Modeling blood-brain barrier formation and cerebral cavernous malformations in human PSC-derived organoids

Abstract

The human blood-brain barrier (hBBB) is a highly specialized structure that regulates passage across blood and central nervous system (CNS) compartments. Despite its critical physiological role, there are no reliable in vitro models that can mimic hBBB development and function. Here, we constructed hBBB assembloids from brain and blood vessel organoids derived from human pluripotent stem cells. We validated the acquisition of blood-brain barrier (BBB)-specific molecular, cellular, transcriptomic, and functional characteristics and uncovered an extensive neuro-vascular crosstalk with a spatial pattern within hBBB assembloids. When we used patient-derived hBBB assembloids to model cerebral cavernous malformations (CCMs), we found that these assembloids recapitulated the cavernoma anatomy and BBB breakdown observed in patients. Upon comparison of phenotypes and transcriptome between patient-derived hBBB assembloids and primary human cavernoma tissues, we uncovered CCM-related molecular and cellular alterations. Taken together, we report hBBB assembloids that mimic the core properties of the hBBB and identify a potentially underlying cause of CCMs.

Keywords: assembloids; cerebral cavernous malformations; human PSC-derived organoids; human blood-brain barrier; neuro-vascular development; neuro-vascular interactions; single-cell transcriptomics; spatial transcriptomics.

Figure 1.. Assembly of hPSCs-derived cerebral cortical and blood vessel organoids mimicking neurovascular development.

A. Schematic protocol for assembling cerebral-blood vessel assembloids to reconstruct the human blood-brain barrier in 3D. B. Confocal images of cerebral cortical organoids expressing Sox2 and Tuj1 at Day 26, and Ctip2 and Tbr1 at Day 61 (n = 4 cultures). C. Induced astrocytes of human cerebral cortical organoids displaying extensive cell soma and processes (arrowhead) and end-feet labeled by AQP4 (arrow, n = 3 cultures). D. Whole-mount staining of blood vessel organoids for endothelial cells (CD31, red) and pericytes (PDGFR-β, magenta) on Day 15 derived from the H9-GFP line (n = 4 cultures). E. Representative images of cerebral-blood vessel assembloids at 7 days after assembling Day 15 H9-RFP vessel organoid (VO) and Day 75 H9-GFP cerebral organoid (CO) with excessive angiogenic front (arrows, red) at the periphery of fusion (n = 5 cultures). F. Extended vascular network (green) in cerebral organoid (highlighted by dashed line) from 14 and 28 days (n = 4 cultures), indicative of actively growing brain vessels. G. Time course analysis of vessel length in cerebral organoids after assembling with blood vessel organoids. H. Vascular cells (green) stained for CD31 (gray) extend complex networks (arrow) into a cerebral organoid (red) on Day 21 assembly (n = 4 cultures), suggesting the formation of human brain capillaries. Purple color indicates merging of green and white colors. Scale bars: 10 μm in (C) right, 50 μm in (B) and (C) left, 100 μm in (D), (E) high magnification and (H), 500 μm in (E) low magnification and (F). See also Figure S1.

Figure 2.. BBB assembloids recapitulate molecular, cellular, functional, and transcriptomic signatures of the in vivo human BBB.

A-C. Majority of ECs express BBB markers GLUT1, ZO-1, and Claudin-5 by day 30 post-assembling with brain organoids (n = 3 cultures). Arrows indicate junctional-like structures. D. Quantification of BBB marker expressions in blood vessel organoids (Vo), BBB assembloids, and human brain tissues (HB). E. Brain endothelium (CD31) covered by continuous Collagen IV basement membrane (n = 3 cultures). F, G. Orthogonal sections showing GFAP-labeled astrocytic processes (red) and AQP-4-labeled end-feet (red) ensheathing capillary wall (arrow, green, n = 3 cultures). H. Vascular network (green) stained for CDH5 (grey) extending capillaries (arrow) into cerebral organoid (red), showing tight innervation with PDGFR-β-labeled pericyte processes (blue) on Day 20 (n = 3 cultures). Purple indicates merging of green and white. I. Schematic of cellular components and structure of BBB assembloids. J. TEER measurement in blood vessel organoids and hBBB assembloids (n = 3, three independent batches). K. Uniform manifold approximation and projection (UMAP) of single-cell transcriptome of human BBB assembloids (n = 3 batches, 28,062 cells) colored by clusters. L. Transcriptome-wide comparison of ECs in BBB assembloids with organ-specific ECs by Tabula Muris Consortium. Scale bars: 50 μm in (A)-(C) and (E)-(H). *p < 0.05, ***p < 0.001. NS indicates no significance. Data: mean ± SEM. See also Figure S2.

Figure 3.. Endothelial acquisition of BBB-specific features via Wnt signaling activation.

