Brain Chimeroids reveal individual susceptibility to neurotoxic triggers

Abstract

Interindividual genetic variation affects the susceptibility to and progression of many diseases^1,2. However, efforts to study how individual human brains differ in normal development and disease phenotypes are limited by the paucity of faithful cellular human models, and the difficulty of scaling current systems to represent multiple people. Here we present human brain Chimeroids, a highly reproducible, multidonor human brain cortical organoid model generated by the co-development of cells from a panel of individual donors in a single organoid. By reaggregating cells from multiple single-donor organoids at the neural stem cell or neural progenitor cell stage, we generate Chimeroids in which each donor produces all cell lineages of the cerebral cortex, even when using pluripotent stem cell lines with notable growth biases. We used Chimeroids to investigate interindividual variation in the susceptibility to neurotoxic triggers that exhibit high clinical phenotypic variability: ethanol and the antiepileptic drug valproic acid. Individual donors varied in both the penetrance of the effect on target cell types, and the molecular phenotype within each affected cell type. Our results suggest that human genetic background may be an important mediator of neurotoxin susceptibility and introduce Chimeroids as a scalable system for high-throughput investigation of interindividual variation in processes of brain development and disease.

Extended Data Fig. 1.. PSC Chimeroids display uneven donor composition but proper cell type composition.

a, Schematic of the PSC-Chimeroid protocol. b, Stacked bar plots showing donor composition, as measured via Census-seq from low-pass whole-genome sequencing, in PSC-Chimeroids from various mixes and timepoints, labeled as the number of days since the day of the seeding (DIV0). For the DIV0 and DIV1 time points, n=4 Chimeroids have been pulled together. c, Left panel: brightfield image of 3-mo PSC-Chimeroids; scale bar: 1 mm. Right panel: immunolabelling of 3-mo PSC-Chimeroid with SATB2, TBR1 and MAP2. Scale bar: 500 μm. d, UMAP of integrated PSC-Chimeroids at 3 mo, colored by annotated cell type (left panel) and donor line as determined with demuxlet (right panel). e, UMAPs split by donor for two different replicates. f, Barplots of cell type proportions in PSC Chimeroids (n=2), demultiplexed by donor. g, Immunolabelling of 1-mo PSC-Chimeroids showing early progenitors (SOX2), cortical progenitors (EMX1); rosette centers are lined with ZO-1, a tight junction protein present in the endfeet of radial glia cells, and NESTIN, indicating correctly-polarized neuroepithelium. CFuPNs neurons are marked by TBR1, and the neuronal dendrites by MAP2. h, Immunolabelling of 2-mo PSC-Chimeroids showing early progenitors (SOX2), IP (TBR2) and oRG (HOPX). CTNNB1 marks the center of the neural rosettes, indicating correctly-polarized neuroepithelium. i, Immunolabelling of 3-mo PSC-Chimeroids showing cortical markers such as SATB2 (upper layers cortical neurons), and CTIP2 (deep layers cortical neurons). White arrows point to the neural rosettes (shown at higher magnification in the lower panels). Scale bar: 100 μm.

Extended Data Fig. 2.. NSC-Chimeroids created by aggregating different numbers of cells display variable development.

