Programmable DNA integration with New-to-Nature tools using Computational Protein Design

Abstract Programmable integration of large DNA cargo (≥ 2 kb), without inducing double-strand breaks, remains challenging for genome editing technologies. Current approaches have limitations in programmability, depend on co-delivery of multiple components, require multiple enzymatic steps, or have variable on-target editing outcomes. Here, we address this challenge using de novo protein design to create highly active, new-to-nature RNA-guided transposons. Our strategy exploits the modular architecture of CRISPR-associated transposons (CASTs), reconfiguring their conserved transposition machinery to interface with widely adopted Cas9. The resulting system, which we call NovoCAST, simplifies the CAST architecture from eight distinct proteins to four, establishing the simplest CAST described to date. NovoCAST exhibits sharply defined integration profiles, a 500-fold increase in activity relative to the parental PmcCAST, and general programmability. Using structural and biochemical analyses, we confirmed that the designed proteins fold and function as intended. Finally, we demonstrate robust programmable genomic integration in human cells highlighting its broad potential applications in research and therapeutics. Together, these results establish de novo protein design as a powerful strategy for engineering efficient genome-editing systems and for coupling CRISPR-mediated DNA recognition to heterologous functions through de novo designed protein interfaces.