Juan Carlos Lopez

Andy Marshall

Around 1 in 5000 people live with a maternally inherited mitochondrial disease like MELAS, Leber’s Hereditary Optical Neuropathy (LHON) or MIDD, for which there are limited or no treatment options. Gene- and base-editing therapies for mitochondrial DNA (mtDNA) have lagged behind CRISPR–Cas9-based approaches targeting nuclear genes. Whereas there is already a CRISPR–Cas9-based product on the market and >150 different active trials of investigational therapies, the company closest to the clinic with an I-CreI (mitoARCUS) meganuclease targeting a mtDNA point mutation in MELAS/mitochondrial myopathy (Precision Biosciences) announced last month that it was pausing development for commercial reasons.

Despite this disparity, there is reason for optimism as a flurry of different types of optimized cytidine and adenine base editors for mtDNA are now available, with base conversion efficiencies of 50% now achievable, and some newer formats reaching efficiencies as high as 82%.

The development of mtDNA editors is not without challenges. First, editors must dispense with the targeting guide RNA, as mitochondria possess a double membrane that lacks any RNA transport system, effectively thwarting CRISPR-based gene or base editors (instead, a mitochondrial targeting sequence is used to ferry-in editor proteins). Second, unlike nuclear DNA with two copies of a gene, every human cell contains thousands of mitochondria — oocytes contain a whopping 193,000 mitochondria on average — and each organelle contains an average 10 mitochondrial genomes. Those ~10,000 genomes per cell may not all have the same sequence, with mutations existing in a state known as heteroplasmy, in which both mutant and wild-type genomes co-exist in the same organelle. Disease only occurs when the percentage of mutant mtDNA exceeds a particular threshold, typically between 70% and 95%.

Heteroplasmic mitochondrial diseases, like MELAS and MIDD, could be treated using I-Crel/FokI meganucleases or restriction enzymes linked to either transcription activator-like effector (TALE) domains or zinc fingers (which introduce double-strand DNA breaks into target sequences, leading to elimination of mutant mtDNA and repopulation of wild-type mtDNA); other conditions like LHON are predominantly mutant homoplasmic, which means they can only be treated using base editors or supplemental gene therapy.

One key concern with base-editing technology has been its propensity for off-target and bystander changes. This has led to various strategies to increase specificity, such as engineering the deaminases to narrow the editing window or use of nuclear exclusion sequences to stop nuclear sequence editing. Now, two papers in Nature Biotechnology represent important advances that could speed up translational studies of mitochondrial diseases.

Liang Chen, Dali Li and their colleagues of ShanghaiTech University, China report the engineering of highly efficient mitochondrial adenine base editors (eTd-mtABEs) by introducing mutations into the TALE TadA-8e deaminase for greater activity and specificity. These editors achieved up to 87% editing efficiency in human cells and over 50% in vivo, with reduced off-target effects compared to earlier tools.

In the first study, the researchers used eTd-mtABEs to introduce mutations in the human ND6 gene, encoding a subunit of the oxidative phosphorylation (OXPHOS) system linked to LHON and Leigh syndrome. They found reduced levels of ATP and more reactive oxygen species in the edited cells compared with controls, consistent with disease phenotypes. Next, the team used this adenine TALE base editor to introduce two pathogenic T-to-C mutations in the mitochondrial TRNS1 gene of rat zygotes, a gene linked to childhood-onset sensorineural hearing loss. The resulting offspring showed sensorineural hearing loss, which was transmitted to the F1 generation, providing proof of concept that eTd-mtABEs can be used to create animal models of disease.

In the companion paper, Chen, Li and their colleagues used the adenine TALE base editor to model Leigh disease in rats using a similar strategy. The resulting rats showed reduced motor coordination and muscle strength, defects that were obtained with editing efficiencies of only 54% on average. To test if the abnormalities could be reversed, the authors then used a cytosine TALE base editor in zygotes from the mutant rats. On average, the editing efficiency was only 53%, but this was enough to rescue the disease phenotypes.

This is the first report of direct correction of mtDNA mutations via a TALE base editor in an animal model. The next step will be to show feasibility in a model after disease onset (only the UK and Australia allow maternal spindle transfer therapy for mitochondrial diseases; no country has permitted mitochondrial base editing in human zygotes).

Achieving effective therapeutic mitochondrial base editing in affected target tissues will thus require efficient AAV delivery. For LHON, an already approved FDA AAV-2 product transduces the optic nerve and retinal ganglion cells, providing a translational path; ​GenSight Biologics also recently published 5-year outcome data for its AAV-2 therapy Lumevoq (lenadogene nolparvec) in LHON. But AAV delivery in other mitochondrial conditions will not be as simple: MELAS patients, for example, require efficient transduction of the CNS, kidney, skeletal muscle and cardiac muscle; MIDD patients need AAV delivery to the pancreas, inner ear, retina and kidney. Although a commercial AAV vector (AAVrh74) is available for muscle (Sarepta’s Elvidys), vectors that reach many of these other tissues have yet to be commercialized and may require next-generation AAV capsids and/or refinement of machine-guided design of cell type-specific synthetic promoters to reach target organs.

It is encouraging that the roughly 50% base conversion rate achieved in these new studies exceeded the heteroplasmy threshold required for disease manifestation and therapeutic rescue. At the same time, despite this remarkable success, concerns remain about off-target effects — both in mitochondrial and nuclear genomes — and narrow therapeutic windows. And with base editing approaches so far behind conventional gene therapies like Lumevoq in development, compelling commercial and clinical advantages benchmarked against best-in-class gene therapy will be needed to convince investors to back these approaches.

One parting thought: the past year has seen a noticeable uptick in publications on mitochondrial base editing technology from labs outside of the US. TALEN specialist Cellectis, headquartered in Paris, France, acquired 19% of equity in the mitochondrial base editing company Primera Therapeutics in 2022, ostensibly for its rapid TALE assembly platform (FusX System), which streamlines TALE repeat construction. In South Korea, Jin-Soo Kim at the Korea Advanced Institute of Science and Technology (KAIST) recently co-founded startup Edgene with Myriad Partners to develop mitochondrial base editors based on his seminal work on TALE-linked deaminases (TALEDs) enabling A to G conversion, which he has continued to optimize. According to Biocentury, 8 out of 13 base editing studies published in 27 translational journals over the past year came from labs in China. Wensheng Wei’s group at Peking University, a founder of Edigene in Beijing, continues to work on mitobase editors, with two recent patents on strand-selective mitochondrial editing. And Jia Chen of ShanghaiTech University, China, and his collaborators Li Yang and Bei Yang, are scientific advisors to Correctseq in Shanghai, which is developing transformer base editors for ex vivo and in vivo applications. It seems that mitochondrial base editing may be another area where US biotech may soon be finding itself chasing the dragon. David Liu and Beam Therapeutics may have something to say about that.