a. Brightfield images of single-donor H1 NSC-Chimeroids seeded with 9,000 (upper) and 12,000 (lower) cells/well at day 3 and day 16 after reaggregation (left panels), and brightfield images of single donor PGP1 NSC-Chimeroids seeded with 9,000 (top left), 12,000 (bottom left), and 20,000 (top right) cells/well at day 3 and day 16 after reaggregation (right panels). Scale bar: 500 μm. Barplot showing growth over time of PGP1 NSC-Chimeroids compared across all different cell counts at aggregation (bottom right). Bars show median values, whiskers show upper and lower quartiles (n=170 Chimeroids across all 4 timepoints). Significance was calculated using ANOVA followed by two-sided Tukey post-hoc pairwise tests. P-values: *** < - 0.001; **** - <0.0001. b. Immunolabelling of PGP1 single-donor NSC-Chimeroids made by aggregating the indicated number of cells, at DIV35. Upper images display whole organoids; lower images, enlargement of indicated portions. Lefthand panel, immunolabelling for SOX2 (progenitors), MAP2 (neuronal dendrites), and TBR1 (deep layer neurons). Righthand panel, immunolabelling for SOX2, ZO-1 (tight gap junctions), and EMX1 (dorsal cortical progenitors). Scale bar: 500 μm (upper) and 125 μm (lower). Arrowheads indicate non-cortical (grey) and cortical (white) regions. c. Left panel: whole-organoid brightfield images of H1 single-donor NPC-Chimeroids seeded with 100,000 cells at aggregation, at 18 and 35 days after mixing. Scale bar: 500 μm. Right panel: immunolabelling of H1 single-donor NPC-Chimeroids showing SOX2, ZO-1, and EMX1. Scale bar: 500 μm and 125 μm (zoom-in).

Extended Data Fig. 3.. Multi-donor NSC Chimeroids display appropriate differentiation markers.

a, Immunolabelling of 1-mo MD-NSC Chimeroids showing early progenitors (SOX2), cortical progenitors (EMX1); rosette centers are lined with ZO-1, a tight junction protein present in the endfeet of radial glia cells, and NESTIN signal, indicating correctly-polarized neuroepithelium. CFuPNs neurons are marked by TBR1, and the neuronal dendrites by MAP2. b, Immunolabelling of 2-mo MD-NSC Chimeroids showing early progenitors (SOX2), IP (TBR2) and oRG (HOPX). CTNNB1 marks the center of the neural rosettes, indicating correctly-polarized neuroepithelium. c, Immunolabelling of 3-mo MD-NSC Chimeroids showing cortical markers such as SATB2 (upper layers cortical neurons), and CTIP2 (deep layers cortical neurons). White arrows point to the neural rosettes. Scale bar: 100 μm.

Extended Data Fig. 4.. NSC-Chimeroids develop appropriate cell-type composition across multiple donors.

a, Stacked barplots showing the donor composition, as measured via Census-seq from low-pass whole-genome sequencing, of multi-donor NSC-Chimeroids from various mixes and at various timepoints, labeled as the number of days since reaggregation (DIV; for the DIV0 and DIV1 time points, n=4 chimeroids have been pulled together). b-e, 3-mo NSC-Chimeroids UMAPs split by donor (b, d) and stacked barplots of donor contributions for each cell type (c, e) for Mix 1 (4 donors) and Mix 2 (5 donors).

Extended Data Fig. 5.. NSC-Chimeroids present uniform spatial distribution across donors.

a, Slice #1. Spatial plot of Slide-seq data from 2-mo MD-NSC Chimeroids (Mix 5), colored by maximal donor contribution via Census-seq. b, Spatial plots showing donor prediction weights. c, Spatial plot of Slide-seq data from 2-mo NSC-Chimeroids (Mix 5), colored by RCTD-assigned cell type. d, Spatial plots showing cell type prediction weights. e, Slice#2. Same as a. f, Slice#2. Same as b. g, Slide#2. Same as c. h, Slide#2. Same as d. i, Representative organoid placed on a Multielectrode Array (MEA). j, Representative activity map of a MD-NSC Chimeroid network burst at 4mo, where each pixel of the map represents an electrode (left) and example action potential traces from the electrodes highlighted in the activity map (right). k, Representative raster plot showing the firing activity of active electrodes (139, defined as electrodes with a mean firing rate >0.15 spikes/s inside the outline of the organoid) over a 15 min recording. Network bursts start to appear around 5mins. l, Effect of AMPA and NMDA blockers (D-AP5, 150μM; DNQX 60μM) on neuronal activity/network bursts (left panel); network bursting activity of 4 recorded NSC-Chimeroids (right panel).

