NEW YORK (GenomeWeb) – A team of Broad Institute researchers led by David Liu today reported that CRISPR-based adenine base editors (ABEs) have been found to generate low but detectable levels of widespread adenosine-to-inosine editing in cellular RNAs.
In order to solve this problem, the researchers developed new ABE variants that retain their ability to edit DNA efficiently but have significantly lower RNA editing activity, as well as lower off-target DNA editing activity and reduced indel byproduct formation, in three mammalian cell lines.
"By decoupling DNA and RNA editing activities, these ABE variants increase the precision of adenine base editing by minimizing both RNA and DNA off-target editing activity," the authors wrote in their study in Science Advances.
ABEs have been shown to enable precise and efficient conversion of target A-T base pairs to G-C base pairs in genomic DNA, and have been reported to exhibit minimal off-target DNA editing, the researchers noted. Because the mutation of G-C base pairs to A-T base pairs is the primary form of de novo mutation, ABEs have the potential to correct almost half of known human pathogenic point mutations, they added. The most recent version of the ABE base editor is called ABEmax.
However, recent studies have also shown that despite their reputations for generating low levels of off-target DNA edits, base editors may create unexpected problems. In March, an international team of researchers reported a new method to detect off-target mutations created from editing one blastomere of two-cell mouse embryos using either CRISPR-Cas9 or one of two base editing technologies. The investigators found that the cytosine base editor 3 induced SNVs with frequencies more than 20-fold higher than the spontaneous mutation rate.
And in April, a team of researchers from Massachusetts General Hospital led by Keith Joung reported in Nature that both ABEs and cytosine base editors could cause extensive transcriptome-wide off-target RNA editing in human cells.
For their new study, Liu and his colleagues measured, with high sensitivity, A-to-I editing that could be attributed to overexpression of ABEmax. They began by transfecting HEK293T cells with a plasmid expressing ABEmax and isolating genomic DNA and RNA after 48 hours. They then performed high-throughput sequencing on 220 to 250 nucleotide regions of three mRNA amplicons: CTNNB1, IP90, and RSL1D1.
They found that in all three transcripts, ABEmax had generated low but detectable levels of RNA editing above the endogenous level of A-to-I editing from cellular deaminases. For example, ABEmax generated an average 22-fold increase in the A-to-I conversion among all sequenced adenosines in the RSL1D1 mRNA, compared to the A-to-I conversions generated by a Cas9 nickase-only control in the same transcript.
Further experiments revealed that, on average, ABEmax overexpression induced 14,959 additional high-confidence A-to-I edits compared to the Cas9 nickase-only control. Although ABEmax overexpression added only 28 percent more detected A-to-I edits than the 53,334 endogenous cellular A-to-I edits observed in the Cas9 nickase-only control, these additional ABEmax-induced RNA edits were widespread throughout the transcriptome, the researchers noted.
In order to reduce these unwanted RNA edits, the investigators then worked to design ABE mutants, starting with a construct that had an inactivated wild-type Escherichia coli tRNA-specific adenosine deaminase (TadA) domain. All ABEs reported to date are single polypeptide chains containing a wild-type E. coli TadA monomer that plays a structural role during base editing, a laboratory-evolved E. coli TadA monomer (TadA*) that catalyzes deoxyadenosine deamination, and a Cas9 nickase.
Their experiments found that RNA editing across the transcriptome was reduced by inactivating either the TadA or TadA* monomers. The catalytically inactivated ABEmax resulted in 53,917 A-to-I edits, similar to the 53,334 A-to-I edits detected in the Cas9 nickase-only control.
The construct with the inactivated wild-type TadA domain maintained its strong DNA base editing activity. And when the team installed mutations into the evolved TadA* monomer, it found that certain mutations not only greatly reduced the levels of RNA editing and maintained DNA editing levels similar to those of ABEmax, but they also reduced the levels of off-target DNA edits compared to ABEmax. They named this version of the editor ABEmaxAW.
"Collectively, these results indicate that mutations that reduce the tolerance of ABEmax for RNA editing also increase the DNA specificity of base editing, likely by reducing DNA binding interactions that support productive editing of off-target loci," the authors wrote.
Finally, the researchers generated and tested two additional ABEmax mutants in order to improve on the editor's DNA on-target editing frequencies. They found that mutation of both the wild-type and the evolved TadA monomers is required for the most effective reduction in RNA editing and indel frequencies.
They named this version of the editor ABEmaxQW, noting that it performed as well as or slightly better than ABEmaxAW at on-target DNA base editing, and displayed similarly low levels of off-target RNA editing.
"Although we note that even ABEmax-mediated RNA editing is both low level (averaging 0.21 percent across all transcripts) and transient given the short half-life of most cellular RNAs, the extent to which low-level RNA editing may interfere with base editing biological studies or therapeutics development efforts will depend strongly on features of the specific applications, including the duration of exposure to the base editor," the authors concluded. "We recommend that researchers use ABEmaxAW or ABEmaxQW for adenine base editing applications that require minimizing RNA editing, off-target DNA editing, and/or indel formation."