAS-deficient cells continue to proliferate and do not exhibit SASP. The anti-proliferative roles of cGAS-STING may serve as anti-cancer mechanisms, as cGAS expression levels positively correlate with the survival of human lung adenocarcinoma patients⁵⁹.

In summary, cGAS-STING signaling upon activation induces not only cytokine responses but also other consequences affecting cellular homeostasis. Below, we will address the types of DNA ligands that activate cGAS and STING.

Sources of substrate DNA that activate cGAS-STING signaling

Exogenous DNA as an IFN-inducing cGAS substrate

As expected from the role of cGAS in the innate immune responses, microbial DNA serves as an IFN-inducing cGAS ligand (Figure 1). Several DNA viruses carrying dsDNA activate cGAS-STING signaling in in vitro and in vivo studies. These DNA viruses include adenovirus⁶², vaccinia virus (VACV)³⁴, and African swine fever viruses (ASFV)⁶³, and herpesviruses such as cytomegalovirus (CMV)^{64, 65}, MHV68³⁴, and HSV-1⁶⁶. cGAS-STING pathway has also been implicated in the cellular responses to some RNA viruses. RNA retroviruses such as human immunodeficiency virus 1 (HIV-1), murine leukemia virus (MLV), Simian immunodeficiency virus (SIV), and human T lymphotropic virus type 1 (HTLV-1), activate cGAS by providing cytosolic DNA as reverse transcription intermediates (RTIs)⁶⁷. In addition to those retroviruses, cGAS-STING signaling is involved in the responses to RNA viruses such as West Nile virus (WNV) and dengue virus, although cGAS senses the mtDNA leaked during viral infection, not the viral nucleic acids^{34, 68}.

DNA from bacteria also activate cGAS-STING signaling. Microbial DNA released from extracellular bacteria such as Pseudomonas aeruginosa, Klebsiella pneumoniae, and Staphylococcus aureus, and intracellular bacteria such as Listeria monocytogenes, Francisella spp., Neisseria gonorrhoeae, Mtb, and Rickettsia parkeri bind to cGAS to elicit type I IFN responses^{37, 52, 69–72}. While these DNA-induced IFN responses are beneficial to some facultative bacteria such as Listeria and Francisella, as shown in vivo^{71, 73}, obligate bacteria such as Rickettsia are sensitive to IFN-mediated killing, and cytosolic DNA released from lysed Rickettsia subpopulation induces inflammasome-mediated host cell death that masks cGAS-STING-dependent IFN production to protect the remaining population⁵². Some bacteria utilize secretion systems and toxins to destabilize phagosomal membranes and release their contents, including DNA, into the cytosol. For example, Mtb uses its ESX-1 secretion system, a subtype of Type VII Secretion System (T7SS), to deliver virulence factors into macrophages³⁷. These secretion systems are important for cGAS activation, as their mutations lead to weak type I IFN responses during infection³⁷.

In addition to DNA, some bacteria produce cyclic dinucleotides (CDNs) that directly activate STING without stimulating cGAS. c-di-GMP is synthesized from two GTP monomers by the act of bacterial diguanylate cyclase, and regulates bacterial metabolism, virulence, and biofilm formation⁷⁴. STING recognizes c-di-GMP to restrict bacterial growth by inducing type I IFN and other cytokine responses¹⁵. Similarly, c-di-AMP produced by bacteria such as Chlamydia trachomatis stimulates STING⁷⁵. However, during an infection, most of these bacteria also release DNA that is sensed by cGAS. The relative contribution of bacterial DNA and CDN to host inflammatory responses is largely undefined.

Lastly, cGAS has been implicated in the defense against parasites and fungi. For example, cGAS responds to the genomic DNA of Plasmodium falciparum, and cGAS-deficient mice permit higher parasite burden after infection than wild type mice⁷⁶. The cGAS-STING pathway also contributes to the type I IFN responses in Aspergillus fumigatus induced keratitis, although whether cGAS directly binds to fungal DNA is unclear⁷⁷.