Extended Data Fig. 6.. NPC-Chimeroids display appropriate differentiation markers.

a, Immunolabelling of 1-mo NPC-Chimeroids showing early progenitors (SOX2), cortical progenitors (EMX1); rosette centers are lined with ZO-1, a tight junction protein present in the endfeet of radial glia cells, and NESTIN signal, indicating correctly-polarized neuroepithelium. CFuPN neurons are marked by TBR1, and the neuronal dendrites by MAP2. b, Immunolabelling of 2-mo NPC-Chimeroids showing early progenitors (SOX2), IP (TBR2) and oRG (HOPX). CTNNB1 marks the center of the neural rosettes, indicating correctly-polarized neuroepithelium. c, Immunolabelling of 3-mo NPC-Chimeroids showing cortical markers such as SATB2 (upper layers cortical neurons), and CTIP2 (deep layers cortical neurons). White arrows point to the neural rosettes. Scale bar: 100 μm. d, Brightfield images of 3-mo Chimeroids across all protocols (scale bar: 1 mm).

Extended Data Fig. 7.. Nascent Chimeroid input to aggregation steps is composed of expected cell-type populations.

a, UMAP of integrated dataset containing DIV23 organoids from our reference map of Velasco-protocol organoid development (“Ref. Map”), and MD-NSC and MD-NPC-Chimeroids 1 day after aggregation (INPUT), color-coded by annotated cell type. b. UMAPs split by protocol. c, UMAP of the INPUT for MD-NSC Chimeroids (mix 2), color-coded by annotated cell type. d, UMAPs split by donors. e, UMAP of the INPUT for MD-NPC Chimeroids (Mix 6), color-coded by annotated cell type. f, UMAPs split by donors. g, Donor demultiplexed after sc-RNA-seq. h, Bar plots of cell-type proportions in INPUT MD-NSC and MD-NPC Chimeroids for each donor.

Extended Data Fig. 8. Single-donor Chimeroids display appropriate differentiation markers.

a, Immunolabelling of 1-mo 11a SD-NSC Chimeroids showing early progenitors (SOX2), cortical progenitors (EMX1); rosette centers are lined with ZO-1, a tight junction protein present in the endfeet of radial glia cells, NESTIN and CTNNB1 signal, indicating correctly-polarized neuroepithelium surrounded by early new born CFuPN neurons. b, Same as panel a, for PGP1 SD-NSC Chimeroids. c, Immunolabelling of 3-mo 11a SD-NSC Chimeroids showing cortical markers such as SATB2 (upper layers cortical neurons), and CTIP2 (deep layers cortical neurons). d, Same as panel c, for PGP1 SD-NSC Chimeroids. White arrows point to the neural rosettes. Scale bar: 100 μm.

Extended Data Fig. 9.. Chimeroids comparison across protocols and donors.

