Disentangling the inflammatory and antiviral arms of cytoplasmic DNA sensing

When double-stranded DNA leaks into the cytoplasm of a mammalian cell — from a virus, a ruptured nucleus, or damaged mitochondria — the cell does not mount one response but at least two competing ones. The better-known arm runs from the enzyme cGAS through its adaptor protein STING to the transcription factors IRF3 (which drives antiviral interferons) and NF-κB (which drives inflammatory cytokines). A second, recently discovered arm runs through Apaf-1, a protein previously known only as a mediator of programmed cell death, which binds the stray DNA directly, recruits the signaling protein RIP2, and pushes the cell toward NF-κB-driven inflammation instead of interferon output or death (Apaf-1 is an evolutionarily conserved DNA sensor that switches the cell fate between apoptosis and inflammation). These two arms disagree about what the cell should do, and every study we could find measures only one arm at a time, in separate assays or separate papers. Nobody has watched both outputs in the same cells under the same DNA stimulus, so we cannot currently answer the clinically central question: if you pharmacologically inhibit NF-κB — the arm responsible for cytokine storms and chronic inflammation — does cytoplasmic DNA signaling collapse, or does the signal simply reroute into the antiviral interferon arm?

This question matters because broad NF-κB shutdown is immunosuppressive and toxic, and because the same STING signal is protective in one setting (USP7 Inhibits Osteoclastogenesis via Dual Effects of Attenuating TRAF6/TAK1 Axis and Stimulating STING Signaling) and destructive in another (KAE ameliorates LPS-mediated acute lung injury by inhibiting PANoptosis through the intracellular DNA-cGAS-STING axis). The unresolved issue is not whether these pathways exist but how a cell arbitrates between them.

Core Claim

The inflammatory (NF-κB) and antiviral (IRF3/interferon) transcriptional responses to a single cytoplasmic DNA stimulus are set by competition between the Apaf-1–RIP2 and cGAS–STING sensor arms over a shared, limiting pool of downstream signaling proteins (candidates include TRAF6 and the IKK kinases). This competition makes the system redirectable: selectively inhibiting the NF-κB-convergent nodes should shunt dsDNA-driven signaling into IRF3/interferon output rather than extinguish it. We will test this with a dual-arm ratiometric reporter cell that measures both transcriptional arms simultaneously in the same living cells, with co-immunoprecipitation (Co-IP — an antibody-based method that pulls a bait protein out of lysed cells together with whatever physical partners it holds) to test the arms' physical competition for shared transducers, and with a panel of NF-κB-pathway inhibitors to test redirectability.

Research Program

The first deliverable is a reporter cell line carrying three integrated transcriptional reporters at low copy number: firefly luciferase under NF-κB response elements, Renilla luciferase under ISRE (the interferon-stimulated response element that reports IRF3 activity), and a constitutive third luciferase for well-level normalization. This is the dual-luciferase logic established by Measuring Transcriptional Activity with Dual Luciferase Assays and Wei Sun, who showed the firefly-to-Renilla ratio is required to separate true regulation from transfection noise, drug toxicity, and temperature artifacts — but extended, for the first time in this system, to two inducible arms rather than one inducible plus one constitutive. Integration rather than transient transfection avoids the episomal and transcription-factor-titration artifacts that the field itself flags as distortions of reporter assays. We will build this in cGAS-competent A549 cells, the line in which the Apaf-1 paper established cGAS-independent cytokine induction, and cross-validate in the 293-Dual hSTING-R232 reporter cells used by Distinct pathogenic mutations in ARF1 allow dissection of its dual role in cGAS-STING signalling.