A. Schematic illustrating brain and vascular cell isolation from BBB assembloids for transcriptomic analysis. B. Heatmap showing statistical distance between RNA-seq datasets. Three replicates from each group: cerebral organoid (CO), blood vessel organoid (VO), brain cells from Day 30 BBB assembloids without GFP (BBB-GFP⁻), and vessel cells from Day 30 BBB assembloids with GFP (BBB-GFP⁺). C. Volcano plots depicting differential gene expression analysis between BBB-GFP⁺ cells and VO (fold change > 2, adjusted P-value < 0.01). D. Bubble map of KEGG pathway enrichment analysis of differentially expressed genes (DEGs) between BBB-GFP⁺ cells and VO, showing top 30 enriched pathways. Color indicates adjusted P-value; dot size represents number of DEGs in pathway. E. Heatmap of differential expression of Wnt signaling genes between VO and BBB-GFP⁺ cells. Red/blue shadings denote higher/lower relative expression levels. F, G. Confocal images demonstrate elevated GLUT1 and ZO-1 expression in VOs without neural tissue post CHIR99021 treatment (n = 3 culture). Scale bars: 100 μm in (G), 200 μm in (F). ***p < 0.001. Data: mean ± SEM. See also Figure S3 and Tables S1 and S2.

Figure 4.. Spatial transcriptomics analysis of BBB assembloids.

A. SpaceFlow identifies spatial domains in the ROI, listing enriched cell types (Cell type 1, Cell type 2) for each domain. B. Spots with detected RNA reads of marker genes, color-coded by spatial domains. C. Statistical association between spatial domains (rows) and cell types (columns), showing odds ratio (dot radius) and significance (−Log10(p-value), Chi-square test) from blue (insignificant) to orange (highly significant). Boxes highlight significant associations between spatial domains and cell types. D, E. Confocal images demonstrate co-localization of endothelium (CD 31) with GABAergic neurons (GAD 67) and smooth muscle cells (SMA) with glutamatergic neurons (CaMKII), respectively. Scale bars: 50 μm in (D), (E). Data: mean ± SEM. See also Table S3.

Figure 5.. Modeling CCMs using patient-derived BBB assembloids.

A. Schematic illustrating the generation and comparison of BBB assembloids from unaffected controls and CCM patients. B. Confocal images showing brain endothelium (CD31) in control and CCM BBB assembloids, with clusters of enlarged endothelial channels in CCMs indicating cavernous malformations. Quantification of vessel width and lesion area (n = 3 cultures). C. Images comparing ZO-1 expression in control and CCM BBB assembloids, with reduced ZO-1 observed in CCMs. Arrows highlight brain endothelium (CD31 staining). D. Quantification of ZO-1, Claudin-5, and Coll IV intensity in control and CCM BBB assembloids (n = 3 cultures). E. Procedure for resecting primary cavernomas tissue, with confocal image showing enlarged endothelial channels in primary cavernomas (n = 3 cultures). F. Vascular network (CD31, red) and Claudin-5 (green) in postmortem brain tissue and primary cavernomas, with reduced Claudin-5 in cavernomas indicating BBB breakdown (n = 3 brain sections and 3 primary cavernomas sections). G. Confocal images of brain endothelium (CD31) and basement membrane (Coll IV) in brain tissue and primary cavernomas, with disrupted Coll IV expression in cavernomas indicating basement membrane disassembly (n = 3 brain sections and 3 primary cavernomas sections). H. Quantification of vessel parameters and Claudin-5/Coll IV intensity (n = 3 brain sections and 3 primary cavernomas sections). Scale bars: 50 μm in (B), (C), and (E)-(G). ***p < 0.001. Data: mean ± SEM. See also Figure S4.

Figure 6.. Single-cell transcriptomics reveals altered profiles in vascular cells and neurovascular interactions in control and CCM BBB assembloids.

A, B. UMAP visualization of cell states from control (n = 3 batches, 23,848 cells) and CCM (n = 3 batches, 22,445 cells) BBB assembloids, colored by cell clusters (A) and groups (B). C. Violin plots showing characteristic marker genes of each identified cell population. D. Heatmap displaying known CCM lesion markers, showing average log fold change (padj<0.001) in CCM versus control BBB assembloids. E. Heatmap showing average log fold change (padj<0.001) in CCM versus control BBB assembloids for selected tip cell and tumor tip cell markers. F. Representative plots illustrating ligand-receptor pairs (colored connections) between neural and vascular cells predicted by CellPhoneDB, with significant changes in expression and responsive vascular cells in CCM BBB assembloids compared to controls. Rectangle size indicates interaction scores, and connection line width represents probability scores. G. Top 10 ligand-receptor pairs between neural and vascular cells, visualized with sizes and colors corresponding to cell-cell communication (CCC) scores. See also Figures S5 and S6 and Tables S6 and S7.

Figure 7.. Developmental loss of vascular smooth muscle cells in CCMs.

A. UMAP plots of neuro-vascular cell clusters in BBB assembloids, with vSMCs indicated by arrows. B. Differential cell composition between control and CCM BBB assembloids, highlighting the loss of vSMCs in CCMs (arrow). C. STREAM visualization of developmental trajectories of mural cells, color-coded by cell types and groups. D. Schematic depiction of vSMC loss from mesenchymal progenitors in CCMs. E-G. Confocal images and quantification revealing reduced smooth muscle actin expression in primary cavernomas tissue compared to healthy brain tissue (n = 3 samples). H-J. Images and quantification showing comparable NG2 expression between primary cavernomas tissue and healthy brain tissue (n = 3 samples). Scale bars: 50 μm in (E), (F), (H), and (I). ***p < 0.001. NS denotes no significance. Data: mean ± SEM. See also Figure S7.