Stroke is a leading cause of death and adult disability, yet effective treatments are scarce. Acid-sensing ion channels (ASICs), particularly the human isoform 1a (hASIC1a), play a key role in the pathophysiology of ischemic stroke, where tissue acidification leads to pathological activation and subsequent cell death. Genetic knockout or pharmacological inhibition of hASIC1a with venom-derived peptides such as PcTx1 or Hi1a markedly reduces infarct volumes. However, hASIC1a remain an underexploited drug target due to the lack of potent and selective inhibitors that are easily produced. To address this, we computationally designed miniproteins targeting hASIC1a using a deep learning-based pipeline based on RFdiffusion . The miniproteins were designed to bind to the extracellular domain of hASIC1a, aiming for inhibition or potentiation, subtype selectivity, and high affinity. With ProteinMPNN, an amino acid sequence was fitted onto the miniprotein backbone, followed by filtering via AlphaFold2 metrics. The miniproteins were subcloned, expressed, and purified before being probed in automated patch-clamp experiments. A screen of 77 miniproteins identified five inhibitors and potentiators, some with nanomolar potency. To expand the pool of potential therapeutics, computational binder optimization was conducted based on the successful miniprotein inhibitors. This strategy yielded 72 additional miniproteins, 32 of which exhibited inhibitory activity. The most promising candidates are currently tested in randomized, double-blinded murine stroke models. Notably, this pipeline integrating computational design, scalable expression and purification, and functional validation can be adapted to other membrane protein targets, providing a versatile platform for ion channel research tools and therapeutic discovery.
Sörmann et al. (Sun,) studied this question.