81implemented into our sgRNA selection procedure to ensure effective genome edit-
ing in rodent embryos.
- Selection of candidate guide RNAs according to on- and off-target scoring web
tools. The current best practice for genome editing in animals is to carefully
select guide RNAs to minimize potential off-target effects and test 2–3 per target
site to ensure that at least one of them has sufficient on-target activity for the
experiment. The choice of the guide RNA sequences can be facilitated by on-line
scoring algorithms for the prediction of the off-target effects [ 23 – 34 ] and the on-
target activity [ 15 – 18 , 20 , 35 , 36 ]. Conveniently, Haeussler, et al., built the web
tool, CRISPOR (http://crispor.tefor.net), that integrates several existing algo-
rithms for a comprehensive assessment of the candidate guide RNAs [ 37 ].
Authors also recommend referring the scores from Moreno-Mateos, et al. web tool
[ 18 ], known as CRISPRscan, if guide RNAs are used for mouse zygotes editing,
because the algorithm was derived from the experiments in zebrafish using in
vitro transcribed sgRNAs (similar to what we used in mouse zygotes), as opposed
to those collected data from the U6 promoter-driven in vivo expression in other
algorithms. To test the prediction accuracy, we compared the guide RNAs’ scores
from Moreno-Mateos, et al. system and their editing efficiency in mice (Fig. 4.2).
We found that Moreno-Mateos, et al. scoring system predicts better than Doench,
et al. algorithms [ 16 , 17 ] that were based on in vivo sgRNAs transcription driven
by a U6 promoter. The latter ones show a higher rate of false negatives and are
harder to define a cut-off. We recommend selecting the guide RNA that has the
highest scores across all algorithms and uses a score of 30 from Moreno-Mateos’s
algorithm as a cut-off. For targeting the non-coding region, we also take chroma-
tin accessibility into consideration. We choose the guide RNAs that target the
open chromatin regions, based on the DNaseI hypersensitivity map.- Careful design of sgRNA for the specific type of genetic modification desired.
- For basic gene knockout, sgRNA validation is not necessary. Because there is
no location restraint, we typically look into all the available exons that are
early in the coding sequence and shared by all transcript isoforms to be
deleted. We then pick two sgRNAs per gene with high predictive on- and off-
target scores from the CRISPOR or other algorithms. We also avoid GC-rich
regions that are known to be resistant to guide RNA and Cas9 targeting [ 17 ,
38 , 39 ]. We routinely target up to 4 genes with up to 8 sgRNAs per zygote
injection. - For large DNA fragment deletion, inversion, or duplication, we recommend
using two pairs of close-by sgRNAs (four sgRNAs total); each pair cuts the
start and end point of the sequence. Although two single sgRNAs are suffi-
cient for this type of DNA manipulation in many cases [ 40 – 42 ], including
ours, using a pair of sgRNAs on each side further increases efficiency, as well
as the range of the DNA length to be mutated [ 43 ]. - For smaller gene knock-in using a single-strand donor oligo (e.g. point muta-
tions, epitope knock-in, etc.), we generally select 2–3 sgRNAs near the inser-
tion site (<20 bp to the cut site is preferred) and validate their activity in the
- For basic gene knockout, sgRNA validation is not necessary. Because there is
4 A Transgenic Core Facility’s Experience in Genome Editing Revolution