De novo and scaffold-based design of GDF15 binders for cancer cachexia diagnostics and therapeutics

Abstract Growth differentiation factor 15 (GDF15), a stress-responsive cytokine of the TGF-β superfamily, is elevated in cancer cachexia, chemotherapy-induced nausea, and hyperemesis gravidarum, making it both a biomarker and a therapeutic target. Here, we developed high-affinity GDF15 binders using an AI-driven protein design framework. To achieve this, we systematically explored three complementary scaffold-generation strategies: scaffold grafting (SG), diffusion-based de novo design, and scaffold-search and grafting (SSG), identifying distinct advantages - SG rapidly optimized receptor-derived motifs to subnanomolar affinity; de novo diffusion produced topologically novel binders; and SSG enabled access to GDF15’s concave site B by repurposing evolutionary structural analogs from natural complexes. The designed GDF15 binders were translated into two functional modalities. First, a one-step, wash-free luminescent biosensor was created by coupling a de novo binder to split-luciferase fragments, enabling the rapid and sensitive quantification of GDF15. Second, the highest-affinity binder was engineered as an Fc-fusion decoy receptor, thereby effectively neutralizing GDF15 signaling in cell-based assays (IC 50 = 7.2 nM), demonstrating comparable in vitro potency to ponsegromab, a monoclonal antibody currently undergoing Phase II clinical trials. Together, this work establishes a versatile AI-driven binder design pipeline with broad potential for next-generation diagnostics and therapeutics in cancer cachexia and other GDF15-mediated diseases.