Genome editing

Every programmable editing platform of the first generation — zinc-finger nucleases, TALENs, CRISPR-Cas9 — does one thing to DNA: it makes a double-strand break at a chosen position. What distinguishes them is only how the position is chosen. Zinc fingers and TALE arrays read the sequence with protein modules; Cas9 reads it with a guide RNA that base-pairs to one strand, provided an adjacent short motif, the PAM, is present. The programmability is real and it is the reason the field moved fast. It is also the entire contribution of the nuclease. The edit itself is written by the cell.

The outcome belongs to the repair pathway

A broken chromosome is a lethal lesion, so cells repair it immediately, by whichever pathway is available. Non-homologous end joining ligates the two ends back together and is active throughout the cell cycle; it is approximate, and repeated cutting and rejoining eventually leaves small insertions or deletions at the junction. In a coding exon those indels shift the reading frame, which is why NHEJ is the reliable route to a gene knockout and why the first therapeutic successes were disruptions rather than corrections.

Writing a specified new sequence requires homology-directed repair, which needs a donor template with homology arms — and needs a sister chromatid, so it operates essentially only in S and G2 phase. That single restriction explains most of the difficulty in the clinic. Neurons, cardiomyocytes and the bulk of resting hepatocytes are not dividing, so the pathway that could install a correct base is largely unavailable in exactly the tissues where a genetic disease does its damage. Editing chemistries that avoid the break altogether exist because of this constraint, not in spite of it: the sibling pages on base editing and prime editing describe them.

Off-target editing is a measured quantity

Guide RNAs tolerate mismatches, especially distal from the PAM, and can accommodate small bulges, so a guide has a set of near-cognate sites in the genome. This is measurable rather than assumed. GUIDE-seq captures a short double-stranded oligonucleotide at breaks in living cells; CIRCLE-seq and similar in vitro methods cut naked genomic DNA to nominate candidate sites with high sensitivity. Both nominate; targeted deep sequencing confirms. High-fidelity engineered variants and delivery as a ribonucleoprotein, which limits how long the nuclease is present, reduce the burden.

The under-reported risk is on target

The more serious damage often occurs exactly where the cut was intended. Reported consequences of a single break include kilobase-scale deletions, loss of heterozygosity across the surrounding region, loss of a whole chromosome arm, and chromothripsis — catastrophic shattering and misassembly of the cut chromosome. Cutting at two sites in one cell adds translocations. None of these are visible to a short amplicon assay spanning the cut site, which reports only what its primers still amplify; a lost allele looks like an unedited one. Karyotype-scale and long-read assays are required to see them, and their absence from an editing report is not evidence of their absence.

This page covers what all editors share. Delivery, containment and the specific chemistries are treated on their own pages.