← Back to all posts

The ferroptosis–cGAS-STING connection: how cell death activates immunity, and when it doesn't

Disease Biology and Mechanism Intelligence Anika Shivhare · · 6 min read
Share:
The ferroptosis–cGAS-STING connection: how cell death activates immunity, and when it doesn't

Ferroptosis is an attractive idea for oncology because it kills cells that survive apoptosis. The immunological corollary has been just as attractive: a cell dying by iron-dependent lipid peroxidation ruptures, spills its contents, and those contents should alert the innate immune system. The cytosolic DNA sensor cGAS and its adaptor STING are the obvious receiver, since they detect misplaced DNA and drive type I interferon.

The mechanistic links between the two pathways are real and now reasonably well mapped. What the literature does not support is the clean story that ferroptosis is an immunogenic form of death. Several well-controlled studies point in opposite directions, and the contradictions are informative for anyone designing a ferroptosis-inducing therapy.

The two pathways, briefly

Ferroptosis was named in 2012, when erastin was shown to kill RAS-mutant cells through a route distinct from apoptosis, necrosis and autophagy, dependent on intracellular iron and blocked by ferrostatin-1 (Dixon et al., 2012). Mechanistically it comes down to a failure of phospholipid hydroperoxide clearance, with GPX4 as the central enzyme.

cGAS-STING is the cytosolic DNA surveillance arm of innate immunity. cGAS binds double-stranded DNA, synthesises cGAMP, and cGAMP activates STING at the ER. STING then has to be palmitoylated and trafficked to the Golgi before it can signal, a detail that turns out to matter here.

Ferroptotic cells release DNA that STING can see

The most direct connection is that ferroptotic damage produces oxidised nucleic acid in the extracellular space.

In a pancreas-specific Gpx4 knockout mouse crossed onto a Kras G12D background, loss of GPX4 or a 1% carbonyl iron diet increased release of 8-hydroxy-2′-deoxyguanosine (8-OHG) and accelerated PanIN formation, stromal response and metastasis (Dai et al., 2020). The 8-OHG, not the lipid peroxidation product 4-HNE, induced macrophage migration and IL-6/NOS2 expression at pathological concentrations. Knocking out Tmem173 (STING) abolished the macrophage response and protected the mice; knocking out TLR9 did not. So the sensor was STING, the ligand was oxidised DNA, and the mediating cell was the macrophage.

The route by which that DNA reaches the cytosol has since been worked out for the mitochondrial pool. Opening of the mitochondrial permeability transition pore, dependent on cyclophilin D, drives the mitochondrial swelling seen in ferroptosis and lets oxidised mtDNA escape into the cytosol, where it activates cGAS-STING (Zhou et al., 2026). In that study the downstream consequence was not interferon signalling but ferritinophagy, which liberates more labile iron and amplifies the ferroptotic signal. Blocking mtDNA repair sensitised cells to ferroptosis and synergised with a ferroptosis inducer against mouse xenografts.

That is a feed-forward loop, and it is worth noting what it is not. The STING output here served the death pathway, not an immune response.

STING also promotes ferroptosis, without sensing anything

The traffic runs both ways, and at least one direction appears to be independent of DNA sensing altogether.

In PANC1 pancreatic cancer cells, erastin relocates STING1 from the ER to mitochondria in a mitochondrial-ROS-dependent manner (blocked by MitoTEMPO and MitoQ). There STING1 binds MFN1 and MFN2, promotes mitochondrial fusion, and raises mtROS and lipid peroxidation. STING1 knockdown reduced erastin-induced death; MFN1/2 knockout in MEFs reduced death, mtROS, lipid peroxidation and HMGB1 release (Li et al., 2021). In xenografts, the erastin analogue IKE suppressed STING1- or MFN1/2-knockdown tumours less effectively.

The authors did not test cGAS or interferon output, and erastin caused no appreciable mtDNA damage in their hands. Read conservatively: this is a structural, non-immunological role for STING1 at the mitochondrion that happens to modulate ferroptosis sensitivity. Whether it generalises past pancreatic lines is open.