Self-DNA as an IFN-inducing cGAS substrate

In addition to DNA from microbes, self-DNA can access the cytoplasm and activate cGAS (Figure 1). During mitosis within cells with the genomic instability, which is a hallmark of many cancers⁷⁸, DNA may segregate from the main chromosome to form perinuclear compartments called micronuclei. Nuclear envelope (NE) rupture after mitosis makes the DNA in micronuclei accessible to the cGAS immunosurveillance, therefore leading to the production of type I IFN responses^79–81. This DNA damage-induced exposure of nuclear DNA to cGAS in the cytosol also occurs during senescence, anti-mitotic chemotherapy, and dysregulation of epigenetic modification such as DNA hypomethylation that leads to skin inflammation^{59, 60}. Recently, however, Flynn et al. have proposed that chromatin bridges rather than micronuclei are the platforms of cGAS activation during mitotic errors, as IFN-inducing abilities of the antimitotic drugs correlated with the generation of cGAS-coated chromatin bridges, not micronuclei⁸². These findings provide important insights that not all DNA damages activate cGAS in a same manner, and the source of cGAS ligand depends on the context of DNA damages. Cytosolic DNA, when not optimally metabolized, stimulates the cGAS-STING pathway. Loss-of-function mutations in the genes encoding nucleases that degrade cytosolic DNA, such as three prime repair exonuclease 1 (TREX1), lead to cGAS-STING-dependent production of ISGs, which can lead to the development of an autoimmune disease called Aicardi–Goutières syndrome (AGS)⁸³.

As mentioned in the section above, mtDNA is also a cGAS ligand. The perturbation of mitochondrial function either by infectious or non-infectious stimuli leads to mitochondrial damages that release mtDNA into the cytosol. RNA viruses such as measles virus (MeV), which do not possess cGAS ligands, induce mtDNA-dependent cGAS activation in addition to activating RNA sensing pathways⁸⁴. During intrinsic apoptosis, Bax/Bak-mediated mitochondrial outer membrane permeabilization (MOMP) triggers the cytosolic release of mtDNA that activates cGAS and STING, which is inhibited by apoptotic caspases (caspase-9 and caspase-3/7) that are downstream of MOMP^{85, 86}. The mechanism by which these caspases inhibit cGAS-STING activation is likely via the cleavage of cGAS, as described later⁸⁷. Recent studies have found that inflammatory cytokines such as TNFα and IL-1β treatment leads to mtDNA release into cytosol, which triggers cGAS-dependent type I IFN responses^{88, 89}. Also, even in the absence of exogenous stimuli, heterozygous knockout of the mtDNA-binding protein transcription factor A, mitochondrial (TFAM) in murine cells leads to the cytosolic release of mtDNA, resulting in the production of ISGs that made cells protective against HSV-1 infection⁹⁰.

Extracellular self-DNA can also stimulate cGAS. Macrophages and other myeloid cells internalize self-DNA from the extracellular space through phagocytosis. In principle, phagocytosed cargo should be degraded in lysosomes, with luminal DNA being hydrolyzed down to single nucleotides by DNases present in the intra-lysosomal environment. Early studies, prior to the discovery of cGAS or STING, revealed that mice and cells deficient in the lysosomal enzyme DNase II are prone to express high levels of IFNs and other ISGs⁹¹. Subsequent work revealed that STING pathway activation was responsible for these responses⁹². These results suggested that in the absence of efficient lysosomal nuclease activity, DNA can somehow leak into the cytosol and activate cGAS-STING dependent responses.

