Evolution-guided phage therapy for antimicrobial-resistant Acinetobacter baumannii: from resistance evolution and genomic plasticity to receptor-associated fitness trade-offs and clinical translation

Abstract

Carbapenem-resistant Acinetobacter baumannii (CRAB) is sustained by environmental persistence, genomic plasticity, biofilm-associated tolerance, and rapid adaptation under antibiotic pressure. These features constrain conventional antibiotic regimens and make CRAB a relevant test case for evolution-guided antimicrobial control. This review first synthesizes how resistance architecture and genomic plasticity sustain multidrug resistance and clinical persistence. It then evaluates bacteriophages as selective agents that can alter the adaptive routes available to bacterial populations. We distinguish direct phage-mediated killing from evolutionary steering, in which receptor-dependent escape may create collateral costs in capsule or envelope function, biofilm formation, virulence, immune evasion, or antibiotic susceptibility. Receptor-associated claims are organized into three evidence categories: direct matched phage-escape evidence, indirect receptor/function evidence, and cross-species hypotheses. Direct matched evidence is concentrated in defined capsule-associated systems and supports context-dependent trade-offs, not a universal class effect. Animal studies support activity in specific systems, whereas a case-level clinical evidence table shows heterogeneous formulations, routes, co-interventions, and escape measurements; these reports primarily establish feasibility and treatment-associated observations, not general efficacy or clinical steering. Delivery and engineering strategies, rapid molecular profiling, and mechanistic models are therefore assessed as indirect evidence or testable design hypotheses rather than as equivalent support for steering. Translation also requires product quality, co-localized exposure, and serial measurement of burden, escape, phenotype, and coverage. The central challenge is to identify, exploit, and monitor a measurable and potentially transient window of collateral vulnerability rather than assume a stable evolutionary trap.