Mitochondria, the cellular powerhouses, harbor their own genome (mtDNA). Point mutations in mtDNA are directly linked to hundreds of genetic disorders like Leigh syndrome and Leber's hereditary optic neuropathy (LHON), with point mutations accounting for approximately 95% of pathogenic variants. In the drug discovery pipeline, accurate genetic disease models that can faithfully replicate the human disease mutation threshold (typically >60% heteroplasmy) and multisystem phenotypes are core infrastructure for preclinical research.
 

However, due to the unique circular structure, multicopy nature, and strict maternal inheritance of mtDNA, traditional technologies have struggled to generate precise, stably heritable models. This has left numerous candidate drugs without reliable in vivo validation platforms, severely hindering the translational progress of related therapies.
 

Differences from Traditional Gene Editing
 

Building mitochondrial disease models differs fundamentally from traditional gene editing models, which is the root of its technical challenges. Conventional gene editing (e.g., for point mutations) targets nuclear DNA. It uses nucleases to create DNA double-strand breaks, relying on cellular repair mechanisms to achieve gene modification. This is suitable for studying monogenic disorders caused by nuclear gene mutations and follows Mendelian inheritance patterns. Mitochondrial base editing targets cytoplasmic mtDNA, which is circular, exists in multiple copies, and follows strict maternal inheritance. Because mitochondria lack an efficient DNA double-strand break repair system, this technology employs a strategy of fusing engineered deaminases with targeting proteins. It directly achieves base conversion without cutting the DNA backbone, thereby precisely mimicking pathogenic mutations. Such models are specifically designed to study diseases related to mitochondrial energy metabolism dysfunction. They require driving the proportion of mutant mtDNA above the 60% disease threshold, posing higher demands on the precision and efficiency of the editing tools.
 

Current Technical Bottlenecks and Industrialization Challenges
 

The generation of animal models for mitochondrial diseases faces two core bottlenecks. The first is off-target effects leading to impure genetic backgrounds. Early tools based on DddA exhibited significant nuclear genome off-target risks. This non-specific editing, stemming from the protein's inherent interaction properties, makes it difficult to establish a reliable "genotype-phenotype" causal relationship, severely undermining the model's credibility for drug evaluation. The second is insufficient editing efficiency leading to the absence of phenotypes. Human mitochondrial disease onset requires a high proportion of mutant mtDNA (typically >60%), but early tools struggled to achieve and maintain this pathogenic load in animals. Consequently, the generated models either did not develop disease or exhibited weak, unstable phenotypes, failing to provide reproducible pathological platforms for preclinical research.
 

Technological Breakthrough: Next-Generation Mitochondrial Base Editor System
 

On January 22, 2025, the Wei Wensheng group from Changping Laboratory​ and Peking University​ published a research paper online in Naturetitled "Precise modelling of mitochondrial diseases using optimized mitoBEs." The study reported breakthrough progress in using an optimized system based on mitoBEs (mitoBEs v2) for generating mitochondrial disease mouse models.

mitoBEs is a novel mitochondrial base editing tool combining nickase and single-strand DNA deaminase activities, enabling C-to-T and A-to-G bidirectional editing of mitochondrial DNA (mtDNA). Compared to earlier tools (e.g., DdCBEs and TALEDs), it offers superior strand specificity and lower off-target effects, theoretically enabling precise modeling of approximately 87% of pathogenic mitochondrial point mutations.
 

The research team successfully generated multiple mouse models with high mutation frequencies using the optimized mitoBEs v2. These models not only exhibited typical phenotypes corresponding to human diseases but also yielded precise genetic background models with 100% mutation load or containing only a single-base mutation through breeding experiments.

To construct high-fidelity models capable of accurately establishing genotype-phenotype associations, the research team prioritized eliminating off-target effects, systematically optimizing the mitoBEs system. The team first evaluated early RNA-encoded mitoBEs, finding they still exhibited non-specific editing at the transcriptome (mitoABE) and mitochondrial genome (mitoCBE) levels. Therefore, the study focused on engineering the core deaminase components: Through large-scale screening, the TadA8e-V106W-V28F variant was obtained for the A-to-G editor (mitoABE v2), reducing transcriptome off-targets to background levels. Simultaneously, the TaddA-derived cytidine deaminase CBE6d was applied to the C-to-T editor (mitoCBE v2), significantly reducing mitochondrial genome off-target effects. Comprehensive testing confirmed that the optimized mitoBEs v2 did not induce detectable off-target events at the nuclear genome level, validating its high safety for in vivo editing.
 

Image source: Precise modelling of mitochondrial diseases using optimized mitoBEs
 

Building on ensured precision, the study further evaluated its editing coverage and efficiency. By mapping 85 human pathogenic mtDNA point mutations to their mouse homologs, the team identified 70 editable sites and successfully validated 68 of them at the cellular level. Editing efficiency was significantly enhanced using a circular RNA delivery system. In mouse embryo injection experiments, mitoBEs v2 achieved highly efficient editing in F0 generation models, with mutation frequencies reaching 82% at the mt-Nd5 A12784G site, and whole-genome sequencing confirmed the absence of off-targets. The editing outcomes were widely distributed and stably present across multiple mouse tissues, proving the system's capability for generating genetically clean, stably heritable mitochondrial disease models. Through breeding and screening, the team successfully obtained mt-Nd5 A12784G mouse models with 100% mutation load.

 

Image source: Precise modelling of mitochondrial diseases using optimized mitoBEs
 

In disease phenotype validation, the mt-Atp6 T8591C mice, corresponding to Leigh syndrome, exhibited significant cardiac insufficiency, while the mt-Nd5 A12784G mice, corresponding to LHON, showed typical retinal electrophysiological abnormalities and visual impairment. These phenotypes closely matched human clinical manifestations, confirming that models built with mitoBEs v2 can faithfully reproduce key pathological features of mitochondrial diseases. In summary, this study developed an efficient, precise, and genetically clean mitochondrial base editing system, providing a reliable modeling tool for dissecting the pathogenesis of mitochondrial diseases and for preclinical drug development.

 

Image source: Precise modelling of mitochondrial diseases using optimized mitoBEs
 

MingCeler looks forward to transforming this efficient and accurate modeling platform into a stable supply of model resources, providing preclinical research tools for global scientific research and drug development teams, and jointly accelerating the development process of treatment strategies for mitochondrial diseases.
 

References:

1、https://mp.weixin.qq.com/s/bH4Fmn5QGyW-u_jWVGjFZw 

2、Zhang, X., Zhang, X., Ren, J. et al. Precise modelling of mitochondrial diseases using optimized mitoBEs. Nature 639, 735–745 (2025). https://doi.org/10.1038/s41586-024-08469-8