Spatio-temporal transcriptome of the human brain

Abstract

Brain development and function depend on the precise regulation of gene expression. However, our understanding of the complexity and dynamics of the transcriptome of the human brain is incomplete. Here we report the generation and analysis of exon-level transcriptome and associated genotyping data, representing males and females of different ethnicities, from multiple brain regions and neocortical areas of developing and adult post-mortem human brains. We found that 86 per cent of the genes analysed were expressed, and that 90 per cent of these were differentially regulated at the whole-transcript or exon level across brain regions and/or time. The majority of these spatio-temporal differences were detected before birth, with subsequent increases in the similarity among regional transcriptomes. The transcriptome is organized into distinct co-expression networks, and shows sex-biased gene expression and exon usage. We also profiled trajectories of genes associated with neurobiological categories and diseases, and identified associations between single nucleotide polymorphisms and gene expression. This study provides a comprehensive data set on the human brain transcriptome and insights into the transcriptional foundations of human neurodevelopment.

Figure 1. Global spatiotemporal dynamics of gene expression

a, Venn diagrams representing total number of genes considered to expressed and the number of spatially and temporally DEX genes for brain regions (top) and NCX areas (bottom). b, MDS plot showing transcriptional similarity, colored by periods (top) and regions (bottom). Euclidean distance of log₂-transformed signal intensity (nonmetric, stress=18.9%) was used to measure pairwise similarity. c, Heat map matrix of pairwise Spearman correlations between brain regions (top) or NCX areas (bottom) during fetal development (periods 3–7), postnatal development (periods 8–12), and adulthood (periods 13–15).

Figure 2. Sex-biased gene expression

a, Number of sex-biased DEX genes in brain regions/NCX areas during fetal development (periods 3–7), postnatal development (periods 8–12), and adulthood (periods 13–15). b, PCDH11Y exon array signal intensity (left) and qRT-PCR validation (right) (N=5 male brains per period). c, IGF2 exon array signal intensity (left) and qRT-PCR (right) validation in NCX (N=4 per sex and period). P-values were calculated by unpaired t-test. Whiskers indicate 5^th and 95^th percentile.

Figure 3. Sex-biased differential exon usage

a, Gene structure and probe set composition of NLGN4X. Depicted by yellow and green arrows are primers used for qRT-PCR validation. b, Heat map of the log₂ male/female signal intensity ratio of each exon for fetal development (periods 3–7), postnatal development (periods 8–12), and adulthood (periods 13–15). Differences in expression of exon 7 (yellow frame) and 3′UTR (green frame) in adult NCX are highlighted. Note that exons 2 and 3A did not meet our expression criteria and are not represented. c, qRT-PCR validation of exon 7 and 3′UTR expression in adult NCX (N=4 per sex). P-values were calculated by unpaired t-test. Whiskers indicate 5^th and 95^th percentile.

Figure 4. Global co-expression networks and gene modules

a, Dendrogram from gene co-expression network analysis of samples from period 3 to 15. Modules of co-expressed genes were assigned a color and number (M1 to M29). b, Heat map of genes in M8 (left) showing the spatiotemporal expression pattern after hierarchical clustering. The expression values for each gene are arranged in the heat map, ordered first by brain regions, then by age and last by NCX areas. Spatiotemporal pattern of M8 (middle) summarized by the first principal component (PC1) for expression of genes in the module across age. Top 50 M8 genes (right), defined by the highest intramodular connectivity, Top 10 hub genes are in red. c, Same analyses were performed for M15. Results for other modules are available in Supplemental Information.

Figure 5. Trajectories of genes associated with neurodevelopmental processes

a, Comparison between DCX expression in HIP and the density of DCX-immunopositive cells in the human dentate gyrus. b, Comparison between transcriptome-based dendrite development trajectory in DFC and Golgi method-based growth of basal dendrites of layer 3 (L3) and 5 (L5) pyramidal neurons in the human DFC. c, Comparison between transcriptome-based synapse development trajectory in DFC and density of DFC synapses calculated using electron microscopy. For (b) and (c), PC1 for gene expression was plotted against age to represent the developmental trajectory of genes associated with dendrite or synapse development. Independent datasets were centered, scaled, and plotted on a logarithmic scale. d, PC1 value percentage of maximum PC1 value for given set of genes plotted against age to represent general trends and regional differences in several neurodevelopmental processes in NCX, HIP, and CBC.

Figure 6. Association between SNPs and gene expression

a, b, SNP distribution around TSS (a) and TES (b) of the associated genes based on several Wald test P-value cut offs after gene-wide Bonferroni correction. c, GLIPR1L2 expression association with rs10785190 genotype, a SNP located in exon 1. The horizontal solid and dashed LOWESS curves are the developmental trend of gene and exon 1 and 2 expression, respectively. d, qRT-PCR validation of exon 1 and 2 expression in NCX for each genotype (N=14 GG, 14 AG, 8 AA) plotted relative to the local regression smooth curve to facilitate comparison across developmental periods. P-values were calculated by unpaired t-test. Whiskers indicate 5^th and 95^th percentile.