Initial intuition
Duncan-Lowey et al. (Cell 2023) characterized the bacterial RADAR supramolecular complex by cryo-EM—an RdrA ATPase associated with an RdrB deaminase that edits double-stranded RNA (A→I, identical to human ADAR1/ADAR2). This is one of the clearest functional parallels in the entire anti-phage arsenal: the same deamination chemistry, the same substrate (dsRNA), and structurally superimposable catalytic folds.
ADAR1 is a therapeutic target of major interest:
- Its editing of human Alu transcripts masks endogenous RNA from MDA5, avoiding interferon-induced autoimmunity.
- Reducing its activity in infected cells exposes viral RNAs to MDA5, potentiating the type I interferon response—this is a clean host-directed antiviral mechanism.
- Selective ADAR1 inhibitors are actively sought (recent Phase I clinical trials: Synthekine, AIRNA, but few validated hits).
Hypothesis
The RdrB-ADAR1 structural conservation (Duncan-Lowey 2023) is sufficient for a virtual screening on bacterial RdrB to identify chemical scaffolds transferable to human ADAR1. The target is not the direct catalytic site (notoriously difficult to drug for deaminases) but allosteric pockets:
- RdrA-RdrB interface in the supramolecular complex—ADAR1 has an equivalent dsRNA binding domain
- dsRNA recognition site—competitive modulators
- Dimerization surface of the deaminase domain
An RdrB hit that transfers to ADAR1 with > 5× selectivity vs ADAR2 (the cerebral paralog, whose inhibition would be toxic) would be a first-class pharmacological probe for the community.
In silico plan
-
Reference structures:
- AlphaFold RdrB (
A0ABU8PPW6,A0AAN0NNX4)—true RADAR positives according to UniProt - Human ADAR1 (PDB 6VFF, 5ED1, 5ED2)—crystallographic structures of the deaminase domain + dsRBD
- Human ADAR2 (PDB 6VFF in complex with dsRNA—selectivity)
- Complete RADAR cryo-EM (PDB 8UV4, Duncan-Lowey 2023) if available
- AlphaFold RdrB (
-
Allosteric pocket identification: FPocket / SiteMap on the non-catalytic surfaces of RdrB. Specifically target regions conserved with ADAR1 (structural alignment via Mol*).
-
Virtual screening: drug-like ChEMBL library (~2M compounds) AutoDock Vina docking on 2-3 selected allosteric pockets. Top 1% redocking on the equivalent ADAR1 pockets.
-
Selectivity and toxicity filters:
- Docking score on ADAR2 (ratio > 5)
- ADMET profile (DEREK, ADMET-AI) excluding hERG cardiotoxicity
- No overlap with APOBEC inhibitors (other essential human deaminases)
-
100 ns MD on the top 30 hits to confirm the conformational stability of the binding modes.
Limitations and risks
- ADAR allostery is poorly characterized: the few published inhibitors are either competitive (thus non-selective) or speculatively allosteric without mechanistic validation. In silico prediction of active allosteric pockets is notoriously noisy.
- The RdrB → ADAR1 transfer assumes structural conservation of allosteric sites beyond the catalytic site. It is plausible but not guaranteed—enzymes can diverge more on regulatory surfaces than on active sites.
- ADAR1 inhibition carries a significant autoimmunity risk (cf. ADAR1-deficient patients who develop Aicardi-Goutières type interferonopathies). An inhibitor must be used in short courses, in the context of an acute infection—not in chronic modulation. The screen should prioritize reversible hits with a short duration of action.
Links to the Bactaegion scope
V1 Family RADAR (high translational priority, host-directed modality). Seminal paper Duncan-Lowey 2023 (Cell, DOI 10.1016/j.cell.2023.10.013). Strategy aligned with the family sheet: RdrB-ADAR1 structural alignment, prioritize compounds modulating host RNA editing to exacerbate viral recognition.
Note: this lead is explicitly host-directed (human ADAR1 target), not pathogen-directed. It falls into the same category as the Schlafen → SLFN11 lead, where the bacterial protein serves as a chemical template to reach a homologous human target.