Human Cortical Organoids Occupied Most of a Mouse Cortex

Stanford researchers created a living mouse model in which human stem-cell-derived cortical organoids occupied most of the measured cortical volume. Published on September 16, 2026, the study transplanted four organoids into each immunodeficient newborn mouse after genetically depleting most dorsal and medial pallium-derived tissue, including much of the neocortex and hippocampus.

The result is a substantial human-derived cortical graft, not a complete human brain inside a mouse. The tissue formed several human cortical cell types, connected with the host nervous system and generated measurable electrical activity. It also remained structurally and functionally incomplete.

What the Stanford study changed

The researchers used an engineered host called an apallial mouse. Conditional deletion of the cohesin factor Esco2 in Emx1-expressing cells removed most glutamatergic neurons from the mouse neocortex and hippocampus. The model retained some pallial derivatives rather than eliminating every related structure.

The mice were also immunodeficient, reducing the ordinary immune-rejection response to transplanted human tissue. The team implanted human cortical organoids generated from induced pluripotent stem cells into pups between 5 and 17 days old. The organoids had been grown in the laboratory for 30 to 60 days before transplantation; the median recipient age was 10 days.

This design gave the human tissue room to develop inside a living nervous system while the host mouse was still in an early postnatal stage. It also created a way to observe the graft alongside circulation, sensory inputs and behavior rather than in a dish alone.

How much human tissue integrated

The numbers are striking, but they describe a specific experimental population and time point:

The graft contained deep- and superficial-layer glutamatergic neurons, progenitor cells, oligodendrocyte progenitor cells and astrocyte-lineage cells. Researchers also identified a small population of human GABAergic neurons and cells with von Economo neuron-like morphology.

The human-derived tissue did not simply remain isolated. Projections from the graft reached the mouse cervical spinal cord, while host inputs came mainly from the paleocortex, thalamus and pallidum. Within the graft, researchers recorded coordinated calcium bursts and synchronized local-field-potential events. Some activity correlated with orofacial motion, showing that the tissue participated in measurable neural and behavioral circuits.

Why the recovery remained partial

Large volume and electrical activity do not equal a mature cortex. At the examined stages, the graft lacked a continuous canonical cortical laminar structure, clear cortical arealization and a mature transcriptional profile. It also lacked a full complement of mature inhibitory interneurons.

The developmental state remained human-like in an important but limited sense: at 24 weeks after differentiation, graft glutamatergic neurons had a transcriptomic profile approximately equivalent to late second-trimester human cortical development. That is a developmental comparison, not evidence of an adult human brain operating inside the mouse.

Behavioral results were similarly specific. Gross locomotion was broadly preserved, but the xenocortical mice differed from comparison groups in gait organization, spontaneous activity, center exploration, rearing, associative conditioning and some avoidance behaviors. The study therefore produced a set of assay-specific behavioral changes rather than a general transformation of the animal's abilities.

What the model may be useful for

The platform is designed to bring human cortical cells into contact with living circuits, making it possible to study development, injury responses and circuit-level behavior in ways that isolated organoids cannot provide. In the study, exposure to 5% oxygen for five hours, after a one-hour graded descent from normal atmospheric oxygen, produced hypoxia-related cellular and behavioral responses in the human graft.

That result gives researchers a measurable stress paradigm. It does not turn the platform into an established treatment or a validated model for a named human neurological disease. The study presents xenocortication as a research system for future work on neurodevelopment, disease modeling and therapeutic development.

The next scientific challenge is to connect specific graft pathways to specific behaviors. The presence of projections, synchronized activity and behavioral differences shows that the tissue is biologically active; it does not by itself identify which graft-derived functions caused each observed outcome. Pathway-specific recordings and manipulations are needed for that level of explanation.

The ethical question follows the biology

The work received approval through Stanford's institutional stem-cell research oversight and animal-care committees. The researchers also consulted Stanford bioethicists and an external independent ad hoc ethics committee, obtained donor consent for in vivo transplantation and used locomotor and cognitive testing to assess altered capacities.

Those safeguards matter because the experiment changes more than the cell inventory of the animal. Human-derived tissue occupied most of the measured cortical volume, developed multiple neural and glial cell types and extended projections into the host spinal cord. As these models become more integrated, questions about animal welfare and the boundary between human and nonhuman biology become part of the experimental design itself.