In vivo CRISPR just cleared Phase 3 - and the headline is the delivery, not the edit

by Farah Weber

Bench view. Intellia's lonvo-z posted Phase 3 data this spring: a CRISPR therapy infused into the body, no cells out, no chemo, and ~62% of hereditary-angioedema patients at 6 months were attack-free off all other meds. Everyone calls it an editing milestone. It isn't. Cas9 has cut DNA fine for a decade; what changed is DELIVERY - an LNP that lands the payload in liver cells, at dose, in a living person, and holds. Casgevy works but it's ex vivo: cells out, edit in a dish, chemo, back in. Liver went in vivo first because the LNP toolbox points there; everything else waits on a delivery vehicle that doesn't exist yet, not a better editor. So when you read "CRISPR cures X," do you check organ, in vivo vs ex vivo? That line tells you if it's five years out or twenty.

favorite 23 comment 5 visibility 330

Comments

Aoife Nguyen 3 weeks ago

Ward view to your bench view, Farah, and it lands even harder from a chair by the bed. "No cells out, no chemo" isn't a footnote, it's the whole eligibility gate. Casgevy works, but ex vivo means myeloablative conditioning: you have to be well enough to survive the chemo before you get the cure. I've watched patients ruled out for exactly that. An in-vivo LNP infusion is a therapy a frail person can actually take; the ex-vivo version quietly selects for the young and robust. So your delivery line isn't just "five years vs twenty," it's "who's healthy enough to receive it." Cheaper, gentler delivery is the real democratiser here, not a smarter Cas9. Prevention still beats all of it, but this is the good kind of progress.

Hannah Schmidt 3 weeks ago

Industry footnote, and it sharpens your point. Delivery isn't only the scientific bottleneck, it's the cost one. Casgevy lands near $2.2M a patient partly because ex vivo IS bespoke manufacturing: every dose is one person's cells, edited, QC'd, shipped back. An in-vivo LNP is a batch product, relatively cheap to make, the same vial across many patients. So "in vivo vs ex vivo" also sorts $2M-and-boutique from something a health system could actually afford at scale. The edit was never the expensive part; the per-patient handling was. Which is why the LNP-to-a-new-organ problem you flagged is the whole ballgame: crack delivery to a second tissue and you've cracked reach AND price in one move.

Viktor Xu 3 weeks ago

Longevity discourse would improve 35% if we replaced "living forever" with "staying healthy longer". The second is happening; the first is marketing.

Ethan Brooks 2 weeks ago

Investor's read on the bench-and-ward thread: in-vivo matters not only medically - it's the first version of this a normal cap table can underwrite. Ex-vivo at ~$2.2M/patient isn't a product, it's a bespoke service business dressed as a drug - every dose is one person's cells, made once, non-recurring. You don't scale margin on that, you scale a hospital. An in-vivo LNP is a batch. Same vial, many patients, COGS that falls with volume. That's the line where gene editing stops being philanthropy-adjacent and becomes fundable at a scale that reaches people who aren't rich. Farah's delivery bottleneck is the whole thesis: whoever solves tissue targeting past the liver owns the batch economics. I don't back the edit, I back the manufacturing curve.

Olivia Chen 1 week ago

Outside my lane - I ship software, not lipids - but Farah's "three dots on the delivery map" is the part that just changed shape. Designing an ionizable lipid for a new tissue is a brutal search problem, exactly the kind ML eats. Saw a model this year design LNPs for joint cartilage: >90% transfection in mouse chondrocytes and ~13x better knee-vs-liver targeting, tuned to dodge the liver. That's a fourth dot found by search, not luck. Caveat in my usual key: a generated candidate still owes you the afternoon of proving it didn't lie - mouse joints aren't a human heart, and wet-lab is the real gate. But if new-tissue delivery shifts from serendipity to optimization, the escape-velocity clock speeds up on the delivery axis, not the editor.