STING Signaling in The Brain: Molecular Threats, Signaling Activities, and Therapeutic Challenges

SUMMARY

Stimulator of interferon genes (STING) is an innate immune signaling protein critical to infections, autoimmunity, and cancer. STING signaling is also emerging as an exciting and integral part of many neurological diseases. Here, we discuss recent advances in STING signaling in the brain. We summarize how molecular threats activate STING signaling in the diseased brain, and how STING signaling activities in glial and neuronal cells cause neuropathology. We also review human studies of STING neurobiology and consider therapeutic challenges in targeting STING to treat neurological diseases.

IN BRIEF

Yang et al review the current state of knowledge about innate immune system STING signaling in neurological diseases and discuss mechanisms of STING signaling activation, activities in different brain cell types, human relevance, and development of STING antagonists.

INTRODUCTION

STING is a multi-pass transmembrane protein localized at the ER with its ligand-binding domain exposed to the cytosol. STING is best known for its role in mediating the innate immune response to microbial infections. Molecular mechanisms of STING activation during infection are well understood.¹ Microbial DNA is the main “molecular threat”, also known as pathogen-associated molecular pattern, that activates the DNA sensor cGAS. cGAS catalyzes the production of the cyclic dinucleotide 2′,3′-cGAMP, which binds STING. cGAMP binding triggers STING trafficking first before activating downstream signaling. STING migrates from the ER to the Golgi, where it recruits the kinase TBK1 and the transcription factor IRF3. After phosphorylation, IRF3 dimerizes and then translocates into the nucleus to promote the expression of type I interferon (IFN-I) and IFN-stimulated genes (ISGs).

Recent studies on several STING-associated inflammatory diseases have expanded our knowledge of STING biology beyond its core mechanism.² We now know that at homeostasis, STING moves continuously through the vesicular trafficking pathway instead of “docking” at the ER, and that STING is constantly being degraded by the lysosomes. In addition to ligand-mediated activation, trafficking-mediated STING activation, originally discovered while studying inborn error diseases due to STING gain-of-function mutations, is also becoming more common. Moreover, STING has many other signaling activities besides IFN-I, including NF-κB signaling that produces inflammatory cytokines and chemokines, antiproliferation activity, cell death, autophagy, and ER stress. Many of these new IFN-independent activities are physiologically important in STING-associated human diseases, although the mechanisms are less well understood compared to STING-mediated IFN signaling.

Although the current understanding of STING signaling is based on studies of the immune system outside the brain, the first identified human disease caused by excessive STING signaling is a neuroinflammatory condition called Aicardi-Goutières Syndrome (AGS).³ In recent years, there has been much excitement in elucidating the neuroscientific aspect of STING biology. STING is now associated with many neurological diseases, including neurodegenerative diseases that affect millions of people. In this review, we take a STING biology-focused look at the current literature in the field of neuroscience. We will summarize how STING is activated in the brain. Based on prior knowledge of STING biology, what are the “molecular threats” detected by STING? What are possible activation mechanisms? We will also review STING signaling activities in the brain. What chain of events leads each activity to a neuropathology? Finally, we will discuss therapeutic strategies and challenges for the development of STING antagonists.

MOLECULAR THREATS IN THE BRAIN AND MECHANISMS OF STING ACTIVATION

The STING pathway has been implicated in many neurological disorders in recent years, including neurodegeneration, neuroinflammation, brain injury, pain, and others (Figure 1). Multiple physical and molecular threats are involved in the pathogenesis of these diseases. It seems impossible for all these conditions to activate STING through a singular mechanism that produces a singular signaling activity. Taking neurodegenerative disease as an example, which itself is a heterogeneous collection of disorders. Each neurodegenerative disease involves different genetic factors and molecular threats that lead to the progressive loss of different neurons. Pathological features of neurodegenerative diseases include inflammation, protein aggregation, defective proteostasis, defects in DNA and RNA homeostasis, altered energy metabolism, synaptic and neuronal network dysfunction, cytoskeletal abnormalities, and neuronal cell death.⁴ In mouse models, deletion of the Sting1 gene or pharmacological inhibition of STING protein amazingly reduced neuropathology in many types of neurodegenerative diseases. These results are exciting because they identify STING as a new drug target that likely has a very different mechanism of action compared to current approaches to treat neurodegeneration. One cannot help wondering: Is STING a central hub that drives neuropathology? How does STING detect and interoperate different molecular threats emerging in different brain diseases?

Figure 1. STING is associated with many neurological diseases associated with a wide range of cellular and molecular defects.

AD, Alzheimer’s disease. PD, Parkinson’s disease. ALS, amyotrophic lateral sclerosis. FTD, frontotemporal dementia. TBI, traumatic brain injury. A-T, ataxia-telangiectasia. NPC, Niemann-Pick disease type C.

mtDNA exposure: mitophagy defects, protein aggregates, tissue injury

The most commonly observed molecular threat directly activating cGAS-STING in the brain is mitochondrial DNA (mtDNA, Figure 2). Neuronal cells in the central nervous system (CNS) are energetically demanding cells in the human body. Efficient oxidative phosphorylation in mitochondria is crucial for the generation of ATP, the key molecule of energy metabolism in the brain. Therefore, mitochondrial dysfunction underlies the pathogenic process of several neurodegenerative diseases. More direct evidence using genetic disruption of the mitochondrial respiratory complex, specifically in dopaminergic neurons, demonstrates that mitochondrial dysfunction is sufficient to trigger progressive parkinsonism.⁵ Beyond energy production, mitochondria can release nucleic acids such as mtDNA and mtRNA, both of which are potent activators of innate immune signaling.^6-8 In order for mtDNA to engage DNA sensor cGAS, it must enter the cytosol. Mechanisms of mtDNA leakage include mitophagy defects, protein aggregates, and tissue injury (examples below).

Figure 2. Molecular threats in the brain and mechanisms of STING activation.

Key evidence supporting mtDNA as a molecular threat that activates cGAS-STING includes: 1) increased copy number of mtDNA in the cytosol and increased concentration of the STING ligand 2′,3′-cGAMP; 2) enrichment of mtDNA sequences in cGAS immunoprecipitation; 3) microscopic detection of mtDNA outside the mitochondria compartment, 4) reduced inflammatory phenotype after mtDNA depletion; 5) increased pSTING levels or decreased inflammatory phenotype after cGAS or STING knockdown or knockout. Not all mitochondrial abnormalities lead to activation of the cGAS-STING signaling. Therefore, it is important to establish multiple lines of evidence to fully support mtDNA as a molecular threat that triggers STING signaling.

Mitophagy defects.

Mitophagy is an important cellular quality control machinery that selectively removes defective or aged mitochondria through autophagy, thus maintaining mitochondrial homeostasis. Mitophagy defects leave damaged mitochondria in the cytoplasm that would spontaneously expose mtDNA to cGAS. Mutations in several mitophagy-related genes have been implicated in various neurodegenerative diseases. For example, mutations in the genes PINK1 (encodes PTEN-induced kinase 1) and PRKN (encodes Parkin) are associated with early-onset of familial Parkinson's disease (PD).^9,10 The ubiquitin kinase PINK1 deposits on damaged mitochondria. This promotes the recruitment of Parkin, an E3 ubiquitin ligase that marks dysfunctional mitochondria for degradation. Defects in PINK1-PRKN-mediated mitophagy result in the accumulation of damaged mitochondria that contribute to the pathogenesis of Parkinson's disease ¹¹. In Pink1^−/− and Prkn^−/− mice, acute mitochondrial stress during exhaustive exercise elicits a strong inflammatory response which is completely abolished by loss of STING.¹² Similar STING-dependent activation of the innate immune system was also observed in mitophagy-defective mutator mice, a chronic model of mitochondrial stress due to the accumulation of mutations in mtDNA.

