Differentiation of cortical brain organoids and optic nerve-like structures from retinal confluent cultures of pluripotent stem cells

Advances in the study of neurological conditions have been possible due to induced pluripotent stem cell technologies and the generation of neural cell types and organoids. Numerous studies have described the generation of neural ectoderm-derived retinal and brain structures from pluripotent stem cells. However, the field is still troubled by technical challenges, including high culture costs and organoid-to-organoid variability. Here, we describe a simple and economical protocol that reproducibly gives rise to the neural retina and cortical brain regions from confluent cultures of stem cells. The spontaneously generated cortical organoids were isolated and cultured in suspension conditions for maturation and are transcriptionally comparable to organoids generated by other methods and to human foetal cortex. Furthermore, these organoids show spontaneous functional network activity with proteomic analysis and electron microscopy demonstrating the presence of synaptic components and maturity. The generation of retinal and brain organoids in close proximity also enabled their mutual isolation. Further culture of this complex organoid system demonstrated the formation of optic nerve-like structures connecting retinal and brain organoids, which might facilitate the investigation of the mechanisms of neurological diseases of the eye and brain.


47
The rapidly progressing field of human pluripotent stem cells (hPSCs), including embryonic (ES) and 48 induced pluripotent stem (iPS) cells, and their derivative organoids continue to provide new insights into 49 basic biology, human development, modelling of human diseases and discovery of innovative 50 treatments. Neural differentiation, in particular, has been extensively studied, improving our 51 understanding of the mechanism of neurodevelopmental conditions 1,2,3 . Large numbers of neurons and 52 astrocytes are able to be generated using two-dimensional (2D) cultures derived from hPSCs 4,5 and 53 these classical differentiation cultures were later optimised to three-dimensional (3D) suspension 54 methods that better recapitulate the physiological niche and environment of the developing human 55 brain 6 .

57
In the developing central nervous system (CNS), the eye and the brain form as an extension of the 58 forebrain diencephalic and telencephalic region, respectively 7 . Developmental biology studies have 59 helped infer the molecular basis of this patterning process and facilitated the establishment of numerous 60 differentiation protocols, which generate miniaturised versions of 3D organoids from PSCs. Brain 61 organoids replicate specific brain regions or whole cerebral areas, with both occasionally developing hypothesised that these were brain vesicles. To fully characterise these structures, we manually excised 115 these regions to be grown in 3D suspension for maturation in culture, as previously described for their

144
Representative image of a retinal vesicle and brain organoid in 2D culture. d. Bright-field image showing 145 floating brain organoids, with typical neural rosettes, following manually excision from 2D culture. e-j.

146
Immunohistochemistry images of 6 -9 weeks old brain organoids. e, f. Brain organoids showing Ki67 147 proliferative NCAD positive neuroepithelium and CASPASE 3 apoptotic cells. g-j. Cortical origin of brain 148 organoids is confirmed by the presence of SOX2, FOXG1 and PAX6 positive neural precursors markers.

150
Cortical organoids differentiate into Cortical neurons and Glial cell types 151 Next, IHC confirmed the presence of cortical plate markers TBR1 which co-expressed with CTIP2-152 positive cells in the cortical organoids (Figure 2a-c). By 10 weeks of culture, CTIP2 regions also 153 contained SATB2-positive cells (Figure 2e, f). From 8 to 12 weeks in culture, organoids increased in 154 size significantly from 2000μm (± 560) to 2600μm (± 474) in diameter (Figure 2g, n=20-25 organoids, 155 N=3 differentiation batches; mean ± SD, p=0.0002, unpaired two-tailed t test). This differentiation 156 protocol also supported the differentiation of S100B- (Figure 2g) and GFAP-positive astrocytes that 157 were present in similar percentages to TUJ1-positive neurons (Figure 2h  cells (e, f; high magnification image of inset in e shown in f) and S100B and GFAP glial cells (g, h). i.

177
Cortical Organoids from 2D/3D confluent cultures have similar cell type compositions to other 178 brain organoids 179 Next, we performed scRNA-seq analysis using the 10X genomics platform to further investigate the cell-180 type composition of these dorsal cortical organoids and to establish their similarity to other published 181 brain organoid scRNA-seq datasets, which were derived from various directed and undirected protocols.

182
We used scClassify 23 , a machine learning-based method, to annotate cell types that were present in our 183 scRNA-seq cortical organoids (Figure 3a) and compared the composition of cell types in 3-month-old 184 cortical organoid (CO) we generated against those from 3-and 6-month-old brain organoids and cell 185 types from gestational week 12 (GW12) pre-frontal lobe (PFL) foetal brain (Figure 3b).

187
First, we compared the composition of cell types by using the cell type labels from the original authors

194
Next, we utilised scClassify to annotate cells in our CO scRNA-seq dataset by training the classification 195 model using either each individual public dataset as a reference or using all datasets jointly (refer to as 196 joint training) (Figure 3d). The predicted cell-type composition in CO scRNA-seq data was plotted with 197 respect to the training data, and, irrespective of the training dataset, a similar cell-type composition was 198 observed in our COs (Figure 3d). In agreement with this, the expression profiles of key marker genes 199 for each cell type population were also largely consistent, irrespective of the training dataset (

207
Organoid cell types were clustered in a tSNE plot showing that cortical neurons were the most abundant 208 cell type (Figure 3e). We then measured the agreement between different cell-type compositions across 209 the multiple datasets using intraclass correlation (ICC) and visualised these results (Figure 3f). Three     showing the agreement between cell-type compositions among different organoids and a fetal brain.

