Why cutting the double helix carries a price
As shown in Plate 01 (CRISPR), first-generation genome editing works like molecular scissors: Cas9 cuts both strands of the DNA double helix, and the cell's hurried repair machinery introduces small insertions or deletions that disable the target gene. In approved ex vivo medicines such as Casgevy, that deliberate cut transforms care for sickle-cell disease.
Yet cutting the genome carries biological overhead. Double-strand breaks can trigger large deletions, chromosomal translocations, or p53 DNA-damage responses, particularly when editing multiple loci at once. And once a nucleotide letter is deleted from a chromosome, there is no undo button.
That limitation prompted a different engineering question: since human cells already silence thousands of genes in every tissue using DNA methylation, can we send a temporary messenger to install that exact chemical padlock on a single disease-driving gene and leave the underlying letters untouched?
An editor that lasts days, a silence that lasts a year
For a decade, the bottleneck was permanence: if a transcriptional repressor is delivered safely as messenger RNA, the protein disappears in a few days. Early designs lost their effect as soon as the editor cleared.
In 2021, Jonathan Weissman, Luke Gilbert, and colleagues showed in Cell with CRISPRoff that fusing three epigenetic effector domains (DNMT3A, DNMT3L, and KRAB) to a catalytically dead Cas9 (dCas9) wrote a methylation state that persisted across more than 450 cell divisions in vitro, even as induced pluripotent stem cells differentiated into neurons.
The decisive in vivo test arrived in February 2024 in Nature. Angelo Lombardo, Martino Alfredo Cappelluti, and colleagues at the San Raffaele Telethon Institute for Gene Therapy engineered evolved transcriptional repressors built on zinc-finger DNA-binding proteins (EvoETR) targeting hepatic Pcsk9 in mice, the master regulator of LDL cholesterol removal.
A single intravenous infusion of lipid nanoparticles carrying the editors' mRNA cut circulating PCSK9 levels by roughly half for nearly one year (330 days), while the optimized all-in-one EvoETR architecture maintained up to a 75% reduction with high genome-wide specificity. To prove that the silent state was genuinely heritable across cell division in a living organ, the team performed partial hepatectomy to force liver regeneration: newly divided hepatocytes copied the CpG methylation marks and kept Pcsk9 switched off.
Cut, swap a base, rewrite, or silence
Four programmable genetic tools now sit on the bench. They solve different problems: if a patient carries a broken letter that makes a toxic protein, base or prime editing can fix the spelling; if the clinical goal is turning down a normal gene whose output is too high (such as PCSK9 in hypercholesterolemia or the MYC oncogene in liver tumors), epigenetic silencing avoids nicking or cleaving the DNA backbone altogether.
| Modality | Cuts DNA strands? | Alters A-T-G-C sequence? | Reversible? | Regulatory status (Oct 2026) |
|---|---|---|---|---|
| CRISPR-Cas9 nuclease | Yes, double-strand break (DSB) | Yes (indels during repair) | No (permanent sequence change) | 1 approved ex vivo (Casgevy, 2023); in vivo BLA accepted (lonvo-z, PDUFA Mar 2027) |
| Base editing | Single-strand nick (nickase) | Yes (converts 1 base: C to T or A to G) | Only via re-editing | Phase 1/2 clinical trials (e.g. VERVE-101/102 on PCSK9); 0 approved |
| Prime editing | Single-strand nick (nickase) | Yes (writes short stretches via reverse transcriptase) | Not in clinical practice | Early Phase 1 trials; 0 approved |
| Epigenetic silencing (EvoETR / CRISPRoff) | No (0 cuts) | No (0 bases changed) | Yes (targeted demethylation via TET1) | Phase 1/2 in humans (OTX-2002 on MYC, NCT05497453); PCSK9 in preclinical |
A 2026 review in Clinical Epigenetics benchmarks three milestones that moved epigenetic editing toward the clinic between 2024 and 2026: mechanistic durability (silencing maintained across at least 450 cell divisions in vitro and 12 months in vivo without continued editor expression), tissue-selective delivery above 50% efficiency in liver, muscle, and brain via lipid nanoparticles and AAV vectors, and first-in-human trials led by OTX-2002 (the MYCHELANGELO Phase 1/2 study targeting MYC in hepatocellular carcinoma, NCT05497453).