Leucine-rich repeat kinase 2 (LRRK2), the most frequently mutated gene in familial and sporadic Parkinson’s disease, also plays a role in mitophagy. Pathogenic mutations in LRRK2 impair the induction of PINK-PARKIN-dependent mitophagy.^13-16 Lrrk2 knockout macrophages exhibit increased basal IFN-I and ISG signature, which is attenuated in Lrrk2/Cgas double knockout cells.¹⁷ Basal IFN-I expression in Lrrk2 knockout macrophages was enhanced by cytosolic sensing of mtDNA by cGAS. Interestingly, pathogenic mutations of LRRK2 in Parkinson’s disease actually increase kinase activity and cause an autosomal dominant form of Parkinson’s disease. Whether PD-associated LRRK2 gain-of-function mutations affect the cGAS-STING DNA sensing pathway needs further investigation. Many other mitophagy- or autophagy-related genes, including PARK1 (encodes α-Synuclein), PARK7 (encodes DJ-1), PARK15 (encodes FBXO7, which acts in Parkin-mediated mitophagy),¹⁸ PARK17 (encodes VPS35), OPTN (encodes Optineurin, which is an autophagy receptor for damaged mitochondria),¹⁹ TBK1, have also been implicated in neurodegenerative diseases, including Parkinson’s disease, amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). (Table 1) It will be interesting to examine whether these gene mutations converge on the mtDNA-cGAS-STING axis and how much STING signaling contributes to their neuropathology.

Table 1.

Mechanisms of STING activation by gene mutations associated with neurodegenerative diseases.

Disease	Mutated Gene	Mechanism of STING Activation
		Mitochondrial DNA leakage
Parkinson’s Disease (PD)	PINK1, PRKN	Defective PINK1-Parkin-mediated mitophagy.12
	LRRK2	Mitochondrial stresses, including oxidative stress and fragmentation.17
	VPS13C/PARK23	mtDNA release likely caused indirectly by defective lysosomal function.80
Alzheimer’s disease (AD)	APP/PSEN1	Accumulation of cytosolic mtDNA via unknown mechanism.23,51
	MAPT/Tau	Mutant tau mislocalizes to mitochondria and induces mtDNA leakage.22
Amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD)	TDP-43	TDP-43 invades mitochondria and triggers mtDNA leakage via mPTP.20
	SOD	Misfolded mutant SOD1 damages mitochondria and induces the release of mtDNA and RNA:DNA hybrids into the cytosol.21
Hungtington’s disease (HD)	HTT	mtDNA leakage through an unknown mechanism.
		Nuclear genomic DNA leakage
Ataxia-telangiectasia (A-T)	ATM	Nuclear DNA leakage due to defective DNA damage response.35,44,45
Aicardi-Goutières Syndrome (AGS)	RNASEH2A/B/C	Accumulation of micronuclei due to extensive DNA damage.36,39,62
AD	APP/PSEN1	DNA damage through an unknown mechanism.
		Endolysosomal Dysregulation
ALS/FTD	C9ORF72	Impaired lysosomal degradation of STING in C9orf72−/− myeloid cells.89
Niemann-Pick disease type C1 (NPC1)	NPC1	Impaired lysosomal degradation of STING due to lack of NPC1 on the lysosome and lysosomal defects.75
PD	VPS13C/PARK23	VPS13C maintains lysosomal lipid homeostasis. VPS13C−/− cells exhibit impaired STING lysosomal degradation and mtDNA leakage.80

Protein aggregates.

Pathogenic protein aggregates can directly impair mitochondrial membrane integrity and trigger mtDNA release. For example, the TAR DNA-binding protein (TARDBP, also known as TDP-43), encoded by a gene associated with ALS, forms cytoplasmic aggregates that invade mitochondria. Then, mtDNA is released into the cytoplasm through the mitochondrial permeability transition pore (mPTP).²⁰ Mutant superoxide dismutase 1 (SOD1), encoded by another ALS-associated gene, also causes mitochondrial damage and the release of mtDNA and mt(RNA:DNA) hybrid into the cytoplasm, activating cGAS-STING and DDX41-STING signaling pathways, respectively.²¹

In Alzheimer’s disease (AD), pathogenic Tau interacts with mitochondrial proteins and impairs mitochondrial function, resulting in neurotoxicity. The cGAS-STING-IFN pathway is activated in Tau-P301S mice, a mouse model of tauopathy. cGas^−/− rescues tauopathy-induced IFN signaling, synaptic plasticity, and memory deficits without affecting Tau loading.²² Elevated levels of cytosolic mtDNA have also been reported in the brain tissue of 5×FAD mice, a mouse model of amyloid pathology, although whether and how β-amyloid damages mitochondria is unknown.²³ Mice expressing mutated huntingtin (HTT), that can form huntingtin protein aggregates, also have increased mtDNA release and cGAS-mediated inflammation in the striatum and cortex.²⁴

It is fascinating to note that protein aggregates of diverse biochemical nature all tend to target mitochondria. What is the mechanism for this? Mislocalization of protein aggregates into mitochondria has been demonstrated in some cases but not all. It is also possible that protein aggregates overwhelm cellular protein quality control machinery such as the proteasome, thereby inducing proteotoxic stress that could indirectly compromise the integrity of mitochondrial membrane. Further research is needed to address these possibilities. In addition, protein aggregates do not always lead to leakage of mtDNA. They can also cause nuclear DNA damage and the release of other molecular threats such as nuclear genomic DNA (gDNA, below).

Tissue injury.

mtDNA can also be released upon tissue injury and then activates cGAS-STING.^25-28 Elevated cytosolic mtDNA levels are found in damaged brain tissue in both traumatic brain injury (TBI) and ischemic stroke mouse models.^29-31 The exact mechanism of the mtDNA leakage is not known. Three mechanisms are proposed: 1) opening of the mPTP pore at the mitochondrial inner membrane (detected in cells overexpressing the mutant TDP-43);²⁰ 2) the Bax/Bak macropores followed by inner membrane rupture (during apoptosis activation);^6,7,32 and 3) pores formed by oligomerization of voltage-dependent anion channel (VDAC) at the mitochondrial outer membrane.³³ As we will discuss below, cell death in tissues can produce another molecular threat, namely genomic DNA (gDNA), which can also activate cGAS-STING. The respective contribution of mtDNA and gDNA to cGAS-STING activation during tissue injury needs further investigation.

Nuclear gDNA exposure: DNA damage and defective repair, retrotransposable elements, and micronuclei

The original discovery of cGAS as a cytosolic microbial DNA sensor made perfect sense: the mammalian cell compartmentalizes gDNA in the nucleus to avoid self-reactivity in the cytoplasm. Any violation of this physical separation could result in gDNA leaking into the cytosol and activating cGAS-STING. However, recent findings from several groups indicate that cGAS is predominantly found in the nucleus but is bound and inhibited by nucleosomes (reviewed in ³⁴). Why is the cytosolic DNA sensor located in the nucleus? We do not have a clear answer yet. According to current knowledge, a large part of the cGAS in a cell is inactive in the nucleus; a small portion of free cGAS resides in the cytoplasm and actively senses DNA. cGAS can potentially switch between “inactive” and “free” states, although the mechanism is unclear. Nonetheless, there is ample evidence that gDNA poses a molecular threat that can activate cGAS-STING directly. gDNA exposure can occur under cellular conditions such as DNA damage, defective DNA repair that causes genomic instability, and DNA replication stress that exposes micronuclei.^35-40

DNA damage and defective repair.