233
Proteome analysis of brain organoids highlights increased abundance of proteins related to 234 synaptic transmission

235
The proteomes of iPSCs and cortical organoids were surveyed to a depth of 6,244 and 5,719 proteins, 236 respectively (after filtering and counting only unique genes). There were 4,444 proteins present shared 237 between the iPSC and cortical organoid lists. Gene ontology enrichment analysis was performed with a 238 focus on biological process and KEGG pathway terms (Figure 4a-d). Figure 4a shows the top thirty 239 biological process terms that had the largest difference in significance between iPSCs and cortical 240 organoids. Terms related to the cell cycle, chromosomes and DNA were more likely to be enriched for 241 iPSCs, whereas terms related to the synapse, synaptic vesicles, vesical transport, axon/dendrite 242 development and neuron projection were more enriched for cortical organoids. Since we are interested 243 in neuronal differentiation in organoids, we extracted the top five synaptic and development terms (not 244 already shown in Figure 4a) from the list of significant biological process terms (Table S2)       Having demonstrated the formation of dorsal cortical organoids from 2D/3D confluent differentiation 336 cultures, we next aimed to test whether the 3D suspension culture of complex organoids comprising 337 both retinal and brain organoids promoted the formation of connections between the two organs.

339
In early 2D differentiation cultures, retinal and brain organoids spontaneously developed in proximity

345
The connection of the retina to the brain to process the visual information is established by retinal 346 ganglion cells (RGCs), the first-born cell type of the retina, which form neuronal outputs connecting to 347 the brain through the optic nerve. In the complex retinal-brain organoids, at 10 weeks of development, 348 the maturation of each individual organoid was evident. In retinal organoids NEUN and HuC/HuD 349 colocalised with THY1 and MAP2 RGC markers delineating these cells' axonal projections towards the 350 centre of the organoid (Figure S6a, b). Similarly, these two markers were also expressed in neurons

389
Here, we adapted a previously-described simple and robust 2D/3D differentiation approach derived from

410
Despite tremendous progress in the field, the lack of disease-relevant functional assays in organoids 411 hinders their ability to test for new treatment efficacy. However, disease molecular signatures and 412 biomarkers can be determined using integrative analysis of omics as well as computational or 413 bioinformatic methodologies. In this study, cortical organoids generated by 2D/3D confluent method 414 were extensively characterized using scRNA-seq and proteomics analysis, setting a baseline for future 415 studies using disease organoids. In agreement with a previous correlation study comparing scRNA-seq 416 datasets of brain organoids and foetal human brain 12 , our organoids showed a high correlation with brain 417 organoids derived by other methodologies, particularly with dorsally patterned brain organoids

421
The proteome of brain organoids has been seldom investigated. Previous proteomic analysis of early-422 staged (day 45) brain organoids generated using a whole-brain spontaneous 3D aggregation method

448
In this study, we demonstrated the spontaneous formation of retinal-and-brain complex organoids in the 449 2D/3D confluent cultures. When in 3D suspension, these retinal-brain organoids maintained their 450 proximity and were connected through an optic nerve-like structure mimicking the neuronal projections 451 that connect the eye and brain. Our confluent method of differentiation enabled the precise isolation of 452 these two organoids, overcoming variability within 3D directed and whole brain protocols that 453 sporadically generate eye structures 8,10 and the forced fusion of different organoids to form assembloids.

454
The latter have been successfully used to model in vivo neuronal interactions between different brain

480
Peripheral blood mononuclear cells (PBMCs) were isolated from the whole blood of a healthy donor 481 using density gradient centrifugation. Briefly, 25ml of whole blood diluted 1:1 with phosphate buffered

540
Primary antibody (Supplemental Table 1) diluted in blocking solution was incubated overnight at 4°C.

544
For immunohistochemistry of whole cortical organoids in 3D view using Lightsheet, a clearing protocol 545 was performed. Briefly, cortical brain organoids were fixed in 4% paraformaldehyde, suspended in 20% 546 sucrose and stored in 4°C as described above. Samples were incubated in clearing reagent 1 (Urea,

574
Cortical organoids were transferred to Eppendorf tubes then washed with 1 ml of PBS. Enzyme mix was 575 prepared as per manufacture instructions and 500 ul added to each tube and incubated at 37°C for 10 576 minutes and flicked every 2 minutes. Organoids were then pipetted up and down with p1000 pipette to 577 mechanically dissociate the big clumps of cells. The tubes were returned to 37°C for another 5 minutes

659
Organoid electrical activity was measured using an MEA2100-lite system with TC01 temperature control

724
The loop count was 12, the isolation window was 1.2 m/z, the first mass was fixed at m/z 140 and the 725 normalized collision energy was 28. Singly charged ions and those with charge >8 were excluded from 726 MS/MS and dynamic exclusion was for 35 s.

748
The list of proteins detected by mass spectrometry were filtered to remove proteins that were not 749 detected in all three biological replicates. The number of "razor plus unique" peptides was required to 750 be two or greater using an average value across the three biological replicates. The filtered protein list 751 was converted to a gene list. Duplicate genes arising from multiple protein accession and isoforms were 752 filtered by removed the gene with the least number of "razor plus unique" peptides. Over-representation 753 analysis was conducted using WebGestalt. Significant enrichments for biological process and KEGG 754 pathway terms were determined by comparing to the human coding genome. The probability of 755 significant enrichment was adjusted using a false discovery rate of 5%.

757
AAV transduction of brain organoids 758 AAV viral vectors (1-3.6x10E11 vg/organoid) were added to a total volume of 375 μl using fresh CODM 759 media used to culture the cortical and whole brain organoids. The organoids were then transferred to 760 low binding 24 well plates (Costar, Corning) and media was completely replaced with CODM containing 761 the AAV vectors. Cortical and whole brain organoids were incubated at 37 °C for half a day before