a, UMAP showing overlapping neighborhoods of cells, as calculated using Milo. Red and blue colors indicate neighborhoods with significant enrichment for cells from single-donor or multi-donor Chimeroids, respectively. Point size indicates the number of cells in a neighborhood, and edge thickness indicates the number of cells shared between pairs of neighborhoods. b, Beeswarm plot showing shifts in the composition of neighborhoods of cells, grouped by the cell-type identity of those neighborhoods. Each point represents a neighborhood of 50-200 cells with similar gene expression profiles. The vertical axis indicates the enrichment of single-donor cells within a neighborhood, with positive log fold change values indicating more than expected single-donor cells, and negative values indicating fewer than expected single-donor cells. Neighborhoods are colored based on statistical significance of that enrichment: grey, not significantly different from random; red, significant over-enrichment; blue, significant under-enrichment. c-d, Volcano plots showing DEGs, overall and for each donor (c), and for each major cell type (d), between the MD- and SD-NSC Chimeroid protocols. Positive log2 fold change indicates higher expression in the SD protocol. e, Correlation plots comparing the DEGs between each pair of donors in MD-NSC Chimeroids to those between the same pair of donors in the SD-NSC Chimeroids. The x- and y-axies shows log2 fold change in MD and SD, respectively. Point size and color indicate statistical significance in MD and SD, respectively. f, Heatmap showing the mean absolute value of the log2 fold change between each donor in the MD-NSC Chimeroid protocol (columns) and each donor in the SD-NSC Chimeroid protocol (rows). Lower values imply more similarity; samples from each donor were transcriptionally most similar to samples from the same donor in the other protocol. g, Aitchison distance measuring the dissimilarity in cell type composition between replicates within each protocol, split into comparisons within batches and between batches, and limited to cells derived from the PGP1 and Mito210 donors (as only these two donors are present in all organoid/Chimeroid protocols assayed here). (n=23 PGP1, 23 Mito210, and 10 fetal samples; Boxes show upper and lower quartiles and median, whiskers show highest/lowest values within 1.5 interquartile range (IQR) of the nearest hinge). The dissimilarity in cell type composition between samples within each of three fetal cortical datasets is also shown, indicating natural variability between individuals. Dotted lines represent mean inter-sample distances within each fetal dataset^,,.

Extended Data Fig. 10.. Chimeroids display similar cell type complexity as single-donor organoids and fetal tissue.

Rank-Rank Hypergeometric Overlap (RR-HO) plots comparing the expression signatures of cell types in multi-donor Chimeroids to all cell types in single-donor Chimeroids, cortical organoids from our reference map of Velasco-protocol organoid development, or endogenous human fetal tissue. The horizontal axes represent lists of marker genes for multi-donor NSC-Chimeroid cell types compared to all other Chimeroid cells, ranked from most upregulated to most downregulated; the vertical axes represent similarly-ranked lists of marker genes for single-donor NSC-Chimeroids, reference map organoids, or human fetal cells. Color at a given position represents the significance (negative log p-value) of the overlap of the gene lists up to that point, as calculated by Fisher’s exact tests. High significance (i.e., red color) in the lower left and upper right quadrants indicates strong concordance between the expression profiles which define the compared cell types. a, b, c, d and e represent different cell types analyzed. f, Explanatory schematic for the RR-HO plots.

Extended Data Fig. 11.. Chimeroids display similar developmental trajectories and metabolic hallmarks across donors and across protocols.

a, Density plot showing the distribution of scaled pseudotimes assigned to each donor within each organoid/Chimeroid protocol. Pseudotime was calculated independently for each protocol using Monocle3, with aRGs in each dataset set as root cells. b, Glycolysis module scores calculated in each cell type/donor/Chimeroid with linear MEMs (using lme4’s lmer) on the PGP1/Mito210 cells using donor as a random effect (n=40 replicates across 4 protocols; boxes show upper and lower quartiles and median, whiskers show highest/lowest values within 1.5 interquartile range (IQR) of the nearest hinge.). P-values were generated via one-sided F-test comparing models with and without “protocol” as a covariate. c-d, Expression of the glycolysis geneset is similar across donors (12,887 cells from n=45 donor/Chimeroids) and protocols (20,425 cells from n=40 donor/Chimeroids and organoids). Violin plots showing the module scores for the MSigDB Hallmark Glycolysis gene set across cell types, donors, and protocols; boxes show upper and lower quartiles and median, whiskers show highest/lowest values within 1.5 interquartile range (IQR) of the nearest hinge. Module scores were calculated with Seurat’s addModuleScores function.