Additional instances of extracellular DNA activating cGAS-STING have been reported. For example, non-apoptotic cell death caused by tissue injury and tumor irradiation can release DNA into the extracellular space, as can the production of DNA-containing neutrophil extracellular traps (NETs). NETs are released by neutrophils during a cell death process called NETosis, and consist of chromatin and antimicrobial molecules⁹³. The abundance of extracellular DNA at the sites of inflammation or tissue injury may result in a phenocopy of DNase II deficiency. In this regard, phagocytes that internalize NETs or chromosomal fragments released at sites of irradiation may contain such high amounts of lysosomal DNA that DNase II is overwhelmed, resulting in DNA leakage in the cytosol and the stimulation of cGAS⁹⁴. Consistent with this idea, myeloid cells are the primary source of cGAS-mediated IFN responses at sites of tissue injury⁹⁴. Overall, there are several sources of DNA that can activate the cGAS-STING pathway. This promiscuity of DNA detection raises the question of how this inflammatory network is regulated to ensure efficient distinction between self and non-self DNA. In the next sections, we discuss our current knowledge of how DNA-induced cGAMP synthesis by cGAS is regulated.

The biophysical mechanisms of cGAS activation and cGAMP synthesis

cGAS is a 522 amino-acid protein that consists of a basic unstructured N-terminal domain, and a conserved C-terminal Mab21 domain with nucleotidyltransferase (NTase) activity. While cGAS monomers cannot synthesize cGAMP, cGAS dimers formed upon interactions with the sugar phosphate backbone of dsDNA display enzymatic activity. Binding of a parallel-aligned pair of dsDNA to the two DNA-binding sites (A-site and B-site) on each of two cGAS molecules stabilizes 2:2 cGAS-DNA complex, accompanied with the cGAS conformational changes that generate ATP and GTP pockets near the catalytic residues^{95, 96}. The cGAS dimerization and conformational changes occur independently on the sequence of dsDNA, but neither single stranded DNA (ssDNA) nor RNA induces cGAS conformational changes and subsequent cGAMP synthesis although they can bind to cGAS^{6, 10}. The cGAS dimers further align adjacently to form a ladder-like structure⁹⁷. In addition, a newly discovered DNA binding site in human cGAS (C-site) allows multivalent interaction between cGAS and DNA to form the mesh-like structure⁹⁸. The formation of these higher order cGAS-DNA complexes requires the certain length of DNA (> 45 bp), and the shorter DNA does not lead to human cGAS oligomerization and efficient cGAMP production⁹⁹. Mouse cGAS, on the other hand, can oligomerize upon binding to short DNA and produce cGAMP, as the amino acids in the DNA binding sites allow more stable interaction with DNA⁹⁹. Indeed, when the residues at this position are substituted for those in mouse cGAS, human cGAS gains the ability to synthesize cGAMP in response to short DNA⁹⁹.

Recently, it has been discovered that cGAS-DNA complexes further assemble into micrometer-sized condensed liquid droplets through the process called phase separation¹⁰⁰. These higher-order complexes have liquid-like properties—the droplet formation is reversible, and the droplets can fuse to form larger spherical units. These membrane-less granules consisting of cGAS and DNA provide a platform for cGAMP production, where cGAS encounters its substrates more efficiently. This liquid droplet formation is dependent on the concentration of cGAS and DNA¹⁰⁰, suggesting that cGAS is activated only when cytosolic DNA levels reach a certain threshold, such as in the case of infection or cellular stresses with DNA damage. Moreover, under the physiological salt concentration, the formation of cGAS liquid droplets is dependent on intracellular ion concentration, mostly zinc¹⁰⁰. Indeed, addition of zinc (Zn²⁺) to cultured cells has been reported to activate cGAS to induce type I IFN responses in the presence of dsDNA treatment¹⁰¹. Zn²⁺ is not the only ion that contributes to cGAS activation. Divalent cations such as magnesium (Mg²⁺) and manganese (Mn²⁺) are required as cofactors of cGAS that catalyze the conversion of ATP and GTP into cGAMP¹⁴. Mn²⁺ directly activates cGAS even in the absence of dsDNA by inducing the noncanonical conformational changes of cGAS to mimic DNA-activated cGAS, which has a unique η1 helix in the catalytic pocket allowing substrate entry and cGAMP synthesis¹⁰¹.

