RNA editing exclusively modifies cellular transcriptional transcripts without permanently altering genomic DNA, offering superior overall biosafety compared with conventional DNA editing platforms. Since 2019, Professor Wensheng Wei’s team has developed the original LEAPER (leveraging endogenous ADAR for programmable editing of RNA) technology, which recruits endogenous cellular ADAR enzymes via customized ADAR-recruiting RNAs (arRNAs) to achieve targeted RNA editing. Operating without exogenous protein delivery, this strategy exhibits minimal immunogenicity and a low delivery burden, conferring distinct advantages over canonical CRISPR-Cas systems. In 2022, the team upgraded the platform to LEAPER 2.0 by employing circular arRNAs to enhance RNA stability, editing efficiency and functional duration, while substantially mitigating off-target bystander editing. Nevertheless, several critical technical bottlenecks remained unresolved: ADAR’s intrinsic sequence preference restricts editing efficiency at numerous clinically relevant disease-associated loci, adjacent bystander editing persists, and poorly defined ADAR–RNA interaction mechanisms have long confined tool development to empirical screening, impeding further clinical translation. Given that LEAPER harnesses human endogenous ADAR proteins, enzyme engineering is infeasible, rendering RNA structure optimization the only viable strategy to overcome existing technical limitations.

On June 10, 2026, the research team led by Professor Wensheng Wei at Peking University/Changping Laboratory published a landmark study in Cell, titled “RNA structure programs endogenous ADAR for precise and efficient editing.” For the first time, this work systematically delineates the core structural rules governing how endogenous ADAR enzymes interact with double-stranded RNA (dsRNA) substrates in human cells. Building on these critical mechanistic insights, the team developed LEAPER 3.0, a next-generation endogenous RNA editing platform with markedly improved editing efficiency, precision and applicability. This study establishes a comprehensive structure-guided framework for the rational design of RNA editing tools, advancing endogenous RNA editing from experience-based optimization to precise mechanism-driven engineering.

In recent years, AI-powered structural prediction technologies have achieved transformative advances. In particular, AlphaFold 3 supports high-accuracy structural prediction of proteins, DNA, RNA and multi-molecular complexes, providing a robust tool for dissecting sophisticated biological systems. In this study, the team utilized AlphaFold 3 to construct high-confidence structural models of ADAR1/ADAR2–dsRNA complexes. Combined with systematic biochemical validation and high-throughput functional screening, the research group comprehensively deciphered the core principles underlying ADAR–RNA recognition.

The structural and functional analyses reveal that ADAR–dsRNA interactions extend beyond well-characterized catalytic domains and RNA-binding interfaces. The team identified a previously understudied functional segment termed the ADAR-covered peripheral region, which substantially modulates overall editing activity. Further validation confirmed that engineered bulge structures introduced at precise dsRNA positions can drastically reshape ADAR–RNA interaction patterns, creating a new avenue for the rational optimization of RNA editing tools.

The team systematically evaluated how bulges with varied positions, sizes and structural configurations regulate ADAR catalytic activity. The results demonstrate that rationally designed bulges inserted at defined regions significantly strengthen ADAR-mediated recognition and catalysis at target sites, effectively boosting editing efficiency at traditionally refractory loci. Meanwhile, strategically positioned bulges broadly suppress unwanted bystander editing and enable fine-tuning of adjacent editing activities, markedly improving overall editing precision. Notably, ADAR1 and ADAR2 exhibit distinct structural preferences for bulge architectures. The synchronous integration of optimized dual bulges stably enhances editing performance across diverse cellular contexts, regardless of endogenous ADAR expression patterns. Based on this design principle, the researchers engineered a new generation of ADAR-recruiting RNA termed arRNAβ and established the advanced LEAPER 3.0 technology platform.

The iterative technological upgrades of the LEAPER series can be clearly illustrated by structural evolution: the original LEAPER employs linear arRNAs that perfectly base-pair with target transcripts to recruit endogenous ADAR for editing. LEAPER 2.0 remodels this linear scaffold into a closed circular RNA loop to prolong RNA stability and functional duration. LEAPER 3.0 further integrates programmable structural “knots” (bulge structures) into the circular backbone. By remodeling RNA spatial conformation, these precisely engineered structures constrain ADAR’s functional scope, enabling robust editing at recalcitrant target sites while efficiently suppressing undesired off-target editing.

The research team validated the superior performance of LEAPER 3.0 in multiple disease models. In cellular and animal models of Duchenne Muscular Dystrophy (DMD), Usher syndrome and Alpha-1 Antitrypsin Deficiency (AATD), LEAPER 3.0 exhibited substantially improved editing outcomes compared with its predecessors. Specifically, for the AATD pathogenic PiZZ genotype that requires single-nucleotide-resolution editing, LEAPER 3.0 achieves approximately 40% sustained and accurate restoration of amino acid sequences, which markedly ameliorates liver pathological phenotypes in preclinical disease models.

The development of RNA editing tools has largely relied on high-throughput screening and empirical trial-and-error, lacking standardized rational design principles. This study systematically delineates ADAR–RNA interaction rules from structural and mechanistic perspectives and establishes a theoretical framework to guide the rational design of next-generation RNA editors. By optimizing RNA secondary structures rather than modifying protein effectors, the team achieved simultaneous improvements in editing efficiency and accuracy, offering new insights for the iterative development of endogenous RNA editing technologies. It also provides an innovative therapeutic strategy for a broad spectrum of previously intractable genetic mutations. These findings confirm that RNA structures function not merely as passive substrates for ADAR catalysis but as programmable functional elements that actively tune RNA editing behaviors.

This study fully demonstrates the profound integration of artificial intelligence and life science research. It validates that AI-empowered structural prediction tools can not only interpret complex biological mechanisms but also guide the rational design and optimization of cutting-edge biotechnological tools. In the future, LEAPER 3.0 holds broad application potential for treating diverse genetic disorders, neurological diseases and metabolic diseases. It lays a solid foundation for the clinical translation and industrialization of precision RNA editing and sets a new benchmark for AI-driven biotechnological innovation.

The corresponding author of this study is Professor Wensheng Wei from Peking University and Changping Laboratory. Deli Song, Gexing Liu, Wei Zhang, Jiwu Ren, Xuanxuan Jin, Yanglong Sun and Zexuan Yi serve as co-first authors. This research was supported by the National Natural Science Foundation of China, Changping Laboratory, the Peking-Tsinghua Center for Life Sciences, and the State Key Laboratory of Gene Function and Modulation Research at Peking University, among other funding institutions.

Dual-bulge Architecture of LEAPER 3.0 Enables Efficient and Precise A-to-I RNA Editing

Paper link: https://doi.org/10.1016/j.cell.2026.04.047