An early cell shape transition drives evolutionary expansion of the human forebrain

Abstract

The human brain has undergone rapid expansion since humans diverged from other great apes, but the mechanism of this human-specific enlargement is still unknown. Here, we use cerebral organoids derived from human, gorilla, and chimpanzee cells to study developmental mechanisms driving evolutionary brain expansion. We find that neuroepithelial differentiation is a protracted process in apes, involving a previously unrecognized transition state characterized by a change in cell shape. Furthermore, we show that human organoids are larger due to a delay in this transition, associated with differences in interkinetic nuclear migration and cell cycle length. Comparative RNA sequencing (RNA-seq) reveals differences in expression dynamics of cell morphogenesis factors, including ZEB2, a known epithelial-mesenchymal transition regulator. We show that ZEB2 promotes neuroepithelial transition, and its manipulation and downstream signaling leads to acquisition of nonhuman ape architecture in the human context and vice versa, establishing an important role for neuroepithelial cell shape in human brain expansion.

Keywords: ZEB2; brain; brain expansion; cell shape; chimpanzee; evolution; gorilla; neural stem cells; neuroepithelium; organoids.

Graphical abstract

Figure S1

Human and ape stem cells and organoids are highly comparable in terms of identity and morphology, related to Figure 1 A. RT-PCR shows expression of pluripotency markers (OCT3/4, NANOG, SOX2) and GAPDH loading control in IMR-90, G1 and G2 cell lines. –ve is the water negative control. B. 5-week organoids stained for neural progenitor marker SOX2 (red), dorsal telencephalic/intermediate progenitor marker TBR2 (gray), neuronal markers TUJ1 (human) and HuCD (gorilla, chimpanzee) in yellow, and DAPI (blue) showing human (H9) derived organoids become larger in overall size than gorilla (G1) and chimpanzee (Chmp) organoids. Scale bar: 1mm. C. Immunofluorescent staining of representative 5-week ape organoids for neural progenitor marker SOX2, intermediate progenitor marker TBR2 and DAPI show cortical lobules of similar thickness and relative proportions of cell types. Scale bar: 100 μm. D. Quantification of the area occupied by individual organoids (left) and the perimeter of visible neuroepithelial buds (right) from brightfield images of day 10 organoids generated from all cell lines tested: human (H9, IMR-90), chimpanzee (Chmp), and gorilla (G1, G2), revealing human organoids grow consistently larger than nonhuman ape organoids. Mean organoid area: IMR-90 = 175,708 μm², G2 = 106,589 μm²; H9, G1, Chmp as reported in Figure 1C. Mean neuroepithelial bud perimeter: IMR-90 = 307 μm, G2 = 211 μm; H9, G1, Chmp as reported in Figure 1C. ^∗p < 0.05, ^∗∗∗∗p < 0.0001, Kruskal-Wallis and post hoc Dunn’s multiple comparisons test, n (IMR-90) = 15 organoids and 105 neuroepithelial buds from 5 independent batches, n (G2) = 15 organoids and 94 neuroepithelial buds from 5 independent batches, n (H9, G1 and Chmp) = reported in Figure 1C. E. Representative brightfield images of pluripotent stem cells and organoids taken throughout the differentiation protocol. Note the more rounded neuroepithelium observed in day 5 organoids generated from nonhuman ape (Chmp, G1 and G2) cells versus day 5 organoids generated from human (H9, IMR-90) cells. Scale bar: 1 mm. F. Representative immunofluorescence images of the center of whole mount human (IMR-90) and gorilla (G2) day 5 organoids with staining for ZO1 and SOX2 showing polarized neural progenitor cells organized around rounded (gorilla) and more convoluted (human) ZO1 positive apical lumens. Scale bar: 100 μm. G. Quantification of the surface area of the largest apical lumen per day 5 organoid showing calculations for organoids derived from all cell lines tested. The data reveals the same trend of more expanded buds in human versus non-human ape. Mean luminal surface area: IMR-90 = 78,463 μm²; G2 = 27,426 μm²; H9, G1, Chmp as reported in Figure 1F.

