CRISPR Gene Editing Moves In Vivo as AI Designs New Editors

CRISPR moves inside the body

In the decade since CRISPR left the lab, the main challenge has been getting it into the patient. The only approved CRISPR medicine, Vertex and CRISPR Therapeutics' Casgevy, works ex vivo. Doctors collect a patient's blood cells, edit them outside the body and infuse them back.11 Several developments in the past few weeks point to the field leaving that model behind. Editing is moving directly into the body, and artificial intelligence is starting to design the editing tools themselves.

The overall picture is consistent. In vivo editing has reached a pivotal clinical success. Delivery science is starting to reach cell types beyond the liver. AI has gone from helping plan experiments to producing working enzymes. Safety, cost and patient uptake are still the open questions.

Intellia's approval would be the first in vivo CRISPR medicine

The biggest clinical milestone of 2026 came in April. Intellia Therapeutics said its in vivo therapy lonvoguran ziclumeran (lonvo-z) met its goals in a Phase 3 trial in hereditary angioedema, a rare disorder that causes potentially life-threatening swelling attacks.1115 In the 80-patient study, one infusion cut attacks by 87% compared with placebo over six months. Sixty-two percent of treated patients were free of both attacks and other therapies, against 11% on placebo.1215 Intellia described the result as the first positive Phase 3 readout for any in vivo CRISPR therapy.1719

The company has started a rolling submission to the FDA. It aims to finish the filing in the second half of 2026 and launch in the first half of 2027 if approved.1115 In an October interview, CEO John Leonard said the company's first FDA approval was approaching and that it was now planning for patient uptake and liver safety.2 If approved, lonvo-z would be the second CRISPR medicine and the first that edits DNA inside the body.12

The safety record in the pivotal trial was clean. The most common side effects were infusion reactions, headache and fatigue, and the treated group had no serious adverse events.15 One asymptomatic grade 2 rise in liver enzymes went away on its own within a week.15 On durability, Leonard said that in almost six years the company has not seen the effect fade in a single patient.11

Those liver readings matter because of what happened elsewhere at Intellia. In its separate transthyretin amyloidosis program, nexiguran ziclumeran (nex-z), a patient had a grade 4 liver injury and later died of septic shock. The FDA then put both late-stage trials on clinical hold.20 Reports on how far the holds have been lifted differ. A January report said the FDA freed only one of the two studies, after Intellia added tighter liver monitoring and raised enrollment from 50 to 60.20 A later analysis said both trials were allowed to restart with stricter liver testing, steroid guidance and narrower eligibility.16 Taking the coverage together, nex-z appears to be moving forward again, but with patient-selection limits that will stay in place.2

Beam, Verve and CRISPR Therapeutics are close behind

Intellia's lead is clear, but other companies are not far back. Beam Therapeutics uses base editing, which changes a single DNA letter without cutting both strands. Its alpha-1 antitrypsin deficiency program corrects the disease-causing mutation itself rather than switching a gene off.14 One analysis reported that at the chosen 60-milligram dose, BEAM-302 raised functional protein above the commonly used protective threshold, and mutant protein fell by 84%.16 Beam has agreed with the FDA on a possible accelerated-approval route based on biomarkers measured over a year.14 Assessments of the race broadly agree: Intellia has stronger clinical-outcome evidence from a randomized trial, while Beam has the more technically ambitious correction, shown in a smaller open-label study.16

Cardiovascular disease is the potential mass market. Verve's PCSK9-editing program, which Eli Lilly acquired in June 2025, has lowered LDL cholesterol.1314 The reported size of that drop depends on which dataset is cited. A 14-person dataset showed an average 59% reduction at the highest dose.14 A later 35-participant update put the top-dose average at 62%, and Lilly committed to starting Phase 2 by the end of 2026.16 Results from different cohorts and times explain most of the gap, but readers should not treat these as one settled number.

