Benchmarking brain organoid recapitulation of fetal corticogenesis

Brain organoids are becoming increasingly relevant to dissect the molecular mechanisms underlying psychiatric and neurological conditions. The in vitro recapitulation of key features of human brain development affords the unique opportunity of investigating the developmental antecedents of neuropsychiatric conditions in the context of the actual patients’ genetic backgrounds. Specifically, multiple strategies of brain organoid (BO) differentiation have enabled the investigation of human cerebral corticogenesis in vitro with increasing accuracy. However, the field lacks a systematic investigation of how closely the gene co-expression patterns seen in cultured BO from different protocols match those observed in fetal cortex, a paramount information for ensuring the sensitivity and accuracy of modeling disease trajectories. Here we benchmark BO against fetal corticogenesis by integrating transcriptomes from in-house differentiated cortical BO (CBO), other BO systems, human fetal brain samples processed in-house, and prenatal cortices from the BrainSpan Atlas. We identified co-expression patterns and prioritized hubs of human corticogenesis and CBO differentiation, highlighting both well-preserved and discordant trends across BO protocols. We evaluated the relevance of identified gene modules for neurodevelopmental disorders and psychiatric conditions finding significant enrichment of disease risk genes especially in modules related to neuronal maturation and synapsis development. The longitudinal transcriptomic analysis of CBO revealed a two-step differentiation composed of a fast-evolving phase, corresponding to the appearance of the main cell populations of the cortex, followed by a slow-evolving one characterized by milder transcriptional changes. Finally, we observed heterochronicity of differentiation across BO models compared to fetal cortex. Our approach provides a framework to directly compare the extent of in vivo/in vitro alignment of neurodevelopmentally relevant processes and their attending temporalities, structured as a resource to query for modeling human corticogenesis and the neuropsychiatric outcomes of its alterations.


Supplementary Figure 8
Overlap between genes retrieved as differentially expressed by stage-wise differential expression analysis in cortical brain organoids (CBO) compared to external brain organoid dataset (Minimally Guided Organoids in panel A; Forebrain Organoids in panel B: Telencephalic Aggregates in panel C). P-value and Odds Ratio are reported for the overlap with P-value < 0.01 and Odds Ratio > 3. Cell color is assigned according to Odds Ratio values. (D) Heatmap depicting the results of the whole-transcriptome correlation analysis between CBO data and external BO protocols.

Supplementary Figure 9
Lollipops showing the expression levels of gene signatures related to brain areas, off-target tissues and cell stress in BS (A), CBO (B), MGO (C), FO (D) and TA (E). Expression levels (Log2Fpkm or Log2Rpm) along timepoints are reported in each dataset as the mean value across replicates. Each bar colour corresponds to a specific signature, as reported in the plot legend. For brain organoids (B-E), asterisks identify genes that are retrieved as significantly changed in stage-wise differential expression analyses (comparison of each differentiation stage with the previous) reported in Figure 2 for CBO and Figure 4 for the other protocols. **: FDR < 0.05; *: Conventional pvalue < 0.05.

Supplementary Figure 10
Lollipops showing the expression levels of gene signatures related to neurotransmission systems in BS (A), CBO (B), MGO (C), FO (D) and TA (E). Expression levels (Log2Fpkm or Log2Rpm) along time-points are reported in each dataset as the mean value across replicates. Each bar colour corresponds to a specific signature, as reported in the plot legend. For brain organoids (B-E), asterisks identify genes that are retrieved as significantly changed in stage-wise differential expression analyses (comparison of each differentiation stage with the previous) reported in Figure 2 for CBO and Figure 4 for the other protocols. **: FDR < 0.05; *: Conventional P-value < 0.05.

Supplementary Figure 11
Overlap between BrainSpan and Cortical Brain Organoid gene modules identified by WGCNA. Dot plot displaying the overlap between BS (y-axis) and CBO (x-axis) gene modules identified by WGCNA. Numbers represent shared genes and are shown for overlaps with odds ratio (OR) > 1, while dots are reported for those having also P-value < 0.05. Dot colour is assigned according to OR values and dot size according to Pvalue. for overlaps with odds ratio (OR) > 1, while dots are reported for those having also P-value < 0.01. Dot colour is assigned according to OR values and dot size according to P-value as shown by each legend.

Supplementary Figure 13
Behaviour of the module eigengene for Cortical Brain Organoids gene co-expression modules in fetal cortex and brain organoids datasets. (A) Visualization of CBO_Turquoise and CBO_Black eigengene module in prenatal fetal cortex as well as in the brain organoids' datasets. PCA was performed to calculate the module eigengene on the CBO dataset and then applied to BS prenatal cortex, as well as each of the other brain organoid datasets. Each dot represents a data point, while the line connects the median value for each postconceptional week (BS) or differentiation day (BO). The same analysis and visualization are applied to CBO_Brown and CBO_Blue (B, decreasing during differentiation) and CBO_Green and CBO_Red (C, nonmonotonic behaviour).

Supplementary Table 2
Neurotransmission signature genes with the specification of their role in the relative neurotransmission system (Enzyme, receptor or transporter).

Supplementary Table 3
Results of Gene Ontology analyses.  Year 4 Year 8 Year 11 Year 13 Year 15 Year 18 Year 19 Year 21 Year 23 Year 30 Year 36 Year 37 Year 40 Year 0.3 Year 0.8