Regulation of DNA sensing and cGAS activation within cells

Despite the increasing knowledge about the mechanism of cGAS activation in vitro, the intracellular behavior of cGAS is more complicated and remains an active area of investigation. Here, we will focus on the subcellular distribution and posttranslational modification of cGAS as the major factors that regulate DNA sensing and activation of cGAS within cells (Figure 2).

Figure 2. Regulation of cGAS activity in cells.

cGAS activation is positively and negatively regulated by various mechanisms in cells. cGAS is localized in not only in cytosol but also in the nucleus and on the plasma membrane, which restricts the access to the cytosolic DNA. cGAS reactivity to the chromosomal DNA is restricted in the nucleus. cGAS tethering to the acidic patch of nucleosomes and BAF binding to DNA prevents cGAS from binding DNA. Besides these mechanisms restricting cGAS accessibility to DNA, cGAS is hyperphosphorylated at its N-terminal region and catalytic activity is attenuated. Post-translational modification such as K27-linked polyubiquitination, monoubiquitination, Neddylation, and SUMOylation positively regulate cGAS activity, whereas K48-linked polyubiquitination, acetylation, and Caspase-mediated cleavage of cGAS negatively regulate the activity. Moreover, cofactor proteins including PQBP1, G3BP1, and ZCCHC3 promote DNA binding, oligomerization, and phase separation of cGAS. Phase separation-induced condensates restricts TREX1-mediated degradation of DNA and therefore accelerate cGAMP synthesis.

Subcellular localization of cGAS

cGAS was originally thought to reside in the cytosol to avoid nuclear self-DNA detection⁵. Despite this general idea, soon after cGAS was identified, Knipe and colleagues reported that cGAS can be detected in the nucleus of human fibroblasts and keratinocytes¹⁰². Other studies have reported cGAS localization in micronuclei, chromatin bridges, and the nucleus. cGAS activation in micronuclei and chromatin bridges are discussed in the section above. Yang et al. revealed that cGAS translocates from the cytosol to the nucleus during mitosis and relocates to the cytosol to react with DNA fragments in the presence of DNA damage⁵⁹. Liu et al. found that cGAS translocates from cytosol to the nucleus upon DNA damage to suppress DNA repair⁵⁵. More recently, Sun et al. identified a nuclear export signal (NES) within cGAS 169–174 a.a. (LEKLKL), which is responsible for chromosome maintenance region 1 (CRM1)-dependent cGAS translocation from the nucleus to the cytosol upon DNA stimulation¹⁰³. Other groups reported that cGAS does not translocate to the nucleus in response, but is rather a nuclear resident protein¹⁰⁴. Finally, work from our lab examining a tagged allele of cGAS revealed different localization patterns of this protein in different cells¹⁰⁵. For example, cGAS was localized to the plasma membrane in macrophages, but was distributed in the cytosol and/or the nucleus in non-phagocytic cells¹⁰⁵. It is unclear why different groups have found cGAS to be localized to distinct regions of the cell, and much more work needs to be done to understand the behavior of cGAS within cells.

The collective data described above could be used to suggest that cGAS is a nuclear protein under several experimental (and perhaps physiological) conditions. However, this conclusion would have significant functional implications for immunosurveillance, as nuclear cGAS is well-recognized to be an inactive protein. Elegant work from Gentili et al. has shown that the nucleus-localized pool of cGAS is ~200-fold less responsive to DNA than extra-nuclear cGAS¹⁰⁶. Within the nucleus, the protein barrier-to-autointegration factor 1 (BAF) operates as the negative regulator of cGAS sensing of chromatin DNA upon nuclear rupture¹⁰⁷. BAF outcompetes cGAS for DNA binding, thereby inhibiting the formation of cGAS-DNA oligomers in the nucleus. Moreover, while cGAS is able to bind to histone-containing DNA and uncoated DNA, only the latter is able to stimulate cGAMP production efficiently^{106, 108, 109}. Cryo-electron microscopy revealed the mechanism underlying chromatin-mediated inhibition of cGAS activity, as an acidic patch of the histone H2A-H2B contacts cGAS and inhibits cGAS oligomerization upon DNA binding^110–114. Thus, an increasing body of evidence supports the idea that the exclusive localization of cGAS to the nucleus would result in the inactivation of its ability to drive type I IFN responses to infection. As such, it is likely that an extra-nuclear pool of cGAS exists and may be responsible for pathogen detection.

