They Cleared Most of a Mouse Cortex — Human Organoids Filled 91.9% of It and Reached the Spinal Cord

Subtitle: A Stanford Nature report introduces xenocortication: genetically apallial mice whose neocortex and hippocampus are depleted early, then filled neonatally with human cortical organoids that grow to ~92% of cortical volume, form rare cell types, send axons to the cervical spinal cord, and show a human-selective hypoxia injury readout.

The hardest problem in human neuroscience is still access. You can sequence genomes. You can culture organoids. You cannot ethically watch a living human cortex wire itself, fail under oxygen stress, or host the rare cell classes that vanish in frontotemporal dementia — not at the resolution a circuit paper demands.

On September 16, 2026, Nature published a platform that tries to shrink that gap without pretending a dish is a brain. Konstantin Kaganovsky, Kevin W. Kelley, Tilo Gschwind, Paul M. Harary, and colleagues — senior author Sergiu P. Pașca at Stanford Medicine — report developmental xenocortication: a genetic strategy that depletes glutamatergic neurons from the mouse dorsal and medial pallium (apallial mice), then neonatally engrafts the emptied cortical cavity with human stem-cell-derived cortical organoids (hCOs) to make xenocortical (XCX) mice (DOI 10.1038/s41586-026-11032-2). The shareable claim is blunt: human tissue occupying most of the cortical volume, integrating with the host nervous system, and supporting circuit- and behavior-level readouts — including a perinatal-hypoxia injury model.

This is not a brain-computer interface story. It is not a Parkinson’s cell-therapy trial. It is a developmental transplantation platform — and that is why it belongs in a Sunday VOLUME after gene-editing methods and fertilizer photochemistry. The bottleneck it attacks is spatial competition: prior organoid grafts had to fight an intact rodent cortex for territory.

What “apallial” actually means

The team used Emx1-Cre to delete the cohesion factor Esco2 in dorsal/medial pallium progenitors on a SCID (immunodeficient) background. MRI and whole-brain single-nucleus RNA-seq confirmed the intended loss: adult apallial brains held roughly 2% of a normal mouse’s cortical content and about half the overall brain tissue volume; snRNA-seq (~880,000 nuclei) showed ~sevenfold depletion of dorsal-pallial glutamatergic classes while ventral/lateral populations were comparatively spared.

Those animals still walked. Locomotion was broadly preserved; fine paw coordination and some memory-linked tasks were selectively altered; behavioral variability rose. Pașca’s Stanford Medicine framing is careful here: when cortex is lost early, other regions may compensate for functions textbooks casually pin on the cortical sheet. The apallial mouse is not a blank slate. It is an emptied cavity with a surviving subcortical chassis.

Into that cavity, shortly after birth, went human cortical organoids — typically four per pup, aged ~30–60 days, each on the order of ~10⁵ cells, derived from consented hiPS donors across three lines. Among 29 recipients, graft survival was 86.2%. Between two and three months post-transplant, MRI measured a 4.7-fold graft-volume increase. At three months, human-derived tissue accounted for 91.9% of the combined cortical tissue volume — density on the order of ~32,000 neurons per cubic millimeter.

That is the headline number. Prior Pașca-lab rat transplants (2022) could claim roughly a third of one hemisphere. Xenocortication claims most of the cortical volume.

What the human tissue became

Volume alone would be a curiosity. The paper’s scientific weight is cell-type and connectivity.

Grafts generated a diversity of human cortical identities, including layer 5 extratelencephalic (L5-ET) projection neurons — a class rarely matured in dish organoids, and enriched here more than threefold versus earlier crowded transplant setups. Anterograde tracing from pre-labeled organoids found sparse human axons in the host cervical spinal cord in every animal examined (n = 3 mice, two hiPS lines), co-localized with the human-specific cytoplasmic marker STEM121. Retrograde work pulled host inputs from palaeocortex, thalamus, and pallidum. Diffusion imaging favored a dorsoventral microstructural bias consistent with organized tissue, not a scrambled blob.

And then the rare cells. In all three sampled XCX mice, the team identified von Economo neuron (VEN)–like cells — elongated, cigar-shaped morphologies associated in humans with fronto-insular and anterior cingulate social/decision circuits, and known to be vulnerable in some frontotemporal dementias. These cells have been essentially impossible to generate in culture and did not turn up in the earlier intact-cortex rat grafts. The paper is honest: VEN-like by morphology, not a claim of fully mature human VENs with every molecular badge stamped. That honesty is what makes the finding usable rather than hype.

