Human cortical organoids transplanted into apallial mice: grafts make up 91.9% of cortical volume and grow about 4.7-fold between 2 and 3 months

Background and open questions

Human brain tissue is hard to obtain directly, and human stem cell-derived cortical organoids (hCOs) can model cortical neurogenesis but lack circuit- and behaviour-level readouts in vitro. Transplanting hCOs into neonatal rat cortex has previously promoted integration and maturation of human neurons, but grafts occupied at most about one third of one rat cortical hemisphere.

The paper argues that mouse neurons develop faster than human neurons, and that a rapidly maturing host environment and pre-existing rat connectivity may limit graft proliferation and projection, which matters especially for modelling neurodevelopmental disorders involving widespread circuit dysregulation and behavioural change. This study's approach is to genetically eliminate dorsal and medial cortex in mice first, then transplant hCOs into the vacated cortical cavity.

Study design

Apallial mice were generated on an immunodeficient SCID background by deleting Esco2 in Emx1-expressing cells (Emx1-cre+/− Esco2fl/fl Prkdcscid/scid), killing dividing cells. Apallial offspring made up 23.4% (n = 91 litters, 46 breeding cages), consistent with the 25% Mendelian expectation; MRI showed about 50% less whole-brain tissue than Esco2fl/fl Prkdcscid/scid controls.

Apallial mice aged 5 to 17 days (median 10 days) received 2 hCOs per hemisphere, 4 per animal, derived from 3 hiPS cell lines, and are termed xenocorticated (XCX) mice after transplantation. Controls were littermate Esco2fl/fl Prkdcscid/scid mice and focal grafts transplanted into intact cortex. The paper states that sample sizes were estimated empirically from previous studies and that the study was not powered to detect small sex effects.

Key results

Primary readout: graft growth and volume fraction

The transplantation success rate was 86.2% (29 mice, 3 hiPS cell lines), consistent with the earlier rat t-hCO protocol and indicating that hCO survival does not require an intact host cortex. MRI showed grafts growing about 4.7-fold between 2 and 3 months after transplantation (n = 14, paired t test, t = 8.219, P = 1.66×10−6). At 3 months, grafts made up 91.9% of total cortical tissue volume (n = 7), averaging about 32,000 neurons/mm3; this fraction is a descriptive value.

Extent of cortical ablation and L5-ET neurons

Whole-brain snRNA-seq (880,149 nuclei, 2 mice per group) showed roughly sevenfold reductions across classes of dorsal cortex-derived glutamatergic neurons, with olfactory bulb, non-cortical, amygdalar and piriform cortex-derived classes preserved, so "apallial" is only an operational label. In XCX mice, 85% of layer 5 extratelencephalic (L5-ET) neurons had UCell scores >0.1, versus only 3 of 213,753 neurons in the HNOCA organoid atlas; the L5-ET proportion in XCX was more than threefold higher than in previous rat and adult mouse transplants.

Behaviour: MoSeq and CatWalk

MoSeq (60 syllables, 99% of variance) placed XCX mice between controls and apallial mice but closer to apallial, with the greatest session-to-session drift (Kruskal–Wallis P = 0.019; XCX versus control Bonferroni-corrected P = 0.027), so transplantation did not normalize them. CatWalk showed no difference in run speed (ANOVA P = 0.8467); swing time was shorter in apallial mice than controls (P = 0.0019), with XCX versus control at P = 0.2303; diagonal paw synchrony was altered in both apallial and XCX groups.

Y maze and fear conditioning

In Y maze spontaneous alternation, controls (n = 39, one-sample t test against the 50% chance level P = 2.08×10−7) and XCX mice (n = 18, P = 0.0004) performed above chance while apallial mice (n = 16, P = 0.3016) did not; the effect replicated in two independent cohorts months apart. Apallial and XCX mice both showed reduced freezing during trace fear conditioning training and cued testing, along with contextual freezing deficits.

Cellular and gait readouts after hypoxic injury

After XCX and control mice were exposed to 5% O2 for 5 hours (including a 1-hour ramp-down), HIF1α immunoreactivity was visible in the human grafts of 3/3 XCX mice, with no clear signal in adjacent host archicortex or in identically exposed control mice. In CatWalk testing, XCX mice showed the largest increase from baseline in the proportion of 3- to 4-paw support (ANOVA P = 0.0047; control versus XCX P = 0.0120, XCX versus apallial P = 0.0076), while mean speed did not change (P = 0.3845).

Mechanistic interpretation

Demonstrated in the paper: XCX grafts at 4 months after differentiation (labelled with hSYN1-GCaMP8s) showed graft-wide synchronous calcium bursts lasting tens of seconds and recurring every few minutes, propagating across the entire dorsal surface within about 100 ms; bursts correlated with orofacial movement (paired t test P = 0.021, n = 3 mice), and 32-channel probe recordings captured LFP bursts occurring synchronously along the full extent (n = 3 mice).

All 3 XCX mice contained VEN-like bipolar or spiral cells, making up 0.16% of labelled somata, with a median soma size of 29.8 µm versus an overall median of 9.251 µm; sparse GCaMP8s+STEM121+ human projections were seen in the cervical spinal cord in all 3 XCX mice, with no signal in the spinal cords of focal graft controls or wild-type controls.

Author hypotheses: The authors suggest that genetic ablation of host cortex may allow enrichment of L5-ET neurons, including corticospinal neurons; behavioural output may reflect interaction between human grafts and preserved mouse subcortical circuits, but whether grafts are necessary or sufficient for specific behaviours remains unclear. Gait changes after hypoxia may be driven by longer hindpaw contact times and larger contact areas, an interpretation that was not tested causally.

Limitations and uncertainties

• Grafts reached at most 177 days after differentiation, with a transcriptional age at 24 weeks roughly equivalent to the late second trimester, lacking canonical cortical lamination, full regionalization, sufficient GABAergic interneurons and a mature transcriptional profile; the authors suggest that a mismatch in maturation rate between host and graft may limit integration.
• The study included no non-cortical organoid transplantation control, so cortical lineage effects cannot be distinguished from general transplantation effects; whether grafts are necessary or sufficient for specific behaviours remains unclear, and functional integration requires pathway-specific recording and manipulation.
• Whole-brain snRNA-seq used only 2 mice per group, and calcium imaging and LFP 3 each; VEN-like cells made up only 0.16% and were identified morphologically; the apallial group in the hypoxia behaviour experiment was female only (n = 9); the 91.9% fraction (n = 7) and the fold enrichment of L5-ET neurons were not statistically tested, and MoSeq also discriminated apallial from XCX mice only weakly.

Clinical and industry implications

If these results are reproduced across more cell lines and at later maturation time points, xenocorticated mice could provide a platform for assessing human cortical neurons in vivo in which grafts make up 91.9% of cortical tissue volume, while reading out circuit activity and behaviour at the same time. The gait changes in the hypoxia model suggest the platform could be used to model perinatal hypoxia and related neurodevelopmental disorders and to test therapeutic interventions on larger volumes of human cortical tissue.

The authors also note that if grafts become more mature, ethical guidance should be established early. All of the above are preclinical mouse data with no patient-level validation.

Authors, source and verification

Evidence level: Full text; verification record: Read the Nature open-access full-text HTML abstract, Results, Discussion, legends for Figs. 1–5 and Methods