One infusion, a permanent edit: what gene editing for cholesterol has measured so far

The idea is easy to picture: one infusion, and your liver stops making a protein that keeps your LDL cholesterol high. No daily tablet, no injection every two weeks, nothing to remember again. In 2026 that idea finally has human numbers in a journal rather than a press release. They are early, from a few dozen carefully selected patients, and nobody has yet counted a single heart attack prevented.

This is an informational article. It is not a diagnosis, not a treatment plan, and it does not replace a consultation with your doctor.

What exactly is being switched off?

Two genes, two companies, two different kinds of editing.

PCSK9 is the protein that pulls LDL receptors off the surface of liver cells; fewer receptors means more LDL cholesterol left in the blood. VERVE-102 carries an adenine base editor and a guide RNA inside a lipid nanoparticle dressed with N-acetylgalactosamine, so liver cells take it up. The edit changes a single base at a splice site, intended to stop the PCSK9 protein from being made — it rewrites one letter rather than cutting the DNA strand in two.

ANGPTL3 is the other target. CTX310, from a different company, uses CRISPR-Cas9 messenger RNA and a guide in a lipid nanoparticle — and this one cuts. VERVE-201 is a base editor aimed at the same gene.

The infusion is given once either way, and the change to liver DNA is meant to be permanent. That is both the appeal and the problem: there is no dose to skip if something turns out to be wrong.

What has actually been measured in people?

VERVE-102 was reported from the Heart-2 trial — phase 1, open-label, single ascending intravenous dose — published in the New England Journal of Medicine in 2026. Participants were adults with heterozygous familial hypercholesterolaemia or premature coronary artery disease. The interim analysis covers 35 people with at least 28 days of follow-up, data cut 27 February 2026, across six dose levels from 0.3 to 1.0 mg of total RNA per kilogram, four to eight people in each.

Mean PCSK9 protein fell by 51% at the lowest dose and 88% at the highest; mean LDL cholesterol by 9% at the lowest and 62% at the highest — in absolute terms a mean of 78 mg/dL at 1.0 mg/kg, ranging from 54 to 117. Follow-up had reached at least a year in 15 of the 35, with no dose-limiting toxicity. What was seen: mild-to-moderate infusion reactions, transient ALT rises, and aspiration pneumonitis in one participant with reflux.

Read the next part carefully, because headlines drop it. The paper itself states that no formal statistical testing was done and that the analysis was not prespecified as a final test. A sponsor summary released alongside it lists the dose-by-dose means as 9%, 44%, 45%, 33%, 51% and 62%; the journal abstract prints only the two ends of that range. At 0.7 mg/kg the mean reduction (33%) came out below both neighbouring doses, 0.6 and 0.8 mg/kg (45% and 51%). Why, these data cannot say: no formal test was performed, and across dose groups of four to eight people a spread between neighbours is expected.

CTX310 was a phase 1 ascending-dose study, single infusion, in 15 adults whose cholesterol or triglycerides stayed high on maximally tolerated lipid drugs. Its primary endpoint was adverse events, including dose-limiting toxicity — not cholesterol. ANGPTL3 changed by +9.6% and +9.4% at the two lowest doses, then by −32.7%, −79.7% and −73.2% at 0.6, 0.7 and 0.8 mg/kg. At day 60 the four people at 0.8 mg/kg had mean LDL cholesterol 48.9% lower and triglycerides 55.2% lower. A one-year follow-up letter in 2026 reported those same four at a mean of 52.5% below their starting LDL cholesterol — range −84.2% to −24.4%.

No dose-limiting toxicity was attributed to CTX310. Two of the 15 had serious events: a spinal disc herniation, and a sudden death on day 179 after the lowest dose, 0.1 mg/kg. The published abstract does not assign a cause to that death.

VERVE-201, the base editor against ANGPTL3, has been recruiting since October 2024. No human lipid result for it was found in the sources checked.

What happened to the first version of this?

Worth knowing, because a clean set of 35 results can hide it. VERVE-101 was the same base editor and the same PCSK9 guide in an earlier lipid nanoparticle, without the GalNAc ligand. Thirteen people were dosed. Among the five of six at 0.45 mg/kg with at least 28 days of follow-up, time-averaged LDL cholesterol reductions ran from 21% to 73%, averaging 46%. The sixth had an asymptomatic grade 3 ALT rise and a grade 3 fall in platelets within four days, without bleeding; both resolved in days. Enrolment was paused in April 2024, and the sponsor moved its priority to VERVE-102, attributing those laboratory events to the nanoparticle.

