Inside the Nature Paper: Mice with Half-Human Brains

The paper behind the headlines appeared in Nature on 16 September 2026, from the laboratory of Sergiu Pașca, professor of psychiatry at Stanford University, and was reported the same day by Ian Sample in The Guardian. Its central claim is simple to state and startling to picture: mice engineered to have almost no cortex and hippocampus, given transplants of lab-grown human brain cells, end up with brains that are roughly half human tissue by volume.

The numbers in the paper

• ~100,000 human cells per injection. Each mouse received brain organoids grown from reprogrammed skin cells — small balls of human neural tissue, injected into the cavity left by the animal's underdeveloped cortex.
• ~14 million mouse cells absent. The genetic engineering left the cortex and hippocampus stunted, removing those structures' contribution to the brain.
• ~4 million human cells after three months. The graft expanded to fill the cavity and hooked itself into the mouse's blood supply — the make-or-break step for any transplant.
• ~50% of brain volume. By three months after surgery, the human graft accounted for about half the brain.

What the graft did inside the animal

The human neurons did not just survive; they integrated. Some connected to the mouse's own cells, and some sent wiring toward the spinal cord. The tissue remained immature — Pașca describes it as roughly halfway through pregnancy-equivalent development — and not structured or wired like a human brain. Gait and cognition in the mice improved slightly, but the researchers are explicit that the animals are not enhanced.

The two experiments that explain why this matters

Low oxygen. Exposing the mice to five hours of low oxygen selectively harmed the human nerve cells far more than the mouse ones. That vulnerability mirrors the cellular damage seen in human cerebral palsy — a disease that is nearly impossible to study directly in human brain tissue. Suddenly there is an animal model where the vulnerable cells are human.

Von Economo neurons. The grafts contained von Economo neurons, large spindle cells found in a few human brain regions and among the first to die in frontotemporal dementia. Pașca hopes the model will let researchers watch that cell type develop and die — and test how to protect it.

"We've been trying really hard as a community... fewer therapeutics than every single branch of medicine." — Sergiu Pașca, explaining why neuropsychiatry needs new models.

Why not just grow organoids in a dish?

That is the field's own preferred endpoint. Professor Madeline Lancaster of the MRC Laboratory of Molecular Biology in Cambridge, quoted in The Guardian's report, noted that many researchers keep organoids entirely in vitro, which reduces animal use, and said the work should only go ahead for a "very good reason," with the field still aiming for fully in vitro solutions. Her caveat: the in-vivo model is best for whole-animal questions, and how well it will inform human understanding of a "rather artificial" setup is less clear. The Pașca paper's answer is that some questions — oxygen injury, circuit integration, vascularization — need a living body, and this is the most controlled living body yet.

Where this fits in the field

This is the third major step in the arc we trace in A Brief History of Humanized Brains: human neural progenitors in mouse cortex (2017), vascularized human organoids in mouse brains (2018), human cortical organoids in rats with full sensory wiring (2022), and now xenocortical mice with half-human brains by volume (2026). Each step traded a little more biological control for a lot more biological reality.