Mixed haematopoietic chimerism (immune tolerance induction)

Almost every cure approach on this site has the same unsolved problem underneath it: put insulin-producing cells into someone with type 1 diabetes and two different immune reactions come for them. Rejection, because the cells are foreign. And autoimmunity, the original attack that caused the diabetes, which does not care whose cells they are. Today's answer to both is the same blunt one — chronic systemic immunosuppression, taken for as long as the graft is meant to last.

Mixed haematopoietic chimerism is the oldest and most ambitious attempt to do something else entirely. The idea is to transplant the donor's blood-forming (haematopoietic) stem cells into the recipient first, so that the recipient's bone marrow ends up a stable mixture — part self, part donor. A chimera. Once that mixture is established, the immune system's own education machinery does the work: developing T cells that would attack donor tissue are deleted in the thymus, and peripheral tolerance mechanisms mop up the rest. The recipient's immune system stops seeing the donor as foreign — not because it has been suppressed, but because it has been re-taught. Islets from that same donor then engraft as if they belonged. And because chimerism also resets the recipient's education against their own beta cells, it is one of the very few strategies that plausibly addresses rejection and autoimmunity with a single intervention.

If it worked in people, the anti-rejection drugs could be withdrawn entirely. Not calcineurin-inhibitor-free. Not "reduced". None.

What has actually been shown

In mice, the results are close to complete. A Stanford group (Bhagchandani, Kim, Shizuru, Meyer and colleagues) reported in the January 2026 Journal of Clinical Investigation that a chemotherapy-free conditioning regimen produced durable mixed chimerism across a full MHC mismatch in NOD mice — the standard autoimmune-diabetes model. In prediabetic mice, chimerism prevented diabetes in 100% of animals. In mice with established diabetes, conditioning followed by combined haematopoietic-cell and islet transplantation durably corrected the diabetes in 100% of chimeric mice, without chronic immunosuppression and without graft-versus-host disease. The mice stayed immunocompetent — they still rejected third-party islets, which is the control that shows this is targeted tolerance and not general immune damage. Adoptive-transfer experiments confirmed the underlying autoimmunity had been corrected.¹ A follow-up in JCI Insight in 2026 pushed the conditioning gentler still, adding baricitinib, venetoclax and a CD47 antibody so that the radiation dose fell to 10 centigray.²

In humans with T1D, nothing has been shown, because nothing has been tried. No trial of chimerism-induced islet tolerance in type 1 diabetes is registered on ClinicalTrials.gov. The closest human step is a Stanford phase 1 (NCT05973734) in which one arm infuses deceased-donor vertebral bone marrow alongside an islet transplant, hoping to reduce how much immunosuppression is needed. It lists no chimerism endpoint and it does not attempt to withdraw immunosuppression.⁵

Why it has never reached a person with T1D

The blocker has always been the conditioning, and it is worth being precise about why, because it is a genuinely hard ethical constraint rather than a technical footnote.

To get donor stem cells to engraft, you have to make room for them in the recipient's bone marrow. Historically that meant myeloablation — chemotherapy and irradiation heavy enough to destroy the existing marrow. That is a treatment with real mortality. It is justifiable against leukaemia. It is not justifiable against a disease that people live with for decades using insulin, however hard that living is. The mismatch between the toxicity of the cure and the survivability of the disease is exactly why an idea this powerful has struggled to reach people with T1D. The invited JCI commentary on the Stanford work takes up the same theme — safer conditioning enabling murine islet allograft tolerance.³

That is why the antibody-based conditioning matters so much. An anti-CD117 antibody clears the stem-cell niche by targeting the receptor on the recipient's own haematopoietic stem cells, rather than by poisoning everything that divides. Swap chemotherapy for antibodies, add a JAK inhibitor, and the irradiation drops to a token dose. If that regimen turns out to be tolerable in humans, the long-standing objection loosens. If — nobody has run that trial.

The human evidence that does exist, and what it is not

The mechanism is not pure speculation in humans. In a phase 3 randomised trial published in 2025, kidney-transplant recipients given a donor cell product (MDR-101) under non-myeloablative conditioning achieved donor chimerism, and 19 of 20 came off all immunosuppression roughly a year after transplant. Fifteen — 75% — were still completely immunosuppression-free more than two years later, with no deaths, no graft losses, and no graft-versus-host disease.⁴

That is a striking result and it is genuinely relevant. It is also not a result about type 1 diabetes, and the differences are not cosmetic:

• It was a kidney, not islets.
• The donors were two-haplotype HLA-matched living siblings. Most people do not have one.
• It solved rejection. It says nothing about whether the same chimerism would durably switch off a pre-existing autoimmune attack on beta cells — the thing that makes T1D T1D. Only the mouse work speaks to that, and mice are not people.

This distinction is the whole ballgame. A therapy that stops rejection is not a therapy that stops autoimmunity. Mixed chimerism claims both — but it has shown both only in a mouse.

How to read our scores

We hold one bar for this category and we hold it here too: a mouse result is not a human result, and an intention is not an achievement. Everything in the mouse data points the right way, and the tolerance logic is more complete than almost anything else on this site — it is the only approach that would let a transplant recipient stop the drugs altogether rather than merely swap them. That is why the durability rationale gets credit for the human kidney data, and why nothing else does.

But maturity is scored at 8 and beta-cell preservation at 12, because the honest human numbers for this approach are zero patients, zero C-peptide, zero years of follow-up. It sits below every clinical-stage immunotherapy on this page — behind teplizumab, behind the Treg programmes, behind anything with a person in it. The idea is not new; the community around it runs back through transplant-tolerance work at Columbia (David Sachs, Megan Sykes) and Stanford (Sam Strober).⁶ What is new in 2026 is a conditioning regimen gentle enough that the old ethical objection might finally be answerable.

Watch for one thing and one thing only: a first-in-human conditioning-safety trial in type 1 diabetes. No such trial was found among the registry records searched; until one exists and reports, this is a beautiful mechanism with no patients.