Evolution- and structure-guided redesign of the Stachel motif yields soluble and potent modulators of adhesion GPCRs

Adhesion GPCRs are activated by an intramolecular agonist known as the Stachel sequence. In the intact receptor, this sequence interacts with both the GPCR autoproteolysis-inducing (GAIN) domain and the seven transmembrane domain (7TMD). In contrast, synthetic Stachel-derived peptides used in experimental settings only need to bind to the 7TMD to activate the receptor. We therefore hypothesized that natural Stachel sequences have evolved under dual structural constraints, interacting with both domains, and that removing the GAIN domain constraint could allow the rational design and optimization of more effective ligands. Using a curated dataset of 13,611 vertebrate adhesion GPCR sequences, we performed large-scale conservation and entropy analyses that revealed strong evolutionary constraint at specific Stachel core positions, particularly X^0.47, X^0.48, and X^0.50. Structure-guided in silico design with ProteinMPNN, applied separately to GAIN- and 7TMD-bound conformations, generated thousands of Stachel variants and demonstrated that single-domain optimization explores sequence space not sampled in natural orthologs. Guided by these analyses, we designed two 11-mer peptides targeting ADGRG2. Both peptides were readily soluble and exhibited substantially increased potency in cAMP assays compared with WT-derived Stachel peptides. Structural modeling suggests that backbone constraint at position X^0.45, polar interactions at X^0.48, and hydrophobic anchoring at X^0.50 contribute critically to enhanced 7TMD engagement. Interestingly, analogous design for ADGRF5 did not yield active peptides. Together, our findings support the concept that separating GAIN- and 7TMD-imposed constraints can enable receptor-selective optimization of Stachel-derived ligands. This strategy provides a framework for exploring previously inaccessible sequence space, while highlighting receptor-specific trade-offs between potency and specificity.