cGAS has been reported to be present on the plasma membrane of macrophages, through interactions between the basic N-terminal domain of cGAS and phosphatidylinositol-4,5-bisphosphate (PtdIns(4,5)P₂; also known as PIP₂)¹⁰⁵. The deletion of the cGAS N-terminal domain released cGAS to the cytosol, resulting in the detection of self-DNA and aberrant type I IFN induction. These data suggest that the N-terminal domain of cGAS is required to prevent self-DNA recognition, perhaps by tethering this protein to the cell surface. The finding that N-terminus of cGAS is required to prevent self-DNA reactivity was recently validated by Li et al., who further demonstrated that the self-DNA detected by cGAS is likely to be mtDNA¹¹⁵. Indeed, N-terminal deletion mutants of cGAS were found to be localized to the microchondria¹¹⁵. These data are interesting to consider in the context of the in vitro behaviors of cGAS, which suggested that DNA-induced liquid droplet formation is potentiated by the cGAS N-terminal domain. As droplet formation in vitro is associated with enhanced cGAMP production, one would have expected that deletion of a driver of droplet formation (the N-terminal domain) would impair IFN responses within cells. Yet the opposite results were obtained independently^{105, 115}. More work is needed to understand the relative importance of the cGAS N-terminal domain in liquid droplet formation and localization within cells.

A recent study by Zhou et al. revealed that cGAS-DNA phase separation is not required for the intrinsic ability of cGAS to produce cGAMP in vitro¹¹⁶. Rather, phase separation may be important to restrict DNA degradation by exonuclease TREX1. Cell-free studies demonstrated that TREX1 is restricted to the outer shell of a cGAS-DNA containing liquid droplet. It is likely that the phase separation resists other negative regulators than TREX1, as the droplet-deficient cGAS mutant signals weaker than wild type cGAS even in the absence of TREX1¹¹⁶. In addition to cytosolic DNA, TREX1 has been reported to degrade micronuclear DNA to prevent cGAS activation¹¹⁷. The access to the ruptured micronuclei is achieved by ER localization of TREX1 and does not depend on DNA-binding function of TREX1. It is unclear whether cGAS-DNA phase separation occurs in micronuclei to resist TREX1-mediated DNA degradation.

Altogether, DNA sensing by cGAS is regulated not only by the physical separation of receptor and ligand within cells, but also by several safeguard mechanisms offered by DNA- or cGAS-binding proteins in the nucleus and the cytosol. These may be the main factors that underline the difference of cGAMP synthetic activity of cGAS in vitro and in cells. However, there is still a lack of consensus about cGAS localization, possibly due to the difference in the experimental settings such as cell types and imaging protocols used in each study. Future investigation that compares these factors will shed more light on the cell biology of cGAS.

Post-translational modifications of cGAS

In addition to regulation by subcellular localization, interaction with other molecules regulate cGAS through post-translational modifications (PTMs). Phosphorylation of cGAS has been reported to inhibit cGAS activity. Akt phosphorylates cGAS at S305 in human cGAS (S291 in mouse cGAS) within the C-terminal catalytic domain, thereby negatively regulating cGAMP synthesis in response to exogenous dsDNA¹¹⁸. Cyclin-dependent kinase 1 (CDK1)-cyclin B complex phosphorylates the same serine residues of cGAS upon mitotic entry to suppress cGAMP synthesis upon self-DNA in the nucleus¹¹⁹. These residues are dephosphorylated by type 1 phosphatase (PP1) upon mitotic exit to allow cGAS reactivity to exogenous DNA in the cytosol. Furthermore, multiple serine residues within the N terminus of cGAS are phosphorylated by Aurora kinase B (AurB) and other kinases during mitosis¹¹⁵. This hyperphosphorylation, along with chromatin tethering, blocks cGAS activation by self-DNA in the nucleus during cell cycle transition¹¹⁵. In addition to serine phosphorylation, cGAS is phosphorylated at tyrosine residues. B-lymphocyte kinase (BLK) phosphorylates cGAS at Y215, thereby inhibits the cGAS nuclear translocation during DNA damage mentioned in the previous section⁵⁵.

