BackgroundGram-negative infection is a global health concern. We present a custom-engineered, novel, patented, dextrin-colistin conjugate that can mask the toxicity of colistin; this conjugate allows controlled, enzymatically mediated dextrin degradation (unmasking) and reinstatement of colistin activity at the infected site. In the presence of amylase, the lead candidate EA-4 (1·0 mol% succinoylation, 11·2 % w/w colistin, molecular weight ∼10 300 g/mol) maximally unmasked and was equally effective to the clinical colistin formulation against strains of multidrug resistant Acinetobacter baumannii, Escherichia coli, and Klebsiella pneumoniae (minimum inhibitory concentration ≤4 g/L), but suppression of bacterial viability was six times longer. Here we report ex-vivo and in-vivo activity, pharmacokinetics and pharmacodynamics, and toxicity. MethodsWe used a newly validated two-compartment static dialysis bag model under infinite sink conditions to unmask EA-4 directly in six samples of infected burn wound fluid. We assessed wound versus plasma amylase partitioning with a Phadebas amylase test. We established pharmacokinetics (with high-performance liquid chromatography and ELISA) and toxicity (clinical variables) in an intravenous, escalating-dose Sprague Dawley rat model (35 rats). FindingsSignificant partitioning of amylase to the infected burn fluid occurred, despite normal plasma amylase (∼6·8:1, p<0·05), and controlled EA-4 unmasking was observed (48 h peak 68%, p<0·05). EA-4 showed no in-vivo toxicity even at maximum protocol dose (0·5 mg/L). However, severe clinical toxicity limited colistin dosing to 0·1 mg/L. Substantial improvement in pharmacokinetic variables compared with colistin were recorded—namely, plasma half-life (1·6 vs 0·9 h), clearance (2·8 vs 13·4 mL/min per kg), and last measurable concentration (>24 vs 4 h). InterpretationThese findings support proof of principle for the first bioresponsive macromolecular antibiotic class, which is equally potent to the current clinical alternative, but presents substantially improved toxicity and pharmacokinetic characteristics. Moreover, the ability for localised targeted, enzyme-mediated reinstatement of activity has implications for future design of controlled drug release polymer therapeutic systems. Imminent scale-up studies and testing in man are the next steps. FundingEuropean Union (European Social Fund), STEPS scheme (Government of Malta, EU), Welsh Government.