Precise temporal regulation of alternative splicing during neural development

Abstract

Alternative splicing (AS) is one crucial step of gene expression that must be tightly regulated during neurodevelopment. However, the precise timing of developmental splicing switches and the underlying regulatory mechanisms are poorly understood. Here we systematically analyze the temporal regulation of AS in a large number of transcriptome profiles of developing mouse cortices, in vivo purified neuronal subtypes, and neurons differentiated in vitro. Our analysis reveals early-switch and late-switch exons in genes with distinct functions, and these switches accurately define neuronal maturation stages. Integrative modeling suggests that these switches are under direct and combinatorial regulation by distinct sets of neuronal RNA-binding proteins including Nova, Rbfox, Mbnl, and Ptbp. Surprisingly, various neuronal subtypes in the sensory systems lack Nova and/or Rbfox expression. These neurons retain the "immature" splicing program in early-switch exons, affecting numerous synaptic genes. These results provide new insights into the organization and regulation of the neurodevelopmental transcriptome.

Fig. 1

Modular organization of dynamic splicing switches during cortex development. a The number of non-redundant cassette exons with differential splicing (|ΔΨ| ≥ 0.2, Benjamini FDR ≤ 0.05) in each pairwise comparison of developmental stages. The numbers of exons with increased inclusion at later stages are shown above the diagonal (top right), and exons with decreased inclusion at later stages are shown below the diagonal (bottom left). b Mouse cassette exons with developmental changes are highly conserved in human, as measured by the percentage of exons with conserved splicing in human (left) or the percentage of exons under strong evolutionary selection pressure (right). c An example of developmental splicing regulation in exon 8a of the Kdm1a gene. Inclusion of this microexon peaks between postnatal days P0–P7. d Four modules of developmentally regulated exons identified by WGCNA analysis with distinct temporal patterns during cortex development. A non-redundant set of 2883 known and novel cassette exons was included for this analysis, and their mean-substracted inclusion levels across developmental stages are shown in the heatmap. Exons in each module were ranked based on their correlation with the eigenvector of the module, and those with the strongest correlation are defined as core members (black bars on the right). Exons in each module are further divided into two groups (e.g., M1+ and M1−) depending on positive or negative correlation with the eigenvector. e Enrichment of gene ontology (GO) terms in exons showing splicing switches with specific timing. The timing of developmental splicing switches is parameterized by sigmoidal curve fitting, and exons are ranked based on the timing. Exons in each sliding window (with a window size of 300 exons) were compared to all cassette exons with sufficient read coverage in the brain to identify significant GO terms. Only GO terms significant in at least one sliding window are shown (Benjamini FDR ≤0.05). Broad categories and top GO terms in each category are highlighted on the right

Fig. 2

The modular organization of the developmental splicing program is pan-neuronal. a The splicing profile of module exons in different neuronal subtypes. Exons are shown in the same order as in the cortex reference. DIV: days in vitro. For differentiation of glutamatergic neurons from ESCs, cells on DIV 0 are enriched in radial glia committed to the neuronal fate, which becomes post-mitotic on DIV 1. b Quantification of developmental splicing switches among module exons in different neuronal subtypes. In each dataset, M1+/M1− and M2+/M2− core module exons also showing differential splicing (|ΔΨ| ≥ 0.2, Benjamini FDR ≤0.05) in each pairwise comparison were counted. The number of exons showing increased inclusion at later time points is shown above the diagonal (top right), and the numbers of exons showing decreased inclusion is shown below the diagonal (bottom left)

Fig. 3

Prediction of neuronal maturation stages based on splicing profiles using Splicescope. Principal component analysis (PCA) of splicing profiles in the cortex reference was used to project high-dimensional data into a 2D space for visualization. Different types of samples are indicated by different marker shapes with border color representing the predicted maturation stage using a regression model and filled color representing the true stage (when available). The reference cortex samples are shown in large filled circles with color representing the developmental stage. Highlighted are spinal motor neurons isolated from E12.5, P1, and adult mice, ESC-differentiated glutamatergic neurons at different days, and purified cPNs between E15.5 and P1. DIV: days in vitro

