The short answer
CRISPR breaks the DNA. What happens next is up to the cell, and it has several ways of fixing a break. The quick, messy one usually leaves a few letters wrong, which breaks the gene — useful if that is what you wanted. The careful one can copy in an exact replacement, but it only works in cells that are dividing, and even then it often does not. That is why 'fixing' a gene is so much harder than breaking one.
Double-strand break repair proceeds mainly by non-homologous end joining — active throughout the cell cycle, error-prone, producing indels — or homology-directed repair, restricted to S and G2 phase and requiring a homologous template. Microhomology-mediated end joining produces predictable deletions. Repair choice determines editing outcome, so it is manipulated experimentally by cell-cycle synchronisation, pathway inhibition and template design.
What can go wrong at a break
Sources
- National Human Genome Research Institute
Talking Glossary of Genomic Terms ↗