The naive question: you keep seeing the word 'endonuclease' in gene-editing articles. What is it? The answer is refreshingly literal. 'Endo' means inside, 'nuclease' means a thing that cuts nucleic acid. An endonuclease is a protein that cuts DNA from the inside of a strand.

Think of it like scissors versus a paper-shredder that only feeds from the edge. An exonuclease trims from the ends; an endonuclease snips in the middle, wherever you steer it. For editing a gene buried in the middle of a chromosome, you need the middle-cutting kind. The hard part — and the part the patents are really about — is aim.

“Many studies have shown that CRISPR-Cas nucleases can tolerate up to five mismatches and still cleave; it is hard to predict the effects of any given single or combination of mismatches on activity.”— U.S. Patent No. 10,544,433 source

That quotation, drawn straight from the patent's own background, names the central engineering problem. A standard CRISPR-Cas9 cutter is guided to its target by a matching RNA, but the match doesn't have to be perfect — the enzyme will tolerate several mismatches and still cut. Off-target cuts are the result, and they are hard to predict in advance. So the whole art of gene editing is making an endonuclease cut only where you want.

The General Hospital Corporation's grant US10544433B2 answers that with a clever structural trick, and the claim spells out the mechanism. It fuses the cutting domain of FokI — a nuclease that only works as a pair — to a catalytically dead Cas9 (dCas9), one carrying inactivating point mutations at the residues that normally do the cutting (positions corresponding to D10 and H840, among others listed). On its own this fusion can't cut at all. It only cuts when two of them dock side by side: two guide RNAs steer two FokI-dCas9 units to neighboring sites on opposite strands, spaced a defined 10–20 nucleotides apart (a dependent claim narrows it to 13–17), and only then do the paired FokI domains nick both strands to make a clean double-strand break. Because cutting now requires two correct addresses at the right spacing instead of one, a single off-target match isn't enough to trigger a cut — the specificity is multiplied. The claims even fix the linker between the two domains (a short Gly4Ser tether) and the exact FokI fragment used. That is precision engineered into the protein's architecture, not just the guide.

Researchers also repurpose the cutting machinery for things other than cutting, and the 2020 record shows that too. Gunma University's US10612044B2 covers a DNA methylation editing kit — using the targeting parts of the system to change a chemical mark on DNA rather than break the strand. The same address system that steers a cutter can steer an enzyme that writes or erases methyl groups instead. This is the broader pattern in the field: the guide-and-target machinery is a delivery platform, and the "warhead" bolted onto it can be a nuclease, a methylation editor, or something else entirely.

Here's the 'so what.' When a patent says it covers an 'endonuclease-based' method — as Seattle Children's Hospital's US10563226B2 does for enhancing gene editing in primary cells — it's telling you the invention is built on a steered middle-cutting protein, and that the claimed improvements are about making that cut land reliably in real patient-derived cells rather than easy laboratory cell lines. Primary cells are harder to edit, so a claim aimed specifically at them is staking out the practical, not the idealized, version of the problem.

The "dead Cas9" idea at the center of the General Hospital claim is worth pausing on, because it shows how thoroughly the field separates two functions that nature combined. A wild-type Cas9 both finds a sequence and cuts it. Disabling the cutting (via the D10 and H840 mutations the claim recites) leaves a protein that still finds its target perfectly but does nothing to it. That inert finder becomes a programmable anchor onto which you can bolt any activity you like — here, a FokI cutting domain that only works in pairs. The requirement for two anchors at a defined spacing is what converts a tolerant, off-target-prone cutter into a stringent one, because the odds of two wrong addresses lining up correctly are far lower than one.

The Gunma University methylation-editing kit makes the same separation visible from the other direction. There, the targeting machinery is kept but the cutter is replaced entirely by an enzyme that edits a chemical mark rather than the DNA sequence. Read together, the two 2020 grants describe a single underlying logic: the valuable, hard-won part is the steering, and the "endonuclease" or other warhead is a module you attach to it. That is why the word endonuclease in a claim is informative but not the whole story — it tells you the warhead is a middle-cutting protein, while the surrounding claim language tells you how that warhead is aimed and controlled.

The short version: an endonuclease is the scissors at the heart of gene editing. Everything clever in the 2020 patents is about aiming it — pairing two dead-Cas9 cutters so both addresses must match before anything is cut, repurposing the same aim to edit chemical marks instead of strands, and getting it all to work in stubborn primary cells. That one word, endonuclease, tells you the core mechanism; the claims tell you how hard people worked to point it.