Several rare inherited neurological disorders are caused by deficiencies in the DNA damage response or DNA repair factors (Figure 2).^41,42 The best example is ataxia-telangiectasia (A-T), a multisystem disorder with diverse neurological features, including notable cerebellar atrophy, early-onset ataxia, oculomotor apraxia (inability to control eye movements), progressive cognitive decline, and dysarthria (slurred speech). A-T is caused by mutations in ATM (ataxia telangiectasia mutated), a kinase that initiates and coordinates double-strand DNA (dsDNA) break response. Loss of ATM leads to dsDNA breaks, resulting in the release of gDNA into the cytoplasm and activation of the cGAS-STING pathway.^35,43 Cytosolic accumulation of gDNA and STING-dependent neuroinflammation were also observed in both primary microglia of Atm^−/− mice and in inducible pluripotent stem cell (iPSC)-derived cortical brain organoids of human A-T patients.^44,45

An increased level of dsDNA breaks and a decrease in DNA repair capacity have long been described in the brain tissue of Alzheimer’s disease patients as well as in mouse models.^46-50 In the brain of an APP/PS1 mutant transgenic mouse, increased DNA damage was observed in both cortex and hippocampus as well as concomitant cytosolic DNA-induced inflammation via the cGAS-STING pathway.⁵¹ In the intrastriatal α-synuclein-preformed fibril (αSyn-PFF) mouse model of Parkinson’s disease, αSyn-PFF injection caused microglial DNA damage and STING signaling.⁵² Traumatic brain injury and ischemia stroke also cause DNA damage in addition to mtDNA release, both of which could activate cGAS-STING.^53-55

Retrotransposable elements.

Transposable elements are “jumping” genes in the mammalian genome. These include DNA transposons that mobilize using a “cut-and-paste” mechanism or retrotransposons that self-amplify via an RNA intermediate. The most abundant retrotransposon in humans is LINE-1. In monogenic AGS patients carrying defective nucleases such as TREX1, LINE-1 cDNA accumulates and then activates cGAS-STING signaling and neuroinflammation.^56,57 AGS patients carrying gain-of-function IFIH1 mutations (IFIH1 encodes cytosolic RNA sensor MDA5) become overly sensitive to cellular Alu retroelements duplex dsRNA, which is likely another route to neuroinflammation.⁵⁸ In addition, transposable elements have also been implicated in other neurodegenerative diseases, although the causal connections are less clear.⁵⁹

Micronuclei.

Micronuclei are small nuclei containing fragments of chromatin. They often arise due to mis-segregation of DNA during mitotic progression after dsDNA breaks.^36,37 The rupture of micronuclear membrane exposes chromatin to cytosol where gDNA can be recognized by free cGAS in the cytosol. There is clear microscopic evidence for the binding of cytoplasmic cGAS to micronuclei. Mutations in RNase H2, a ribonuclease that removes RNA-DNA hybrids in genomic DNA, cause extensive DNA damage and micronuclei that engage cGAS.^36,39 Patients carrying RNase H2 mutants are associated with AGS, an autoinflammatory disease characterized by elevated IFNα in the cerebrospinal fluid, white matter calcification, and systemic autoimmunity.⁶⁰ Germline RNaseh2-deficient mice develop embryonic lethality dependent on both p53 and cGAS-STING.⁶¹ Neuronal deletion of Rnaseh2b in mice promotes accumulation of cytosolic DNA, activation of cGAS-dependent IFN-I response, and cerebellar atrophy.⁶² In addition, numerous micronuclei and activation of cGAS-STING signaling have also been found in mouse striatal neuronal cells of Huntington’s disease (HD) and in human embryonic stem cells (hESC)-derived neurons.⁶³

The exact mechanism by which dsDNA breaks activate cGAS requires further investigation. The default hypothesis is gDNA leakage into the cytoplasm, although the nature of gDNA as a molecular threat and how it gets to the cytoplasm are unclear. Whether the abundant inactive cGAS bound to the nucleosome helps detect DNA damage is also unclear. Many questions remain unanswered. For example, what type of DNA damage and at what level is required to activate cGAS? Besides micronuclei, how can we best detect or visualize this molecular threat, particularly in pathological brain tissues?

Intracellular Vesicle Trafficking and Endolysosomal Dysregulation

A recently emerging molecular threat that can activate STING signaling directly is defective vesicle trafficking (Figure 2).² In contrast to DNA-mediated activation of cGAS, which is relatively easy to envision, trafficking-mediated activation of STING is less obvious. This new mechanism was discovered through studies of rare inborn error diseases of chronic STING signaling: SAVI (STING-Associated Vasculopathy with Onset in Infancy), the COPA syndrome, and Niemann-Pick disease type C1 (NPC1). The implication for neurological disease is exciting because vesicle trafficking defects are frequently observed in these conditions. Two main concepts are: 1) STING travel on vesicles that continuously move through the intracellular vesicle trafficking pathway from the ER to the Golgi then to the lysosome at homeostasis. STING also travel from the Golgi back to the ER. 2) Promoting STING trafficking to the Golgi or delaying STING trafficking to the lysosome can activate STING signaling.

SAVI.

The majority of SAVI cases are caused by de novo heterozygous gain-of-function mutations in STING (e.g. N154S, V155M) that promote STING trafficking from the ERto the Golgi via COPII (coatomer protein II) vesicles independent of cGAS and cGAMP.^64,65 SAVI patients develop interstitial lung disease, cutaneous vasculopathy, and systemic features of type I interferonopathy. Surprisingly, mouse models of SAVI develop lung disease and T-cell lymphopenia independently of IFN-I signaling. Although SAVI patients report no significant neurological symptoms, SAVI mice show increased microglia and astrocytes in the adult brain and decreased dopaminergic neuron density.⁶⁶ Neurological abnormalities in SAVI mice are also independent of IFN-I signaling. It is important to note that IFN-dependency is a key difference between human SAVI patients and mouse models. Human SAVI patients do exhibit high levels of IFN-I signature in the peripheral blood. ⁶⁷ IFN-I cytokines bind to IFN receptors (IFNAR1 and IFNAR2) on the cell surface and then recruit kinases JAK1 and TYK2, which phosphorylate and activate transcription factor STAT1 and STAT2. JAK inhibitors are FDA-approved for treating several inflammatory and autoimmune diseases, and SAVI patients also benefit from JAK inhibitor treatments.⁶⁸

COPA.

