Annual Research & Review in Biology

The crystalline pesticidal proteins of Bacillus thuringiensis occupy an unusual position in applied biology. They are simultaneously the most extensively deployed insecticidal molecules in agriculture, the subject of a mechanistic dispute that has remained unresolved for more than two decades, and the template for most attempts to engineer proteinaceous insect control. This review examines the state of knowledge across three linked domains: the molecular mechanism by which Cry proteins kill susceptible larvae, the historical and taxonomic decisions that shaped how the protein family is conceptualised, and the record of deliberate protein engineering. Literature was identified through structured searching of open scholarly databases and indexes, supplemented by citation tracking and examination of authoritative institutional sources, with critical appraisal focused on design adequacy, independent replication and the correspondence between experimental system and field relevance. Three findings emerge. First, the sequential pore-formation model and the receptor-mediated signalling model are not equally supported; the balance of independent evidence favours pore formation, although several component claims of the dominant model rest on a narrow range of cell-based systems and remain inadequately replicated. Second, resistance genetics constitutes the most informative and least exploited test of mechanistic claims, because loss-of-function mutations in candidate receptors provide causal evidence that binding assays cannot supply. Third, protein engineering has succeeded most consistently where it altered proteolytic processing or spectrum rather than where it attempted to circumvent resistance, and the durability of engineered variants under field selection remains largely untested. The most consequential gaps concern the physical fate of crystals within the larval gut, the quantitative contribution of host immune and regulatory responses to survival, and the near-absence of long-term field data on engineered and spectrum-expanded proteins. Confidence in current mechanistic consensus should be regarded as moderate rather than settled.