The inconvenient result: lipid peroxidation shuts STING down

Here is the finding that complicates every "ferroptosis activates STING" narrative.

In primary mouse macrophages, Gpx4 deletion increased lipid peroxidation and inhibited cGAS-STING signalling. The mechanism is specific: 4-HNE carbonylates STING at Cys88, blocking the palmitoylation required for ER-to-Golgi trafficking and therefore activation (Jia et al., 2020). Functionally, GPX4-deficient mice mounted weaker antiviral responses and supported more HSV-1 replication. The same system was probed with the GPX4 inhibitors RSL3 and FIN56 and with erastin, so the effect is not an artefact of genetic deletion alone.

So the same biochemical event that defines ferroptosis, the accumulation of lipid peroxidation products, covalently disables the adaptor the field expects it to activate. The two results are not formally incompatible. Dai et al. looked at a recipient macrophage sensing oxidised DNA released by a dying neighbour, while Jia et al. looked at STING inside the peroxidising cell itself. But it means the net effect in a tumour depends on which compartment the lipid peroxidation is in, and that is not something you can infer from a ferroptosis marker panel.

Ferroptotic cells make poor vaccines

The immunological bottom line has been tested directly. Using MCA205 fibrosarcoma with inducible Gpx4 knockdown to synchronise death, ferroptotic cells failed to protect mice in prophylactic vaccination at any stage of the death process, despite releasing ATP, HMGB1 and chemokines (Wiernicki et al., 2022). Early-stage ferroptotic cells actively suppressed dendritic cell maturation and were not engulfed; DCs that did take up ferroptotic cargo showed reduced NF-κB and Ccr7 and impaired cross-presentation. Adding ferroptotic cells to an immunogenic apoptotic vaccine reduced its protective effect.

The authors' conclusion is blunt: ferroptosis is not an immunogenic cell death modality in this system. Their proposed explanation, that oxidised phospholipids and lipid droplet accumulation suppress antigen-presenting function, is a hypothesis carried over from prior work rather than something they tested directly.

What this adds up to

A defensible synthesis, stated as reasoning rather than established fact:

  • Ferroptosis reliably generates STING ligands. Both extracellular 8-OHG and mPTP-released oxidised mtDNA are documented.
  • Whether STING activation follows depends on where the lipid peroxidation sits. In the dying cell, 4-HNE adducts suppress STING. In a bystander macrophage receiving oxidised DNA, STING fires.
  • STING output is not necessarily interferon. One documented consequence is ferritinophagy that amplifies ferroptosis; another is MFN-dependent mitochondrial fusion. Neither is an immune response.
  • Context determines whether this helps the host. In Kras-driven pancreatic disease, the ferroptosis/8-OHG/STING/macrophage axis promoted tumorigenesis, and GPX4 behaved as protective.

For translational work, the practical caution is that "induces ferroptosis" and "turns the tumour hot" are separate claims requiring separate evidence. A ferroptosis inducer combined with a STING agonist is not obviously additive: the inducer may be carbonylating the target of the agonist. Measuring interferon output, not just lipid peroxidation markers, is the minimum.

The open questions are concrete. Does the Cys88 carbonylation mechanism operate at the lipid peroxidation levels reached in tumours treated with clinical-stage inducers, or only at Gpx4-null extremes? Does the MFN1/2 axis hold outside pancreatic cells? And given that the Wiernicki vaccination result used one synchronised model, does slower, asynchronous ferroptosis in a real tumour behave the same way? The reviews of this crosstalk are accumulating (Archives of Toxicology, 2025), but the decisive experiments are compartment-resolved ones that the field has largely not done yet.

Explore Purna's Molecular Intelligence Platform

AI-powered workspace for biology teams to accelerate drug discovery from target identification to lead optimization.

Try Purna AI →

Also Read

Stay Updated

Get the latest insights on molecular intelligence and AI-driven drug discovery delivered to your inbox.

We email once every two weeks. No spam.