RND efflux pump inhibitor-like natural products as antibiotic adjuvants against multidrug-resistant gram-negative bacteria: a graded-evidence review

Abstract

Multidrug-resistant (MDR) Gram-negative bacteria—particularly carbapenem-resistant Enterobacterales, Pseudomonas aeruginosa and Acinetobacter baumannii—remain among the hardest infections to treat, and Resistance-Nodulation-Division (RND) efflux pumps are central to that resistance, making them a recurring target for antibiotic-adjuvant strategies. Natural products offer chemically diverse starting points for efflux-modulating adjuvants, but a fall in MIC is hard to attribute to pump inhibition: membrane permeabilisation, proton-motive-force (PMF) dissipation, efflux-gene downregulation, intrinsic antibacterial activity and general stress responses all give the same readout. We review natural product-derived and semisynthetic RND efflux pump inhibitor-like (EPI-like) candidates in Gram-negative bacteria, organising the evidence by target pump, potentiated antibiotic, proposed mechanism, strength of supporting data and translational feasibility. In place of chemical structures, candidates are read through mechanism-relevant structure-feature associations and the extent of pump-specific validation. Throughout, we separate three routinely conflated claims—antibiotic synergy, functional efflux modulation and direct target binding—and grade them with a pragmatic five-level scheme aimed at curbing over-reading of MIC-based synergy. One pattern dominates: synergy is reported widely, yet candidates reaching pump-associated validation are scarce, and higher-grade pump-dependent evidence remains concentrated on Escherichia coli AcrB. Camalexin is the clearest AcrB-dependent phenotype, but only provisionally—a Level 4a grade resting on a single unreplicated study—while plumbagin and nordihydroguaiaretic acid also reach provisional Level 4a because their potentiation and efflux phenotypes were AcrB-dependent and were not explained by outer-membrane permeabilisation. Direct target engagement remains unproven; the rare Level 4 cases beyond AcrB rely on a MexXY-OprM chemical probe or on expression change alone. Comparable pump-specific support for MexB, MexY and AdeB is largely absent. The proposed roadmap should therefore be viewed as an AcrB-anchored hypothesis that requires independent validation across these clinically important pumps. We propose a mechanism-informed structure-feature model and a staged validation roadmap coupling pump-specific strains, membrane and PMF controls, direct target engagement, feature-guided optimisation, cytotoxicity/PK assessment and in vivo infection models.