Calcium imaging and local-field recordings showed organized, immature-circuit dynamics: large synchronous bursts lasting tens of seconds, recurring every few minutes, events that start locally and sweep the graft in ~100 ms, sometimes correlating with orofacial movement. That is developmental network behavior — not proof that the mouse “thinks with human cortex.”

A human-selective hypoxia readout

The disease application they actually ran is perinatal oxygen injury — the biology behind much cerebral palsy risk — not a late-onset dementia cure. Mice got a graded descent to 5% O₂ for five hours. HIF1α lit up prominently in the human graft, not appreciably in adjacent mouse palaeocortex or in identically exposed control mice. Susceptibility-weighted MRI later showed vascular changes; microglia and astrocytes reacted inside the graft; CatWalk gait testing captured the largest increase in three- to four-paw support in injured XCX animals. Normal and apallial mice were comparatively resilient under the same protocol.

That species-selective vulnerability is the platform’s clinical argument in miniature: human cortical cells, in a living body, failing in a way mouse tissue next door does not — with a behavioral correlate you can measure two days later.

What this is not

It is not a humanized “thinking mouse,” and the authors do not sell it as one. Anatomical connectivity ≠ pathway-specific functional necessity for any particular behavior. Y-maze performance above chance in XCX (vs apallial below chance) is interesting; it does not prove restored human working memory. Grafts remain developmentally immature: incomplete layering and arealization, limited GABAergic interneuron representation, immature network statistics, and an intrinsic tempo mismatch between slow human neurons and a fast rodent host. Ethics consultation was explicit — bioethicists, an independent committee, years of Asilomar-style debate — because maturity and organization will only increase. Stanford holds related organoid and transplant IP; Pașca is an inventor. Say “platform for neurodevelopment, disease, and therapeutics research,” not “mind transplant.”

Why Sunday readers should care

Schizophrenia, epilepsy, profound autism, cerebral palsy, and frontotemporal dementia share a cruel constraint: the tissue that matters is human, developmental, and usually unreachable. Dish organoids lack blood, immune dialogue, long motor axons, and behavior. Intact-cortex transplants hit a turf war. Xenocortication’s bet is architectural — make space first — and the September data say the bet pays in volume, rare cell classes, spinal projections, and a hypoxia injury phenotype you can score.

The next experiments write themselves: older grafts, better interneuron balance, patient-derived lines for Timothy syndrome / FTD / profound autism cohorts, protective-drug screens under controlled hypoxia, and continued ethical oversight as circuits mature. None of that is promised by one Nature paper. All of it becomes fundable once the field can say, without inventing science fiction: human cortical tissue can occupy most of a living cortical volume, wire past the skull’s old limits, and fail like human tissue when oxygen runs out.

Pașca’s lab spent a decade teaching stem cells to self-organize into cortex-like tissue. Kaganovsky, Kelley, Gschwind, Harary, and colleagues just taught that tissue what empty developmental space feels like — and watched it grow into 91.9% of a mouse’s cortical volume, sprout VEN-like cells no dish had offered, and send human axons into a rodent spinal cord.

Not a BCI. Not a cured patient. A peer-reviewed cavity, filled.

Sources

Kaganovsky, K., Kelley, K. W., Gschwind, T., Harary, P. M., Kochalka, J., White, A. D., Lerma-Usabiaga, G., Chen, X., Revah, O., Gore, F., Aoyama, A., Shadrach, J. L., Yoon, S.-J., Valencia, A., Ogawa, S., Reis, N., Vogel, H., Wandell, B., Kaltschmidt, J. A., Soltesz, I., Deisseroth, K. & Pașca, S. P. (2026). Developmental xenocortication using human-derived organoids in mice. Nature. Published 16 September 2026. https://doi.org/10.1038/s41586-026-11032-2
Goldman, B. (2026, September 16). Stanford Medicine team creates advanced model for studying brain development, disorders. Stanford Medicine News. https://med.stanford.edu/news/all-news/2026/09/model-to-study-brain-development.html
Nature News (2026, September 16). Human brain cells transplanted into mice in ‘most extensive’ integration ever. Nature d41586-026-02912-8. https://www.nature.com/articles/d41586-026-02912-8
Malesu, V. K. (2026, September 18). Transplanted human cortical organoids filled most of a mouse cortex. Then they began to connect. News-Medical. https://www.news-medical.net/news/20260918/Transplanted-human-cortical-organoids-filled-most-of-a-mouse-cortex-Then-they-began-to-connect.aspx
Nature Portfolio press release (2026). Neuroscience: Human brain organoids form neural networks in mice. https://www.natureasia.com/en/info/press-releases/detail/9442