Results
Generation of 3D brain organoids with primordial eye fields
We previously described a protocol of inducing differentiation into neural epithelium directly from iPSCs (Gabriel and Gopalakrishnan, 2017; Gabriel et al., 2016, 2017; Ramani et al., 2020). These brain organoids expressed retina- and eye-related genes (data not shown) but did not develop into visible optic vesicles. We therefore modified our protocol, starting with a low-density cell number (1 × 104 iPSCs as starting number for one organoid). Our intention was not to force development of purely neural cell types at the earliest stages of organoid differentiation. Therefore, we provided retinol acetate ranging from 0–120 nM to the culture medium at a time point when the neuroectoderm expands (for details, see STAR Methods). Retinoic acid is critical for early eye development because it is a paracrine inhibitor of the mesenchyme around the optic cup (Rosen and Mahabadi, 2020; Cvekl and Wang, 2009; Janesick et al., 2015). Addition of 60 nM of retinol acetate reproducibly induced formation of pigmented structures, possibly primordial eye fields at around day 30 (Figure 1B). Interestingly, a region proximal to the pigmented area often displayed an invagination, suggesting that the organoids attempted to assemble optic vesicle-like structures. Organoids differentiated from iPSCs reprogrammed from adult retinal Müller glial cells generated strongly pigmented structures within 20 days (Slembrouck-Brec et al., 2019; Figure S1A). Pigmented areas restricted to one pole of the organoid suggested the presence of the forebrain-like region where the primordial eye field develops.
Testing for the presence of eye field patterning markers in these organoids, our immunostaining revealed the presence of RAX, Pax6, and FOXG1-positive progenitor cells in this region (Figures 1C and 1D; Furukawa et al., 1997; Mathers et al., 1997; Stigloher et al., 2006). FOXG1 is initially expressed in the prosencephalic neuroepithelium and later involves the telencephalon and eye field segregation (Stigloher et al., 2006). We also observed prominent SOX2-positive invaginating regions that displayed VSX2 and gradients of FOXG1, suggesting that the optic field in these organoids is segregated from the forebrain region (Figures 1E–1G).
To investigate cell diversity, we performed single-cell RNA sequencing (scRNA-seq) of 30-day old organoids and analyzed the transcriptomes of 3,511 single cells. Embedding the cells in a uniform manifold approximation and projection (UMAP) revealed the presence of a group of cycling progenitor cells developing toward multiple neuronal cell fates. Cluster analysis revealed eight main transcriptionally distinct cell populations (clusters C1–C8) (Figures 2A–2C; S1B; Table S1). The organoids contained cell clusters of neural progenitors, progenitors of the ventral forebrain mixed with the anterior neural tube, dorsal cortical progenitors, and dorsal cortical neurons (C1–C4). We also identified cell clusters of developing optic vesicles (C6) and the pre-optic area (C7), which were segregated from neuronal cell types (Figures 2A–2C). Clusters C1–C5 contain radial glial (as defined by SOX2 expression) and ventral forebrain progenitors (as defined by NKX6-1 and SP8), dorsal cortical progenitors (as defined by EMX2 and HES1), dorsal cortical neurons (as defined by DCX, NCAM1, MAP2, LHX2, and LHX9) and cycling cells (as defined by AURKB), which occupied a significant proportion of total cells (Rétaux et al., 1999; Peukert et al., 2011; Chou and Tole, 2019; Camp et al., 2015; Kanton et al., 2019; Figures 2A, 2B, and S1B).
Figure 2. scRNA-seq of 30-day-old brain organoids containing primitive eye fields
(A) UMAP plot showing cells from 30-day-old organoids clustered in eight groups based on transcriptome similarity.
(B) UMAPs showing marker genes used for annotation of the clusters.
(C) Heatmap showing expression of the top 20 markers calculated by Student’s t test for each cluster. See Table S1 for the complete list of markers.
