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.

Autoimmune disease · A research question · November 2026

The Great Reset

Can one infusion, in early and mostly uncontrolled reports, leave people with lupus, systemic sclerosis or inflammatory muscle disease in remission after they stop their drugs? That is the question. It is not a finding that a cure exists.

Most approved treatment for autoimmune disease has the same shape: a drug that suppresses the immune system, taken continuously, with infections and other harms as the price of control. A separate body of human reports describes a different shape. Immune cells are depleted deeply, the compartment grows back, and in the strongest case series the disease has not returned with it. Immunologists call that an immune reset. The name is a hypothesis. The evidence, graded below, is real, early, and almost entirely without a control group.

Grades. A, controlled human evidence. B, early human evidence. C, preclinical only. D, a press release or no relevant human data. Negative, controlled evidence against. Where a programme has a peer-reviewed core and a sponsor-reported layer on top, both are named.

The mechanism, as the papers describe it

Antibody-driven autoimmune diseases run in large part on B cells, the cells that become antibody factories. Drugs such as rituximab deplete B cells that carry CD20. They miss earlier precursors and many tissue-resident cells, and autoreactive memory can persist in lymph nodes, marrow and inflamed tissue. When the drug wears off, that population can expand again.

CD19 CAR-T, in the studies below, is aimed deeper. A patient’s own T cells are collected, engineered with a receptor for CD19, and returned after a short course of lymphodepleting chemotherapy. In the reported cases the engineered cells drive B cells down, then contract. Months later the compartment regrows, and the papers describe it as dominated by naive cells, with autoantibodies gone or sharply reduced. The claim is not ongoing suppression. It is that the repertoire which returns is not the one that caused the disease. That claim is what the uncontrolled series can illustrate and what they cannot prove.

Erlangen

Grade B. Early human evidence, peer-reviewed, uncontrolled, consistent.

In 2022 Mackensen, Müller, Schett and colleagues reported in Nature Medicine that five patients with severe, treatment-refractory systemic lupus erythematosus received a single infusion of anti-CD19 CAR-T cells. All five reached DORIS remission within three months. Anti-double-stranded DNA antibodies became undetectable. B cells returned after roughly three to four months, and the remission held after they returned, without ongoing immunosuppression. doi:10.1038/s41591-022-02017-5.

In 2024 the same group extended the series in the New England Journal of Medicine: fifteen patients, eight with lupus, three with idiopathic inflammatory myositis and four with systemic sclerosis. Every lupus patient reached DORIS remission. Every myositis patient achieved a major clinical response. Every systemic sclerosis patient improved on the EUSTAR activity index. Immunosuppressive therapy was stopped in all fifteen. Median follow-up was 15 months. Cytokine release syndrome was grade 1 where it occurred. Mean B-cell absence was about 112 days.

Fifteen patients, no control group, selected for severity and for fitness to undergo conditioning chemotherapy. The design can show that deep B-cell depletion coincided with drug-free remission in diseases where spontaneous remission of this depth is rare. It cannot show how much of the effect was the CAR-T, how much was the chemotherapy, or how long remission lasts across years.

CASTLE

Grade B. Prospective Phase 1/2 basket trial, peer-reviewed, still uncontrolled.

CASTLE, published in Nature Medicine in 2026 (volume 32, pages 1142 to 1151; NCT06347718; doi:10.1038/s41591-025-04185-6), tested zorpocabtagene autoleucel (zorpo-cel, MB-CART19.1), an autologous CD19 CAR-T from Miltenyi Biomedicine, in treatment-resistant lupus, systemic sclerosis and inflammatory myopathies.

Twenty-four patients were enrolled: ten with lupus, nine with systemic sclerosis and five with myositis. At 24 weeks, 22 of 24 met their predefined disease-specific endpoints. Nine of ten lupus patients reached DORIS remission. Nine of nine with systemic sclerosis showed no progression of interstitial lung disease. Four of five with myositis reached a major or moderate response. Every patient remained off glucocorticoids and other immunosuppression during observation. No cytokine release syndrome above grade 2, and no ICANS, was reported. B-cell aplasia lasted a median of 83 days, and the returning cells were mostly naive.