The COPA syndrome is caused by loss-of-function of COPI (coatomer protein I) vesicle trafficking. Anterograde ER-to-Golgi vesicle transport relies on COPII vesicles and retrograde Golgi-to-ER vesicle transport relies on COPI vesicles. Heterozygous mutations in the COPA gene, which encodes COP-α subunit of the COPI complex, are the underlying causes of the COPA syndrome.⁶⁹ COPA patients share many similar clinical manifestations to SAVI patients. In COPA mutant cells, STING constitutively localizes to the Golgi, resulting in chronic STING signaling.^70-72 The disease pathology of a Copa-deficient mouse can be rescued by Sting1^−/−, establishing STING as the major disease driver for COPA. Although we group COPA under “trafficking-mediated” STING activation, there is evidence suggesting that cGAS is necessary for STING activation in COPA.⁷² This is likely because homeostatic STING trafficking is supported by cGAS sensing of a sub-pathological level of self-DNA.⁷³ In addition, defects in ER-to-Golgi or COPII trafficking have been linked to several neurological disorders, as neurons are particularly susceptible to defects in ER export (reviewed in ⁷⁴). It will be interesting to investigate whether STING plays a role in these COPII-associated neurological disorders.

NPC1.

STING is degraded by the lysosome at homeostasis and after ligand activation. Blocking or delaying STING degradation by the lysosome enhances STING signaling. NPC1 is a lysosomal membrane protein that facilitates the intracellular transport of cholesterol and regulates cholesterol homeostasis. NPC1-deficiency causes a lysosomal storage disorder called Niemann-Pick disease type C1. Recently, we found that NPC1 also serves as an adapter protein that recruits STING to the lysosome for degradation.⁷⁵ Npc1^−/− mouse cerebellar tissue exhibits an increased IFN signature and decreased number of Purkinje cells, both of which are rescued by Sting1^−/−, but not cGas^−/−75 Lysosomal storage disorder is a collection of about 70 monogenic diseases affecting lysosomal hydrolases, ion channels, and membrane proteins. Despite different genetic defects and storage materials in the lysosome, disease pathology in humans and mice commonly involves splenomegaly, increased IFN-I signature in the cerebellum, Purkinje cell loss, and pulmonary disease.^76,77 All of these are familiar traits associated with chronic STING signaling. Future research is ongoing to determine whether STING plays a role in other lysosomal storage disorders.

VPS13C and C9ORF72.

Delayed lysosomal degradation of STING is also observed in cells lacking VPS13C, whose mutations are associated with familial early-onset Parkinson’s disease.^78-80 VPS13C is a putative lipid transfer protein that localizes to contact sites between the ER and later endosomes/lysosomes.⁸¹ Loss of VPS13C disrupts lysosomal lipid homeostasis and impairs STING degradation without affecting global protein degradation.⁸⁰ This raises an interesting question regarding substrate specificity: How does lysosomal dysfunction specifically affect STING degradation? Another example is C9ORF72 (hexanucleotide repeat expansions). Mutations in the C9ORF72 gene are the most common genetic cause of familial ALS and FTD.^82,83 Although the precise cellular function of C9ORF72 is still unclear, C9ORF72 is involved in the autophagy-lysosome pathway at multiple levels, including autophagosomal and lysosomal biogenesis, endosomal trafficking, autolysosome acidification, and endolysosomal degradation.^84-88 STING degradation by the lysosome is also delayed in C9orf72^−/− cells and mice, resulting in STING-dependent systemic inflammation in these mice.⁸⁹ It’s important to note that specific mutations associated with C9ORF72-associated ALS involves the expansion of a hexanucleotide repeat sequence (GGGGCC) in the non-coding region of the gene. STING-dependent inflammation need to be further tested in this model.

Other endolysosomal genes.

Endolysosomal trafficking is essential for neuronal cell function. Many genes encoding proteins involved in this process are associated with neurodegenerative diseases, such as SNCA, LRRK2, SYNJ1, DNAJC6 and DNAJC13, HTT, ATXN3 (reviewed in ^90-92). In Alzheimer’s disease, most of current knowledge is based on patients with early-onset Alzheimer’s disease caused by highly penetrating mutations in PSEN1, PSEN2, and APP. However, the majority cases of Alzheimer’s disease cases are late-onset Alzheimer’s disease. Many endolysosomal genes are risk factors associated with late-onset Alzheimer’s disease, such as APOE4, BIN1, CD2AP, PICALM, PLD3, TREM2.⁹³ The primary impact of an endolysosomal defect in neurodegenerative diseases is likely to be on protein processing or secretions. This has been demonstrated for some of the late-onset Alzheimer’s disease genes. Given the recent advances in STING biology, we should also consider a direct link between endolysosomal defect and immune activation. Unfortunately, there is a general lack of mouse models for late-onset Alzheimer’s disease. This is a challenge that needs to be addressed in the future. In addition, other immune sensors or immune signaling proteins could also be dysregulated when endolysosomal function is disrupted. For example, Toll-like receptor (TLR) 7 and 9 rely on trafficking to endosomes to function.⁹⁴ TLR4 can be endocytosed and degraded by the lysosome.⁹⁵ Therefore, more examples of trafficking-mediated STING or other immune sensor activations in neurological diseases will certainly emerge in the future.

Double threat.

Defective lysosomes often lead to defective mitochondria, particularly in age-related and metabolic disorders (reviewed in ^96-100). These scenarios pose a “double threat” for activating the cGAS-STING pathway. For example, a genetic disruption of lysosomal acidification by knocking out acid α-glucosidase (Gaa) in mice induces pervasive inflammation in the brain associated with mtDNA instability and decreased mitochondrial biogenesis and function.¹⁰¹ Depletion of the lysosomal enzyme β-glucocerebrosidase (GBA), the major genetic risk factor for Parkinson’s disease, could prolong mitochondria-lysosome contacts thus disrupting mitochondrial distribution and function.^102-104 Loss of lysosomal lipid homeostasis in VPS13C^−/− cells not only interferes with STING degradation, but also causes mtDNA release into the cytosol (via an indirect mechanism since VPS13C is not localized to mitochondria).⁸⁰ Similarly, in addition to dysfunctional mitophagy and mtDNA leakage discussed above, LRRK2 deficiency also disrupts endolysosomal trafficking, leading to impaired protein degradation and accumulation of α-synuclein.^105-108 In these cases, it is often difficult to decipher the individual contribution of each molecular threat, mtDNA leakage or lysosomal defect.

Other emerging mechanisms of STING activation in the brain

PQBP1.

A study showed that polyglutamine binding protein 1 (PQBP1) recognizes polymerized tau and then recruits and activates cGAS-STING-mediated NF-κB signaling in microglia (Figure 2).¹⁰⁹ A similar mechanism has been described in HIV-1 infection, where PQBP1 recognizes the HIV-1 capsid and recruits cGAS.¹¹⁰ Then, cGAS is activated by HIV-1 DNA in the capsid. Whether polymerized tau fdaments contain cellular DNA that can activate cGAS remains unclear. In addition, PQBP1 binds several other polyglutamine-containing proteins such as huntingtin, ataxin-1, and ataxin-3. Whether the same mechanism can be extended to other neurological diseases needs further investigation.

cGAMP transfer.

The concept of cell-to-cell communication via small molecules such as neurotransmitters is common in the brain. Recent studies have shown that cGAMP can be transferred from cell to cell and acts as an “immunotransmitter” (Figure 2). To date, several cGAMP importers and one exporter have been identified and they function in a cell-type specific manner ^111-117. Many of these cGAMP transporters are known ion channels that also support cGAMP transfer. In addition, cGAMP can also pass through gap junctions.¹¹⁸ In SOD1 mutant cells (associated with ALS), the cGAS-STING-dependent IFN response spreads to surrounding cells by transferring cGAMP across gap junctions.²¹ The concept of cGAMP transfer resembles one of the intriguing aspects of neurodegenerative diseases, namely the ability of protein aggregates to seed and propagate the pathological process. Very little is known about cGAMP transfer in the brain, which presents an exciting opportunity.

