In a Kyoto lab, a single mouse model is helping rewrite the future of gene therapy—one precise edit at a time. Professor Akitsu Hotta and his team at the Department of Clinical Application have developed a groundbreaking framework to evaluate the safety of CRISPR-Cas9 genome editing, a technology poised to cure devastating genetic diseases like Duchenne muscular dystrophy. Their work, published in Molecular Therapy Nucleic Acids, offers a new gold standard for detecting both intended and unintended genetic changes, especially when CRISPR is delivered via lipid nanoparticles (LNPs)—a method showing cleaner, safer results than traditional viral vectors.
CRISPR’s potential is immense: by correcting faulty genes such as the DMD gene responsible for muscular dystrophy, it could restore production of the vital dystrophin protein and halt disease progression. But with great power comes great risk—permanent DNA alterations mean even rare off-target mutations could have serious consequences. Until now, the safety of LNP delivery, which carries CRISPR components as RNA rather than DNA, hadn’t been systematically assessed at the whole-genome level. Hotta’s team stepped in to fill that gap with a three-pronged strategy combining computational modeling, in vitro cleavage assays, and deep whole-genome sequencing in human induced pluripotent stem (iPS) cells.
The results were striking. When compared to adeno-associated virus (AAV) delivery, LNPs produced fewer insertions at the target site and—critically—showed no trace of vector-derived DNA integration, a known risk with viral methods. Even after multiple doses, LNP delivery maintained stable editing efficiency, suggesting lower immune activation. To pinpoint off-target effects, the researchers tested 13 widely used prediction tools and found many generated excessive false positives. By layering in experimental cleavage data, they refined the search and then applied a novel "indel cluster" method to distinguish real edits from background noise in the genome.
This integrated approach revealed only a small number of high-confidence off-target sites—most linked to just one guide RNA and often located in repetitive, hard-to-sequence regions where artifacts are common. Crucially, very few of these sites overlapped with functional genes, and those that did showed no meaningful biological impact. The study not only confirms the safety advantages of LNPs but also delivers a scalable blueprint for evaluating next-generation therapies.
As genome editing inches closer to widespread clinical use, Hotta’s framework offers a way to move forward with both confidence and caution—ensuring that the cures we develop do no hidden harm.