Ubiquitination and other ubiquitin-like protein conjugation events affect cGAS-STING signaling. K27-linked polyubiquitination of cGAS by ER-resident E3 ligase RNF185 positively regulates cGAS-mediated innate immune responses to HSV-1 infection¹²⁰. Additionally, E3 ligase TRIM56-induced monoubiquitination of cGAS at K335 is necessary for cGAS oligomerization upon DNA binding¹²¹. Moreover, cGAS is subject to K48-linked polyubiquitination at K271 and K414, which targets cGAS for proteasomal degradation. As a positive feedback loop of cGAS-STING signaling, this K48-linked ubiquitination is inhibited by tripartite motif (TRIM) E3 ligases. TRIM38 conjugates small ubiquitin-like modifier (SUMO) to cGAS that inhibits DNA binding and K48-linked polyubiquitination of cGAS¹²², while TRIM14 recruits deubiquitinating enzyme (DUB) USP14 to cleave K48-linked ubiquitin chains of cGAS¹²³.

Neddylation is the covalent conjugation of neural precursor cell expressed, developmentally downregulated 8 (NEDD8), another ubiquitin-like protein. cGAS is neddylated on K231 and K421 by the roles of E2 enzyme Ube2m and E3 ligase Rnf111, inhibition of which abolishes type I IFN induction during HSV-1 infection¹²⁴.

Acetylation of cGAS has been reported to modulate cGAS-STING innate immune signaling. cGAS K384, K394, and K414 are acetylated in the steady state, and DNA binding of cGAS stabilizes the interaction with histone deacetylase 3 (HDAC3), which deacetylates cGAS to activate the signaling¹²⁵. Aspirin, a non-steroidal anti-inflammatory drug (NSAID), inhibits cGAS activation by directly binding and acetylation on these residues. Treatment of aspirin suppresses inflammation in AGS patient cells and in an AGS mouse model, suggesting cGAS acetylation prevents self-DNA-induced cGAS-STING signal activation. On the other hand, the cGAS acetylation in the N-terminal domain at K47, K56, K62, and K83 by the lysine acetyltransferase 5 (KAT5) potentiates cGAS-STING signaling during DNA virus infections¹²⁶.

The cleavage of cGAS also modulates is inflammatory activities in cells. The canonical and non-canonical inflammasome activation leads to caspase-1- and caspase-11-mediated cGAS cleavage, respectively, to dampen cGAS-STING signaling¹²⁷. Similarly, activated caspase-3 cleaves cGAS during apoptosis to suppress DNA-induced cytokine responses, keeping apoptotic cells immunologically silent⁸⁷.

Besides post-translational modification, several proteins enhance cGAS-DNA complex formation by physically interacting either the enzyme or the substrate. GTPase-activating protein SH3-binding protein 1 (G3BP1) was identified as a protein that interacts with cGAS in the cytosol to promote DNA binding and subsequent activation¹²⁸. Recently, the interaction with G3BP1 has been found to promote primary condensation of cGAS even in the resting cells, which primes cGAS for its rapid response to free DNA¹²⁹. Although G3BP1 is required for stress granule (SG) assembly and therefore plays an important role in RNA-sensing innate immune signaling, cGAS activation by G3BP1 is independent on the SG assembly¹²⁸.

