Programmable Klebsiella pneumoniae Phage Tropism Enabled by Scalable Receptor-Binding Protein Mining and Modular Assembly.

Phage therapy is an attractive countermeasure to multidrug-resistant pathogens, but clinical deployment is limited by the narrow and poorly predictable host range of most phages. Public repositories now house tens of thousands of phage genomes, yet there is no systematic route for turning this sequence space into designer phages with defined receptor specificities. Here we present a scalable, data-driven framework that converts receptor-binding protein (RBP) diversity into a modular toolkit for programmable phage engineering. Using Klebsiella pneumoniae as a model, we mined 280 non-redundant Przondovirus RBP sequences and resolved them into >50 discrete clusters. Functional screening of the Prz_RBPs from 41 previously uncharacterized Przondovirus phages expanded the number of experimentally validated capsular locus (KL) targets from 14 to 32 in Przondovirus. Each cluster was primarily associated with a dominant KL type, enabling construction of a genotype-to-phenotype map that accurately predicts receptor tropism. The identification of conserved anchor motifs enabled combinatorial pairing of Prz_RBP1 and Prz_RBP2 as plug-and-play modules for programmable phage tropism. Supplying exogenous Prz_RBP2 variants that assemble with Prz_RBP1 further and predictably expands host range, providing broad and tunable coverage. This framework transforms raw genomic diversity into customizable antibacterial agents and offers a general blueprint for precision phage therapy.