cGAS-STING Signalling: Pathway, Regulation and Key Assays

Double-stranded DNA in the cytosol is always a mistake. A virus has uncoated, a mitochondrion has ruptured, a micronucleus has burst, or a chromosome has been broken and left unrepaired. cGAS treats all of these identically: it binds the sugar-phosphate backbone without reading a base, makes a cyclic dinucleotide, and hands the alarm to STING. What follows — interferon, inflammation, autophagy or death — is decided by trafficking and by a set of brakes as important as the sensor itself.

Sensing DNA where it should not be

cGAS binds B-form double-stranded DNA through two basic surfaces without reading a single base. Two enzymes and two duplexes assemble into a 2:2 ladder, and that geometry — not simple occupancy — reorganises the catalytic pocket. Long DNA nucleates cooperative laddering and phase separation, so cGAS acts as a length-dependent, switch-like sensor. Its product, 2′3′-cGAMP, carries one 2′–5′ and one 3′–5′ bond, a mixed linkage that binds STING roughly a hundred-fold more tightly than bacterial cyclic dinucleotides.

cGAS is not alone. DDX41 feeds the same STING module, while AIM2 diverts cytosolic dsDNA into an inflammasome instead of interferon, so the two arms compete for one ligand. cGAMP also leaves the cell, crossing gap junctions and entering neighbours via the folate carrier SLC19A1, arming bystanders that never saw DNA. Tumours counter here, using ENPP1 to hydrolyse extracellular cGAMP into immunosuppressive adenosine. Upstream, TREX1 clears cytosolic DNA before cGAS meets it.

STING: the trafficking is the signal

STING sits in the ER as a dimer with a V-shaped cleft. cGAMP closes a lid over it and rotates the ligand-binding domain about 180 degrees relative to the transmembrane region, relieving autoinhibition and permitting oligomerisation. Release requires disengaging its retention factors — STIM1, better known as the ER calcium sensor, moonlights as the anchor holding resting STING in place. STING then travels by COPII to the ERGIC and Golgi, where palmitoylation clusters it into lipid rafts.

At the Golgi the STING C-terminal tail recruits TBK1, which trans-autophosphorylates and marks STING at Ser366. That phospho-serine docks IRF3, which TBK1 phosphorylates on its C-terminal serine cluster; the dimer enters the nucleus and drives IFN-β. IKKε is partly redundant and itself interferon-induced, contributing a second wave through IRF7. Activation and shutdown share one itinerary: ULK1 phosphorylates inhibitory sites and p62/SQSTM1 delivers ubiquitinated STING to the lysosome.

The second output: NEMO, IKK and NF-κB

Trafficking STING also assembles a TRAF6-dependent ubiquitin scaffold recruiting NEMO (IKKγ) with IKKα and IKKβ. IKKβ phosphorylates IκBα on Ser32/Ser36, marking it for proteasomal degradation and freeing NF-κB p65/p50. The output is inflammatory: TNF-α, IL-6 and, in primed myeloid cells, pro-IL-1β. This arm is not a side reaction — the IFN-β enhanceosome needs IRF3 and NF-κB bound cooperatively, so IKK inhibition lowers interferon even when IRF3 is phosphorylated normally.

Kinetics decide the biology. An acute STING pulse in a tumour bed gives an IRF3-dominant interferon burst, dendritic-cell licensing and CD8 priming. Chronic low-grade signalling in the same tissue drifts toward an NF-κB-dominant output that recruits suppressive myeloid cells and supports metastasis. Reading only IFN-β, or only TNF-α, gives opposite conclusions about the same intervention; both arms belong in any agonist study, measured from the same sample.

Amplification: IFNAR, ISGF3 and the ISG output

Secreted IFN-β acts back on the producing cell and its neighbours through IFNAR1/IFNAR2. IFNAR2 carries the high-affinity site and constitutive JAK1; IFNAR1 contributes TYK2 and sets the step discriminating IFN-β from the IFN-α subtypes. Dimerisation brings the kinases into range to trans-phosphorylate and create docking sites. STAT1 and STAT2 are then recruited, phosphorylated and joined by IRF9 to form ISGF3, which reads ISREs rather than the GAS elements bound by STAT1 homodimers.

The map shows the informative products: the ubiquitin-like modifier ISG15, the antiviral GTPase MX1, the RNase-L activator OAS1, the chemoattractant CXCL10, and IRF7, which converts one IFN-β pulse into a self-amplifying multi-subtype response. This matters for quantification. IFN-β protein is transient and often already cleared before tissue is harvested, whereas CXCL10 integrates the response over hours, sits comfortably in ELISA range, and tracks effector T-cell recruitment.

Brakes, inflammasome crosstalk and where this goes wrong

The clinical relevance is almost entirely a story about failed brakes. Biallelic TREX1 loss causes Aicardi–Goutières syndrome, an interferonopathy fully rescued in mice by deleting cGAS or STING; heterozygous variants recur in lupus, and ligand-independent STING alleles cause SAVI. The other route is self DNA from the wrong compartment: BAX/BAK macropores release mtDNA, and micronuclei rupture after mis-segregation — the link between genome instability, radiotherapy and interferon in tumours.

The same DNA feeds the inflammasome branch. AIM2 nucleates ASC filaments directly, while NLRP3 responds indirectly and is transcriptionally primed by the very NF-κB and interferon signals STING produces. Either route activates caspase-1, which matures pro-IL-1β and cleaves gasdermin D into membrane pores; pyroptosis then releases more DNA and closes the loop. Whether a cell makes interferon or dies is largely a race between sensors for one ligand.

Key targets and matching reagents

Studying cGAS-STING in vivo

Most of this pathway is intracellular, so in vivo intervention happens at the secreted outputs. Experiments fall into three groups.

1. Neutralising the secreted output

Functional-grade anti-mouse IFN-β, anti-mouse TNF and anti-mouse IL-1β let you ask which arm a phenotype depends on. Blocking IFN-β while leaving TNF-α intact separates the IRF3 branch from the NF-κB branch far more cleanly than any kinase inhibitor, since TBK1 and IKK compounds cross-react at working concentrations. Use low-endotoxin, azide-free preparations: residual LPS primes NF-κB and the inflammasome directly, manufacturing the signal you are trying to measure.

2. Quantifying the two arms in parallel

Run both axes on the same samples. For the IRF3 branch: IRF3, TBK1 and IKKε upstream, then STAT1, STAT2, IRF9, JAK1 and TYK2 at the receptor, with ISG15, MX1, OAS1 and CXCL10 as durable output. For the NF-κB branch: NEMO, IKKβ, IκBα and p65, with TNF-α and IL-6 secreted. Add IFNAR1 and IFNAR2 when a model looks interferon-resistant rather than interferon-poor.

3. Mapping the brakes and the inflammasome branch

The regulatory nodes are where unexplained phenotypes live. Assay TREX1 and ENPP1 to see how much DNA and cGAMP the system disposes of, SLC19A1 for cGAMP import, and ULK1 with p62/SQSTM1 for the autophagic shutdown limb. STIM1 and DDX41 cover ER retention and the parallel sensor. On the death side, pair AIM2, NLRP3, ASC, caspase-1, gasdermin D and BAX to tell an interferon-dominant response from one tipped into pyroptosis.

Featured products for this pathway

Frequently asked questions

For research use only. Not for use in diagnostic or therapeutic procedures.