In addition, polyglutamine binding protein 1 (PQBP1) and CCHC-type zinc-finger (ZF) protein ZCCHC3 are co-sensors of cGAS, which bind to both DNA and cGAS to promote cGAS recognition of DNA^{130, 131}. While PQBP1 is specific receptor for the reverse-transcribed retroviral DNA such as HIV-1, ZCCHC3 binds to viral DNA such as HSV-1 and synthetic dsDNA in the cytosol. Because the affinity of cGAS alone for DNA is low and the recognition is promiscuous^{95, 99}, these co-receptors are essential for optimal cGAS-DNA complex formation. Absence of these co-receptors in the nucleus may explain the low reactivity of cGAS to the self-DNA in the nucleus in addition to the presence of BAF as mentioned above, which needs further examination. Moreover, PQBP1 interacts with extrinsic tau, transmissible neurodegenerative disease protein, and then triggers cGAS-STING-dependent NF-κB activation¹³². As this finding bridges cGAS-STING signaling to tau-related neurodegenerative disorders, identification of cGAS-STING signaling regulators that are known to play important roles in a disease pathogenesis will indicate the unexpected involvement of cGAS and STING in that disease.

Altogether, cGAS-STING signaling is regulated by various mechanisms in cells including subcellular localization and post-translational modifications. In the next section, we will discuss how cGAS-STING pathway is targeted by pathogens.

Viral targeting of cGAS as an immune evasion strategy

Viruses and bacteria utilize several strategies to inhibit cGAS-STING signaling including the regulation of transcriptional levels and degrading CDNs, but here we focus on the modifications of cGAS and its activation reported for some viral proteins. Some DNA and RNA viruses target cGAS for degradation. The NS2B protease complex of dengue virus (DENV) cleaves cGAS to block its activation by mtDNA released from stressed mitochondria during infection¹³³. Other viruses take indirect strategies to degrade cGAS. NS1 protein of Zika virus (ZIKV) stabilizes caspase-1, which cleaves cGAS as described above¹³⁴. F17 protein in poxviruses binds to Raptor and Rictor, regulators of mammalian target of rapamycin complexes mTORC1 and mTORC2, to hyperactivate mTOR leading to the proteasomal degradation of cGAS¹³⁵.

Other viral proteins physically perturb DNA sensing and activation of cGAS. VP22 and UL37 from HSV-1^{136, 137}, UL31, UL42, and UL83 from human cytomegalovirus (HCMV)^138–140, and ORF52 and LANA from Kaposi’s sarcoma-associated herpesvirus (KSHV)^{141, 142}, ORF9 from Varicella-Zoster virus (VZV) are known examples of cGAS-inhibiting viral proteins¹⁴³. Among these viral proteins, VP22, ORF52, and other structurally similar homolog tegument proteins have been recently found to compete DNA binding and phase separation with cGAS, thereby inhibiting cGAS-DNA phase separation¹⁴⁴. In contrast, the HSV-1 UL37 tegument protein inhibits cGAS through PTM. UL37 deamidates N210 within the activation loop of human cGAS, which blocks cGAMP synthesis by impinging on the catalytic site¹³⁷. This deamidation residue of cGAS is conserved in mice but not in many other non-human primates, which accounts for the species-specific permissiveness of HSV-1.

Conclusion

cGAS function has been implicated as a driver of inflammation in a variety of cellular processes. Studies from diverse laboratories have unified our understanding of the intrinsic enzymatics that govern DNA-induced cGAS activities. However, much less unanimity exists regarding the regulation of these intrinsic activities within cells. The cell biology of the cGAS-STING pathway will therefore likely remain at the forefront of research in this area. The open questions in this regard include how diverse the regulation of cGAS activation is in different cell types and different species. These questions are even more demanded by our recent discovery of three classes of cGAS with different self-DNA reactivity in mammals¹⁴⁵. Also, the mechanisms of cGAS activation by non-self- and self-DNA in cells are regarded to be the same, but there is still a possibility that these are different, which warrants further investigation. Since the dysregulation of cGAS binding of self-DNA and subsequent aberrant cytokine production underlies a wide variety of diseases, future studies will likely be focused on filling the gap between our knowledge of cGAS activities in vitro and those within cells.