Extended Data Figure 12.. VPA and EtOH treatment differently affect cell-type composition in NSC-Chimeroids.

a, Stacked barplot showing cell type and donor composition for control Chimeroids; the width of each bar corresponds to the proportion of the indicated donor in Mix 1 (4 donors, left panel) and Mix 2 (5 donors, right panel). b, Brightfield images of whole cortical NSC-Chimeroids at 3 mo, in the EtOH treatment condition. Scale bar: 1 mm. c, Stacked barplot showing cell type and donor composition for EtOH treated Chimeroids; the width of each bar corresponds to the proportion of the indicated donor in Mix 1 (4 donors, left panel) and Mix 2 (5 donors, right panel). d, UMAPs of EtOH treated NSC-Chimeroids, split by mixes and replicates. e, Brightfield images of whole cortical MD-NSC Chimeroids at 3 mo, in the VPA treatment condition. Scale bar: 1 mm. f, Stacked barplot showing cell type and donor composition for EtOH treated Chimeroids; the width of each bar corresponds to the proportion of the indicated donor in Mix 1 (4 donors, upper panel) and Mix 2 (5 donors, lower panel). g, UMAPs of VPA treated NSC-Chimeroids, split by mixes and replicates.

Extended Data Figure 13.. VPA treatment cause protein and transcriptome alterations in single-donor Chimeroids.

a, Immunolabelling of multi donor NSC-Chimeroids (a, Mix 1, upper images and b, Mix 2, lower images). Left panel, MAP2 (neuronal dendrites), EMX1 (cortical progenitors), and DLX2 (GABAergic cells). Right panel, DCX (migrating neurons), SATB2 (upper layers cortical neurons), and CTIP2 (deep layers cortical neurons). Scale bar: 100 μm. c, Stacked barplot showing cell type composition for control and treated single-donor NSC-Chimeroids. d, Left panel: cell-type specific changes in single-donor Chimeroids treated with VPA (yellow) vs control (grey). Right panel: UMAP showing overlapping neighborhoods of cells, as calculated using Milo. Red and blue colors indicate neighborhoods with significant enrichment for VPA-treated cells or control cells, respectively. Point size indicates the number of cells in a neighborhood, and edge thickness indicates the number of cells shared between pairs of neighborhoods. e, Beeswarm plot showing shifts in the composition of neighborhoods of cells in response to VPA treatment in single-donor Chimeroids, grouped by the cell-type identity of those neighborhoods. Each point represents a neighborhood of 50-200 cells with similar gene expression profiles. The vertical axis indicates the enrichment of VPA-treated cells within a neighborhood, with positive log fold change values indicating more than expected VPA-treated cells, and negative values indicating fewer than expected VPA-treated cells. Neighborhoods are colored based on statistical significance of that enrichment: grey, not significantly different from random; red, significant over-enrichment; blue, significant under-enrichment. If most neighborhoods within a cell type collectively shift up or down, it implies an overall gain or loss, respectively, of that cell type in VPA-treated Chimeroids. Cell types with neighborhoods that form long tails of both over-enrichment and under-enrichment are likely to have treatment-induced changes in expression profile, without necessarily changing in abundance. f, GSEA of donor specific genes from MD-NSC Chimeroids ranked on the corresponding single-donor datasets. g, GSEA of donor specific genes from SD-NSC Chimeroids ranked on the corresponding multi-donor datasets (lower panels). P-values calculated via two-sided Kolmogrov-Smirnov test.

Figure 1.. NSC-Chimeroids maintain donor representation across cell classes.

a, Immunolabelling for MAP2, EMX1, and NESTIN in cortical PSC-Chimeroids at 1 month (mo). Scale bar: 500 μm (whole) and 250 μm (zoom). b, Donor contribution in 3-mo PSC-Chimeroids (demultiplexed single-cell RNA-seq), color-coded by donor. c, Schematic of Neural Stem Cell (NSC)-Chimeroid protocol. d, Donor contribution in 3-mo NSC-Chimeroids, as in c. e, Immunolabelling of 1-mo (scale bar: 100 μm) and 3-mo cortical NSC-Chimeroids (scale bar: 500 μm). f, Slide-seq analysis of 2-mo MD-NSC Chimeroids (Mix 5), colored by RCTD-assigned cell type (upper panel) and by donor contribution via Census-seq (lower panel). g, UMAP of integrated NSC Chimeroids (showing total cells per dataset), color-coded by annotated cell type (upper panel) and donor line, and quantification of donor contribution (lower panel). h, UMAPs split by donor for two different mixes (Mix 1 and Mix 2). i, Bar plots of cell-type proportions in NSC-Chimeroids, demultiplexed by donor. Solid bar: median; error bars, upper and lower quartiles across replicates (n=7 for CW, H1, Mito210, and PGP1; n=3 for 11a). j, Stacked bar plots of cell-type proportion for each donor within Mix 1 and Mix 2.

