A virtual screen of 59,319 peptides turned up one that binds STING and turns its alarm down. That is the opposite of what most people want from STING, and that is the point.
STING, short for stimulator of interferon genes, is a sensor inside cells. When stray DNA shows up where it should not be, an upstream enzyme called cGAS makes a small signal molecule, cGAMP, that switches STING on. STING then drives interferon and inflammatory genes, the body's response to viruses and to damaged, DNA-leaking cells. That response is useful in short bursts. When it runs constantly, it becomes a driver of autoimmune and inflammatory disease, from lupus to rare inherited conditions like STING-associated vasculopathy (SAVI) and Aicardi-Goutières syndrome, and a brake on STING is something drug developers have wanted for years.
The screen, reported in the Journal of Enzyme Inhibition and Medicinal Chemistry ↗, ranked the whole peptide library against STING by molecular docking, a computational method that estimates how well each candidate fits the target's pocket. The four best-scoring peptides were made and tested for real binding by microscale thermophoresis, an assay that measures how a molecule moves along a heat gradient when a partner is bound. The lead, called Peptide-1, held onto STING with a dissociation constant of 0.15 micromolar, meaning it binds tightly enough to matter at the low concentrations a real inhibitor would use.
Binding is only the first bar. In mouse macrophages, both the RAW264.7 line and fresh bone-marrow-derived primary cells, Peptide-1 cut cGAMP-driven production of two inflammatory messengers, interferon-beta and interleukin-6, and did so in a dose-dependent way at both the protein and the messenger-RNA level. It also lowered the phosphorylation of STING itself and of IRF3, the transcription factor STING hands the signal to, which is the readout you would expect if the peptide is genuinely damping the pathway rather than doing something off-target. Up to 10 micromolar it did not appear to kill the cells.
The mirror image of a peptide we covered
Peptidemodel ran a piece last month on a designed peptide built to do the reverse. That molecule, a six-residue sequence wrapped around a manganese ion, forced cancer cells to leak their own mitochondrial DNA and trip the cGAS-STING alarm on purpose ↗, making an immunotherapy work better in mice. Same pathway, opposite direction. One side wants STING roaring so the immune system attacks a tumor. The other wants it quiet so it stops attacking healthy tissue. That both goals are now being chased with peptides, rather than small molecules or antibodies, says something about how much of this target's surface is a flat protein-protein interface, the kind of terrain peptides handle better than pills.
Early, and honest about it
This is a discovery paper, and it stops at the cell dish. There are no animal results, the potency numbers come from a lab assay and not a living system, and the lead is still called Peptide-1 rather than a named clinical candidate. Docking-first screens also carry a known risk: a molecule that scores well against a static model of the pocket does not always behave in a cell, which is exactly why the binding and macrophage steps here matter more than the screen that pointed to them.
What the work adds is a concrete, submicromolar peptide handle on a target that has mostly been approached with small molecules, plus a clean demonstration that a large virtual library can be narrowed to a functional STING inhibitor without a high-throughput wet screen. Neither Peptide-1 nor its three runners-up has a card on peptidemodel, which is expected for sequences this new. STING sits on the immune ↗ target shelf.