Developmental trajectories of thalamic nuclei revealed by single-cell transcriptome profiling and Shh perturbation
The thalamus is the principal information hub of the vertebrate brain, with essential roles in sensory and motor information processing, attention, and memory. The complex array of thalamic nuclei develops from a restricted pool of neural progenitors. We apply longitudinal single-cell RNA-sequencing and regional abrogation of Sonic hedgehog (Shh) to map the developmental trajectories of thalamic progenitors, intermediate progenitors, and post-mitotic neurons as they coalesce into distinct thalamic nuclei. These data reveal that the complex architecture of the thalamus is established early during embryonic brain development through the coordinated action of four cell differentiation lineages derived from Shh-dependent and independent progenitors. We systematically characterize the gene expression programs that define these thalamic lineages across time and demonstrate how their disruption upon Shh depletion causes pronounced locomotor impairment resembling infantile Parkinson’s disease. These results reveal key principles of thalamic development and provide mechanistic insights into neurodevelopmental disorders resulting from thalamic dysfunction.
Steps to reproduce
Processed count matrices will be made available on GEO. Except where otherwise specified, we processed and visualized the scRNA-seq counts with the following Seurat-based pipeline, using Seurat v3.0.2. We filtered out cells with less than 2,000-5,000 UMIs based on the inflection point of the log-transformed barcode rank plot of each sample, or more than 15% of the UMIs coming from mitochondrially encoded genes. In total, 249,071 cells passed these filters, with a median of 4,891 UMIs/cell, (interquartile range: 3,795 – 6,372 UMIs/cell), 2,466 genes/cell (interquartile range: 2,092 – 2,943 genes/cell), and 2.7% of UMIs in each cell coming from mitochondrially encoded genes (interquartile range: 2.1 – 3.5%). We next scaled and centered the UMI counts and used the default vst method to identify the top 2,000 variable genes. We removed all genes from the X and Y chromosomes to reduce the effect of unequal male and female mouse replicates between conditions. To correct for non-biological batch effects between conditions and time points, we used the Harmony algorithm with its Seurat integration, run on the top principal components (PCs) of the variable genes. Harmony outputs a batch-corrected representation of the scRNA-seq data of same dimensionality as the input PCs. We ran Louvain clustering and UMAP on the output of Harmony to visualize this consolidated gene expression space. The full dataset was visualized using 20 PCs and 3 UMAP dimensions. We used the differentially expressed genes according to edgeR in each cluster to annotate it based on cell type, differentiation stage, or area of the brain according to published literature and ISH images from the Allen Developing Mouse Brain Atlas.