BALTIMORE – Researchers at the Center for Genome Engineering at the Institute for Basic Science (IBS) in South Korea have demonstrated the first programmable base editors that can convert adenine (A) to guanine (G) in the human mitochondrial genome.
In a study published last month in Cell, the scientists detailed the base editing technology and showed it was highly efficient in human cells, achieving A-to-G conversion frequencies of up to 49 percent at 17 target sites within various mitochondrial genes.
“This study is the first demonstration of A-to-G base editing in mitochondrial DNA (mtDNA), which, in principle, can correct almost half of pathogenic mtDNA mutations,” Jin-Soo Kim, director of IBS’s Center for Genome Engineering and senior author of the paper, said in an email.
According to Kim, various mutations in mtDNA, which encodes essential genes for oxidative phosphorylation and ATP production, are not only associated with heritable genetic disorders that collectively affect one in 5,000 individuals but are also linked to aging and age-related diseases such as cancer. Therefore, precise mtDNA base editing could pave the way for making disease models in cell lines and animals in the short term and for therapeutic editing to correct the mutations down the road, he said.
He and his colleagues have filed a patent on the technology and have founded a company, called Edgene, that focuses on developing gene therapy for mitochondrial genetic disorders.
Other methods to target mtDNA mutations already exist, including engineered nucleases such as zinc-finger nucleases (ZFNs) and transcriptional-activator-like effector nucleases (TALENs). These can be deployed to cleave targeted mutant mtDNA, causing mtDNA degradation and thereby reducing mutant mtDNA copies among the wild-type mtDNA molecules.
However, because the nuclease-based methods rely on wild-type mtDNA as a template to replicate new DNA molecules, these approaches cannot be used to correct homoplasmic mutations — mutations that are universally present in the mitochondrial genome — or to induce mutations in mtDNA de novo, according to the authors.
The first targeted human mitochondrial base editors appeared in 2020, developed by David Liu’s group at the Broad Institute. Named DddA-derived cytosine base editors (DdCBEs), they are composed of an interbacterial toxin derived from Burkholderia cenocepacia, coined DddAtox, which catalyzes cytidine deamination within double-stranded mtDNA; transcription activator-like effector (TALE) array proteins that can be programmed to bind target DNA sequences; and an uracil glycosylase inhibitor (UGI), which inhibits the removal of uracil, a product of cytosine deamination.
But the technology is limited to cytosine (C) to guanine (G) editing, said Kim, which accounts for less than 13 percent of all possible point mutations. Meanwhile, point mutations are known to be culpable for 90 out of the 95 clinically confirmed pathogenic mtDNA mutations listed in Mitomap, a public database of human mitochondrial DNA variation.
To create base editors capable of A-to-G conversion in human mtDNA, the IBS researchers first adapted Liu’s base-editing method by fusing a TALE protein custom-designed to bind to the ND1 or ND4 gene in human mitochondria; an adenine deaminase, named TadA8e, that was previously engineered from the Escherichia coli TadA protein for nuclear adenine base editing by Liu’s group; and a mitochondrial targeting sequence. After testing the design in human embryonic kidney cells, the researchers found the system to be “poorly active,” inducing A-to-G conversions with frequencies of 0.7 to 1.2 percent that they wrote “were impractically low but clearly above noise levels caused by sequencing errors.”
The group next experimented fusing the TadA adenine deaminase with the C-to-T mitochondrial base editors previously designed by Liu’s group, DdCBEs. The authors reasoned that the DddAtox present within the DdCBEs, which operates on dsDNA, might make DNA more accessible to the TadA adenine deaminase.
Structurally, DdCBEs are composed of a pair of TALE arrays, each fused to a catalytically deficient half split DddAtox to avoid cytotoxicity of the full-length protein, and an UGI. Interestingly, by replacing the UGI on either subunit of the DdCBEs with a TadA deaminase, the researchers found the new system could induce simultaneous A-to-G and C-to-T edits in human mtDNA, with A-to-G editing frequencies of up to 19 percent in human embryonic kidney cells.
Encouraged by the results, the researchers then sought to further enhance the base editor so it would only catalyze A-to-G conversions without C-to-T conversions in mtDNA. They hypothesized that by removing the UGI from both TALE subunits, the system would avoid C-to-T conversions while retaining A-to-G editing capability.
Deep sequencing results confirmed this, showing that the UGI-free editors designed to target the ND1 site within the mtDNA were able to induce A-to-G editing with at least 40 percent frequency without causing C-to-T substitutions or unwanted indels. Additionally, they constructed UGI-free editors targeting the ND4 gene, which achieved up to 34 percent A-to-G conversion frequency without causing unwanted C-to-T edits or indels.