Figure 2.. NSC- and NPC-Chimeroids control hyperproliferative states of starting pluripotent stem cells.

a, 3-mo NSC-Chimeroids analyzed by single-cell RNA-seq and demultiplexed by donor for Mix 3 and Mix 4. b, 3-mo NSC-Chimeroid UMAPs for Mix 3 and Mix 4 (total cells for each dataset). c, Bar plot with cell type and donor composition; bar width corresponds to donor proportion. d, UMAPs split by donor for Mix 3. e, Density plot of pseudotime by cell line, calculated using Monocle3 with all Cycling cells as roots. f. Schematic of NPC-Chimeroid protocol. NPCs: Neural Progenitor Cells g, Immunohistochemistry of 3-mo NPC-Chimeroids showing SOX2, CTIP2 and MAP2. Scale bar: 500 μm. h-i. UMAP representation of NPC-Chimeroids integrated (h) and split by donor (i) (showing total cells for each dataset). j, Barplots of cell type and donor composition.

Figure 3.. Chimeroids are reproducible and display appropriate cell-type composition

a, Schematic of cell populations in cortical organoids. b, Expression of marker genes across major cell types in Ref. Map organoids, NSC-Chimeroids, single-donor (SD) NSC-Chimeroids, and NPC-Chimeroids. c, Pearson correlation coefficients for normalized marker gene expression between human fetal cortical cell types (43,935 cells, n=4 replicates) and corresponding cell types in Ref. Map organoids (134,282 cells, n=21 replicates, 7 batches), NSC-Chimeroids (95,123 cells, n=10 replicates, 2 batches), SD-NSC-Chimeroids (48,119 cells, n=25 replicates, 2 batches), and NPC-Chimeroids (3,711 cells, n=2 replicates). Box: median with upper and lower quartiles; whiskers: highest/lowest values within 1.5 interquartile range (IQR) of the nearest hinge. d, Concordance between CFuPN expression signatures in MD-NSC Chimeroids vs Ref. Map organoids or endogenous fetal tissue, by Rank-Rank Hypergeometric Overlap (RR-HO) (also see Extended Data Fig. 10f). e, Average abundance of each cell type for each donor line, split by organoid protocol. Asterisks indicate cell types that have significantly altered abundances compared to the same donor within multi-donor NSC Chimeroids (edgeR two-sided quasi-likelihood F test on a negative binomial model) (0.05 > * > 0.005 > ** > 0.0005 > ***). f, Similarity in cell type composition between donor-matched replicates (Aitchison distance), within each protocol and between MD-NSC Chimeroids and each other protocol (n=58 replicates across 4 protocols and 12 batches). Boxes: median and upper and lower quartiles; whiskers: highest/lowest values within 1.5 IQR of the nearest hinge. Values and averages (dotted lines) for three different datasets of endogenous human fetal cortex^,, are shown for comparison (total, n=16 samples). g, Cell type annotations assigned to Chimeroid cells using two different reference datasets: Ref. Map organoids and endogenous human fetal cortex cells^,,. h, Distribution of scaled pseudotimes assigned to each major cell type within each organoid/chimeroid protocol. Below, schematic showing the “temporal” appearance of each cell type. i, RNA velocity calculated with velocyto and scVelo for MD- (left) and SD- (right) NSC Chimeroids.