Figure 1

Human telencephalic organoids are larger with extended apical lumens (A) Schematic of the timeline for generating brain organoids from human, gorilla, and chimpanzee stem cells. Colors represent changes in media and stages of the protocol. +MG represents Matrigel embedding. Note chimpanzee organoids have an EB stage that is 2 days shorter than human and gorilla. (B) Bright field images of human (H9, top panels), gorilla (G1, middle panels), and chimpanzee (Chmp, bottom panels) organoids at days 3, 5, and 10. Black arrows indicate neuroepithelial buds, which appear more elongated in the human beginning at day 5. Scale bar, 200 μm. (C) Quantification of bright field images at day 10 show human (H9) neural tissue is enlarged relative to gorilla (G1) and chimpanzee (Chmp). Left: area occupied by individual organoids. Right: visible perimeter of individual neuroepithelial buds. Mean organoid area: H9 = 168,327 μm²; G1 = 84,876 μm²; Chmp = 81,086 μm². Mean neuroepithelial bud perimeter: H9 = 322 μm; G1 = 237 μm; Chmp = 204 μm. ^∗p < 0.05, ^∗∗∗∗p < 0.0001, Kruskal-Wallis and post hoc Dunn’s multiple comparisons test, n (H9) = 22 organoids and 144 neuroepithelial buds from 6 independent batches, n (G1) = 23 organoids and 148 neuroepithelial buds from 5 independent batches, n (Chmp) = 13 organoids and 107 neuroepithelial buds from 3 independent batches, error bars are SD. (D and E) Representative immunofluorescence images of the center of whole mount human (H9), gorilla (G1), and chimpanzee (Chmp) organoids with staining for apical marker ZO1 and neural progenitor marker SOX2 at day 3 (D) and day 5 (E). Note the appearance of less rounded ZO1 positive apical lumens (arrowheads) in human organoids relative to nonhuman ape organoids at day 5. DAPI is in blue. Red background signal outside the organoid comes from nonspecific uneven staining of surrounding Matrigel. Scale bar, 100 μm. (F) 3D MATLAB reconstructions of apical lumens of day 5 organoids showing a representative example used for quantifications in (G). Values on the axes are in μm. Luminal surface area of reconstructed examples: human (H9) = 63,394 μm²; gorilla (G1) = 15,146 μm²; chimpanzee (Chmp) = 19,730 μm². (G) Quantification of the surface area of the largest apical lumen per day 5 organoid reveals significantly expanded luminal surface areas in human versus nonhuman apes. Mean luminal surface area: human (H9) = 51,243 μm²; gorilla (G1) = 15,437 μm²; chimpanzee (Chmp) = 19,632 μm². ^∗p < 0.05, ^∗∗p = 0.0012, ^∗∗∗∗p < 0.0001, one-way ANOVA and post hoc Tukey’s multiple comparisons test, n (H9 and G1) = 11 organoids from 5 independent batches, n (Chmp) = 8 organoids from 2 independent batches, error bars are SD. See also Figure S1.

Figure S2

Ape NE cells undergo cell shape transition before the onset of neurogenesis, related to Figure 2 A. Representative immunofluorescence images through the center of whole mount organoids derived from human (H9) and gorilla (G1) at days 5, 10 and 15 post-neural induction with staining for intermediate progenitor cells (TBR2), newborn neurons (DCX), ZO1 and DAPI, showing that neurogenesis has not started by day 5, is commencing at day 10 and is underway by day 15. Scale bar: 100 μm. B. Representative immunofluorescence images through whole mount organoids showing the morphology of neural progenitor cells revealed by sparse labeling with viral GFP with staining for ZO1, SOX2 and DAPI for human (H9) and gorilla (G1) organoids at days 8 and 10. Note in both species the thin, elongated tNE cell morphologies on day 8 and RG cell morphologies on day 10 with stereotypical elongated narrowed apical and basal processes. Scale bar: 20 μm.

Figure 2

Human NE cells exhibit species-specific differences in cell shape (A) Representative immunofluorescence images through whole mount organoids showing the morphology of neural progenitor cells (SOX2⁺), polarized around apical (ZO1⁺) lumens, revealed by sparse labeling with viral GFP in human (H9, IMR-90), gorilla (G1, G2), and chimpanzee (Chmp) organoids. Day 3 cells (top panels) are columnar and exhibit typical NE morphology. Day 5 (bottom panels) human cells still appear columnar, whereas gorilla and chimpanzee cells show a thinning of apical processes (arrows). DAPI is in blue. Red background signal outside the organoid comes from nonspecific uneven staining of surrounding Matrigel. Scale bar, 20 μm. (B) Immunofluorescent staining for ZO1 on the surface of apical lumens showing the apical surface areas of individual progenitor cells at day 5 in human (H9, IMR-90), gorilla (G1, G2), and chimpanzee (Chmp) organoids. Perimeters of some individual progenitor cells are delineated in white. Scale bar, 10 μm. (C) Quantification of the apical surface area of individual neural progenitor cells of day 5 organoids show significantly smaller apical surface sizes of gorilla (G1, G2) and chimpanzee (Chmp) progenitor cells compared to human (H9, IMR-90). Measurements were performed on delineated ZO1 cell perimeters as shown in (B). Mean apical surface area/cell: H9 = 9.39 μm², IMR-90 = 8.53 μm², G1 = 4.48 μm², G2 = 4.54 μm², and Chmp = 3.55 μm². ^∗∗∗∗p < 0.0001, Kruskal-Wallis and post hoc Dunn’s multiple comparisons test, n (H9) = 341 cells from 8 organoids from 2 independent batches, n (IMR-90) = 121 cells from 4 organoids of 1 batch, n (G1) = 321 cells from 9 organoids from 2 independent batches, n (G2) = 172 cells from 5 organoids of 1 batch, and n (Chmp) = 277 cells from 6 organoids and 2 independent batches. Boxplots show median with whiskers representing min-max values; data points on the boxplots represent individual cells. (D) Representative immunofluorescence images through whole mount human (H9) and gorilla (G1) organoids with superimposed individual segmented cells (white) showing the 3D morphology of individual GFP⁺ progenitors. Note the thinning of apical processes observed in gorilla at day 5 that becomes pronounced in both species by day 10. Scale bar, 10 μm. (E) Quantification of the volume normalized to surface area of the apical processes of human (H9) and gorilla (G1) neural progenitor cells on day 3 (left), day 5 (middle), and day 10 (right), showing significantly reduced apical volumes in gorilla cells relative to human at day 5. The apical processes of segmented cells directly below the cell body were used for quantification. Mean apical process volume:surface area ratio: human day 3 = 1.21, gorilla day 3 = 1.31, human day 5 = 1.23, gorilla day 5 = 0.76, human day 10 = 0.66, and gorilla day 10 = 0.65. Mann-Whitney U, ^∗∗∗∗p < 0.0001, two-tailed, n (day 3 human) = 10 cells, n (day 3 gorilla) = 8 cells, n (day 5 human) = 14 cells, n (day 5 gorilla) = 16 cells, n (day 10 human) = 8 cells, and n (day 10 gorilla) = 8 cells. Error bars are SD. See also Figure S2.