STING expression.

STING expression is generally low in brain tissue. Increased STING expression in the brain is observed in several neurological diseases (Figure 2). For example, both late trauma human brain tissues and mouse traumatic brain injury models showed upregulation of STING expression in lesion tissue.^29,54,119 Most of these observations remain limited to the tissue level. One study showed increased STING expression in microglia after ischemic stroke in mice.³⁰ STING is typically not induced by IFN-I or inflammatory cytokines. Increased STING expression in brain tissue could be cell-intrinsic or due to an increased number of microglia and STING-positive immune cells infiltrating the affected area. Whether increased STING expression is the cause or consequence of disease pathology requires further investigation.

STING SIGNALING ACTIVITIES IN THE BRAIN

STING mediates IFN and IFN-independent activities

STING’s main signaling activity is mediated via the TBK1-IRF3-IFN axis. In addition to the IFN-I response, STING also activates NF-κB signaling through TBK1 or IκB kinase IKK, leading to the expression of many inflammatory cytokines (e.g. TNFα, IL-1β and IL-6) and chemokines (e.g. CXCL10, CCL5, and others). STING activation also induces autophagy, cell death, cellular senescence, unfolded protein response, and antiproliferation (Figure 3A).^120-125 The IFN-I signaling is the predominant activity of STING in myeloid cells. Many of STING’s IFN-independent activities are cell-type dependent. For example, STING activation in lymphoid cells leads to ER stress and cell death.¹²²

Figure 3. STING signaling activities in the central and peripherical nerves system.

A. STING signaling activities with known and unknown mediators. B. STING signaling activities in microglia. C. STING signaling activity in sensory neurons.

When considering STING signaling activities in the brain, IFN-I signaling is the most reported activity. Elevated IFN-I signaling suggests activation of one or more innate immune pathways. However, whether IFN-I is a key mediator of neuropathology is less clear. We should be careful with making assumption as IFN-I signaling is usually overwhelming on transcriptome analysis, and we may be overlooking other STING activities that are not directly measured. An example of a misperception is the role of STING in the antiviral response during infection with herpes simplex virus (HSV-1, a DNA virus). The conventional logic is as follows: DNA virus activates cGAS-STING, STING drives the IFN-I response, IFN-I is antiviral, thereby STING-mediated IFN-I response restricts HSV-1 infection. Unfortunately, this reasoning is proven wrong. Transgenic mice carrying a single point mutation S365A that selectively abolishes STING’s IFN-I activity can still protect mice from HSV-1 infection, as can wild-type mice. ^122,126,127 Another example is cancer immunotherapy. STING agonist-mediated IFN-I signaling has potent anti-tumor activity. However, STING agonists also stimulate cell death in T cells and other cell types, which may limit their therapeutic efficacy. Therefore, broad consideration of STING signaling activities is important when investigating STING neurobiology mechanisms.

In addition to the numerous STING activities, IFN-I itself could also have different effects in the brain compared to other tissues. Many neurons and glial cells express very high levels of the IFN-I receptor, which plays a crucial role in maintaining brain homeostasis. Therefore, the impact of IFN-I signaling in the brain may differ greatly from its classical functions in peripheral tissues.

STING signaling in microglia

Microglia are a type of glial cells that play an essential role in the CNS development and homeostasis by remodeling synapses and clearing dead cells and cell debris. As brain-resident macrophages, microglia express high levels of STING. There is a broad consensus that microglia are the main cell type in which STING activation occurs in many neurological diseases and that activated STING drives the production of IFN-I and inflammatory cytokines from microglia. What these cytokines do in the brain and whether other STING activities also play a role are unknown.

Key evidence for STING activation in microglia comes from in vitro microglia culture assays as well as immunohistochemical (IHC) staining of microglia and STING activation markers (e.g. CD68, pSTING, pTBK1, p65) in brain tissues. Some studies provide further genetic evidence for STING dependence using germline or microglial specific Sting1 knockout mice. For example, microglial activation in response to pathogenic Aβ42 and secretion of the inflammatory cytokines IL-1β, IL-6 and TNFα are reduced by cGas or Sting1 deficiency or treatment with STING inhibitor.^23,51 Tau activation of primary mouse microglia induces p65 nuclear translocation and upregulation of IFN and ISGs in a cGAS-STING dependent manner.^22,109 Tau fibril treatments of human iPSC-derived microglia also activate cGAS-STING signaling defined by pSTING, which is lost after small molecule inhibition of cGAS or STING.²² In 5×FAD mice, phosphorylated STING and IRF3 are found in CD68+ activated microglia.²³ In Tau-P301S mice, cGAS is required for STAT1 phosphorylation in Iba1-positive microglia.²² cGas^−/− alleviates CNS pathologies and improves cognitive functions in both 5×FAD and Tau-P301S mice.

Activation of microglial STING is also implicated in neurodegeneration caused by other forms of protein aggregates. In the αSyn-PFF injection-induced Parkinson’s disease mouse model, STING activation was detected by measuring phosphorylated TBK1 in microglia in mouse striatum.⁵² STING expression is also upregulated in primary microglial culture after αSyn-PFF treatment. In addition, Sting1^−/− reduces neuroinflammation and alleviates the neurological defect in αSyn-PFF injected mice. In the ME7 prion-induced neurodegeneration mouse model, both germline Sting1^−/− and Ifnar1^−/− reduced IFN-I and inflammatory cytokine induction in microglia. Ifnar1^−/− also rescued prion-induced neuropathology.¹²⁸

In amyloid lateral sclerosis and frontal temporal dementia (ALS/FTD), the strongest evidence for microglial STING activation comes from C9orf72^−/− mice (germline whole-body knockout).⁸⁹ Primary microglia and macrophages from C9orf72^−/− mice produce increased amounts of IL-1β and IL-6. C9ORF72 promotes STING degradation by the lysosome, and its deficiency activates STING due to delayed lysosomal degradation. The inflammatory phenotype in C9orf72^−/− mice is entirely dependent on STING. Myeloid specific deletion of C9orf72 largely recapitulates the inflammatory phenotype in whole-body knockout mice. Since C9orf72^−/− mice do not develop neuropathology, it remains to be elucidated whether microglial STING activation plays a central role in C9ORF72-related neuropathology in humans. In contrast, STING clearly drives neuropathology in two other ALS mouse models. ALS-associated SOD1 mutation elicits an IFN-I response in vitro in multiple cell types, including microglia, via the cGAS-STING pathway.²¹ TDP-43 mutation also activates the cGAS-STING pathway through mtDNA leakage.²⁰ In both SOD1 and TDP-43 mutant mouse models, Sting1^−/− reduces neuropathology and improves motor function in mice.

More recently, cGAS-STING signaling has been shown to drive aging-related inflammation and neurodegeneration.¹²⁹ mtDNA activates cGAS in microglia of naturally aged mice. Microglial-specific expression of a gain-of-function cGas-R241E mutant recapitulates features of aging-related neurodegeneration. STING inhibitor H-151 treatment of aged mice reduces neuroinflammation and improves cognitive function. In addition to neurodegeneration, microglial STING activation is also observed in the context of brain injury. In mice receiving middle cerebral artery occlusion (MCAO), a common model of ischemic stroke, cGAS and STING expression are upregulated concomitantly with activation of STING in microglia.^30,31,55,130 Microglia-specific knockout (Cx3cr1x^CreERT2) of cGAS improves post-injury neuroinflammation and brain pathology.