(D) Top panel: pseudotime trajectory showing radial glia/stem cells as root (blue cells) and all other cells colored according to the computed pseudotime (green to yellow). Bottom panel: the heatmap shows expression of key genes identified by Sridhar et al. (2020) for RGC formation in the early retina. The Heatmap also shows expression of such genes in cells belonging to C1, C7, and C6, sorting them according to the pseudotime. Plotted expression values are convolved in each row by a window of 20.
Interestingly, C2 constitutes cells expressing forebrain progenitors such as OTX2, PAX6, EMX2, SOX2, and LHX2. Pax6 and Emx2 are homeodomain transcription factors (Wurst and Bally-Cuif, 2001). This cluster also contains NKX6-1 and SP8, ventral forebrain progenitor marker-expressing cells that can generate interneurons (Sander et al., 2000). On the other hand, diencephalon progenitors (as defined by TCF7L2) had a broader distribution (Lee et al., 2017; Figure S1B). We also identified a glial cell cluster (C3) with cells expressing markers specific for Müller glia (S100A16, APOE, ITM2B, and COL9A1) and microglia (ICAM1 and AIF1). Both cell types are required at the early stages of retina development, such as regulation of retinal ganglion cell (RGC) subsets and synapse pruning (Li et al., 2019; Figure 2B). Our analysis revealed two more clusters of developing optic vesicles (C6) and a pre-optic region (C7) expressing the early transcription factors TFAP2A/B and the LIM homeobox gene LHX1, essential for optic fissure closure and precise timing of neural retina differentiation (Figures 2A–2B). TFAP2A expression has also been detected in the lens and neural retina (Min et al., 2020). On the other hand, LHX1 is expressed in the forebrain at a time point that parallels formation of the neural retina (Inoue et al., 2013). The optic vesicle cluster (C6) also contained cells that strongly expressed NEFM and NEFL suggestive of developing RGCs. To substantiate that our organoids contain neuroectoderm lineages and an anterior portion of the neural tube, we did analyze for cells expressing OTX2, PAX6, and EMX2, homeodomain transcription factors expressed in the neural tube’s anterior part and developing forebrain in a rostrocaudal manner (Wurst and Bally-Cuif, 2001). OTX2 is also required for RPE specification after an optic vesicle is formed (Ghinia Tegla et al., 2020; Beby and Lamonerie, 2013; Figure 2B). Testing our organoid’s purity, we queried forebrain, midbrain, hindbrain, endoderm, ectoderm, mesoderm neural crest cells-related genes, cadherins, and HOX genes. Our analysis indicated that the rest of the unrelated markers are not expressed considerably except for the forebrain and neural cadherins (Figure S1C).
To compute pseudotime, we performed a trajectory analysis of the cells developing toward the pre-optic cluster. 30-day-old brain organoids containing primitive eye fields express essential marker genes of developing RGCs with the same pattern of appearance, as described previously for the human fetal retina (Sridhar et al., 2020; Figure 2D). These data demonstrate that brain organoids with developing optic vesicles could recapitulate brain- and optic-related cell composition and transcriptomic signatures.
Brain organoids progressively develop optic vesicles over time and reveal brain-, early retina-, and eye-related cell populations
Continued culturing of organoids resulted in progressive development of these pigmented regions forming one or two intensely pigmented optic vesicle-like structures between days 50 and 60. We call these organoids optic vesicle brain organoids (OVB-organoids) (Figures 3A–3C). We could generate OVB-organoids across five independent iPSC donors (Figure 3B). For example, 86 of 95 organoids from the IMR-90 iPSC line could develop easily recognizable bilateral symmetric optic vesicles with an overall success rate of 226 (66%). None of the organoids derived from any of the iPSC donors formed more than two pigmented regions. Intriguingly, these pigmented optic vesicles were restricted to one pole of the organoids near each other, suggesting an area topographically patterned at the forebrain region.
Figure 3. OVB-organoids reveal brain-, early retina-, and eye-related cell populations
(A) 60-day-old organoids with bilaterally symmetric pigmented optic vesicles (i–v). Insets in (ii) show individual pigmented regions. Scale bar, 1 mm. n ≥ 32 organoids from ≥ 3 batches. Cell lines IMR-90 (i and ii), F13535.1 (iii + v), and GM25256 (iv).