Figure 3

Delayed human NE transition is associated with a shorter cell cycle (A) Immunofluorescent staining for ZO1 on the surface of apical lumens of human (H9) and gorilla (G1) organoids revealing apical surface areas of individual neural progenitor cells at days 3, 5, 8, and 10. Perimeters of some individual cells of day 5 organoids are delineated in white. Scale bar, 10 μm. (B) Quantification of the surface area of individual human (H9) and gorilla (G1) NPCs between day 3 and 10 showing a gradual reduction in apical surface area over time in both species by 7-fold. Note gorilla cells are more constricted than human during the transitioning period from days 5 to 8. Measurements were performed on delineated ZO1 cell perimeters as demonstrated in (A). Mean apical surface area/cell: human day 3 = 13.82 μm², gorilla day 3 = 14.80 μm², human day 8 = 3.86 μm², gorilla day 8 = 2.46 μm², human day 10 = 1.92 μm², gorilla day 10 = 2.13 μm², and human and gorilla day 5 are reported in Figure 2C. ^∗∗∗∗p < 0.0001, per time point Mann-Whitney U, two-tailed, n (day 3 human) = 164 cells from 8 organoids and 2 batches, n (day 3 gorilla) = 176 cells from 8 organoids and 2 batches, n (day 8 human) = 171 cells from 4 organoids of 1 batch, n (day 8 gorilla) = 55 cells from 2 organoids of 1 batch, n (day 10 human) = 68 cells from 3 organoids of 1 batch, n (day 10 gorilla) = 74 cells from 4 organoids of 1 batch, and n (day 5) = reported in Figure 2C. (C) Representative immunofluorescence images showing the position of nuclei (DAPI, blue) of neural progenitor cells relative to the apical surface (ZO1, red) in neuroepithelial buds of day 5 human (H9) and gorilla (G1) organoids. Dashed lines in white and yellow represent apical and basal surfaces of the neuroepithelial bud respectively. Some nuclear distances from the apical surface are shown as dashed lines in cyan, and as quantified in (D) and (E). Note the more basal distribution of nuclei in gorilla tissue. Scale bar, 10 μm. (D and E) Quantification of the position of nuclei of human (H9) and gorilla (G1) neural progenitor cells relative to the apical surface as a percentage of apicobasal neuroepithelial (NE) thickness. Measurements were performed on images as shown in (C) (D) Quantification of positions of individual cells showing gorilla progenitors located more basally. (E) Mean nuclear position per neuroepithelial bud showing gorilla cells located significantly more basal. Mean nuclear position: human = 59.34%, gorilla = 67.14%. ^∗∗∗∗p < 0.0001, Student’s t test, unpaired, two-tailed, n (human) = 335 nuclei from 9 organoids and 3 independent batches, and n (gorilla) = 307 nuclei from 9 organoids and 3 independent batches. Error bars are SD. (F) Still frames of live imaging of neural progenitor cells sparsely labeled with GFP (grayscale) in human (H9) (Video S2) and gorilla (G1) (Video S3) organoids covering an entire cell cycle between day 3 and 4.5. Note the presence of neural progenitor cells with columnar NE morphology in both species prior to the first division (00:00, reference time point) and loss of basal process (white arrowheads) during cell mitosis (blue arrowheads). Note daughter cells (yellow arrowheads) show thicker apical processes (magenta arrowheads) in human relative to the more constricted morphology of gorilla apical processes. Time shown in hours:minutes. (G) Quantification of cell cycle duration of human (H9) and gorilla (G1) neural progenitor cells imaged between days 3 and 5 showing significantly longer cell cycles in gorilla. Mean cell cycle length: human = 18.83 h, gorilla = 22.10. ^∗∗∗∗p < 0.0001, Mann-Whitney U, two-tailed, n (human) = 30 cells from 5 independent batches, n (gorilla) = 29 cells from 5 independent batches. Error bars are SD. (H) Growth curve modeling of the predicted effect that differences in cell cycle length, as measured in (G), would have on human and gorilla progenitor numbers (bold lines) and on neuron numbers (dashed lines) with a 1.9-fold expected increase in human for both. (I) Schematic summarizing the morphological changes in neural progenitor cells. Progenitor cells of apes undergo a gradual transition from NE to tNE to neurogenic RG cells. Nonhuman ape cells show tNE morphologies (blue background) earlier than human cells, which show shorter cell cycles leading to increased neuron numbers. See also Figure S3 and Videos S1, S4, and S5.

