NEW YORK – Researchers from the Broad Institute and the University of Washington School of Medicine have discovered a bacterial toxin that allows for base editing of mitochondrial DNA without the need for a CRISPR-Cas system.
As the researchers noted in their study, which was published on Wednesday in Nature, certain interbacterial toxins in the deaminase superfamily have proven useful in gene editing. Previously described cytidine deaminases have operated on single-stranded nucleic acids, but their use in base editing has required the unwinding of double-stranded DNA (dsDNA) with a CRISPR-Cas system.
And because base editing applications in general have thus far required the use of a CRISPR system with a guide RNA (gRNA), base editing of mitochondrial DNA (mtDNA) has been hindered by the challenges associated with delivering a gRNA into the mitochondria, limiting researchers to designer nucleases that destroy the mitochondrial genome.
In their new study, the researchers led by the Broad's David Liu and Washington's Joseph Mougous described an interbacterial toxin — which they named DddA — that catalyzes the deamination of cytidines within dsDNA. They engineered split halves of DddA that were non-toxic and inactive until brought together on target DNA by adjacently bound programmable DNA-binding proteins. Fusions of these split halves with transcription activator-like effector array proteins and an uracil glycosylase inhibitor resulted in RNA-free DddA-derived cytosine base editors (DdCBEs) that catalyzed C-G to T-A conversions in human mtDNA with high target specificity and product purity.
"We realized that the properties of this 'bacterial warfare agent' could allow it to be paired with a non-CRISPR-based DNA-targeting system, raising the possibility of making base editors that do not rely on CRISPR or on guide RNAs," Liu said in a statement. "It could enable us to finally perform precision genome editing in one of the last corners of biology that has remained untouchable by such technology — mitochondrial DNA."
The investigators used DdCBEs to model a disease-associated mtDNA mutation in human cells, resulting in changes in respiration rates and oxidative phosphorylation.
"This is the first time in my career that we've been able to engineer a precise edit in mitochondrial DNA," co-author Vamsi Mootha added. "It's a quantum leap forward — if we can make targeted mutations, we can develop models to study disease-associated variants, determine what role they actually play in disease, and screen the effects of drugs on the pathways involved."
The researchers began by analyzing predicted bacterial deaminases that contain sequences suggesting them to be substrates for intercellular protein delivery systems, such as the type VI secretion system (T6SS). They sought to define the biochemical activity of T6SS-associated deaminases, and focused on a predicted deaminase belonging to the SCP1.201-like family encoded by Burkholderia cenocepacia. Their findings suggested that DddA may act on a previously undescribed deaminase substrate.
In further experiments, they found that expression of sub-lethal levels of DddA in E. coli substantially increased the mutation frequency. They performed whole-genome sequencing on five E. coli lineages that experienced serial DddA exposure and clonal bottlenecking, and five control strains that underwent a similar regimen in the presence of DddA, and observed approximately 50-fold more total SNPs in strains exposed to active DddA than in strains producing the inactive enzyme.
The researchers observed, however, that the expression of intact DddA fused to programmable DNA-binding proteins was toxic to human HEK293T cells. To avoid this toxicity, they proposed splitting the protein into two inactive halves that would reconstitute deamination activity only when assembled adjacently on target DNA. Their experiments also collectively demonstrated that split DddA could be fused to transcription activator-like effectors (TALE) arrays to mediate C-G to T-A conversions in human nuclear DNA.
This resulting optimized architecture of the DddA fusion editor, which they called DdCBE, represented the first agent capable of performing precise genome editing in mtDNA, the researchers said. To explore its generality for mtDNA editing, they engineered or adapted seven pairs of TALE arrays to target five mitochondrial genes: MT-ND1, MT-ND2, MT-ND4, MT-ND5, and MT-ATP8.
Three to six days after treatment, the researchers observed that the mitochondrial base-editing efficiencies of DdCBEs in HEK293T cells varied between 4.6 percent and 49 percent depending on the split type, split orientation, and target cytosine position within the spacing region. They did not detect indels or base editing outside the spacing region.
The investigators also confirmed the durability of mtDNA edits in HEK293T cells over 18 days, and noted that mtDNA editing did not reduce cell viability, produced no large mtDNA deletions, and did not perturb mtDNA copy numbers.
To profile the off-target activity of DdCBE in the human mitochondrial genome, they transfected HEK293T cells with plasmids that constitutively expressed optimized DdCBE or the corresponding dead-DdCBE control in order to distinguish DdCBE-induced single-nucleotide variants from background heteroplasmy. The researchers found that the average frequency of mitochondrial genome-wide off-target C-G to T-A editing by each DdCBE was similar to that of the untreated and dead-DdCBE control, except in the case of the DdCBE targeting MT-ND5, which had a 1.6-fold higher average off-target editing frequency than the untreated control. They attributed the unusually high average off-target editing frequency to the permissive mutant N-terminal domain of TALE, which may have increased the non-specific binding of TALE arrays.
DdCBEs with standard N-terminal domain generally exhibited 150- to 860-fold higher on-target editing relative to off-target editing, with no strong correlation between on-target editing efficiencies and off-target activity, the researchers added.
"Additional research will be needed to fully elucidate the principles that govern the efficiency and specificity of DdCBE," the authors wrote. "Developing in vitro and in vivo strategies to deliver DdCBEs will be essential for exploring their therapeutic potential in other cell types and in animal models of mitochondrial diseases."
In an accompanying New and Views column published on Wednesday in Nature, the Wellcome Centre for Mitochondrial Research's Magomet Aushev and Mary Herbert said this approach has an advantage over CRISPR for the editing of mitochondrial DNA because it obviates the need for a guide RNA, and that the DdCBE construct can be "efficiently imported" into mitochondria.
Aushev and Herbert noted that the reliance of DdCBE on DNA replication to implement the C-G to T-A conversion implied a theoretical maximum editing efficiency of 50 percent, but added that because the activity of DdCBE persists over several days, that could offer an opportunity for further editing during subsequent replication events.
"These caveats mean that DdCBE might cause a reduction in — rather than complete elimination of — mtDNA mutations. But given that the severity of the symptoms of mtDNA diseases increases with mutation load, the ability to reduce the mutation level in itself holds therapeutic promise," they wrote.