Fig. 4

A set of tissue-specific or neuron-specific RBPs regulate the timing of developmental splicing switches. a Dynamic expression of four families of tissue-specific RBPs, including Ptbp1/2, Nova1/2, Rbfox1–3, and Mbnl1/2. RPKM values are normalized based on the maximum expression value in each family separately and shown in color scale. b Integrative modeling to define the target alternative exons regulated by each RBP family. The Venn diagram summarizes target exons regulated by each RBP family. Note 11 exons regulated by all four RBP families and an additional 58 exons regulated by three RBP families. c Regulation of WGCNA module exons by each of the four RBP families. Activation and repression of an exon by each RBP resulting from integrative modeling analysis are indicated in red and blue, respectively. The total number of regulators for each exon is shown in the bar on the right in gray scale (the darker, the more regulators). d–f Gabrg2 exon 9 as an example in module M1 under combinatorial regulation by all four RBP families. The exon inclusion level in developing cotex is shown in d and changes upon depletion of Ptbp2 (P0) and Mbnl1/2 (adult) are shown in e. Inclusion of the exon in wild type (WT) and mutant (MUT) splicing reporters, in combination with overexpression of different RBPs, is shown in (f). Rbfox-binding and Mbnl-binding site sequences are shaded. RBP expression and exon inclusion were measured by immunoblot and RT-PCR, respectively. g RBPs either antagonize (Ptbp2) or facilitate (Nova, Rbfox, and Mbnl) the mature splicing pattern through activation or repression of exon inclusion. h Time of the maximal splicing switch for target exons regulated by specific RBPs (*p < 0.05, **p < 0.001, t test). Only exons showing a more mature (for Ptbp) or embryonic (for Nova, Rbfox, and Mbnl) pattern upon RBP depletion were included for this anlaysis. i Prediction performance of exon module membership based on regulation by each RBP family. The performance is measured by partial area under curve (pAUC) of the receiver operating characteristic (ROC) plot with a cutoff at false-positive rate (FPR) ≤0.1. j Changes of predicted maturation stages of mouse brain tissues upon depletion of RBPs

Fig. 5

A neurodevelopmental splicing code predicts early and late splicing switches. a Performance of prediction as measured by AUC. Four models were trained to predict exons in modules M1 and M2. Exons with increased inclusion or exclusion were predicted by separate models. Different sets of features were used to build models. b Importance of RBP motifs for prediction in each model. Motif sites in the upstream intron (UI), exon (E), and downstream intron (DI) were scored separately. Only motifs ranked among the top 100 features in at least one region are shown. Motifs enriched in specific regions are shown in red and motifs depleted are shown in blue. c Summary of prediction results for exons in module M4 indicating that the two developmental splicing switches of these exons can be predicted separately

Fig. 6

Distinct regulation of early-switch exons in mature sensory neurons. a The PCA scatter plot of splicing profiles in different subtypes of neurons isolated from adult mice. The circles represent the cortex reference samples and the triangles represent the four categories of cell types: non-neuronal sensory receptors, sensory neurons, sensory ganglion neurons, and mature CNS neurons. Samples are colored by the predicted maturation stage. EC enterochromaffin cell, TRC taste receptor cell, OSN olfactory sensory neuron (OMP+), DRG dorsal root ganglia sensory neurons (Nav1.8+ or Avil+), DN dopaminergic neurons, CGN cerebellar granule neurons, PN Purkinje neurons, and MN motor neurons. b The splicing profile of module exons in sensory neuron subtypes, in comparison with non-neuronal sensory receptor cells and mature CNS neurons. Exons are shown in the same order as in the cortex reference. c Statistically enriched GO terms of genes with differentially spliced exons between all sensory cell types and mature CNS neurons. The size and color represent the number and enrichment of genes associated with each term and related GO terms with overlapping genes are connected. d Expression levels of the four RBP families we focused on in our analysis as quantified by RNA-seq data. Note the high abundance of the Mbnl and Ptbp families and lack of Nova1/2 in sensory neurons. Rbfox1-3 are absent or low in sensory neurons, but expressed in ganglion neurons. e Immunofluorescence analysis of Rbfox1-3 expression in OSNs. Red, Calmegin is a marker of mature OSNs; green, Rbfox; blue, DAPI staining the DNA. Scale bar, 20 μm. f Maturation stages of different types of neurons predicted using the overall splicing profile or target exons of each RBP family. g The proposed model that explains the distinct splicing profiles of different neuronal subtypes

Fig. 7

Overexpression of Nova in DRG neurons promotes the “mature” splicing pattern observed in CNS neurons. a Schematic illustrating overexpression of Nova1 in rat primary DRG neuronal culture using a Nova1-GFP-expressing lentivirus. A GFP-expressing lentivirus is used as a Mock control. b qRT-PCR of Nova1 mRNA expression level in DRG neurons transduced with Nova1-GFP-expressing or GFP-expressing lentiviruses. The average relative expression level normalized to Mock-transduced cells is shown. Error bars represent standard deviation (n = 3). c Immunostaining of primary rat DRG neurons transduced with Nova1-GFP-expressing or GFP-expressing lentiviruses. Nova1 protein has predominant nuclear localization. Scale bar, 50 µm. The insert shows the zoom-in view of the cell. d Changes in the alternative splicing of Nova target exons upon Nova1 overexpression in rat primary DRG neurons. For each exon, the exon inclusion level in WT and Nova2 KO cortex, and DRG sensory neurons, as quantified by RNA-seq, is shown in the barplot on the left. Error bars represent the standard deviation (n ≥ 3). A representative gel image on the right shows inclusion of the alternative exon in Mock-transduced and Nova1-transduced cells detected using RT-PCR analysis (n = 3). The bands corresponding to the inclusion (top) and skipping of the exon (bottom) are indicated