Figure S3

Live imaging of human and gorilla cerebral organoids, related to Figure 3 Representative still frames of live imaging of neural progenitor cells sparsely labeled with GFP (greyscale) in human (H9) (Video S4) and gorilla (G1) (Video S5) organoids covering an entire cell cycle between day 4 and 5, showing a significantly longer cell cycle in gorilla, calculated as the period between cell divisions (blue arrowheads). Parent and daughter cells (yellow arrowheads) show more constricted apical processes (magenta arrowheads) in gorilla relative to the apical process of human progenitors which thickens into a NE morphology in the frames following the first division. Note the loss of the basal process (white arrowhead) during mitosis. Time shown in hours:minutes, 00:00 marks the first division and reference time point

Figure 4

Ape organoids exhibit comparable developmental molecular trajectories (A) Schematic of the timeline for human and gorilla brain organoids with RNA-seq collection time points shown in red. 3 biological replicates of organoids derived from human (H9) and gorilla (G1) cells were collected at each of the 7 time points. (B) PCA biplot of PC1 versus PC2 performed on Z-scaled log2-transformed TPMs of the 3,000 most variable genes. Samples are color-coded by time point and species. Note samples separating primarily by time point with a slight separation between samples of different species at days 5, 10, and 25, highlighted by ellipses (black ellipses for human and fuchsia for gorilla). (C) Heatmap with hierarchical clustering based on shared expression pattern (Z scores of log2-transformed TPMs of the 3,000 most variable genes) between samples. The dendrogram shows samples clustering by time point. (D) Pearson’s correlation map using Z-scaled log2-transformed TPMs of all genes. Darker blue depicts stronger positive correlation between samples and darker red a stronger negative correlation. Red boxes highlight the correlation between species at matched time points. Note lower correlation between species at days 5 and 10. Pearson’s correlation coefficient: r = 0.66 (day 0), 0.63 (day 2), 0.66 (day 3), 0.52 (day 5), 0.40 (day 10), 0.62 (day 15), and 0.66 (day 25). (E) Temporal expression pattern (Z-scaled) of characteristic developmental markers show predicted expression dynamics. Non-neural ectoderm markers (GATA3, CDH1) are lost rapidly, followed by a gain in neural progenitor markers (CDH2, VIM), and a later increase in intermediate progenitor (TBR2) and early-born neuron (TBR1) markers. Shaded error bar is SD. (F) GO term enrichment analysis on the top 300 genes driving species variance at days 5, 10, and 25. Shown are the 8 most significant (p < 0.05) enrichments for GO categories molecular function (GO:MF), cellular compartment (GO:CC), and biological process (GO:BP). See also Figure S4 and Data S1.

Figure S4

RNA-seq data analysis pipeline and normalization, related to Figure 4 A. Workflow summarizing the RNaseq analysis pipeline (see methods). B. PCA performed using log2-transformed TPMs. Samples are color-coded by time point and species. Graph of PC1 versus PC2 (left) shows biological replicates grouping together, samples separating by species along PC1 and separating by time point along PC2. Plotting PC2 versus PC3 (right) shows samples separating by time point and not species. C. Pearson’s correlation of all samples using z-scaled log2-transformed TPMs of all genes. Darker blue depicts stronger positive correlation between samples and darker red a stronger negative correlation. Data shows the strongest correlation between biological replicates within a species followed by between species time point-matched samples. Correlation between species is lowest at day 5 and 10. D. Temporal expression pattern (z-scaled) of genes related to synaptic formation and maturation (GRIA2, GRIK2, SNAP25, SNAP91). Shaded error bar is SD.

Figure 5

The human neuroepithelium exhibits differential temporal dynamics of morphogenesis genes (A) Clustering genes by temporal expression dynamics shows species differences in GO:BP term enrichment. Columns from left to right: far left, TCseq clusters with genes in each cluster plotted with their temporal expression (z-scaled) and color-coded by membership value (degree to which data points of a gene belong to the cluster, pink represents high membership values). The 10 clusters are ordered (top to bottom) based on similarity in expression pattern. Middle left: representative GO:BP term from shared (purple), human-exclusive (black), or gorilla-exclusive (fuchsia) terms for each cluster. Middle right: histograms show the number of enriched (p < 0.05) GO:BP terms found in both species (purple), exclusively in human (black) or gorilla (fuchsia) per cluster. Axis range: 0–8 (cluster 2,5,7,8); 0–15 (cluster 9,10); 0–20 (cluster 1); 0–25 (cluster 3); 0–50 (cluster 4); 0–80 (cluster 6). Far right: weighted arc network graph visualizing interspecies differences in the enrichment/membership of specific GO:BP terms per cluster. The bases of the arc are aligned to both a human (black) and a gorilla (fuchsia) bar from the histogram in the adjacent panel, highlighting the species-specific shifts in expression patterns associated with specific GO:BP terms. Weight/thickness of the arc is dictated by the number of GO:BP terms enriched in a species-exclusive manner “moving” between clusters in the defined pattern. (B) Mean temporal expression pattern (z-scaled) of genes in clusters enriched for “cell morphogenesis”-related GO:BP terms (human clusters 1, 9, 10, and gorilla cluster 3). (C) Temporal expression pattern (Z-scaled) of SHROOM3, a gene involved in cell morphogenesis and apical constriction. (D) Immunofluorescent staining of day 5 organoids for SHROOM3 showing strong apical expression in gorilla (G1) neuroepithelium, but not human (H9) at this time point. Scale bar, 40 μm. (E) Immunofluorescent staining of day 5 organoids for OCLN showing expression spread along the apicobasal length of human (IMR-90) progenitor cells (white arrowheads) but more limited apically (yellow arrowheads) in gorilla (G1) progenitor cells. DAPI is shown in blue. Scale bar, 100 μm. (F) Venn diagram summarizing search for cell morphogenesis-related transcription factors with species-specific expression patterns. (G) Mean temporal expression pattern (Z-scaled) of ZEB2, showing peak expression earlier in gorilla (G1) than human (H9) organoids. Shaded error bars are SD. See also Figure S5 and Data S1.

