CRISPR–Cas9 genome editing

Questions to investigate

1. For an early exon, which protospacer has a good GC content, no poly-T, and the fewest close off-target matches? Why should a second independent guide still be tested?
2. How does moving the cut from an early coding exon to a later one or out of the coding sequence change the chance of a loss-of-function knockout?
3. Compare the indel percentage from the T7E1 assay with the exact figure from amplicon sequencing. When does the T7E1 estimate mislead?
4. Add the single-stranded donor. What fraction of alleles carry the point mutation, and why is the HDR rate so much lower than the NHEJ rate? What do the silent PAM- or seed-blocking mutations in the donor achieve?
5. In zygotes, run several founders. How many are mosaic rather than cleanly biallelic, and how should they be bred to obtain a stable line?
6. Amplify the top predicted off-target sites. How does the editing rate there depend on the number and position of the mismatches?

In a real laboratory

A target region is chosen in an early coding exon. A design tool lists candidate protospacers with their GC content, any poly-T stretch (which terminates U6 transcription), the predicted cut site and genome-wide off-target matches. Two independent guides are usually taken forward, so that a shared phenotype cannot be explained by off-target cutting. The guide is delivered with Cas9 into cultured cells by transfection or electroporation, or injected as Cas9 messenger RNA or protein with the sgRNA into one-cell zygotes; a single-stranded donor oligonucleotide is added when a precise point mutation is wanted. Every experiment includes a non-targeting guide and an untreated control. Editing is estimated quickly by a T7 endonuclease I assay and exactly by amplicon sequencing, loss of the protein is confirmed by Western blot, and the top predicted off-target sites are sequenced. Co-injection of Cas9 messenger RNA and sgRNAs into mouse zygotes produced mice with biallelic mutations in two genes at high efficiency, and co-injection with oligonucleotides produced precise point mutations (Wang et al., 2013). Founders are often mosaic and are bred to separate single alleles. The work requires approval by an Institutional Biosafety Committee, and animal work requires approval by an Institutional Animal Ethics Committee (in India, under the CCSEA and the Prevention of Cruelty to Animals Act 1960; elsewhere, the equivalent national body), humane end points and attention to the 3Rs.

How the simulation works

A seeded artificial gene of 300 codons in six exons, with untranslated regions and introns, lies in a simulated genome of about 180 kilobases. Near-matching off-target sites are planted for some of its protospacers, with mismatches mostly at the PAM-distal end. Every SpCas9 protospacer (20 nucleotides followed by NGG) on both strands is listed with its position, GC content, poly-T, cut site three base pairs from the PAM, whether the cut lies in an early coding exon, and off-target counts for zero to three mismatches. The score shown is an illustrative model based on sequence features, not a published efficiency algorithm. Each allele is cut with a probability set by delivery and score, and repaired by NHEJ (mostly small deletions and 1-base insertions, with some larger deletions) or, with a donor, occasionally by HDR. Cells carry two alleles and zygote founders several, which produces mosaicism. The T7E1 estimate is computed from the allele distribution as 100×(1−√(1−cleaved)). Amplicon reads include sequencing errors and are aligned by an exhaustive small-indel search, and the edited coding sequence is translated. Mismatches far from the PAM reduce off-target editing less than those close to it (Hsu et al., 2013). All rates are approximate. Chromatin, guide secondary structure, base and prime editing and the biochemistry of delivery are not simulated.