Two products, two teams, the same pattern. That is replication at this stage of a field: not proof, and not a randomised result. Cabaletta Bio has reported drug-free responses with rese-cel across several diseases in its RESET programme, and Kyverna’s KYV-101 is in Phase 2 trials in myasthenia gravis (NCT06193889) and progressive multiple sclerosis (NCT06384976). Those layers are sponsor-reported where a paper has not yet carried them. Registry numbers for some of Kyverna’s other studies were ambiguous in the materials reviewed and are not cited here.

Three caveats that decide the question

Nobody in the CAR-T series has been randomised. The comparisons are to each patient’s own past. Autoimmune disease fluctuates. Trial patients receive intense attention. The conditioning drugs, fludarabine and cyclophosphamide, are themselves potent immunosuppressants. Until a CAR-T arm beats a real comparator, the size of any reset effect is unknown.

The conditioning is the dangerous part, so far. In the autoimmune CAR-T reports, the cells themselves have mostly caused low-grade cytokine release. The chemotherapy has not been gentle. The clearest warning came from a neighbouring technology: in October 2026 Artiva discontinued AlloNK after a death from leukoencephalopathy assessed as likely caused by fludarabine. That death was not in a CAR-T trial. The drug is the one most CAR-T protocols use. Cartesian Therapeutics has reported a placebo-controlled trial in myasthenia gravis with an mRNA CAR-T given without lymphodepletion (below). Cabaletta has reported dosing some lupus patients with rese-cel without preconditioning. Those are attempts to remove the chemotherapy, not proof that it can be removed.

Time is the endpoint that is still short. A remission of eighteen months is not a cure. The earliest Erlangen patients are years out without relapse in follow-up reports, which is encouraging, and the world experience is still counted in dozens of patients. A published myositis case in which disease relapsed, and a second CAR-T against a different target was used, is a reminder that the returning immune system can relearn the disease.

The same question, disease by disease

The pages above are the founding CAR-T series. The rest of this survey is what happens when the same ambition is graded in other diseases. “Cure-path” here means the stated design goal is durable drug-free remission, immune tolerance, or replacement of destroyed tissue, rather than indefinite suppression. Programmes aimed only at better management are marked as context.

Lupus

Lupus is the proof-of-concept disease for the CAR-T reports: Erlangen and CASTLE, both grade B, both uncontrolled. Nkarta’s allogeneic CAR-NK, NKX019 (Ntrust-1, NCT06557265), had no efficacy data at the time of writing. See the off-the-shelf reset. Fate’s FT819 and FT839 are a different manufacturing bet; FT839 itself is grade D because no person has received it. See FT839.

Type 1 diabetes

Two different problems: stop the autoimmune attack, or replace the beta cells. Zimislecel (VX-880), Vertex, Reichman et al., New England Journal of Medicine 2025: of 12 full-dose patients with at least a year of follow-up, 10 were insulin independent at day 365, all free of severe hypoglycaemia with HbA1c under 7 percent. The product requires ongoing immunosuppression. Two deaths were reported (cryptococcal meningitis; progression of pre-existing cognitive impairment). Grade B for replacement, not for a drug-free reset.

UP421, Sana, one patient, New England Journal of Medicine 2025: hypoimmune donor islets, transplanted without immunosuppression, survived and secreted insulin. The dose was under a tenth of a therapeutic dose by design. The patient still used insulin. A follow-up letter reported function past 14 months. Grade B, n=1. Vertex’s encapsulated product VX-264 was discontinued in March 2025 for lack of benefit. Grade: negative.

Teplizumab is approved to delay stage 3 type 1 diabetes in at-risk relatives, on the TN-10 trial, and later for recent-onset disease in a defined age group. That is delay, not cure. Grade A for delay. Diamyd’s antigen-specific Phase 3 was stopped for futility in 2026. Grade A, negative, for that recent-onset use.