Figure S5

Expression patterns of key factors with differential temporal dynamics, related to Figure 5 A, B. Mean temporal expression pattern (z-scaled) of genes in clusters enriched for: A. ‘ribonucleoprotein complex assembly’-related GO:BP terms (human cluster 4, gorilla cluster 6) B. ‘cell cycle’-related GO:BP terms (human cluster 8, gorilla cluster 4). C. Immunofluorescent staining of day 10 organoids for SHROOM3 shows strong apical expression in neuroepithelium of both species (H9, G1). Scale bar: 40 μm. D, E. Immunofluorescent staining for F-actin (Phalloidin) and DAPI of human (H9) and gorilla (G1) organoids on day 5 (D) and day 10 (E) showing weaker apical accumulation in human versus gorilla at day 5 with strong apical accumulation observed in both species by day 10. Scale bar: 40 μm. F. Mean temporal expression (z-scaled) of EMT-related genes enriched for WP term ‘Epithelial to mesenchymal transition in colorectal cancer’ MAPK12, MEF2D, PIK3R1 and MAPK11, robustly changing pattern between species in the same way enrichment of ‘cell morphogenesis’-related terms changes pattern between species (Figure 5B). G, H. Immunofluorescent staining of organoids (H9, G1) for OCLN at G. day 3, showing expression along the apicobasal length of progenitor cells in both species H. day 10, showing lowered expression limited apically in both species. Scale bar: 40 μm.

Figure 6

Decreased ZEB2 leads to expanded NE with delayed transition (A) Mean temporal expression pattern (log normalized transcripts per million) of ZEB2 across the entire time series, showing peak expression earlier in gorilla (G1) than human (H9) organoids. Shaded error bars are SD. (B) Western blot expression time course from PSCs to day 25 human (H9) and gorilla (G1) organoids reveals a premature onset and higher levels of ZEB2 protein expression in gorilla compared to human. This is accompanied by a premature expression of the radial glial marker vimentin, and premature downregulation of the epithelial markers CDH1 and EpCAM in gorilla relative to human. Bottom panel shows quantification of ZEB2 relative to GAPDH (AU, arbitrary units). (C) Immunofluorescent stain for ZEB2 and DAPI in human (H9) and gorilla (G1) organoids at days 3, 5, and 10 showing neuroepithelial buds with nuclear expression (yellow arrows). Note the interspecies difference at day 3 where gorilla organoids already display nuclear expression compared to a weaker stain in most human cells. Insets show higher magnification of the boxed regions. Scale bar, 40 μm. (D) Immunofluorescence image of a day 25 human (H9) organoid showing a mutually exclusively pattern of expression between ZEB2 and the committed radial glia marker, BLBP. Scale bar, 100 μm. (E) Western blot of H9 wild-type (WT) and ZEB2^+/− organoids at day 16 for ZEB2, the tight-junction protein OCLN, the junction components CDH1 and CDH2, the dorsal telencephalic marker EMX1, and the loading control β-actin. The blots show a sizeable increase in CDH1 and OCLN and a decrease in CDH2, whereas EMX1 levels, and thus dorsal telencephalic identity, appears to be largely unaffected. (F) Representative bright field images of day 12 WT and ZEB2^+/−. Insets show higher magnification of the boxed regions, dashed yellow are representative neuroepithelial bud perimeters quantified in (G), dashed turquois are representative neuroepithelial bud thicknesses quantified in (H). Scale bar, 500 μm. (G) Quantification of neuroepithelial bud perimeters of WT (n = 106) and ZEB2^+/− (n = 116) organoid buds from 27 WT and 28 ZEB2^+/− organoids at day 17, Mann-Whitney U test, two-tailed (^∗∗∗p = 0.0001) from 3 organoid batches. (H) Quantification of neuroepithelial bud thickness of WT (n = 80) and ZEB2^+/− (n = 119) organoid buds from 26 WT and 31 ZEB2^+/− organoids at day 12, Mann-Whitney U test, two-tailed (^∗∗∗∗p < 0.0001) from 3 organoid batches. (I) Quantification of neuroepithelial bud perimeters of two ZEB2^+/−;iZEB2 colonies treated with and without doxycycline. Colony 1: − Dox (n = 108 buds from 25 organoids), + Dox (n = 70 buds from 18 organoids). Colony 2: − Dox (n = 62 buds from 17 organoids), + Dox (n = 84 buds from 19 organoids) across 3 organoid batches. Mann-Whitney U tests, two-tailed (^∗∗∗p = 0.0003 ^∗∗∗∗p < 0.0001). (J) Immunofluorescence images of day 15 WT and ZEB2^+/− organoids showing increased OCLN immunostaining (yellow arrowheads) along the apico-(ZO1) basal (dashed line) axis of progenitor cells and reduced numbers of TBR2⁺ cells in ZEB2^+/− organoids compared to WT. Scale bar, 100 μm. See also Figure S6.

