Comprehensive assessment of on- and off-target mutagenesis via lipid nanoparticle delivery of CRISPR-Cas9 genome editing.

Ensuring the safety of CRISPR-Cas9 genome editing requires comprehensive assessment of off-target mutagenesis, particularly for therapeutic applications under regulatory review. DNA-free lipid nanoparticle (LNP) delivery is expected to minimize insertional risks compared to adeno-associated virus (AAV) vectors, but experimental validation has been limited. Here, on-target amplicon sequencing in mouse muscle demonstrated insertions largely derived from host genomic sequences, with no detectable AAV or transgene integration. Benchmarking 13 in silico prediction tools identified Cas-OFFinder as the most sensitive, though precision remained low, and correlation with in vitro CIRCLE-seq data was modest. To strengthen off-target detection, we employed karyotypically normal human iPSCs and developed an "indel cluster" method using high-depth whole-genome sequencing, enabling discrimination between clustered genome editing-induced indels and background variants. Integration of in silico predictions, CIRCLE-seq cleavage sites, and in cellulo WGS clusters yielded 11 high-confidence off-target candidates, predominantly associated with one gRNA. While sensitivity was maximized, reproducibility across conditions remained limited, and many candidate sites overlapped repetitive or low-mappability regions. Our multilayered framework demonstrates both the utility and current limitations of off-target risk assessment, providing a practical foundation for the safety evaluation of genome editing therapies.