Two cardiac events in that trial are also on the record: a fatal cardiac arrest about five weeks after a 0.3 mg/kg dose, judged unrelated by both the investigator and the data monitoring board; and a heart attack the day after a 0.45 mg/kg dose, judged potentially related on timing, in a person with unreported pre-dose unstable chest pain and critical disease on angiography.

All of those VERVE-101 figures are sponsor-reported, from company releases rather than a peer-reviewed paper. Saying so is not a dismissal — it is the difference between a number that has been through review and one that has not.

What none of this shows

• No trial here used heart attacks, strokes or death as an efficacy endpoint. LDL cholesterol and triglycerides are surrogate markers. A 62% LDL drop sustained for decades is the kind of change that has reduced events for statins and PCSK9 antibodies — but these studies have not shown that an edit lasts for decades, and they have not counted events.

• Both human trials are open-label, with no placebo arm and no active comparator. Heart-2's published analysis is an interim cut without formal hypothesis tests.

• Thirty-five people and fifteen people cannot define rare harm. Gene-editing products are followed for years for that reason.

• The one-year CTX310 LDL figure rests on four people, with a range from −84% to −24%. A mean of −52.5% is not a forecast for the next person.

• VERVE-101's history does not disappear because VERVE-102 uses a different particle. Cleaner early laboratory results in 35 people do not establish that the new particle is safe at scale.

• The populations were heterozygous familial hypercholesterolaemia, premature coronary disease, or lipids that stayed high on maximal therapy — not mildly raised cholesterol in an otherwise well middle-aged reader.

• Off-target editing in humans was not the endpoint reported. The VERVE-102 paper notes that its non-clinical off-target screening was done largely in cells from White donors.

• Permanence cuts both ways. You cannot stop a DNA change by skipping the next dose.

What to do with this now

Read these as phase 1 results, which is what they are: a dose-finding stage whose job is to look for harm, not to establish benefit. None of the four products is approved, and nothing here is prescribable.

Do not confuse them with what already exists. PCSK9 antibodies and the siRNA inclisiran are approved LDL-lowering drugs; evinacumab, an antibody against ANGPTL3, is approved for homozygous familial hypercholesterolaemia. The editing products described here are not those drugs, and the existence of approved options is why an early-phase trial can take its time.

The question actually worth taking to a doctor is not when you can be edited, but whether your own high LDL is familial. That is what the trials above recruit on, and the answer changes how your numbers are followed regardless.

Your LDL cholesterol is the one piece of this you can see today. If you keep your labs in Lonevi, it sits in your card with the rest of your lipid profile, where a trend across years is worth more than any single draw.

Sources

• Vafai SB et al. In Vivo Base Editing of PCSK9 with VERVE-102 for Hypercholesterolemia (Heart-2 interim). N Engl J Med 2026;395:648–659. PMID 42187087 · doi:10.1056/NEJMoa2601283

• Heart-2 (VERVE-102), registry record: phase 1, recruiting. NCT06164730

• Laffin LJ et al. Phase 1 Trial of CRISPR-Cas9 Gene Editing Targeting ANGPTL3 (CTX310). N Engl J Med 2025;393:2119–2130. PMID 41211945 · doi:10.1056/NEJMoa2511778

• Laffin LJ et al. Durability of CRISPR-Cas9 Gene Editing Targeting ANGPTL3 with CTX310 (one-year correspondence). N Engl J Med 2026;395:1343–1345. doi:10.1056/NEJMc2609825

• CTX310 phase 1, registry record. ACTRN12623000809639

• Heart-1 (VERVE-101), registry record: phase 1, 13 participants dosed. NCT05398029. The VERVE-101 figures above, the April 2024 enrolment pause and the grade 3 laboratory events come from the sponsor's releases of 12 November 2023 and 2 April 2024; they have not been peer-reviewed or published as a paper.

• Pulse-1 (VERVE-201), registry record: phase 1b, recruiting since 30 October 2024, no human lipid results published. NCT06451770