Figure S6

ZEB2 expression and targeting for loss of function, related to Figure 6 A. Representative immunofluorescence image showing ZEB2 expression in SOX2+ progenitor cells in day 10 human (H9) organoid. Scale bar: 50 μm. B. Representative immunofluorescence image showing a salt-and-pepper pattern of ZEB2 expression in the ventricular zone at day 25, after the onset of neurogenesis in human (H9) organoid. DCX (Doublecortin) stains newly born neurons. Scale bar: 100 μm. C. Representative immunofluorescence image of a mature day 60 human (H9) organoid revealing ZEB2 expression in CTIP2+ neurons and absence of ZEB2 staining in the ventricular zone. Scale bar: 100 μm. D. Representative immunofluorescence image of a day 25 human (H9) organoid showing a mutually exclusive pattern of expression between ZEB2 and the radial glia marker protein GLAST. Scale bar: 100 μm. E. Schematic representation of the CRISPR-Cas9n editing strategy, where the first coding exon of the ZEB2 gene (exon 2, NCBI ref sequence NM_014795.4:182-323) was targeted by two nickases (dashed lines) and screening was performed by assaying the drop-off frequency of a HEX-labeled probe, binding to one of the nick sites, relative to a FAM-labeled reference probe binding away from the disrupted region. The exon is marked in orange while introns are marked in purple. F. Example ddPCR 2D scatter-plots of a negative control sample (HEK293 cells), showing only a FAM-HEX double positive (red) and an empty droplet cluster (black) and a positive control sample (HEK293 cells expressing WT Cas9 and ZEB2 guides), showing a FAM-only cluster (blue) in the upper-left quadrant of the 2D plot corresponding to edited alleles. ddPCR 2D scatter-plot of the H9 ZEB2^+/− hESC edited line showing a 1:1 ratio between the WT and edited allele. (G). Representative chromatograms of the ZEB2 alleles in the H9 ZEB2^+/− hESC cells. The CRISPR-Cas9 target region was PCR amplified with a high-fidelity polymerase, the PCR product was blunt-end cloned into the pJET1.2 vector and following purification, plasmids from different colonies carrying the insert were sequenced. Sequencing reveals that the edited allele harbors a 23 bp deletion. H. DNA-PAGE analysis of a short PCR amplicon spanning the CRISPR-Cas9 ZEB2 target site in WT H9 and H9 ZEB2^+/− hESCs. The gel reveals the presence of two bands, corresponding to the WT and the edited allele in H9 ZEB2^+/− hESCs. I. Representative images of karyotype analysis on 20 G-banded metaphase spreads from the H9 ZEB2^+/− hESCs used to generate the stock. The cell line displays normal karyotype. J. RT-PCR analysis for expression of ZEB2, the key pluripotency markers SOX2, NANOG, OCT4 and DPPA5 and the loading control GAPDH. PCR shows that upon a ~50% reduction in ZEB2 mRNA levels the mutant stem cells retain expression of pluripotency markers at comparable levels to WT H9 hESCs. WT and ZEB2^+/− were run on the same gel but not adjacent to each other, the dashed line indicates where the gel was spliced. K. Full length western blot for ZEB2 in WT and ZEB2^+/− organoids at day 15 – loading control was GAPDH L. Box and whiskers plot reporting the quantifications of the number of TBR2+ cells per unit area (TBR2+ cells/mm²) in day 16 WT and ZEB2^+/− organoids. Quantifications were performed by manual counting on n = 52 WT and n = 68 ZEB2^+/− ventricles corresponding to 12 organoids from 2 distinct batches. A two-tailed Mann-Whitney U test was used for statistical comparison (^∗∗∗∗p < 0.0001). M. Representative immunofluorescence images of day 55 WT and ZEB2^+/− cerebral organoid buds used for quantifications shown in N. Scale bar: 200 μm. N. Box and whiskers plot reporting the quantifications of the number of TBR2+ cells per unit area (TBR2+ cells/mm²) in day 55 WT and ZEB2^+/− organoids. Quantifications were performed using an automated cell segmentation pipeline on n = 17 WT and n = 17 ZEB2^+/− organoid regions from 3 distinct batches. A two-tailed Mann-Whitney U test was used for statistical comparison (ns, p = 0.1139). O. Plasmid maps of the CRISPR homology-directed repair (HDR) templates used to target the AAVS1 safe-harbor locus in H9 hESC cells – top is the CAG-lox-STOP-lox-ZEB2-GFP-Flag inducible expression construct and bottom is the construct encoding CRE recombinase under the control of a tetracycline responsive promoter and the reverse tetracycline transactivator (rtTA) driven by the CAG promoter. P. UCSC Genome Browser view of the AAVS1 locus and CRISPR-Cas9 targeting strategy of intron 1 of PPP1R12C. The promoter-less splice-acceptor (SA), T2A peptide-linked “gene trap” is such that expression of the promoter-less selection cassette is driven by the endogenous PPP1R12C gene, thus effectively eliminating false-positive background arising from random integration. The panel reports the PCR genotyping strategy – upon successful targeting of the AAVS1 locus, while amplicon 1 is lost due to the size increase following insert integration, amplicons 2 and 3 are gained - see Figure S7A. Q. PCR gel showing successful genotyping of the two rescue clones used for the experiments shown. R. Representative brightfield images of day 15 ZEB2^+/−; iZEB2 cerebral organoids treated with and without doxycycline. Scale bar: 100 μm. S. Representative immunofluorescence images of ZEB2^+/−; iZEB2 treated with and without doxycycline stained for GFP, TBR2 and DAPI. Scale bar: 100 μm T. Box and whiskers plot reporting the quantifications done using an automated cell segmentation pipeline of the number of TBR2+ cells per unit area (TBR2+ cells/mm²) in day 15 ZEB2^+/−; iZEB2 organoids - colony 1: -Dox (n = 17 organoid regions), +Dox (n = 16 organoid regions); colony 2: -Dox (n = 13 organoid regions), +Dox (n = 13 organoid regions) from three independent batches. Mann-Whitney U tests, two-tailed (^∗∗p < 0.01).

