Harnessing diverse tRNAs and AI-guided mining for compact and efficient plant multiplex genome editing

The widespread use of CRISPR-Cas9 (Clustered regularly interspaced short palindromic repeats-CRISPR-associated protein 9) in plants highlights the need for compact and efficient multiplexed genome editing systems. This study optimizes single-guide RNA (sgRNA) expression in CRISPR by leveraging endogenous tRNA processing mechanisms for efficient multiplexed genome editing. Screening in Arabidopsis thaliana and Oryza sativa identified superior tRNAs that outperformed the widely used AtGly-tRgcc. Leveraging tRNA's dual functions in sgRNA processing and their intragenic RNA polymerase III promoter activity, we established a compact multiplexed system for simultaneous editing of at least ten genomic loci in rice and soybean. Moreover, we developed plant tRNA large language models that learn sequence representations to identify both canonical and noncanonical tRNAs, uncovering thousands of tRNAs missed by traditional algorithms and expanding the repertoire for genome editing. This work provides a robust tRNA-based CRISPR platform, an artificial intelligence-guided tRNA mining framework, and a comprehensive tRNA resource for advanced plant genome engineering and germplasm innovation.