Not medical advice. The Journal of Biohacking reports published research and teaching models. Nothing on this page is medical advice, a diagnosis, or a recommendation to use any compound, product, or therapy. Where a dose appears, it is a figure reported in a cited study or by its sponsor, not an instruction. Editorial position on dosing.
Editing the body from inside
In October 2026 a patient with Wilson's disease received an infusion designed to rewrite one wrong letter of DNA inside their own liver. It was the first time prime editing, the most precise gene editor ever built, has been used inside a human body. This dossier maps the science, the company, the rivals, the delivery problem, and the harm ledger, with every claim graded and checked against primary sources.
The short version
Prime editing can rewrite DNA without cutting both strands, and it has already worked in human stem cells outside the body (two CGD patients, NEJM 2025). PM577a is the first attempt to do it inside a patient, by LNP infusion to the liver. There are no human results yet, and the honest grade for the therapy is early signal. The wider in vivo editing field it joins is further ahead and carries the field's first death on its record. Both facts belong in the same breath.
Three editors, three philosophies
Pick an editor and watch what it does to the same stretch of DNA. The highlighted base is the one that needs to change. Nuclease editing cuts. Base editing swaps one letter class and can nick the neighbors. Prime editing rewrites the letter from a template the guide itself carries.
The trial landscape
Every major program editing DNA inside the body, graded. Filter by editor type or search. Harm flags print next to the programs that carry them.
The harm ledger
Deaths and defining safety events in gene therapy and editing, attributed to the modality that actually caused them. The pattern matters: the deaths cluster in high dose viral vectors and one LNP nuclease trial. None involves prime editing, which has dosed one in vivo patient. That is an absence of data, not a safety record.
Adenoviral gene addition, University of Pennsylvania. Jesse Gelsinger, 18, died four days after infusion from a severe immune reaction to the vector and multiorgan failure. Not an editing trial. The field's founding catastrophe.
Custom CRISPR activation on rAAV9, Cure Rare Disease. Terry Horgan died after treatment for Duchenne muscular dystrophy. The NEJM analysis attributed death to an innate immune reaction to the high dose vector causing ARDS, not to CRISPR machinery.
AAV gene addition, Pfizer. A boy in the Phase 2 DAYLIGHT trial died suddenly on May 3, 2024; Pfizer stated the cause was not yet known. The program was later discontinued after the Phase 3 missed its endpoints.
AAV gene addition, Sarepta. Two non ambulatory Duchenne patients on commercial Elevidys died of acute liver failure. FDA added a boxed warning on November 14, 2025 and limited the label to ambulatory patients aged 4 and older. A third death, on the related investigational product SRP-9004, led FDA to request a distribution suspension and revoke Sarepta's platform designation in July 2025.
LNP nuclease editing, Intellia. A MAGNITUDE patient developed Grade 4 liver injury, was hospitalized, and died on November 5, 2025. The investigator attributed death to septic shock from a perforated duodenal ulcer in a course including liver injury and steroid treatment (company 10-K). FDA held both Phase 3 trials, then lifted the holds in January and March 2026 with enhanced liver monitoring.
AAV CRISPR editing, Duchenne program HG302. A fatal serious adverse event at high dose in August 2025, reported as complement and cytokine mediated. Vector dose is the recurring villain of this ledger.
Watchlist: what happens next
Named milestones from November 2026 to late 2028. Each one moves a grade in this dossier when it lands.
CTX310 Phase 1b data scheduled around the American Heart Association meeting, November 6 to 9. CTX320's peer reviewed data remains overdue.
FDA decision date for lonvo-z (hereditary angioedema). Approval makes it the first in vivo CRISPR medicine and sets the template PM577a follows.
First PM577a data, company guided. The numbers that matter: biopsy measured editing percentage and the liver safety profile.
MAGNITUDE readouts under new monitoring: does the Grade 4 liver signal recur, and at what rate.
BEAM-302 pivotal data and a possible accelerated approval filing; VERVE-102 under Lilly moving toward Phase 3; Prime's PM647 (alpha-1 antitrypsin) regulatory filings, unverifiable at this writing.
FDA finalization of the April 2026 draft guidance on sequencing based safety assessment. First human data for targeted delivery beyond the liver would rewrite section grades across this dossier.
Addendum, 8 November 2026: drug-controllable T cells, in the laboratory
This note reports a peer-reviewed paper that was not in the original dossier. It is ex vivo work. No patient has received these cells, and it does not change the grade for in vivo prime editing in humans, which still rests on the separate PM577a trial.