Figure S7

Modulation of ZEB2 and SMAD signaling in human and gorilla cells, related to Figure 7 A. PCR gel showing successful genotyping of the Hum^iZEB2 colony used for all experiments shown based on the PCR genotyping strategy outlined in Figure S6P. The asterisks mark unspecific bands. B. Transgene induction in Hum^iZEB2 cells treated with and without doxycycline for 6 days and assayed by western blot for ZEB2, GFP and β-actin. C. Immunofluorescence images of 6-day induced and uninduced Hum^iZEB2 cells stained for ZEB2 and DAPI, showing that doxycyline induction results in ZEB2 expression and nuclear translocation, without adverse effects on its localization due to tagging with GFP. Scale bar: 20 μm D. Immunofluorescence images of 6-day induced and uninduced Hum^iZEB2 cells stained for DAPI, CDH1 and CDH2. The data reveal a reduction in CDH1 expression and an increase in CDH2 expression following induction. Scale bar: 50 μm. E. Immunofluorescence images of 6-day induced and uninduced Hum^iZEB2 cells stained for GFP, Vimentin and EpCAM. The data reveal a reduction in EpCAM expression and an increase in Vimentin expression following expression of ZEB2-GFP. Scale bar: 50 μm. F. Brightfield images of induced (+ Dox) and uninduced (- Dox) Hum^iZEB2 organoids and gorilla (G1) organoids at days 3, 5 and 10, showing indistinguishable tissue architecture between organoids at day 3, while day 5 and 10 organoids show neuroepithelial buds that are smaller and more rounded in gorilla and ZEB2 induced (+ Dox) versus uninduced (- Dox) organoids. Scale bar: 200 μm. G. Western blot of day 5 WT and ZEB2^+/− organoids revealing both decreased ZEB2 and SHROOM3 levels in ZEB2^+/− organoids compared to WT control β-Actin was used as loading control. H. Representative bright-field images of uninduced and induced Hum^iZEB2 and gorilla neuroepithelial buds at day 5, used for quantification in Figure 7H. Scale bar: 100 μm. I. Immunofluorescent staining for ZO1, SOX2, DCX and DAPI showing normal tissue morphology and onset of neurogenesis (DCX+ neurons) in uninduced (- Dox) and induced (+ Dox) Hum^iZEB2 organoids at day 10. Scale bar: 100 μm. J. Western blot for ZEB2, CDH1, and CDH2, with β-Actin as loading control, of WT and ZEB2^+/− organoids treated with dual-SMAD inhibitors, or treated with vehicle, for 10 days and assayed at day 12. Note the rescued levels of junctional components CDH1 and CDH2. K. Morphological assessment of WT and ZEB2^+/− organoids treated with dual-SMAD inhibitors, or treated with vehicle, for 10 days and assayed at day 12 by brightfield imaging (left panels) and hematoxylin-eosin staining (right panels). Note the elongated neuroepithelial buds (arrows) in mutant organoids that appear rescued (arrowheads) upon SMAD inhibition. Scale bars: 1 mm (left panels), 50 μm (right panels).

