CIT8, a machine learning-optimized antimicrobial peptide, exhibits potent activity against carbapenem-resistant Acinetobacter baumannii

As the burden of antimicrobial resistance continues to rise worldwide, the need for novel antimicrobial therapeutics has become increasingly urgent. Machine learning (ML) has emerged as a promising strategy for accelerating antimicrobial peptide discovery; however, experimental characterization remains essential to define the biological properties and broader applicability of ML-optimized peptides. CIT8 is a 13-residue, citropin-based peptide previously developed through ML-optimization with potent antimicrobial activity against methicillin-resistant Staphylococcus aureus (MRSA). Here, we demonstrate that its activity extends to multidrug-resistant Gram-negative pathogens. Among these, CIT8 displayed the most potent activity against Acinetobacter baumannii, maintaining an MIC of 4 μg/mL against 15 of 16 A. baumannii strains tested, including multiple carbapenem-resistant (CRAB) clinical isolates. At 20 μg/mL (5 × MIC), CIT8 reduced an initial population of approximately 5 × 10⁷ CFU/mL of exponential-phase A. baumannii to below the limit of detection within 15 min. Consistent with its rapid bactericidal activity, CIT8 induced rapid, dose-dependent membrane depolarization and permeabilization, accompanied by leakage of intracellular ATP. In addition, CIT8 and, even more notably, its D-enantiomer CIT8-D exhibited strong antibiofilm activity while maintaining relatively low mammalian cytotoxicity, with therapeutic indices of >32 for CIT8 in both cell lines and 16.2 and 18.5 for CIT8-D in HaCaT and HKC-8 cells, respectively. Moreover, CIT8 potentiated the bactericidal activity of meropenem against CRAB, with the combination exhibiting bactericidal synergy by time-kill assay. These findings highlight CIT8 as a promising scaffold for further optimization and preclinical evaluation against multidrug-resistant A. baumannii.