Effector functions of STING activation in microglia

Few studies examine downstream effectors or mediators of STING signaling in microglia (Figure 3B). STING activation induces hundreds of genes in microglia alone. Which gene products modulate microglial function intrinsically or are secreted and then act on a different cell type is unknown. Take IFN-I as an example: deletion of the receptor (e.g. Ifnar1^−/−) may not rescue neurodegeneration to the same level as Sting1^−/− since many other cell types in the brain respond to IFN-IFNAR1 signaling to maintain homeostasis. The individual contribution of STING-induced inflammatory cytokines is also difficult to define. One proposed model states that STING activation in microglia promotes the secretion of inflammatory cytokines that activate astrocytes into the A1 neurotoxic phenotype, which then kills neurons.²³ Another proposed model states that IFN-I secreted by microglia suppresses the expression of a key transcription factor MEF2C in the neuron, thereby reducing cognitive resilience in AD.²² By and large, the question of how STING activation in microglia leads to neuron death remains a “black box”.

New genetic tools are available to illuminate this “black box”. Selective loss-of-function STING mutant mice (e.g. Sting-S365A, L374A, ΔCTT) can eliminate some STING activities individually or in combination.^122,126,127 Gain-of-function STING mutants (e.g. N153S, V154M) can activate STING signaling “on demand” with cell type specificity without the need for upstream molecular threats. Genetic fate mapping tools can permanently mark brain cells affected by STING activation. In addition, single-cell and single-nucleus RNA-seq analysis can provide a general framework for subpopulations of microglia with different phenotypes. It will be interesting to combine these genetic and single-cell transcriptomic tools to ultimately define how STING activation alters microglia fate and how STING signaling spreads from microglia to other cell types.

STING signaling in astrocytes

There is some evidence for cGAS-STING expression in astrocyte cultures in vitro and that these cells are responsive to exogenous DNA.^131,132 How these expression or response levels compare to other cell types such as microglia is not known. A recent study provided more compelling evidence showing that cGAS-STING is activated and induces senescence in astrocytes of MPTP(1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine)-treated mice (a Parkinson’s disease model); cGAS knockdown in astrocytes in these mice alleviates PD-like pathology.¹³³ Human iPSC-derived astrocytes appear to have a functional cGAS-STING signaling pathway, as iPSC-derived AGS astrocytes spontaneously activate IFN-I and an inflammatory response that can be reduced by a STING inhibitor.^56,134,135 Increased STING expression in astrocytes is observed in mouse brain tissue after traumatic brain injury.¹¹⁹ Overall, comparing to microglia, evidence for STING signaling in astrocytes is less overwhelming and needs further study.

One possibility is that astrocytes “relay” signals from microglia to neurons (Figure 2B). Microglia and astrocytes often communicate via secreted factors in homeostasis and disease. A study using conditioned media in vitro showed that Aβ42 stimulates cGAS-STING signaling in primary mouse microglia, which produces TNFα, IL6 and IL-1α that further activates astrocytes.²³ Then, astrocytes adopt a neurotoxic profile and secrete mediators that cause neuronal cell death. It remains to be tested whether astrocytes are required for relaying STING signaling in vivo. Another example of astrocytes taking on a relay function is HSV-1 infection. HSV-1 activates cGAS-STING in microglia, which then induces the IFN-I response in astrocytes by upregulating TLR3.¹³⁶ Recent scRNA-seq analysis of astrocytes reveals several distinct populations of astrocytes that arise during neuroinflammation. Neurotoxic astrocytes are induced by inflammatory cytokines such as IL-1α, C1q, TNFα and they are capable of killing neurons directly.¹³⁷ A new subset of IFN signature-positive astrocytes is induced by IFN signaling and migrate to the brain periphery where they are likely to protect against further threats.¹³⁸ Therefore, STING-mediated production of soluble mediators can direct different groups of astrocytes to perform different tasks.

Positive and negative effects of IFN-I in the brain

The IFN-IFNAR1 signaling axis can positively or negatively impact on neuronal survival. Mice with constitutively active STING knock-in (Sting^N153S/+) exhibit neuroinflammation and dopaminergic neuron loss that is dependent on both inflammasomes and IFN-I.⁶⁶ Nestin-Cre-mediated neural knockout of Ifnar1 in 5×FAD mice reduces plaque burden and rescues synapse loss.¹³⁹ Blocking IFNAR using blocking antibodies also reduces engulfment of synapses by microglia.¹³⁹ However, mice lacking Ifnb1 and Ifnar1 develop spontaneous Parkinson’s disease-like phenotypes.^140,141 Neurons from these mice develop functional defects and accumulate αSyn-containing Lewy bodies, suggesting that physiological IFN-I levels are neuroprotective. Inflammation-induced IFN-specific astrocytes migrate to the periphery of the brain and are potentially protective.¹³⁸ What determines the difference between a healthy and a pathogenic level of IFN-I signaling needs further investigation.

STING signaling in neurons

Neuronal loss is the hallmark of neurodegeneration, and this loss can be partially prevented by Sting1^−/− in many disease mouse models (including AD¹⁰⁹, PD^12,52, ALS²⁰, NPC⁷⁵, TBI¹⁴², encaphalopathy¹⁴³). Neurons affected by STING activation in the brain include dopaminergic neurons, cortical neurons, Purkinje neurons and possibly others. However, STING expression in neurons is generally low compared to microglia at homeostasis. The one exception is Purkinje neurons, where STING protein is detectable and higher than other neurons in mouse cerebellum.³⁰ Cortical neurons in mice and iPSC-derived human motor neurons do not express STING.⁵² In brain tissue of mice after TBI and chronic cerebral hypofusion, STING protein levels in neurons increase.^119,144 STING activation in neurons, determined by pSTING co-staining with neuronal markers, is observed in TBI and encephalopathy. In principle, many of the IFN-independent activities of STING such as calcium signaling, ER stress and endolysosomal trafficking, could alter neuronal function or survival. The physiological function of STING activation in neurons is unknown.

Interestingly, STING activation and IFN-I signaling are also involved in the regulation of nociception and pain sensing in the peripheral nervous system (Figure 3C). Whether the effect is positive or negative remains controversial. Donnelly et al.¹⁴⁵ found that intrathecal injection of a STING agonist increases mechanical sensory thresholds without affecting motor functions. Intrathecal injection of a STING agonist also shows analgesic potential in chemotherapy-induced peripheral neuropathy and bone cancer pain. Sting1 knockout mice and mice treated with STING inhibitor show hypersensitivity to mechanical and cold stimuli, again suggesting that STING activation is antinociceptive. The peripheral sensory neuron-specific deletion of Sting1 (Nav1.8-Cre) partially reduced the antinociceptive effect mediated by the STING agonist. Contrary to these findings, several other reports suggest that STING activation induces hyperalgesia in neuropathic pain and tumor-induced bone pain.^146-148 STING activation and IFN-I signaling can also reduce opioid-induced itch.¹⁴⁹ These studies raise the exciting possibility that STING signaling could extend from the central nervous system to the peripheral nervous system.

