Краткий ответ
Three things have to happen. First, a short RNA guide finds the exact spot in the DNA. Second, a protein called Cas cuts there. Third — and this is the part most explanations skip — the cell notices the break and repairs it. The repair is where the edit comes from, and the cell chooses how to do it, not you.
CRISPR editing proceeds through target recognition, PAM-dependent R-loop formation, strand cleavage by the HNH and RuvC domains, and endogenous repair. Non-homologous end joining predominates and produces small insertions or deletions; homology-directed repair, active only in S and G2 phase, can install a precise change from a supplied template but is inefficient. Microhomology-mediated end joining and single-strand annealing contribute further outcomes.
Why repair is the whole story
The cell has several ways of repairing a double-strand break, and it chooses between them based on its state and the phase of its cycle. End joining is fast, always available, and sloppy — it usually leaves a few letters added or missing, which typically breaks the gene. That is a fine outcome if breaking the gene is the goal.
If you want a precise correction you need homology-directed repair, which copies from a template you supply. It only works in dividing cells and is inefficient even then. Non-dividing cells — neurons, mature muscle fibres — cannot use it at all. That single fact explains a great deal about which diseases are tractable and which are not, and it is why base and prime editing were invented.
Sources
- National Human Genome Research Institute
Talking Glossary of Genomic Terms ↗