A Clinically Silent Resistance Phenotype That Promotes <i>Acinetobacter baumannii</i> Survival During Colistin Therapy

<i>Acinetobacter baumannii</i> is a major cause of multidrug-resistant nosocomial infections, particularly ventilator-associated pneumonia, for which therapeutic options are increasingly limited. Colistin, a polymyxin antibiotic, is a drug of last resort for <i>A. baumannii</i>, boasting high susceptibility rates. Yet, despite relatively low rates of breakpoint-defined colistin resistance, clinical outcomes are highly variable, and the bacterial strategies that enable survival during colistin therapy remain poorly understood. Here, we integrate supervised machine-learning-guided genomic prioritization with functional, physiological, and in vivo analyses to interrogate the genetic basis of colistin response in <i>A. baumannii</i>. Machine-learning analysis of clinical isolates identified candidate loci associated with colistin survival, many of which did not alter minimum inhibitory concentration (MIC) when disrupted. Instead, growth-dynamic assays uncovered a subset of mutants capable of maintaining fitness upon inhibitory colistin exposure despite classification as susceptible via standardized antibiotic susceptibility testing. We define this phenotype as clinically silent resistance (CSR), a genetically encoded, MIC-independent survival state. Using a murine pneumonia model, we further demonstrate that CSR mutants thrive during colistin therapy in vivo. Together, these findings reveal a hidden layer of colistin survival that is not captured by standard susceptibility testing and highlight fundamental limitations of breakpoint-centric paradigms for predicting treatment outcomes in <i>A. baumannii</i>.