Figure 7

ZEB2-driven junctional remodeling and apical constriction dictate species-specific timing of NE transition (A) Immunofluorescent staining of uninduced (− Dox) and induced (+ Dox) Hum^iZEB2 organoids for GFP and SHROOM3. Note the expression of ZEB2-GFP and apical accumulation of SHROOM3 in induced organoids. Scale bar, 50 μm. (B) Representative bright field images of day 5 Hum^iZEB2 and gorilla organoids. Induced (+ Dox) Hum^iZEB2 organoids show smaller neuroepithelial buds (arrowheads) that are more round in shape, similar to gorilla (G1), while uninduced (− Dox) show more elongated structures typical of human. Scale bar, 200 μm. (C) Immunofluorescence images through day 5 whole mount Hum^iZEB2 uninduced (− Dox), induced (+ Dox) and gorilla (G1) organoids stained for GFP, ZO1, and SOX2. Sparse labeling with viral GFP shows ZEB2 induction triggers the constriction of apical processes (arrows) in progenitor cells, similar to gorilla at day 5. Scale bar, 50 μm. (D) Representative immunofluorescence images through whole mount day 5 uninduced (− Dox), induced (+ Dox) Hum^iZEB2 and gorilla (G1) organoids with superimposed individual segmented GFP+ progenitor cells (white) showing their 3D morphology. Note the thinning of apical processes observed upon ZEB2 induction. Scale bar, 10 μm. (E) Immunofluorescent staining for ZO1 on the surface of apical lumens showing the apical surface areas of individual progenitor cells in day 5 Hum^iZEB2 uninduced (− Dox), induced (+ Dox) and gorilla (G1) organoids. Perimeters of some individual progenitor cells of day 5 organoids are delineated in white. Scale bar, 10 μm. (F) Quantification of the volume as normalized to surface area of the apical processes of induced (+ Dox) versus uninduced (− Dox) Hum^iZEB2 neural progenitor cells on day 5. The apical processes of segmented cells directly below the cell body were used for quantification. Gorilla day 5 measurements from Figure 2E are included for comparison. Mean apical process volume:surface area ratio: Hum^iZEB2 − Dox = 1.11; Hum^iZEB2 + Dox = 0.76. ^∗∗p < 0.01, Mann-Whitney U, two-tailed, n (− Dox and + Dox) = 9 cells. Error bars are SD. (G) Quantification of the surface area of individual delineated ZO1 cell perimeters as shown in (E). Gorilla measurements from two cell lines (G1, G2) combined are shown for comparison. Mean apical surface area/cell: Hum^iZEB2 − Dox = 9.62 μm², Hum^iZEB2 + Dox = 3.08 μm², and gorilla (G1,G2) = 4.50 μm². ^∗∗∗∗p < 0.0001, Mann-Whitney U, two-tailed, n (− Dox) = 180 cells from 8 organoids, n (+ Dox) = 199 cells from 8 organoids, both from 2 independent batches, box and whisker plots show median with min-max values, data points represent individual cells. (H) Quantification of perimeters of neuroepithelial buds from bright field images at days 5 and 10, and overall organoid size at day 10. Gorilla (G1,G2) measurements were combined and included for comparison. Day 5 mean neuroepithelial bud perimeter: Hum^iZEB2 − Dox = 272 μm, Hum^iZEB2 + Dox = 237 μm, and gorilla (G1,G2) = 232 μm. ^∗∗∗∗p < 0.0001, Mann-Whitney U, two-tailed, n (− Dox) = 142 neuroepithelial buds from 41 organoids from 3 independent batches; n (+Dox) = 195 neuroepithelial buds from 38 organoids from 3 independent batches; n (G1,G2) = 555 neuroepithelial buds from 114 organoids from 16 independent batches. Day 10 mean neuroepithelial bud perimeter: Hum^iZEB2 − Dox = 300 μm, Hum^iZEB2 + Dox = 198 μm, and gorilla (G1,G2) = 227 μm. Day 10 mean organoid area: Hum^iZEB2 − Dox = 201,434 μm², Hum^iZEB2 + Dox = 132,325 μm², and gorilla (G1,G2) = 93,447 μm². ^∗∗∗∗p < 0.0001, Mann-Whitney U, two-tailed, n (− Dox day 10) = 15 organoids and 106 neuroepithelial buds from 3 independent batches, n (+ Dox day 10) = 15 organoids and 149 neuroepithelial buds from 3 independent batches, error bars are SD. (I) Representative immunofluorescence images showing the effect of BMP4 on the morphology of neural progenitor cells revealed by sparse viral labeling with GFP on day 5 untreated (− BMP4) and treated (+ BMP4) gorilla (G1) organoids with staining for GFP, SOX2, and DAPI. Arrows indicate the apical process. Scale bar, 40 μm. (J) Immunofluorescent staining for ZO1 showing apical surface areas of individual progenitor cells from BMP4-treated (+ BMP4) and untreated (− BMP4) gorilla (G1) organoids at day 5. Perimeters of some individual progenitor cells are delineated in white. Scale bar, 10 μm. (K) Quantification of individual delineated ZO1 cell perimeters as shown in (J). Mean apical surface area/cell: gorilla − BMP4 = 2.72 μm² and gorilla + BMP4 = 4.13 μm². ^∗∗∗∗p < 0.0001, Mann-Whitney U, two-tailed, n (− BMP4) = 301 cells from 8 organoids from 2 independent batches, n (+ BMP4) = 326 cells from 8 organoids from 2 independent batches, box and whisker plots are median with min-max values, data points represent individual cells. (L) Immunofluorescence images of human (H9) and gorilla (G1) day 5 organoids untreated (− LPA) and treated (+ LPA) with staining for OCLN, ZO1, and DAPI. LPA treatment in gorilla results in increased OCLN distribution along the apicobasal axis of cells (arrowheads) and expanded apical surfaces of cells (ZO1, bottom panel). Scale bar, 40 μm (upper panels), 10 μm (bottom panels). (M) Quantification of individual delineated ZO1 cell perimeters as shown in (L). Mean apical surface area/cell: human − LPA = 5.36 μm², human + LPA = 5.25 μm², gorilla − LPA = 2.44 μm², and gorilla + LPA = 4.73 μm². ^∗p < 0.05 ^∗∗∗∗p < 0.0001, Kruskal-Wallis and post hoc Dunn’s multiple comparisons test, n (human − LPA) = 146 cells from 3 organoids of 1 batch, n (human + LPA) = 200 cells from 3 organoids of 1 batch, n (gorilla − LPA) = 375 cells from 7 organoids and 2 independent batches, n (gorilla + LPA) = 457 cells from 10 organoids and 2 independent batches, box and whisker plots show median with min-max values, data points represent individual cells. (N) Schematic summarizing the morphological changes that occur in neural progenitor cells as they transition from NE to tNE cells (purple background). ZEB2 is highlighted as a driver, which acts through BMP-responsive SMADs to downregulate epithelial features, notably tight-junction proteins (TJs, green), and involves apical constriction through rearrangements in the actin cytoskeleton (actin, magenta). See also Figure S7.