(B) Bar diagram quantifying organoids with pigmented optic vesicles. Donor 1 (IMR90) yields the highest number of organoids with pigmented optic vesicles (91%). Donors 2, 3, 4, and 5 yield 65%, 70%, 43%, and 40%, respectively. Donor 1: 95 organoids, 5 batches; donor 2: 280 organoids, 5 batches; donor 3: 87 organoids; 4 batches; donor 4: 32 organoids, 3 batches; donor 5: 241 organoids, 3 batches. Donors 1–5 are IMR-90, GM25256, F13535.1, F14536.2, and Crx-iPS. Two-way ANOVA followed by Sidak’s multiple comparisons test. Significance within cell lines: IMR-90: n = 5, p∗∗∗∗ < 0.0001; GM25256: n = 5, p = 0.7359; F13535.1: n = 4, p = 0.1852; F14536.2: n = 3, p > 0.9999; Crx-iPS: n = 3, p > 0.9999.
(C) Bar diagram quantifying percentages of two-, one-, or no optic vesicle-containing organoids derived from donor 1 (IMR90) iPSC cells. 95 organoids, 5 batches.
(D) UMAP plot showing OVB-organoids clustered into 12 distinct groups according to their transcriptome. Brain and retinal cells are found.
(E) Heatmap highlighting expression of the top 20 marker genes for each cluster computed by Student’s t test. See Table S2 for the complete list of marker genes.
(F) Dot plot showing mean expression in each group and fraction of cells expressing sets of literature-based marker genes. The plot shows brain-related clusters expressing higher levels of key brain marker genes (light blue boxes). Simultaneously, clusters C5, C7, C8, C10, and C11 show expression of retina development markers or microglia (pink boxes).
(G) Subclustering clusters expressing retinal-related marker in the dataset (pink boxes in F). UMAP shows six subclusters. The matrix plot shows groups of genes related to specific areas of the eye or cell type: lens and cornea components, optic chiasm, developing retina, eye development, lateral geniculate nucleus (LGN), and RGCs (Kanton et al., 2019).
To analyze the cell diversity of OVB-organoids, we analyzed a total of 7,298 high-quality single cells of six organoids. Clusters C1–C12 show distinct transcriptome profiles (Figures 3D and 3E; Table S2). UMAP embedding showed the presence of a main group of cells expressing pan-neuronal markers of DCX and MAP2, the telencephalon marker EMX1, and a smaller subset of cells expressing the diencephalon marker DLX2 (C5 and C7). Cortical excitatory neurons (ENs) were identified by expression of MAP2, neurogenic factors of NEUROD 2 and 6, glutamatergic neurons (SLC17A7), ENO2 containing mature neuronal types including FOXG1, a critical transcription factor for forebrain development (Hettige and Ernst, 2019; Figure 3F). The ganglionic eminence (GE) appears transiently during embryonic nervous system development, ensuring proper guidance of cell and axon migration (Nery et al., 2002; Métin and Godement, 1996). The cluster GE (C5) comprises cells with GE signatures and GABAergic signatures (for example, identity markers of DLX1, DLX2, DLX5, and ISL1). A significant proportion of cells in this cluster expressed endogenous GAD2, which is expressed in intrinsic photosensitive RPGs (ipRGCs) (Sonoda et al., 2020). The diencephalon cluster (C7) comprises cells with signatures of diencephalon developmental genes such as LHX9, EBF1, and NHLH2.