Multiple sclerosis

The strongest randomised reset result in this survey is not CAR-T. It is autologous haematopoietic stem-cell transplant. Burt et al., JAMA 2019, the MIST trial, 110 people with relapsing-remitting MS: progression in 3 patients after transplant against 34 on continued drug therapy, with no deaths in that trial. A historical cohort (Muraro et al., JAMA Neurology 2017) includes treatment-related deaths, so the benefit and the risk travel together. Grade A, for relapsing disease, with that mortality context. CAR-T in progressive MS (Kyverna KYSA-7, NCT06384976) does not yet have a cohort efficacy result. Grade B at most.

The non-inflammatory mRNA vaccine in mice (Krienke et al., Science 2021) is grade C. No human trial of it has reported. Viral claims of an mRNA cure for MS trace back to that paper and are not supported by it. The longer account is on The Dog, the Chatbot, and the Vaccine That Points Both Ways.

Crohn’s disease and ulcerative colitis

This is the honesty check. Hawkey et al., JAMA 2015, the ASTIC transplant trial in refractory Crohn’s: the strict primary endpoint was not met (2 of 23 against 0 of 22). Secondary analyses found more clinical remission, at high toxicity, including one death. Grade A, negative on the primary endpoint. Rosnilimab, a PD-1 antibody, missed a randomised ulcerative colitis trial in November 2025 and the programme was discontinued, even though the drug depleted the cells it was designed to deplete. Grade A, negative. Faecal microbiota transplant has several positive randomised trials in ulcerative colitis, with remission in roughly one in four to one in three patients in the positive studies, and durability that usually needs repeat dosing. That is a modest effect, not a cure. Grade A for that limited claim.

Myasthenia gravis

Descartes-08, Cartesian Therapeutics, Vu et al., Nature Medicine 2026: a randomised, placebo-controlled trial, 26 patients. The primary endpoint was met, a 5-point MG Composite improvement at month 3 in 66.7 percent against 27.3 percent. Six once-weekly outpatient infusions, no lymphodepletion, no cytokine release syndrome or ICANS reported in the programme. Grade A, and the sample is small. A Phase 3 was sponsor-stated, and this page does not treat the timing of its readout as independently verified. It is the only placebo-controlled cell-therapy trial in autoimmunity found in this survey.

Systemic sclerosis

Transplant already has randomised trials. van Laar et al., JAMA 2014 (ASTIS), and Sullivan et al., New England Journal of Medicine 2018 (SCOT), found better event-free survival than cyclophosphamide, with treatment-related mortality that was about 10 percent in the earlier European trial and about 3 percent with later selection. Grade A, for selected severe early diffuse disease, with that risk stated. CASTLE’s 9 of 9 without interstitial-lung-disease progression at 24 weeks is grade B, uncontrolled, and is the opening comparison, not the verdict.

ANCA vasculitis, pemphigus, coeliac, biliary cholangitis, thyroid

ANCA vasculitis has peer-reviewed case reports of CD19 CAR-T, including drug-free remission at 7 months in one granulomatosis case (Minopoulou et al., Annals of the Rheumatic Diseases 2025). Grade B, case level. Pemphigus is the test of deleting only the autoreactive cells: DSG3-CAART was safe in Phase 1 and has not shown a clear antibody drop. Grade B for safety, and not a positive efficacy result. Rese-cel without lymphodepletion in four pemphigus patients is sponsor-reported.

Coeliac disease has a known external antigen, which makes it the tolerance classroom. TAK-101 met a biomarker endpoint under gluten challenge (Kelly et al., Gastroenterology 2021) and has not produced a pivotal clinical-efficacy trial. Nexvax2 was discontinued. Grade: negative for that peptide vaccine. ZED1227 attenuated gluten-induced mucosal injury in a controlled trial (New England Journal of Medicine 2021) by blocking antigen modification, which is disease modification, not restored tolerance.

