CRISPR and editing in vivo

You can now read a genome (M-Med-01) and you know that most of it is a recipe we only partly understand. This module is about the next capability: rewriting it. CRISPR is the tool that made rewriting cheap and precise enough to become medicine — and, like sequencing, it turned out that the headline capability was the easy part and the delivery was the hard part.

Borrowed from bacteria

Established CRISPR did not come from a genetics lab dreaming up an editing tool. It is a bacterial immune system. Bacteria under attack by viruses keep short snippets of viral DNA filed away in their own genome, in regions called CRISPR arrays, as a molecular memory of past infections. When a matching virus returns, the bacterium transcribes those snippets into guide RNAs that direct a protein called Cas9 to the matching viral sequence, which Cas9 then cuts.

The 2012 insight — the one that won a Nobel Prize — was that this natural targeting system could be reprogrammed. Supply Cas9 with a guide RNA of your own design, and it will find and cut whatever matching 20-letter DNA sequence you specify. A billion-year-old bacterial defence became a general-purpose “find this sequence and cut it” machine.

How the targeting works

Established The mechanism reuses the complementarity you met in M-Bio-04. The guide RNA is a ~20-base sequence complementary to the DNA target. Cas9 carries the guide, unwinds candidate DNA, and lets the guide test for a base-by-base match. Where the match is good and a short adjacent motif (the PAM) is present, Cas9 cuts both strands.

What happens next is the cell's own repair machinery, and this is subtle. The cut alone edits nothing; it just breaks the DNA. The cell repairs the break, and how it repairs determines the edit. Sloppy repair (deleting a few bases) is often used to disable a gene. Precise repair using a supplied template can rewrite a sequence, but it is far less efficient. Newer tools — base editors and prime editors — sidestep the double-strand break entirely, chemically converting one base to another or writing in a short new sequence with a template, achieving far greater precision than first-generation CRISPR.

Notice what the tool does and does not give you. It gives you an exquisitely programmable way to reach a chosen sequence. It does not give you control over the cell's repair, control over which cells receive the machinery, or a guarantee that the guide will not also match somewhere else. The programmability is real; the control is partial.

Ex vivo: the easy mode

Established The first approved CRISPR medicine, Casgevy (2023, for sickle-cell disease and beta-thalassemia), works ex vivo: doctors remove a patient's blood stem cells, edit them in a dish, check and expand the correctly-edited cells, and return them to the patient after chemotherapy clears the old marrow. This is a genuine cure for a disease caused by exactly the single-base change you traced in M-Bio-04 — a landmark.

But ex vivo works only for tissues you can remove, edit, and put back — chiefly blood. You cannot extract a brain, a heart, or a pancreas, edit it in a dish, and reinstall it. For the vast majority of the body, editing has to happen in place.

In vivo: where it gets hard

Frontier In 2021, the first in-body CRISPR editing was demonstrated clinically: NTLA-2001, delivered intravenously to edit a disease gene in liver cells, using lipid nanoparticles (the subject of the next module) to carry the machinery. It worked, and it opened the era of in vivo editing. It also made the remaining problems sharp.

The liver is the easy in vivo target, because nanoparticles injected into the blood naturally accumulate there. Almost every other tissue is harder to reach selectively. And in vivo you lose the two safeguards that make ex vivo safe: you cannot select for correctly-edited cells (every mis-edit stays in the patient), and you cannot check before the edit is permanent. Off-target cuts, edits in the wrong cell type, and mosaic results (some cells edited, some not) are no longer discardable — they are consequences.

What actually remains

Frontier Three problems dominate, and it is worth being precise about which is which. Delivery — getting the machinery to the right tissue among trillions of cells — is the biggest, and it is why the next module is about lipid nanoparticles. Precision — ensuring the edit lands only where intended — has improved enormously with base and prime editing but is not solved for whole-body use. Permanence — a DNA edit cannot be undone — is a feature for curing a genetic disease and a liability if anything is wrong.

Speculative A fictional healing system that rewrites cells throughout the body on demand is assuming all three problems solved at once, in every tissue, reliably enough to act autonomously. Reading this module tells you which part of that is a smooth extrapolation of current capability (the cutting and the programmability are largely there) and which part is the genuine chasm (targeted, precise, whole-body delivery). As always, the honest move is to locate the boundary rather than wave at the whole thing.