Chu et al. have recently reported in Nature Biotechnology a detailed description of inhibiting the NHEJ pathway to enhance HDR (homology-directed repair) in cell culture CRISPR editing experiments. This follows on the observation by Singh et al that SCR7, an inhibitor of ligase IV, accomplishes this in mouse embryos although the data set in the Singh paper was limited. This new paper is very detailed in its implementation of a traffic-light reporter system to quantify NHEJ and HDR under various conditions.
Since there's more than one way to potentially inhibit NHEJ, the authors used several approaches. First, shRNA was used to silence ligase IV and/or KU70 or KU80. The latter two proteins make up the KU heterodimer, required for NHEJ. Second, they used SCR7. Third, they coexpressed the adenovirus proteins E1B55K and E4orf6. These are known target ligase IV for degradation.
Before getting into more details, the bottom line was that each of these approaches (in human HEK293 cells) could both inhibit NHEJ and increase the rate of HDR at the same time. However the approaches did this to varying degrees. There was a clear trend that the more NHEJ inhibition occurred, the greater was the increase in HDR efficiency. The most effective treatment was the coexpression of E1B55K + E4orf6. Not only did this reduce NHEJ more than the other methods but it showed the greatest HDR increase as well.
Figs. 1c, 2h, and 2j respectively suggest ~7x, ~8x, and ~3.4-51x increases in HDR were observed with E1B55K + E4orf6 coexpression as compared to the identical CRISPR/Cas experiments without these factors.
Therefore, transient NHEJ inhibition is looking better and better as a way to enhance HDR in CRISPR applications. This will be likely be very important for gene therapy approaches, as well as just plain very useful for any HDR applications including of course, mouse embryo injections. In fact, Chu et al go on to show that by combining flow-sorting of transfected cells with transient selection with an inserted antibiotic gene they were able to obtain essentially HDR-edited cell clones at essentially a 100% rate (e.g. supplemental fig. 13).
Suddenly I am very interested in these adenoviral E1B55K and E4orf6 genes… These are apparently derived from the adenovirus C, a.ka. adenovirus 5, and Origene sells cDNAs and antibodies for these. E4orf6 is also called E4orf6/7.
This is a new paper that is nice as it presents a lot of detailed results across a series of test CRISPR injections in mice.
It's always tough to generate lots of numbers across mouse injections. It's just a lot of work and expensive. Despite some small numbers in the individual experiments, and some repeat description of indel mutations, etc. that others have reported, this group presents some new insights that are good to know. The most important one I got from this paper was an interesting observation concerning their gene repair experiments.
The target gene in this case was Tyr, or Tyrosinase, which of course when completely inactivated causes albinism. This group did several good comparisons of different targets, nickase vs. "native" Cas9, and also injections on C57BL6 ("B6")versus albino C57BL6 ("AB6"). Thus, they could make induce mutations disrupting wild type Tyr in the B6 strain - generating albino mice - while attempting to repair the already-mutated Tyr in the albino AB6 strain.
To do the latter they tried using as the donor a wild-type ds-DNA fragment with 800 bp homology arms flanking the wild type region of the CRISPR target(s). This failed to repair Tyr in AB6 mice, which in of itself was not a tragedy - CRISPR-mediated repair is not as efficient as simple mutagenesis and the sample size was low. But what they also observed is that these injections generated zero indel mutations, even though they showed the same CRISPR reagents generated lots of mutations at the target sites just fine when they were injected WITHOUT the wild type donor DNA molecule. Therefore they repeated the experiments with a modified donor DNA fragment that had sense mutations disrupting the CRISPR target site. This worked well in both generating a few correctly repaired alleles, but also restoring high rates of indels generated by NHEJ, thus proving that the wild-type molecule itself was inhibiting the CRISPR machinery. Since the donor was a double-stranded DNA, it makes sense that it was "soaking up" the Cas9/sgRNA, It was certainly present at large stoichiometric excess to the chromosomal targets.
I counted up the gene repair injections they did with either the WT or modified donor DNA fragment, and the summary was as follows:
Experiment : (Mice with indels/ Mice with HDR-mediated repair / Total number mice)
Wild-type donor DNA: 0 / 0 / 27. Across 3 experiments,
Modified donor DNA: 18 / 3 / 41. Across 5 experiments.
These experiments were a mix of nickase vs. native Cas9, 2 different targets, and B6 and AB6 mice. But the trend seems like it's clear - CRISPR-mediated mutations were strongly inhibited with a wild-type donor DNA that carried a perfect match to the CRISPR target.
Would a single-stranded donor oligo with a wild-type target match also inhibit? It's not clear to me, as I think all the oligo edits I've seen published disrupted the target so that doesn't provide data; the single stranded DNA probably won't complex with Cas9.