CRISPR Therapeutics will present Phase 1b results for CTX310, its in vivo editor targeting ANGPTL3 in severe hypertriglyceridemia, as a late-breaking talk at the American Heart Association meeting on November 9.7 Earlier data published in November 2025 showed reductions of 55% in triglycerides and 49% in LDL.13 Investors have reacted ahead of the readout. One market report linked a 7% weekly gain in the company's shares to the upcoming CTX310 data and autoimmune results for its cell therapy zugo-cel.6

Building CAR-T cells inside the body

The most notable research result this month came from UC San Francisco. Researchers reported in Nature that they had created cancer-fighting CAR-T cells directly inside the body.1 They used two particles. One carried CRISPR-Cas9 and was coated with antibodies against CD3, a protein found on T cells. The other carried the CAR gene and instructions to insert it at a precise spot in the genome that is active only in T cells.1 The team described it as the first time a large DNA sequence has been inserted at a precise location in human T cells without removing them from the body.1

In mice with humanized immune systems, one injection cleared detectable leukemia in nearly all animals within two weeks. The approach also worked against multiple myeloma and a solid sarcoma tumor.1 Today's CAR-T therapies cost roughly $400,000 to $500,000 and take weeks to manufacture. Making the cells inside the body could also remove the need for the harsh chemotherapy patients receive beforehand.1 The researchers say the method is not ready for patients. They have founded Azalea Therapeutics to develop it toward the clinic.1

In our view, this is the more important long-term story. Every in vivo program with clinical results so far targets the liver, where lipid nanoparticles naturally collect.14 Delivering edits to specific immune cells would open up a much larger group of diseases.

AI starts designing the editors

AI and CRISPR are now connected in a practical way. In July, Jennifer Doudna's lab reported in Science that it had used an inverse-folding protein model, guided by evolutionary data, to redesign TnpB. TnpB is a compact enzyme related to the CRISPR proteins.2224 In a bacterial screen, 466 of 1,980 designs showed activity, and about 8% outperformed the natural enzyme. In human cells, two variants reached 46% and 50% editing efficiency, compared with 28% for the original.2230 The new DNA- and RNA-binding regions shared only 83% and 72% sequence identity with their closest natural relatives. Earlier language-model designs had kept binding domains more than 99% identical to natural proteins.22

Size is a big part of why this matters. TnpB is about a third the size of Cas9. The viral vectors often used for delivery hold only about 4.7 kilobases of genetic material.24 A separate Chinese team used AlphaFold 3 to build ContactSeek, which identifies the parts of an editor most linked to off-target errors. Two changes guided by the tool produced a base editor that beat several existing high-fidelity editors.23

AI is also making other parts of the toolkit faster. Monash University researchers used RFdiffusion and ProteinMPNN to design "off-switch" proteins that block Cas13. The process took eight weeks from choosing the target to identifying leads.2527 CRISPR-GPT, a language-model agent, planned and analyzed complete gene-editing experiments in human cancer cell lines.21 Related work points toward drug discovery more generally. One preprint describes a platform that combines CRISPR functional maps with AI to find approved drugs that modulate the NLRP3 inflammasome.28 Separately, Genentech validated a neuroscience target discovered by AI through its partnership with Recursion.24

The outlook here should stay modest. None of the AI-designed editors is a drug. They have been tested in bacteria, plants and cultured cells, and still have to show they can edit enough cells in the body without provoking an immune attack.2324 The coverage agrees that AI's near-term effect will be on research and agriculture. Its therapeutic payoff will take years.2430

What to watch next

The next several months include several clear checkpoints. These include Intellia's completed filing and the FDA's decision on lonvo-z, CRISPR Therapeutics' AHA presentation on November 9, and Lilly's planned Phase 2 for its PCSK9 editor.71116 Regulators have also made room for personalized treatment. A February 2026 FDA guidance allows a single trial framework customized for each ultra-rare patient. It was built on a 2025 case in which an infant was treated within six months of diagnosis.13

Gene editing is also reaching consumer products. A startup has used CRISPR to remove three allergy-causing proteins from peanuts and aims to commercialize them by 2027.10

Our reading is that 2026 is the year in vivo CRISPR moves from proof of concept to product. Durability no longer looks like the biggest uncertainty. The harder questions are whether liver-directed delivery can stay safe across larger, sicker patient groups, and whether doctors and patients will choose a permanent one-time edit over existing treatments.215 AI-designed editors and in-body cell engineering will shape the generation of therapies after that.