Long-Term Data Show Durable, Single-Dose In-Vivo Gene Editing

The central question for in-vivo gene editing — editing a patient's DNA inside the body rather than in harvested cells — has never been whether it works for a few weeks, but whether a single treatment can stay effective and safe for years. Three programs now have human follow-up long enough to begin answering that question, and the early signal is consistent: the editing effect is remarkably durable. What is also clear is that durability raises the stakes on every safety event, because an edit cannot be undone.

Intellia's nex-z: the longest human track record

Nexiguran ziclumeran (nex-z, formerly NTLA-2001) is the most clinically advanced in-vivo CRISPR therapy. Delivered as a liver-directed lipid nanoparticle (LNP) carrying CRISPR-Cas9 mRNA and a guide RNA, a single intravenous infusion permanently knocks out the TTR gene in hepatocytes — the main source of circulating transthyretin. In a Phase 1 trial in transthyretin amyloidosis with cardiomyopathy (ATTR-CM), 36 patients received one dose; mean serum TTR fell 89% at day 28 and 90% at 12 months (Fontana et al., New England Journal of Medicine, 2024).

The durability data have since extended well beyond one year. In a September 2025 disclosure, Intellia reported that among patients dosed at 0.3 mg/kg or higher (n=33), mean TTR reduction was 92% at 24 months, and 90% (mean absolute level ~20 µg/mL) among the 12 patients followed to 36 months. At the 2025 American Heart Association (AHA) Scientific Sessions, longer-term ATTR-CM data (n=9 at 36 months) showed an 87% mean TTR reduction with "no evidence of a waning effect," and stability or improvement in NT-proBNP and high-sensitivity troponin T in 70% and 85% of patients, respectively, at 24 months. Intellia also reported an all-cause mortality rate of 3.9 per 100 patient-years in the nex-z group versus 12.7 in a matched cohort (HR 0.27, p=0.009) — a striking figure, but one drawn from a small, non-randomized comparison that cannot establish causation.

A safety setback that tempers the optimism

Durability makes the safety bar structurally higher than for chronically dosed drugs. That tension became concrete in late 2025. On October 29, the U.S. FDA placed clinical holds on Intellia's Phase 3 MAGNITUDE and MAGNITUDE-2 trials after a patient in MAGNITUDE developed Grade 4 liver transaminase elevations and increased bilirubin; that patient died on November 5. Intellia stated the treating physician attributed the death to complicating comorbidities, and at the January 2026 JPMorgan Healthcare Conference the company said investigation pointed to sepsis/infection rather than liver failure. Grade 4 transaminase elevations had occurred in less than 1% of MAGNITUDE patients. The holds were lifted in early 2026 with added mitigations — enhanced liver-enzyme monitoring, short-course steroids for early elevations, and exclusion of patients with certain liver abnormalities or very low ejection fraction. In the earlier Phase 1 study, liver-enzyme elevations were generally transient and asymptomatic.

Verve: base editing PCSK9 for cardiovascular disease

Verve Therapeutics pioneered in-vivo base editing — a precise A-to-G change in the PCSK9 gene that switches off the protein and durably lowers LDL cholesterol. Its first-generation candidate, VERVE-101 (standard LNP), in the Heart-1 Phase 1b trial (10 patients with heterozygous familial hypercholesterolemia and established cardiovascular disease) produced LDL-C reductions up to 55% and PCSK9 reductions up to 84%. Crucially for the durability question, Verve reported that the single highest-dose participant maintained a time-averaged LDL-C reduction of 58% two years after a single infusion.

Heart-1 also surfaced the delivery-risk theme: enrollment was paused after a treatment-related serious adverse event (Grade 3 transaminase elevations and thrombocytopenia) that Verve attributed to the LNP rather than the editor, and cardiovascular serious adverse events in two patients were judged consistent with underlying disease. Verve then prioritized VERVE-102, which uses a refined GalNAc-LNP that targets hepatocytes more specifically. Initial Heart-2 Phase 1b data (April 2025; 14 patients across three cohorts) showed dose-dependent LDL-C reductions of 21%, 41%, and 53% (maximum 69% in the top cohort), with no treatment-related serious adverse events, no dose-limiting toxicities, and only a single mild-to-moderate infusion reaction.

Beam: base editing as genetic correction

Beam Therapeutics takes a distinct approach — correcting, not merely disabling, a disease-causing mutation. Its lead in-vivo program, BEAM-302, uses adenine base editing to fix the PiZ mutation in SERPINA1 that causes alpha-1 antitrypsin deficiency (AATD), restoring functional antitrypsin. Early Phase 1/2 data (March 2025) in nine patients showed durable, dose-dependent correction and rises in functional AAT to therapeutic levels. A second-quarter 2025 update (17 patients dosed, follow-up ranging from three days to 14 months) reported all adverse events were mild-to-moderate, with no serious adverse events and no dose-limiting toxicities; liver transaminase elevations were Grade 1 and resolved without intervention. The FDA granted BEAM-302 Regenerative Medicine Advanced Therapy (RMAT) designation in June 2025.

What durable in-vivo editing could mean — an original assessment

The following is analysis, not established fact. The convergence of these datasets suggests a plausible inflection point: for a subset of chronic, genetically tractable liver-driven diseases, a one-time in-vivo edit may replace decades of repeat dosing — with ATTR amyloidosis the clearest near-term test case, given nex-z's deep, sustained TTR suppression and the precedent of TTR silencers. The strategic logic extends to cardiovascular risk, where Verve's data hint that a single PCSK9 edit could match or exceed the LDL lowering of lifelong injections. If durability holds in randomized Phase 3 outcomes trials, the value proposition shifts from "effective drug" to "infrastructure" — a treatment administered once and largely forgotten, with major implications for adherence, cost structure, and access in lower-income health systems.

The counterweight is equally real. Irreversibility is a feature only when the edit is correct; an off-target event, an unexpected immune response, or a delayed toxicity cannot be dialed back. The Intellia clinical hold is a reminder that the liver-toxicity signal seen with viral and non-viral delivery alike is not solved, merely managed. Until hard cardiovascular and neurological outcomes — not surrogate biomarkers — are proven in controlled trials, "one-and-done" remains a hypothesis supported by pharmacokinetics, not by patient-relevant endpoints.

Caveats

All programs discussed remain in early-phase or preliminary Phase 3 stages. Patient numbers at the longest follow-up are small (n=9 to n=12 for three-year ATTR data; n=14 for VERVE-102; n=17 for BEAM-302), open-label designs lack control arms, and endpoints are largely biomarkers (TTR, LDL-C, AAT) rather than definitive clinical outcomes. No therapy described here is approved, and none should be characterized as a cure. Long-term safety — including cancer risk from editing and rare delayed adverse events — cannot be excluded with current follow-up and is being tracked under mandated long-term monitoring.