STING signaling in other cells

In the APP/PS1 mouse model of Alzhemer’s disease, endothelial cells adopt an IFN-I siganature.¹⁵⁰ Treatment of brain endothelial cells with IFN-I results in decreased adherent junction protein and increased leakage of the blood-brain barrier (BBB). Whether this pathogenic IFN-I response in brain endothelial cells is triggered by STING activation is unclear. In the developing brain, STING is found to be expressed in the cerebral cortex around E14.¹⁵¹ The role of STING in neurodevelopment is largely unexplored.

STING signaling in brain cancer

Brain tumor is a complex disease state in which tumor-stroma and tumor-immune interactions take place in an unusual environment full of cells with dynamic properties. The most common and deadly brain tumor is glioblastoma (GBM). GBM is an immunologically “cold” tumor and remains highly refractory to currently available immunotherapy.¹⁵² Recent studies suggest that downregulation of the cGAS-STING pathway is an immune evasion strategy exploited by GBM. STING expression is epigenetically silenced in GBM tumor cells, which could be restored by DNA methyltransferase inhibitors.¹⁵³ Most GBM tumors lose cGAS expression due to mechanisms yet to be defined.¹⁵⁴ Intriguingly, STING protein is detected in myeloid cells as well as endothelial and stromal cells in the tumor microenvironment of GBM, but absent in tumor cells.^153,155 Intratumoral administration of STING agonists has demonstrated anti-GBM T-cell or NK cell-mediated immunity in various preclinical models.^155-158 Therefore, STING activation in non-tumor cells in TME is likely critical for eliciting an anti-tumor response.

STING signaling may also increase GBM immunogenicity. Loss of CHEK2 in GBM activates the STING pathway and enhances antigen presentation.¹⁵⁹ CD47 blockade in combination with temozolomide increases tumor cell phagocytosis and antigen presentation via activation of cGAS-STING.¹⁶⁰ Genetic ablation or pharmacological inhibition of the catalytic subunit of protein phosphatase-2A (PP2Ac) increases dsDNA production and cGAS-STING signaling in glioma cells.¹⁶¹ Tumor Treating Fields (TTFields), an approved noninvasive electric field-based modality for GBM treatment, causes focal disruption of the nuclear envelope, resulting in cytosolic release of large micronuclei that activate the cGAS-STING-IFN pathway and enhance anti-tumor immunity in GBM.¹⁶² Therefore, there are two main approaches with great potential for effective immunotherapy of GBM: targeting STING in non-tumor cells within the GBM-TME or restoring the tumor-intrinsic STING pathway.

While primary brain tumors are relatively rare, brain metastases are more common. An estimated 20-40% of patients with advanced cancer develop brain metastases, with a grim prognosis and almost invariably fatal.^163,164 In contrast to its anti-tumor activity, one study found that STING signaling promotes brain metastasis. Infiltrating cancer cells transfer cGAMP into astrocytes across gap junctions, thereby activating STING-mediated production of IFN-I and proinflammatory cytokines that support metastatic tumor growth and chemoresistance.¹⁶⁵ Astrocyte-derived IFN signaling induces the production of chemokine CCL2, resulting in further increased recruitment of immunosuppressive monocytic myeloid cells.¹⁶⁶ When we should activate or inhibit STING signaling in brain tumors and brain metastases needs further study.

HUMAN RELEVANCE

Abnormal DNA accumulation in human samples.

Recent studies have begun to examine the cGAS-STING pathway in postmortem brain tissues or primary cells from human patients. One set of efforts was to look for abnormal DNA accumulation. Cytoplasmic mtDNA is significantly increased in Alzheimer’s disease patients compared to age-matched control fibroblasts.⁵¹ Parkinson’s disease patients with PRKN/PINK1 mutations have higher levels of serum cell-free mtDNA compared to healthy controls.¹⁶⁷ Cytoplasmic micronuclei containing DNA fragments are significantly higher in Huntington’s disease striatal neurons derived from hESCs than in neurons derived from unaffected hESCs.⁶³ These are important evidence demonstrating that molecular threats activating cGAS-STING in mice are also present in human diseased tissues. In addition, it has been found that amyloid fibrils, neurofibrillary tangles and senile plaques also contain nucleic acids.^168-171 Furthermore, plaque-associated nucleic acids strongly correlate with activated ISG-expressing microglia in neuritic plaques in postmortem brains of Alzheimer’s disease patients.¹⁷¹ These observations are highly reminiscent of those seen in autoimmune diseases such as lupus: nucleic acid-containing protein complexes are highly immunogenic and often the cause of tissue inflammation and damage.

cGAS-STING expression and activation in human samples.

Another set of effort was to look for cGAS-STING expression and activation markers. The expression of cGAS and STING is frequently upregulated in the brain of human patients with Alzheimer’s disease, Parkinson’s disease, amyotrophic lateral sclerosis, and traumatic brain injury.^{52,119,172,173} Not only brain glial cells, neurons and brain endothelial cells also have higher levels of STING protein in neurodegenerative brain tissues compared to healthy individuals.¹⁷² Elevated levels of cGAMP are found in spinal cord samples from TDP-43 ALS patients, providing direct evidence of cGAS activation.²⁰ The phosphorylation levels of STING, TBK1, IRF3 and p65 are significantly increased in the brain of human Alzheimer’s disease patients compared to healthy brain tissue.^22,23 Tau stimulation induces STING phosphorylation at Ser366 in human microglia-like cells derived from iPSCs.²² Downstream IFNα and IFNβ expression are also increased in brain tissue of human patients with Alzheimer’s disease, Parkinson’s disease, amyotrophic lateral sclerosis, and traumatic brain injury.^{89,119,171,173-176}

Considerations for future human studies.

The sample size in each of these studies remains very small. Larger scale analyses of human brain tissue is needed to better assess the extent of cGAS-STING activation in any neurological disease. Clinical and gene expression data from the Mt. Sinai Brain Bank show that ISGs are highly upregulated in human Alzheimer’s disease brain tissue.¹⁷¹ Single-cell RNA sequencing studies in human Alzheimer’s disease brains have identified microglial subsets with IFN-responsive gene signatures.¹⁷⁷ It is tempting to postulate a role for cGAS-STING in these elevated IFN/ISG signatures. However, we should remember previous lessons and rigorously establish causality.

THERAPEUTIC DEVELOPMENT AND CONSIDERATIONS

STING antagonists and preclinical studies in mouse models of neurological diseases

STING agonists have been developed by many pharmaceutical companies to treat cancer. In comparison, the development of STING antagonists is still in its infancy. This is largely due to: 1) the lack of a well-defined motif that ensures inhibition since STING is not an enzyme; 2) the lack of a large human patient population that would benefit from STING inhibition. Recent advances in STING biology in autoinflammatory and neurological diseases put the development of STING antagonists in the spotlight. Several small molecule antagonists are reported (reviewed in^178,179). The best example is H-151, which covalently modifies STING palmitoylation residues, preventing its oligomerization and signaling activation.¹⁸⁰

H-151 treatment alleviates inflammatory disease in Trex1^−/− mice, which is a gold standard for STING-mediated systemic autoinflammation. H-151 has also been tested in several mouse models of neurodegenerative diseases. For example, H-151 improves Aβ pathology and reduces reactive gliosis in 5×FAD mice.²³ H-151 and C-176 (a mouse STING-specific covalent inhibitor similar to H-151) also suppresses the IFN-I response and neuroinflammation in several ALS mouse models such as C9orf72^−/−, SOD1^G93A, and Prp-TDP-43^Tg/+ mice.^20,21,89 H-151 treatment of Ataxia-Telangiectasia brain organoids inhibites DNA-triggered senescence-associated secretory phenotype (SASP) and neurodegeneration and improved A-T neuropathology.⁴⁵ Treatment with C-176 alleviates neuroinflammation caused by brain injury in animal models of ischemic stroke,³⁰ traumatic brain injury,¹⁸¹ and subarachnoid hemorrhage.¹⁸²

Challenges in developing a STING antagonist for human use

Target sites.

