NEW YORK (GenomeWeb) – A team of researchers from Massachusetts General Hospital led by Keith Joung has reported in a new study that CRISPR-Cas base editing technology can induce transcriptome-wide off-target RNA editing in human cells, along with off-target DNA edits.
The most widely used base editors — versions BE3 and BE4 of the cytosine base editor (CBE) platform — induce DNA cytosine deamination using the rat APOBEC1 (rAPOBEC1) enzyme, which is targeted by a linked Cas protein guide RNA complex. Previous studies on the specificity of CBEs have focused on and identified off-target DNA edits in human cells. But for their new study published today in Nature, the researchers showed that a CBE with rAPOBEC1 can cause extensive transcriptome-wide RNA cytosine deamination in human cells, inducing tens of thousands of cytosine-to-uracil edits with frequencies ranging from .07 percent to 100 percent in 38 percent to 58 percent of expressed genes. They also observed that newer adenine base editors (ABEs) can also induce transcriptome-wide RNA edits.
The researchers were also able to engineer two CBE variants with rAPOBEC1 mutations that decreased the numbers of RNA edits by 390-fold and 3,800-fold in human cells. These variants also showed more precise on-target DNA editing and editing efficiencies compared to those they observed with wild-type CBEs.
"These results have important implications for the research and therapeutic uses of base editors, illustrate the feasibility of engineering improved variants with reduced RNA editing activities, and suggest the need to more fully define and characterize the RNA off-target effects of deaminase enzymes in base editor platforms," the authors wrote.
To test whether CBEs might deaminate RNA cytosines, the researchers examined the activity of BE3 in human liver-derived HepG2 cells. Quadruplicate experiments confirmed efficient on-target DNA editing by BE3 at the RNF2 gene. To assess RNA editing, the team then used targeted RNA amplicon sequencing to examine cytosines in the human APOB transcript, and found that C-to-U alterations were induced with frequencies ranging from .07 percent to nearly 81.5 percent and were distributed throughout the transcriptome.
Significantly, 43 percent to 52 percent of the genes detected in the RNA-seq experiments had at least one C-to-U edit. The researchers found alterations in coding sequences and non-coding sequences, and found evidence suggesting that BE3 is consistently editing specific cytosines.
"Importantly, using whole-exome sequencing that captures both exons and UTRs, we were able to sequence with 100X coverage (in pooled triplicates) 49 percent of cytosines identified as edited on RNA and found that 98.48 percent of these showed no evidence of DNA, confirming that the edits observed in the RNA-seq experiments are not caused by editing of corresponding DNA sequences," the authors wrote.
The researchers then tested whether transcriptome-wide RNA editing could occur in a non-liver human cell line by examining the editing activity of BE3 with guide RNAs for the RNF2 and EMX1 genes in human HEK293T cells.
They again observed efficient on-target DNA editing alongside tens of thousands of C-to-U RNA edits with editing efficiencies ranging from .07 percent to 66.7 percent. The edits were distributed transcriptome-wide in both coding and non-coding sequences, with 38 percent to 52 percent and 47 percent to 51 percent of expressed genes having at least one C-to-U edit for the RNF2 and EMX1 gRNAs, respectively.
Although CBEs with rAPOBEC1 are now widely used, the researchers also questioned whether recently described ABEs, which induce targeted adenosine to inosine (A-to-I) DNA alterations, might cause RNA edits. They co-transfected HEK293T cells in triplicate with plasmids encoding the ABEmax editor or a negative control and the HEK site 2 gRNA and observed efficient on-target DNA adenine editing at HEK site 2.
RNA-seq analysis, however, revealed that tens of thousands of RNA base positions were altered in cells expressing ABEmax compared to matched negative control cells, with about 99.8 percent of the changes being A-to-G edits on cDNA that was reverse transcribed from RNA. The researchers observed adenine edits in frequencies of ranging from .1 percent to 100 percent, and they were distributed throughout the transcriptome. They found RNA edits in coding and non-coding sequences, and observed at least one adenine edit in 51 percent to 59 percent of the genes with detectable RNA transcripts. Further, the researchers were able to sequence 88 percent of the adenines edited on RNA at 100X coverage using WES, and found that 95.4 percent of them were not edited on DNA.
The team then went on to determine if it could engineer a solution to the problem with CBEs. It developed SElective Curbing of Unwanted RNA Editing (SECURE) variants, with the goal of reducing RNA editing activity but still efficiently editing DNA.
The researchers screened 16 BE3 editors harboring various rAPOBEC1 mutations previously reported to reduce RNA C-to-U editing, and identified two variants (BE3-R33A and BE3-R33A/K34A) that had on-target DNA editing efficiencies comparable to wild-type BE3 but that also showed substantially reduced RNA editing activities. They then performed RNA-seq experiments with the RNF2 gRNA, and found that the variants' on-target DNA editing efficiency was comparable to wild-type BE3 with the RNF2 gRNA in HEK293T cells.
Further experiments with BE3-R33A and BE3-R33A/K34A added to 12 gRNAs designed for various human genes in HEK293T cells showed that the variants' on-target editing efficiency was generally comparable to wild-type BE3, but with higher precision. Comparable or sometimes higher efficiencies of base editing were observed at 10 of the 12 sites with BE3-R33A and at eight of the 12 sites with BE3-R33A/K34A.
Both variants also demonstrated significantly reduced numbers of RNA edits throughout the transcriptome in HepG2 cells, the researchers added.
"Confounding effects of unwanted RNA editing will need to be accounted for in research studies, especially if stable base editor expression (even in the absence of a gRNA) is used," the authors concluded. "For human therapeutic applications, the duration and level of BE expression should be kept to the minimums needed. Our data suggest that safety assessments for human therapeutics may need to include an analysis of the potential functional consequences of transcriptome-wide RNA edits."