Brain Organoid Atlas Maps Distinct Cellular Defects Across Four Neurodevelopmental Disorders
A 6,000+ organoid study links specific cellular phenotypes to microcephaly, polymicrogyria, epilepsy, and intellectual disability with 92% classification accuracy.
Summary
Researchers built the largest brain organoid atlas to date using iPSCs from 352 neurodevelopmental disorder (NDD) patients, studying over 6,000 organoids from 35 patients and 10 healthy controls. Using histology and single-cell RNA sequencing, they found that organoids reliably reproduced disease-specific cellular features. Microcephaly organoids were smaller and showed excessive TTR+ choroid plexus-like cells and cell death. Polymicrogyria organoids had defects in intermediate progenitor cell junctions. Epilepsy organoids showed excessive astrocyte generation and reactive astrogliosis. Intellectual disability organoids also overproduced TTR+ cells. A linear discriminant model classified disease categories from organoid data with up to 93% AUC accuracy, demonstrating that patient-derived brain organoids faithfully capture genotype-to-phenotype relationships across diverse NDDs.
Detailed Summary
Neurodevelopmental disorders (NDDs) collectively affect more than 4% of children and cost the US healthcare system over $400 billion annually, yet targeted treatments remain scarce because human brain tissue is inaccessible for direct study. Brain organoids — self-organizing 3D clusters grown from induced pluripotent stem cells (iPSCs) — offer a window into early human brain development. This landmark study presents a CIRM-funded NDD biobank of 352 genetically diverse patient-derived iPSC lines, now publicly available, and an accompanying organoid atlas that systematically maps cellular phenotypes across four major NDD categories: microcephaly (MIC), polymicrogyria (PMG), epilepsy (EPI), and intellectual disability (ID).
The research team generated more than 6,000 human brain organoids (hBOs) from 35 carefully selected patients and 10 healthy controls (including the H1 embryonic stem cell reference line), harvesting organoids at days in vitro (DIV) 28 and 52 for immunohistology, and at DIV52 for single-cell RNA sequencing (scRNA-seq). The scRNA-seq dataset encompassed 155,323 cells from 26 individuals, resolving 59 cell clusters annotated into 10 major brain cell classes. Label-transfer analysis matched organoid cells to human fetal brain developmental stages spanning gestational week 6 through 8 months postnatal, with particular enrichment in the second trimester (GW18–26), validating the developmental relevance of the model.
Principal component analysis of immunostaining data (SOX2, CTIP2, TBR2, KI67, CC3) demonstrated that patient organoids differed significantly from controls as early as DIV28 (PC1 scores: −0.04 for controls vs. +0.025/+0.01 for structural/non-structural patients). By DIV52, structural and non-structural NDD groups separated further. Crucially, linear discriminant analysis with leave-one-out cross-validation achieved AUC values of 0.88–0.93 for classifying disease categories at both timepoints, far exceeding chance. This performance degraded at DIV80, likely due to increased organoid variability at later stages.
Each NDD category showed distinct cellular signatures. MIC organoids were markedly smaller (1.9 ± 0.29 mm vs. 3.9 ± 0.4 mm at DIV52, p=2.18×10⁻³⁸), with elevated cleaved caspase-3 (CC3) indicating apoptosis, reduced KI67 proliferation, and an unexpected overabundance of TTR+ (transthyretin-expressing) choroid plexus-like cells. PMG organoids displayed disrupted ZO1+ tight junctions in intermediate progenitor cells (IPCs), consistent with aberrant cortical folding. EPI organoids showed striking excess astrocyte generation — including reactive astrogliosis markers — aligning with the known role of reactive astrocytes in seizure propagation. ID organoids, like MIC, also overproduced TTR+ cells but without the survival defect, suggesting a fate-shift rather than a survival phenotype.
The entire biobank, clinical metadata, brain imaging data, whole exome sequencing results, and single-cell transcriptomic datasets are publicly available at brain-org-ndd.cells.ucsc.edu, making this a community resource for drug discovery and disease mechanism studies. While the study is limited by the use of unguided cerebral organoids (capturing only early fetal brain stages) and a relatively small number of patients per disease category, the robust phenotypic concordance across genetically heterogeneous patients within each NDD category is a major advance. This atlas bridges genotype and phenotype at scale, providing a foundation for identifying drug targets and evaluating therapeutic candidates across the full genetic spectrum of NDDs.
Key Findings
- MIC organoids were 51% smaller than controls at DIV52 (1.9 ± 0.29 mm vs. 3.9 ± 0.4 mm, p=2.18×10⁻³⁸), with elevated apoptosis (CC3+) and reduced proliferation (KI67+)
- Linear discriminant analysis classified NDD disease categories from organoid cellular phenotypes with AUC of 0.92–0.93 at DIV52, outperforming DIV28 (AUC 0.88–0.87)
- scRNA-seq of 155,323 cells from 26 individuals resolved 59 cell clusters across 10 major brain cell classes, matching human fetal brain stages from GW6 to 8 months postnatal
- EPI organoids showed a marked increase in astrocyte numbers and reactive astrogliosis markers, linking excessive astrocyte generation to epilepsy pathophysiology
- Both MIC and ID organoids exhibited overproduction of TTR+ (transthyretin+) choroid plexus-like cells, suggesting a shared fate-shift phenotype despite distinct clinical presentations
- PMG organoids displayed disrupted ZO1+ apical junctions in intermediate progenitor cells, consistent with the cortical folding defects seen in polymicrogyria patients
- The CIRM NDD biobank encompasses 352 publicly available patient-derived iPSC lines with matched clinical, brain imaging, and whole exome sequencing data across four NDD categories
Methodology
Over 6,000 brain organoids were generated from 35 NDD patients (9 MIC, 8 PMG, 7 EPI, 11 ID) and 10 controls using a single standardized unguided cerebral organoid protocol, harvested at DIV28, 52, and 80. Histological quantification used immunostaining for SOX2, CTIP2, TBR2, KI67, CC3, and ZO1; scRNA-seq profiled 155,323 cells from 26 individuals at DIV52. Principal component analysis and leave-one-out linear discriminant analysis (with AUC as performance metric) were used to classify disease categories; label transfer matched organoid transcriptomes to a human fetal brain reference atlas.
Study Limitations
The study used unguided cerebral organoids that primarily model early fetal brain development (GW6 to 8 months postnatal), limiting applicability to later developmental processes or mature circuit dysfunction relevant to some NDDs. Patient numbers per disease category are small (7–11 per group), and not all patients have a confirmed genetic cause (only 68.5% had a likely pathogenic variant identified). The paper does not explicitly disclose conflicts of interest, and CIRM funding may represent an institutional interest in validating the biobank resource.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.