Evolutionary scenarios for the specific recognition of nonhomologous endogenous peptides by G protein-coupled receptor paralogs.

Several peptides interact with phylogenetically unrelated G protein-coupled receptors (GPCRs); similarly, orthologous GPCRs interact with distinct ligands. The neuropeptide substance P (SP) activates both NK1R and another unrelated primate-specific GPCR, MRGPRX2. Furthermore, MRGPRX1, a paralog of MRGPRX2, recognizes BAM8-22 (bovine adrenal medulla peptide 8-22), which has no evolutionary relatedness to SP. To elucidate the molecular basis and evolutionary history of this phylogenetically unrelated ligand selectivity, we developed a systematic procedure, the "interaction determinant likelihood score" system, which estimates the amino acid residues responsible for peptide-GPCR interactions predicted by peptide descriptor-incorporated support vector machine, our original machine learning-based peptide-GPCR interaction predictor. An interaction determinant likelihood score-based approach followed by pharmacological validation revealed the determinant residues for the ligand selectivity of SP-MRGPRX2 (F3.24 and G4.61) and BAM8-22-MRGPRX1 (L1.35). Molecular phylogenetic analysis revealed that the MRGPRX1 of common ancestral primates recognized BAM8-22, whereas the ancestral Cercopithecinae MRGPRX1 lost its interaction with BAM8-22 because of the loss of L1.35. The SP-MRGPRX2 interaction emerged in the common ancestors of Euarchonta, and, thereafter, the interaction of MRGPRX2 with both SP and BAM8-22 was acquired via substitution with L1.35 in several lineages. Collectively, the present study unraveled the molecular mechanisms and evolution of ligand specificity in evolutionary unrelated GPCRs.