The paper, and the grade
Early signal. Horton, Lew and colleagues at the Walter and Eliza Hall Institute, Melbourne, “Prime editing enables drug-controllable T-cell therapies with clinical immunosuppression,” Blood 2026;148(19). Blood. Human T cells were edited outside the body. The living-animal work used immunocompromised NSG mice, including humanized models. Nothing in the paper is a human dosing result.
The problem the paper sets is control. Adoptive T-cell therapies are often given to people who also need immunosuppressive drugs, and those drugs suppress the therapeutic cells along with the others. The reported approach uses prime editing twice in the same cells: once to correct a disease mutation, and once to install a second edit that makes only the edited cells resistant to one immunosuppressive drug, by changing the gene that encodes that drug’s target. In the paper’s account, the edited cells stay sensitive to other immunosuppressants, so one drug can favour them and a different drug can still suppress them.
Across disease-associated loci in primary human T cells from healthy donors, including CTLA4, IL2RA, PRF1, STXBP2, UNC13D and HAVCR2, the paper reports intended prime edits at 8.3 percent to 52.1 percent efficiency (mean 22.2 percent), with indels of 0.23 percent to 5.8 percent (mean 1.8 percent). Six of the eight mutations shown are not correctable by base editing, the authors report, because they are transversions or small indels. Those percentages are laboratory editing rates, not clinical doses.
The disease anchor in the paper
Subcutaneous panniculitis-like T-cell lymphoma with haemophagocytic lymphohistiocytosis, associated with germline mutations in HAVCR2, which encodes TIM-3. In T cells from 3 patients, the paper reports HAVCR2 correction at 18.7 percent to 35.8 percent (mean 29.8 percent) and restored TIM-3 on the cell surface in 19.9 percent to 44.4 percent of cells (mean 34.2 percent). Two further examples are reported: CD40LG correction at 2.5 percent in T cells from a patient with X-linked hyper IgM syndrome, and IL2RG correction at 51.3 percent in CD34-positive cells from a patient with X-linked SCID. The low CD40LG figure is the paper’s own caution that a new locus can start inefficient.
Off-target checks in the paper, including deep sequencing of candidate sites and molecular karyotyping, did not find reproducible prime-editing mutations in a protein-coding exon or a known oncogene across the donors tested. The authors also report preserved T-cell subsets and repertoire diversity in the assays they ran. That package is a preclinical signal, not human safety evidence. TIM-3 restoration was measured as surface expression, not by a functional rescue assay.
Drug-resistance edits, as the paper reports them
Three drug-and-target pairs are described. Knocking the target out reduced fitness and did not confer resistance, so the edits are function-preserving substitutions. FKBP1A, for tacrolimus resistance, edited at 21.1 percent to 63.9 percent (mean 42.2 percent). DHFR, for methotrexate resistance, edited at 26.2 percent to 65.6 percent (mean 41.8 percent). IMPDH2, for mycophenolate resistance, edited at 12.5 percent to 60.5 percent (mean 26.9 percent), with resistance showing in cell division rather than viability. Phenotypic resistance varied by variant. These are experimental results in cells, not regimens for any person.
In one mouse experiment, patient T cells edited at both HAVCR2 and FKBP1A were transferred into NSG mice and treated with tacrolimus or vehicle for 8 weeks. The paper reports mean corrected HAVCR2 reads in the spleen of 76.8 percent under tacrolimus, against 9.3 percent without selection. A separate experiment reported that the edited cells remained suppressible by cyclosporin. Group sizes were 4 to 7 mice. Tacrolimus exposure was inferred from the dosing schedule, not measured drug levels.
What this does not show
The paper also reports, in vitro and in mice, that virus-specific T cells and CD19 CAR-T cells edited at FKBP1A kept function under tacrolimus where unedited cells did not. The use cases named by the authors are transplant-related infection and post-transplant lymphoproliferative disorder. No person has received the edited cells, and the authors state that clinical-grade manufacturing metrics are not established.
This work does not make rheumatoid arthritis, or any polygenic autoimmune disease, a near-term prime-editing target. The disease in the paper is monogenic and T-cell restricted. Rheumatoid arthritis has no single letter to correct. Citing the paper as evidence of a prime-editing treatment for rheumatoid arthritis overstates it. The grade for that claim, on this page, stays vapor.
Scope note
What this technology is not
Polygenic autoimmune disease, including rheumatoid arthritis, is not a near term prime editing target. RA comes from hundreds of common variants interacting with environment and immune history. There is no single letter to correct, no liver gene whose repair resets the immune system, and no credible program claims otherwise in the sources reviewed here. Prime editing's autoimmune relevance is indirect and long dated: better tools for engineering immune cells outside the body, and maturing delivery. Anyone selling prime editing as an autoimmune cure on a plannable timeframe is selling vapor, and this journal will grade it that way.