CNP-104 in primary biliary cholangitis is a small randomised trial with a placebo-separated biochemical and stiffness signal at one year after two doses, reported by the sponsor in March 2026 and in meeting abstracts. The full paper was not out. Grade A-minus, with the small sample and the unpublished paper as the limit. If it holds in print, it is the first controlled clinical signal for antigen-specific tolerance in this survey. Diamyd’s 2026 futility result is the large controlled failure beside it.

Thyroid autoimmunity is the gap. No clinical-stage cure-path programme for Hashimoto’s thyroiditis was found. ATX-GD-59 in Graves’ disease has a Phase 1 safety report and no controlled efficacy data; the primary paper was not opened for this survey, so the grade stays provisional. Standard care for Hashimoto’s remains hormone replacement, which treats the deficit and not the autoimmunity.

What the survey can support

  1. CD19 CAR-T has produced drug-free remission in published uncontrolled series across several diseases. It has not been randomised. The first randomised readouts will decide the next reading of those series.
  2. Haematopoietic stem-cell transplant already beat drugs in relapsing MS and beat cyclophosphamide in systemic sclerosis, with a mortality that has fallen as selection improved and has not disappeared. CAR-T’s comparator is that literature, not a blank page.
  3. Conditioning chemotherapy is the tax on most reset protocols. The AlloNK death is the sharpest recent statement of that tax. Descartes-08 is the controlled attempt to stop paying it.
  4. Antigen-specific tolerance has one small controlled clinical signal (CNP-104) and one large controlled failure (Diamyd, 2026). Safety has been the consistent finding. Pivotal efficacy has not.
  5. Delay can be an approved endpoint, as teplizumab shows. Cure is not a label claim any programme in this survey has earned.
  6. In type 1 diabetes, replacing cells under immunosuppression and cloaking cells from the immune system are both early human results. Neither has stopped the original autoimmune process.

What was not verified

Some registry numbers and conference claims were left out on purpose. Kyverna study numbers other than the two cited above were ambiguous in the decks reviewed. Registry identifiers for MIST, BEAT-MS and STAR-MS were not pulled. The Descartes-08 Phase 3 timing, Cabaletta’s MuSK-CAART registry record, TRX103’s status in Crohn’s disease, and a BMS basket programme reported via a secondary source were not graded. ATX-GD-59 is described only at review level. Absence from this page is not a claim that those programmes do not exist.

Related pages: Is there a cure for rheumatoid arthritis?, CAR-NK, FT839, in vivo prime editing.

Sources

  1. Mackensen et al., Nature Medicine 2022;28:2124-2132. doi:10.1038/s41591-022-02017-5.
  2. Müller et al., New England Journal of Medicine 2024;390:687-700.
  3. CASTLE, Nature Medicine 2026;32:1142-1151. doi:10.1038/s41591-025-04185-6. NCT06347718.
  4. Reichman et al., New England Journal of Medicine 2025, on zimislecel. Sana UP421, New England Journal of Medicine 2025, doi:10.1056/NEJMoa2503822, and the July 2026 follow-up letter.
  5. Burt et al., JAMA 2019 (MIST). Hawkey et al., JAMA 2015 (ASTIC). van Laar et al., JAMA 2014 (ASTIS). Sullivan et al., New England Journal of Medicine 2018 (SCOT).
  6. Vu et al., Nature Medicine 2026, on Descartes-08. Kelly et al., Gastroenterology 2021, on TAK-101. Krienke et al., Science 2021.
  7. Artiva Biotherapeutics, October 2026 disclosures on the AlloNK discontinuation. AnaptysBio, 10 November 2025, on the rosnilimab ulcerative colitis miss. Diamyd Medical, 2026, on DIAGNODE-3 futility. Cour Pharmaceuticals and partner disclosures on CNP-104, March 2026, meeting-abstract grade.

Research reporting only. Nothing on this page recommends a therapy, a trial, or a course of action for any person.