NEW YORK (GenomeWeb) – Caribou Biosciences, the Bay Area CRISPR/Cas9 tools and services firm, said it has discovered regular — even predictable — patterns in the ways that DNA is repaired after genome editing.
Working with scientists form the Novartis Institute for Biomedical Research, Caribou's own research team led by Andy May looked closely at DNA repair following CRISPR/Cas9 editing at more than 200 sites in the human genome.
While a CRISPR/Cas9-induced double-strand break could theoretically lead to a multitude of different repair outcomes, cells seem to privilege the same 10 to 20 over all the others, May told GenomeWeb. "It's a very small subset of the possibilities that end up being used and it's the same subset that gets used each time, for a given cut site," he said. Not only that, the repair signature is based on the sequence.
Led by May and first authors Megan van Overbeek and Daniel Capurso, also of Caribou, the scientists published their results last week in Molecular Cell.
"This is a much more predictable, controlled process than we had anticipated," May said. Whether in same cell type or different cell type, the types of DNA repair outcomes you get are very well preserved, regardless of when and how you carry out that experiment. It says that in genome editing, it's not all about the nuclease, it's also about the DNA repair."
In their study the scientists used HEK293 cells transfected with complexed Cas9 and single-guide RNA (sgRNA) ribonucleoproteins (RNP). To analyze DNA repair outcomes, they used amplicon sequencing with reads assigned to specific types of indels.
Indel signatures looked more or less the same in multiple iterations and across two other cell lines, HCT116 and K562 cells.
To make sure the results were actually due to genomic sequence and not genomic context, the scientists designed guides against sequences found multiple times in the genome but in different contexts. In this case, all of the sites that were targeted by the same RNP were more similar to each other than to other sites, May said.
The study represents several years of work for May's team. "This was not a small piece of work to generate all of the data," he said, but it's part of Caribou's effort to drive better CRISPR/Cas9 genome-editing outcomes.
"There have been hints of some sequence dependence associated with DNA repair," May said, but he noted it was an unexpected finding. "People have assumed to date that when Cas9 makes breaks in the genome, the way that's repaired by the cell is somewhat arbitrary," he said, especially in the case of the non-homologous end joining (NHEJ) pathway.
The Caribou paper challenges that notion. It's an early break in the wave of studies using Cas9 to further probe DNA repair mechanisms. At the recent Cold Spring Harbor Laboratory Genome Engineering meeting, an entire session was devoted to the topic of DNA repair.
And in the same week that the Molecular Cell paper came out, University of California, Berkeley scientist Jacob Corn published a study in Nature Communications reporting a way to enhance gene knockout with CRIPSR/Cas9. "If you do something really simple — just feed cells inexpensive synthetic oligonucleotides that have no homology anywhere in the human genome — the rates of editing go up as much as five times," Corn said in a statement.
The study came from an observation made as the lab tried to optimize CRISPR editing using a different DNA repair process, homology-directed repair (HDR), which enables inserting a DNA template into the DSB. "We found that the frequency of error-prone repair outcomes also tended to increase when single-stranded HDR donor DNA was present in the editing reaction," they wrote.
Corn and his co-authors said that the outcome was based on cellular context and that indels formed in a cell line-specific manner, adding that the non-homologous DNA "appears to divert cells towards error-prone instead of error-free repair pathways, dramatically increasing the frequency of gene disruption."
The Caribou scientists were able to show even more specificity. They reported being able to perturb the canonical NHEJ pathway and privilege repairs from the microhomology-mediated end joining pathway.
The results have several consequences for genome editing, May said. Like Corn's study, it could help scientists get better knockout efficiencies.
"If you do a quick screen with a series of gRNAs in a region you want to edit, you can pick ones that have a high frequency of out of frame mutations," May said. "You can select gRNA for which none of the top repair outcomes will result in an in-frame mutation and all of your outcomes will lead to a mutation. If they're random, one in three would be in frame and not result in an actual KO. We can select gRNA that break that rule because the majority of repair outcomes result in a non-functional mutation.
"If you understand what those outcomes look like, you can even pick guides that give you specific mutations," he added.
May described how his lab has been able to correct a disease-causing one-nucleotide deletion using this approach. "If you cut with the right gRNA that has a high frequency of single nucleotide insertion, those repairs will put the sequence back in frame and correct the problems associated with that mutation," he said.
Better knockout efficiency and mutation correction are just two possible applications of this insight. May said he's hopeful that the study will provide even more precision to CRISPR/Cas9 genome editing.
"The system has this level of determinism, and the way cell is repairing breaks is incredibly tightly regulated and controlled," he said. "Long term, we hope we'll have more ability to predict what the outcome would be for a given sequence."