Perhaps the most intriguing clusters are early eye development (C5 and C11), diencephalon (C7), Müller glia (C11), and RGC clusters that exhibit eye development signatures enriched with cells expressing SIX3, a transcription factor that plays a crucial role in eye formation in the forebrain (Loosli et al., 1999; Carl et al., 2002). These clusters also contain OTX2-expressing cells that activate the downstream transcription factors PRDM11 and VSX2 to determine photoreceptor and bipolar cell fates (Goodson et al., 2020; Ghinia Tegla et al., 2020). These cell clusters also harbor factors that regulate development of horizontal cells, ganglion cells, cones, and amacrine cells (for example, ONECUT1/2 and NEFL). Cells in these clusters express progenitor markers of FOXG1, SOX2, and HES1. Among these, HES1 is a transcriptional repressor critical for the timing of retinal neurogenesis (Figure 3F; Lee et al., 2005). The C5, C7, and C11 clusters also contained signatures of non-neural ectoderm (NNE), sparsely preplacodal ectoderm (PPE), and lens ectoderm cell clusters (DLX5, DLX6, BMP7, MAB21L1, MEIS1, MEIS2, and EYA2) (Ealy et al., 2016; Figure S2A). The PPE contains neurogenic and non-neurogenic placodes, which give rise to ocular surface ectoderm (Lleras-Forero and Streit, 2012), suggesting that OVB-organoids could include non-neuronal cell types of optic vesicles, such as the lens and cornea, whose biogenesis begins from the surface ectoderm.
Thus, to further delineate the degree of cellular diversity generated within these OVB-organoids, we closely dissected clusters exhibiting retinal-related expression markers (Figure 3F; C5, C7, C8, C10, and C11) by a second iteration of clustering (SC1–SC6) (Figures 3G and S2B). To identify cells expressing signatures of primitive lens and cornea components (SC1), optic chiasm (SC2 and SC3), developing retina, eye, lateral geniculate nucleus (LGN), and RGCs (SC4–SC6). Examining these subclusters, we noticed CRYAB- and KRT10-expressing cells in a distinct subcluster (red, SC1). These genes signify formation of the mammalian lens and corneal epithelium. Similarly, we noticed SPRY1 and SPRY2 (lens and corneal markers), essential for eyelid closure (purple and brown, SC2 and SC3). Cells within this subcluster express ZIC2 and SLIT2 (brown, SC2), transcription factors that determine the routing of RGC axons at the optic chiasm to the respective hemispheres, a patterning event critical for binocular vision (Lee et al., 2008; Herrera et al., 2003). Other closely related subclusters (SC4–SC6) contained cells expressing signatures of developing retina (turquoise; ONECUT1, ONECUT2, and SIX3), eye development (orange and green, PAX6 and MEIS2), and RGCs (turquoise and orange; NEFM, NEFL, and ISL1). We also identified cells expressing LHX9 and PCP4, molecules critical for forming LGNs; primary sensory thalamic nuclei receive major sensory input from the retina (Iwai and Kawasaki, 2009). Hierarchical clustering based on the subset of genes shows the main distinction between SC1–SC3 and SC4–SC6. These analyses indicate that distinct neuronal and non-neuronal cell types of the brain in OVB-organoids transcriptionally resemble the endogenous counterparts.
To identify similarities between our OVB-organoids, fetal retina (day 59), and pure retinal organoids (days 45 and 60) (Sridhar et al., 2020), we integrated the respective scRNA sequences using a recently described algorithm (Polański et al., 2020). Specifically, we used the subset of cells expressing retina-related markers of 60 OVB-organoids and cross-referred to the raw count matrices of the fetal retina and pure retinal organoids (Sridhar et al., 2020). Sridhar et al. (2020) showed that, at this stage of retina development, the primary cell types present are a pool of glial cells/stem cells, the RGC cluster, photoreceptor precursor cells, and amacrine/horizontal cells. Integration of the three-reference dataset into OVB-organoids shows that both contain a population of glial/stem cells and that the generated pre-neuronal population has similarities to the RGCs of the fetal retina and retinal organoids. This analysis indicates that the subset of OVB-organoid datasets partially resembles what is seen in the fetal retina and retinal organoid (Figure S3A). A substantial overlap with pure fetal or retinal organoids is unexpected because of the significant portion of brain tissue in OVB-organoids.