The second deliverable is a perturbation matrix on this line: CRISPR knockouts of Apaf-1, cGAS, STING, and RIP2, alone and in pairs, crossed with the pharmacological inhibitors already validated in the literature — the STING inhibitor H-151 and the TBK1 inhibitor amlexanox used in the ARF1 study, plus inhibitors of IKK and of the RIP2/NOD2 axis — all under titrated doses of HSV60 dsDNA, the exact ligand whose affinity for Apaf-1 was measured at $K_d = 1.71 \times 10^{-7}\,\text{M}$ (Apaf-1...). The primary metric is the ratio $R$ of normalized NF-κB to ISRE output and its temporal trajectory. The rivalry hypothesis predicts that Apaf-1 loss shifts $R$ toward the interferon arm, and conversely that STING blockade leaves NF-κB output sustained (via Apaf-1) rather than abolished.

The third deliverable is physical: Co-IP of Apaf-1–RIP2 and STING–TBK1 complexes under DNA challenge, probing each for co-association with TRAF6 and IKK subunits, with and without the inhibitors above. Because STING is a multi-pass ER/Golgi protein notoriously vulnerable to detergent denaturation and Co-IP can create post-lysis artifacts, every interaction will be cross-validated in intact cells by proximity ligation assay, the fluorescence method in which antibodies carrying DNA oligos light up only when two targets sit within ~40 nm (Ian T. Foster et al.; Yi Fu). A limiting-pool competition assay — raising Apaf-1 dosage and measuring TRAF6 recovery in STING complexes — is the sharpest test of the shared-transducer claim.

As a secondary use of the platform, the same reporter doubles as a functional readout for discovering post-translational regulators that single-output assays miss: for example, depleting candidate E3 ubiquitin ligases and scoring recovery of reporter activity would attack the open question of which host ligase the hepatitis B viral protein HBx recruits to destroy cGAS, a mechanism established by Co-IP and dose-dependent reporter suppression in HBx Inhibits DNA Sensing Signaling Pathway via Ubiquitination and Autophagy of cGAS but left with the ligase unidentified.

Baselines

The naive reference point is the unperturbed DNA stimulus itself: parental A549 cells challenged with HSV60 dsDNA with no genetic or pharmacological perturbation, which defines the cell's natural NF-κB-to-ISRE set point. No reported result found on Lacuna for this ratio in any paper in the file, which is itself the gap this proposal fills; the closest existing number is the Apaf-1 paper's endpoint cytokine readouts (CXCL1, IL-6 induction impaired by Apaf-1 knockout during vaccinia infection, and HSV-1 survival dropping to 20% in Apaf-1 heterozygous knockouts versus 80% in wild-type, from Apaf-1...), which we will use as orthogonal checks that our reporter tracks endogenous behavior.

The strongest existing competitor is the single-arm dual-luciferase assay as practiced in the two closest studies. The HBx paper quantified dose-dependent suppression of an IFN-β promoter dual luciferase (no reported result found on Lacuna for its exact IC50 or percentage inhibition), and the ARF1 paper quantified ISRE-firefly and ISRE-SEAP activity in 293-Dual hSTING cells (no reported result found on Lacuna for luciferase fold-change values). To make the comparison fair, we will run these exact single-arm assays in parallel wells of the same cells, with the same stimuli, the same DNA doses, the same inhibitor concentrations, and the same number of biological replicates and tuning cycles. Our integrated dual-arm construct beats this baseline only if it detects ratio shifts (for example, the predicted Apaf-1-knockout shift toward ISRE) at effect sizes that the separately-run single arms, pooled across wells, cannot resolve — the ratiometric logic of Wei Sun predicts exactly this variance advantage. The functional-screening baseline is the arrayed CRISPR signalome screen of Signalome-wide mapping of the NFκB pathway in T-cells reveals novel targets for immunotherapy, which measured NF-κB–GFP dynamics over 48 hours across a 706-gene library (no reported result found on Lacuna for its numeric AUC values); our extension to DNA inputs and dual-arm output is defined against it. Finally, the key ablation is the split-well version of our own construct: the same two reporters assayed in separate wells rather than one cell. If the split-well version performs as well, the integration was unnecessary and the claim narrows; if it does not, the ablation isolates simultaneity as the ingredient that matters. As an in-silico pre-screen baseline, current computational protein-interaction models achieve Graph Similarity below 0.5 even on human networks (PRING: Rethinking Protein-Protein Interaction Prediction from Pairs to Graphs), so we treat such predictions strictly as Co-IP hypothesis generators, never as evidence.

