Logo: to the web site of Uppsala University

Endogenous antimicrobial peptides (AMPs) derived from host proteins represent a largely underexplored class of natural products tied to innate immunity. Here, we investigated collagen proteins as a source of latent alpha-helical AMPs encoded within nonfibrous extracellular matrix domains. Using a targeted in silico approach, verified collagen sequences were mined and prioritized based on secondary structure and three essential physicochemical properties: net charge, Boman index, and hydrophobic moment, yielding 107 predicted alpha-helical AMP candidates. The highest ranked peptides were synthesized and experimentally evaluated alongside benchmark AMPs and peptides prioritized by machine learning-based prediction tools. Three collagen-derived peptides identified by the targeted physicochemical approach exhibited broad-spectrum bioactivity against bacterial and fungal pathogens with minimum inhibitory concentrations comparable to those of LL-37 and melittin. In contrast, peptides ranked highly by machine learning predictors showed reduced or no activity. Collagen-derived peptides disrupted bacterial mimicking lipid membranes yet displayed markedly reduced cytotoxicity toward human cells, maintaining high viability at concentrations well above their antimicrobial MICs. These findings demonstrate that nonfibrous domains of extracellular matrix collagens constitute a previously underexplored reservoir of endogenous antimicrobial peptides with favorable biocompatibility, expanding the natural product space of host defense peptides and identifying collagen-derived AMPs as promising scaffolds for future antimicrobial discovery.