Given the base editors contained split DddAtox on both TALE arrays, the researchers named them split TALE deaminases, or sTALEDs. To further investigate whether the catalytically deficient, non-toxic, full-length DddAtox can boost the A-to-G conversions, Kim’s team constructed two more types of TALEDs : monomeric TALEDs (mTALEDs), which consist of only a single TALE array fused to the TadA protein and the full-length DddAtox; and dimeric TALEDs (dTALEDs), which include two DNA strand-specific TALE arrays in a tail-to-tail configuration, with one strand fused to the TadA protein and the other to the DddAtox.
In summary, the team found sTALEDs are the most efficient among the three TALED architectures, said Kim. However, because mTALEDs are smaller than dTALEDs or sTALEDs, they may be more appropriate for viral delivery. In addition, mTALEDs and dTALEDs appear to have higher specificity than sTALEDs.
Based on these results, Kim said mTALEDs might be suited for in vivo experiments moving forward, while sTALEDs are likely to be more efficient and useful for most in vitro experiments.
The study “beautifully shows how integration of our mitochondrial cytosine base editor and the deaminase enzyme we evolved for our nuclear adenine base editors can achieve an important goal in the field — adenine base editing in mitochondrial DNA,” Liu wrote in an email.
He added that even though mtDNA mutations may lead to many serious genetic diseases, mitochondria have previously resisted precision gene editing due to the challenges of delivering CRISPR guide RNAs into the organelle. To circumvent the problem, years ago, his group reported the first CRISPR-free base editors, DdCBEs, that use proteins instead of CRISPR protein-RNA complexes, to target DNA in mitochondria.
By pairing this DdCBE architecture with the deoxyadenosine deaminase that his lab previously developed, this study is “a key advance towards the precise correction of pathogenic mitochondrial mutations,” Liu added.
“The findings are very interesting, and the way they approached this is also quite interesting,” said Magomet Aushev, a genome editing researcher at the Wellcome Centre for Mitochondrial Research at Newcastle University in the UK. “It's not so intuitive to attach two deaminases together and then expect them to work. But they did work — surprisingly, they worked quite well.”
Aushev also emphasized the challenges of mitochondrial genome editing compared to modifying the nuclear genome. For example, while the nucleus only has a single membrane, which can break down during cell division, the mitochondrial genome is surrounded by two membranes that do not normally break down, making it “quite difficult to get things inside the mitochondria,” especially RNA molecules like CRISPR.
In terms of applications, Kim said the base editing technology described in this study can be immediately useful for generating mtDNA mutations in cell lines and animals to create disease models. In the long term, he is hoping for improved versions of TALEDs to be developed to correct disease-causing mtDNA mutations therapeutically in vivo.
Aushev said that in addition to being useful for drug discovery and disease modeling in the short term, he would not be surprised if mtDNA base editing technologies started to be applied in the clinic within the next decade or so, especially given this is a “very fast moving field.”
However, he pointed out several general challenges for base-editing technologies being applied as a therapeutic tool. For one, the efficiency of the base editors can be a holdup. The editors would have to be efficient enough, he said, to modify sufficient mtDNA so that the pathogenic mutations can be pushed below the threshold of clinical manifestations of symptoms.
The other bottleneck is safety, Aushev said. As a rule of thumb, base editors need to have high specificity and accuracy to minimize off-target effects, he noted, which can cause collateral damage if unwanted mutations are introduced.
Kim said the TALEDs investigated in the study are “reasonably specific,” with off-target mtDNA edits rarely induced. However, the authors did acknowledge that bystander editing is still a limitation for the base editors. Therefore, further engineering of the TadA deoxyadenosine deaminase or DddAtox are needed to reduce or eliminate the bystander editing activity of TALEDs.
Additionally, Kim said the study only demonstrated A-to-G base editing in cell lines. As such, further studies are necessary to confirm the method also works in vivo.
Delivery might be the biggest hurdle for base editing technologies as a therapy, Aushev pointed out. “The number of cells in the human body is substantial,” he said. Therefore, “a very big and very important issue is how we get this base editor to a sufficient number of cells in the affected tissues.”
Still, despite the possible caveats, Aushev said the precise base editing of mtDNA achieved by this study is “definitely a very good addition to the existing toolbox” for the mitochondrial genome editing field.
“It's just fantastic for the mitochondrial community to have these tools,” he said. “They really open up a lot of doors.”