https://www.cas.cn/syky/202608/t20260804_5117880.shtml
https://www.nature.com/articles/s44259-026-00253-0
Antimicrobial peptides (AMPs) are promising candidates for anti-infective development, but their translation is often limited by the difficulty of achieving potent microbial membrane activity without damaging mammalian cells. Whole-genome data from *Gastrodia elata* (Tianma) combined with protein resource mining and generative AI-based design has led researchers to a broad-spectrum antimicrobial peptide candidate named GT-2.
The team computationally screened candidate peptide fragments derived from defense-related proteins in *Gastrodia elata*. Using the natural protein fragment GP-1 as a scaffold, they explored sequence space via the DeepAMP generative model and the D-POS optimization workflow to ultimately derive GT-2. *In vitro* experiments demonstrated that GT-2 exhibits inhibitory activity against *Candida albicans*, *Pseudomonas aeruginosa*, *Staphylococcus aureus*, *Escherichia coli*, and methicillin-resistant *Staphylococcus aureus* (MRSA), with minimum inhibitory concentrations (MIC) ranging from 4 to 32 μg/mL; it also disrupts microbial membranes and inhibits biofilm formation. At a concentration of 500 μg/mL, the hemolysis rate of GT-2 against rabbit red blood cells was less than 5%. Animal experiments showed that the peptide effectively cleared pathogenic bacteria from wounds in mice infected with *Pseudomonas aeruginosa*.
Mechanistic studies revealed that GT-2 readily forms helical structures and binds deeply within bacterial- and fungal-like membrane environments. In contrast, mammalian-like cell membrane environments—characterized by phosphatidylcholine and cholesterol—do not favor its effective insertion. This “lipid-dependent membrane binding” property enables GT-2 to effectively distinguish between pathogenic microorganisms and host cells.
By integrating plant genomic resources with AI-driven peptide design, this study offers a novel pathway for discovering antimicrobial lead molecules from unique biological resources and provides new insights into enhancing the pathogen selectivity of membrane-active antimicrobial peptides.