Figure 4.. NSC-Chimeroids model treatment-specific alterations

a, Schematic of protocol. b, Integrated UMAPs colored by annotated cell type for control and treated Chimeroids. c, Bar plots of cell-type proportions in control MD-NSC Chimeroids in control, (grey, n=45 samples from 10 Chimeroids) and treated with EtOH (purple, n=31 samples from 7 Chimeroids) or VPA (yellow, n=49 samples from 11 Chimeroids). Solid bar: median, error bars: upper and lower quartiles. Significance test: one-sided F-test comparing negative binomial generalized linear models. FDR-adjusted p-values: * - <0.1; ** - <0.01; *** - <0.001. d. Changes in cell-type proportions in SD-NSC Chimeroids from the corresponding lines (n=25, 15, and 25 samples for Control, EtOH-treated, and VPA-treated, respectively), as in c. e, Left: UMAP showing the localization of cells from control MD-Chimeroids (grey) and cells from MD-Chimeroids treated with VPA (yellow) and EtOH (violet). Right: UMAP showing overlapping neighborhoods of cells, as calculated using Milo. Red and blue colors indicate neighborhoods with significant enrichment for EtOH and VPA-treated cells or control cells, respectively. Point size indicates the number of cells in a neighborhood, and edge thickness indicates the number of cells shared between pairs of neighborhoods. f-g, Beeswarm plot showing shifts in the composition of neighborhoods of cells in response to EtOH and VPA treatment, grouped by the cell-type identity of those neighborhoods. Each point represents a neighborhood of 50-200 cells with similar gene expression profiles. The vertical axis indicates the enrichment of EtOH and VPA-treated cells within a neighborhood: positive log2 fold change indicates more treated cells than expected, and negative values indicates fewer treated cells than expected. Neighborhoods are colored-coded by statistical significance of enrichment (calculated via Milo’s testNhoods function, using a Quasi-Likelihood F-test): grey, not significantly different from random; red, significant over-enrichment; blue, significant under-enrichment.

Figure 5.. NSC-Chimeroids reveal differential donor-specific susceptibility.

a, UMAPs of control, EtOH-, and VPA-treated Chimeroids split by donor and colored by cell type. b, Bar plot of normalized donor proportions in control NSC-Chimeroids (grey, n=10 replicates) and treated with EtOH (n=7) or VPA (n=11). The y-axis represents the ratio of the measured donor abundance in each Chimeroid to the expected abundance if all donors within a mix were equally represented. Solid bar: median; error bars: upper and lower quartiles. Significance test: one-sided F test comparing negative binomial generalized mixed effects models applied to the cell counts per donor. FDR-adjusted p-values: * - <0.1; ** - <0.01; *** - <0.001 (see Methods). c, Principal Component Analysis (PCA) performed on the gene expression matrix of different clusters from VPA-treated (triangles) and untreated (circles) NSC-Chimeroids, pseudobulked by donor (indicated by color) and replicate, both using all cells (top left, n=94) and subsetted by cell type (n=85, 70, 94, 88, 94 for oRG, aRG, CFuPN, IP, and CPN, respectively). Corresponding right panels: violin plots showing the pairwise Euclidean distance in PCA space between treated and untreated pseudobulked data points from each donor. Boxes: median and upper and lower quartiles; whiskers: highest/lowest values within 1.5 interquartile range (IQR) of the nearest hinge. Larger distances imply that VPA treatment induces larger changes in expression profiles. Significance test: Analysis of Variance (ANOVA) followed by two-sided Tukey post-hoc pairwise tests. d, Heatmap of donor-specific differentially expressed genes from VPA-treated MD-NSC Chimeroids. e, GSEA of donor-specific differentially expressed genes from VPA-treated MD-NSC Chimeroids, ranked on the corresponding SD-NSC datasets. P-values calculated via two-sided Kolmogorov-Smirnov test.