The first challenge in developing a STING antagonist for human use is the lack of STING-specific target sites. The two main strategies for targeting STING are: 1) competitive binding to the ligand-binding pocket, and 2) covalent modification of cysteine palmitoylation residues. The native ligand cGAMP binds to the STING ligand-binding pocket with an extremely high affinity of 4 nM, which makes competitive inhibition very difficult. The two cysteine palmitoylation residues are located in the multi-pass transmembrane domain and lack any distinct structural features to ensure selectivity. Most cysteine-reactive covalent STING antagonists are also likely to react to cysteine residues in many other proteins. This could be a major obstacle to future development.

A recent study identified a cryptic pocket in the transmembrane domain of human STING that promotes oligomer formation once bound to a small molecule agonist C53.¹⁸³ Binding of C53 to this new pocket increases the cGAMP-stimulated IFN response but, unexpectedly, inhibits STING-mediated autophagy. The latter inhibitory activity was based on recent discoveries that STING can act as a proton channel.^184,185 STING activates non-canonical autophagy by inducing proton leakage from the Golgi or Golgi-derived vesicles. C53 binding to STING transmembrane domain blocks the channel for proton leakage. Therefore, this new transmembrane pocket could serve as a target for STING antagonists but it would only inhibit autophagy activity but not IFN activity.

In addition to small molecule inhibitors, targeted protein degradation has also been attempted, albeit with limited success. The first STING proteolysis-targeting chimera (PROTAC) combines a STING inhibitor C-170 with an E3 ligase ligand pomalidomide, which successfully promotes STING degradation in THP-1 cells.¹⁸⁶ The degradation potency of STING PROTAC is moderate (DC₅₀ = 3.2 μM), probably because STING is a transmembrane protein that is not ideal for PROTAC. STING naturally migrates from the ER to the Golgi then to the lysosome for degradation. Therefore, utilizing the autophagy-lysosome degradation pathway by autophagy-targeting chimera (AUTAC) may represent a better strategy for inhibiting STING.¹⁸⁷

Bioavailability and CNS activity.

The second challenge of STING antagonists is bioavailability and BBB permeability. Most STING antagonists must be administered interperitoneally to mice. Oral bioactive compounds are better suited for preclinical studies, especially in neurodegenerative diseases that require treatment over a longer period of time. C-176 and H-151 have been successfully tested in mouse models of several neurological diseases, suggesting that they are active in the brain. However, their pharmacokinetics in vivo and in the brain are unknown. Another competitive inhibitor of STING, SN-011, reduced brain injury and inflammatory gene expression in rats and mice with acute cerebral ischemia, indicating positive CNS activity (patent CN112057443).¹⁸⁸ The BBB permeability of other published STING antagonists is unknown.

Common STING1 variants.

A lesser-known challenge in the development of STING antagonists is the heterogeneity of the human STING1 gene (also known as TMEM173). Most STING antagonists work well in mouse cells that have only one Sting1 allele. In humans this is more complex. There are several STING1 variants that are present at very high frequency in the human population (Figure 4). Two of these, HAQ (R71H-G230A-R293Q) and H232, encode STING proteins with attenuated activity compared to R232 STING (“wild-type” or “reference” allele).^189-192 For example, only 50% of Europeans are homozygous for the STING wild-type allele (R232/R232). The other half of the European population carries either one or two copies of the attenuated STING1 allele. In East Asia, the dominant STING1 genotype is actually HAQ/R232 (34% of the population), with R232/R232 ranking the second (22% of the population) and HAQ/HAQ homozygous present in 16% of the population. In fact, the THP-1 monocyte cell line is homozygous HAQ/HAQ, which, despite its popularity in immunology research, does not fully represent human STING1 alleles. These common STING1 variants have presented challenges for the development of STING agonists. It is not yet clear how they would affect the performance of STING antagonists. We should bear this in mind when screening or developing compounds in human cell lines carrying different STING1 alleles. Success in a preclinical mouse model is necessary, but it also needs rigorous validation against all common STING1 variants before considering human clinical trials.

Figure 4. Human STING common alleles.

A. CryoEM structure of a STING dimer in the cGAMP-bound active state (ref.¹⁹³). The cGAMP structure is not showing to better visualize the R232 residue. Dotted circle indicates cGAMP binding site. B. Frequency of STING alleles in each ethnic group (top). Data is summarized from ref.¹⁸⁹. “OTHERS” include various combinations of non-R232 alleles, such as H232/HAQ, AQ/HAQ, H232/AQ, etc.

Risk of inhibiting STING signaling

A major side effect of any immunosuppressive therapy is an increased risk of infection. Patients with neurological diseases may also need to be treated for a longer period of time that may further increase the risk. There are several reasons for optimism: 1) At the innate immunity level, many of the microbial pathogens recognized by cGAS-STING are also recognized by several other innate immune sensors. There are also many redundancies at the adaptive immunity level. Therefore, treatments with cGAS- or STING-specific inhibitor may not significantly increase the risk of infection. 2) A partial reduction in STING activity in humans is probably safe or at least tolerable. This is based on the relatively high frequency of STING1 alleles that encode STING protein with attenuated activity in the human population. Further, whole-body Sting1^−/− mice have no developmental or growth defects. These observations suggest that target inhibition of STING in humans bear relatively low risk. Nonetheless, the long-term safety studies are essential.

FUTURE PERSPECTIVES

STING signaling has a long history in neurological diseases, since the first report of the TREX1-associated neuroinflammatory disease AGS in 2006 and the discovery of STING in 2008. However, during the first decade of STING research, we did not learn much about STING signaling in the brain, but we have learned a lot about the functions and mechanisms of STING signaling in the innate immune system. Building upon that knowledge, interest in STING neurobiology has exploded in the past 5 years. As we have summarized here, today we have a wealth of compelling evidence that not only is STING activated in many neurological diseases, but that STING signaling is also essential for the development of neuropathologies and neurological symptoms. Mechanistically, many questions remain unanswered. We have some understanding of the molecular threats in the brain that trigger STING signaling in each neurological disease. However, the signaling cascade and cellular relay through which STING drives neuropathology are not yet defined. How different signaling activities of STING affect neuronal and glial cell function also need further investigation. We need more human studies with larger cohorts. It will also be interesting to investigate whether the distribution of common STING1 alleles is skewed in the aging population with neurodegenerative diseases. In addition, the mammalian brain expresses very high levels of IFN-I receptors in many cell types, so how positive and negative functions of IFN-I signaling maintain balance is also an interesting question. Other innate immune pathways that trigger similar IFN-I and inflammatory responses should also be investigated in the brain. From a therapeutic perspective, the STING-stimulated IFN-I or ISG gene signature, if preserved in peripheral blood as in the brain, can be used as a biomarker to aid clinical diagnosis or monitoring. STING also represents one of the few “immune targets” that are shared amongst several neurological diseases, as well as autoimmune and inflammatory diseases. Many industrial and academic groups are very interested in developing STING antagonists. It will be exciting to see how these efforts unfold and are ultimately translated into the clinics.