Why not just…?

Why not just measure endogenous transcripts by RT-qPCR or RNA-seq? The file's own evidence says this fails at precisely the junction we study: HBx dismantled cGAS-mediated immune signaling while cGAS mRNA stayed completely stable, so a transcript-only assay would have missed the whole phenomenon (HBx...), and the Jurkat signalome screen concluded that many critical regulators act post-translationally with no mRNA change (Signalome-wide mapping...). We will still run qPCR as an orthogonal endpoint check, but it cannot see the competition dynamics that define the hypothesis.

Why not just measure released cytokines in the supernatant? Cytokine readouts conflate the two arms — IL-6 can come from the cGAS-STING route or the Apaf-1-RIP2 route — and the therapeutic-suppression literature argues that downstream cytokine neutralization leaves the intracellular NF-κB-to-NLRP3 priming loop intact, so it cannot distinguish routing from shutdown (Therapeutic Suppression of the NF-κB and NLRP3 Pathways). ELISA data are cheap and we will collect them, but they are not decisive for the arbitration question.

Why not just do phosphoproteomics instead of building reporter cells? The signalome screen's authors addressed this directly: signaling fluxes at physiological stimulation are so low and localized that they are nearly invisible to mass spectrometry without massive, non-physiological stimulation, while functional reporters captured the same biology (Signalome-wide mapping...). Phosphoblots of pTBK1 and p-p65 will support specific experiments, but they cannot deliver the quantitative two-arm ratio across time.

Why not just test NF-κB inhibitors in the reporter cells that already exist? This is the most serious objection, because one could put an IKK inhibitor into the ARF1 study's ISRE reporter cells tomorrow. The honest answer is that a single-arm system structurally cannot detect the central phenomenon: whether suppressing NF-κB redirects flux into the interferon arm requires measuring the interferon arm and the NF-κB arm in the same cells under the same stimulus. Our split-well ablation is precisely the experiment that settles whether this added construct is necessary or whether two existing assays run side by side suffice.

Proposed Experiments and Decisive Checks

The decisive checks, in order, are as follows. First, Apaf-1 knockout under HSV60 stimulation should shift the NF-κB-to-ISRE ratio significantly toward the interferon arm, with the magnitude cross-checked against the Apaf-1 paper's own genetic effects (impaired CXCL1/IL-6 induction and the 20%-versus-80% HSV-1 survival benchmark). If the ratio is unchanged, the arms are independent and the rivalry hypothesis is wrong in its strong form — the platform still yields the first joint dose-response atlas of both arms. Second, STING blockade with H-151 should leave NF-κB output partially intact via Apaf-1; total collapse of both arms would falsify the redirectability claim. Third, the Co-IP limiting-pool assay should show that increasing Apaf-1 dosage reduces TRAF6 or IKK recovery in STING complexes, with proximity ligation confirming the effect in intact cells; if the complexes do not share transducers, competition must occur at the DNA or sensor level instead, which the assay will also discriminate. Fourth, the dual-arm-versus-split-well ablation decides whether simultaneous measurement is a genuine methodological advance or an implementation detail. Finally, as a translational endpoint, we will test whether NF-κB-axis inhibition during DNA-virus challenge in the A549 system preserves interferon output, with the in vivo HSV-1 survival benchmark from the Apaf-1 study as the reference point for any future animal extension. Success on checks one through three would establish, for the first time, that the inflammatory and antiviral responses to cytoplasmic DNA are a routed, not a fixed, allocation — and that the routing can be steered.