A comparison of basal and activity-dependent exon splicing in cortical-patterned neurons of human and mouse origin

Abstract

Rodent studies have shown that alternative splicing in neurons plays important roles in development and maturity, and is regulatable by signals such as electrical activity. However, rodent-human similarities are less well explored. We compared basal and activity-dependent exon splicing in cortical-patterned human ESC-derived neurons with that in cortical mouse ESC-derived neurons, primary mouse cortical neurons at two developmental stages, and mouse hippocampal neurons, focussing on conserved orthologous exons. Both basal exon inclusion levels and activity-dependent changes in splicing showed human-mouse correlation. Conserved activity regulated exons are enriched in RBFOX, SAM68, NOVA and PTBP targets, and centered on cytoskeletal organization, mRNA processing, and synaptic signaling genes. However, human-mouse correlations were weaker than inter-mouse comparisons of neurons from different brain regions, developmental stages and origin (ESC vs. primary), suggestive of some inter-species divergence. The set of genes where activity-dependent splicing was observed only in human neurons were dominated by those involved in lipid biosynthesis, signaling and trafficking. Study of human exon splicing in mouse Tc1 neurons carrying human chromosome-21 showed that neuronal basal exon inclusion was influenced by cis-acting sequences, although may not be sufficient to confer activity-responsiveness in an allospecific environment. Overall, these comparisons suggest that neuronal alternative splicing should be confirmed in a human-relevant system even when exon structure is evolutionarily conserved.

Keywords: RNA-seq-RNA sequencing; alternative splicing; calcium signaling; evolutionary conservation and divergence; gene expression; neuronal activity.

FIGURE 1

Comparison of basal exon inclusion in cortical-patterned neurons of human and mouse origin. (A–D) The exon inclusion ratio, otherwise known as “percent spliced in” (PSI) in DIV4 Mus-PRIM^CORT neurons plotted against the corresponding PSI in the indicated cell types (mean PSI, n = 3 independent biological replicates here and throughout the figure). All exons plotted have a 1:1 human-mouse ortholog, the mean of 3 independent biological replicates is shown. (E) Pearson r correlation coefficients for the comparisons made in (A–D), and Supplementary Figure 2C. Error bars indicate the 95% confidence limits and in all cases p < 0.0001. For data relating to this figure see Source_Data.xlsx. (F) Examples of two 1:1 orthologous exons (coordinates relate to this exon), plus flanking exons, showing relative RNA-seq read density. One exon (from ZMYND11) has a similar PSI in human and mouse neurons, while one (from HNRNPAB) has a different PSI in human and mouse neurons. The PSI of the ZMYND11 exon and HNRNPAB exon is highlighted in the scatter graphs A–D with magenta and green circles, respectively. (G) Fold enrichment of exons classed as alternatively spliced in DIV4 Mus-PRIM^CORT neurons in exons which are also classed as alternatively spliced in the indicated neuronal cell types (defined as 80 > mean PSI > 20, n = 3). Error bars indicate the 95% confidence limits of the enrichment factor. In all cases p < 0.0001 (Fisher’s exact test). *P < 0.05 (normal approximation to difference in log odds ratios). (H) A heat map summary showing the correlation coefficients of all possible pairwise comparisons as indicated.

FIGURE 2

Comparison of activity-dependent alternative splicing in human and mouse cortical neurons. (A–E) For the indicated neuronal preparations, PSI of exons in control neurons is plotted against that in KCl-stimulated neurons [(A–D): 4h; E: 3h-data were generated by another lab (Quesnel-Vallières et al., 2016)]. All exons plotted have a 1:1 human-mouse ortholog. Red crosses indicate a significant difference in PSI (p < 0.05, read count for exon inclusion or exclusion in all samples > 5, PSI difference > 10, n = 3). (F) The KCl-induced change in PSI in DIV4 Mus-PRIM^CORT neurons is plotted against the corresponding change in Hum-ESC^CORT neurons. (G–I) The KCl-induced change in PSI in exons in DIV4 Mus-PRIM^CORT neurons is plotted against the corresponding change in the indicated cell types. (J) Correlation coefficients for the comparisons made in (F–I). Error bars indicate the 95% confidence limits and in all cases. p < 0.0001 For data points relating to this figure see Source_Data.xlsx. (K) Examples of two 1:1 orthologous exons (coordinates relate to this exon), plus flanking exons, showing relative RNA-seq read density. One exon (from ARHGAP21) has a similar KCl-induced PSI change in human and mouse neurons, while one exon (from ACIN1) is only subject to activity-dependent alternative splicing in human neurons. The KCl-induced PSI change of the ARHGAP21 exon and ACIN1 exon is highlighted in (F,G,H and I) with a magenta and a green circle, respectively. (L) A heat map summary showing the correlation coefficients of all possible pairwise comparisons as indicated.

FIGURE 3

Comparison of activity-dependent alternative splicing as a percentage of possible change. (A–D) For exons classed as alternatively spliced (80 > PSI > 20) in DIV4 Mus-PRIM^CORT neurons, the effect of KCl stimulation on PSI was calculated as a percentage of the maximum possible PSI change and plotted (x-axis) against the corresponding value for the other cell types (y-axis) as indicated. (E) Correlation coefficients for the comparisons made in (A–D). Error bars indicate the 95% confidence limits and in all cases p < 0.0001. For data points relating to this figure see Source_Data.xlsx. (F) A heat map summary showing the correlation coefficients of all possible pairwise comparisons as indicated.

FIGURE 4

Ontology of genes subject to human-mouse conserved and “human-specific” activity-dependent exon usage. (A–C) Selected GO terms are shown which are enriched (Fisher’s weighted p < 0.05) in genes that have one or more 1:1 human-mouse orthologous exon which are subject to activity-dependent inclusion/exclusion in both Hum-ESC^CORT neurons and in one or more of our mouse cortical neuronal preparations (DIV4 and DIV10 Mus-PRIM^CORT, mESC^CORT-neurons). 782 genes contain exons that qualify as being regulated in a conserved manner by the above criteria, out of a background of 8039 genes (defined as 1:1 orthologous genes possessing ≥ 1 orthologous exons expressed in human and mouse neurons). The nature of the GO term is shown (BP, Biological Process; MF, Molecular Function; CC, Cellular Component). (D,E) Enrichment tests were performed for RBFOX and SAM68 target cassette exon splicing events (Jacko et al., 2018; Farini et al., 2020) (see Methods, *p < 0.0001 (Fisher’s exact test)). For (D) the presence of these target exons was compared between the set of all exons expressed in mouse neurons, and the set of exons expressed in mouse neurons for which there is a 1:1 human ortholog. For (E) the presence of these target exons was compared between the set of 1:1 orthologous exons subject to activity-dependent splicing in both human and mouse neurons [as per (A–C)], and the whole set of expressed 1:1 orthologous exons. (F,G) Selected GO terms are shown which are enriched (Fisher’s weighted p < 0.05) in genes that have a 1:1 human-mouse orthologue which are subject to activity-dependent alternative splicing in human neurons but not mouse neurons. In (G) the genes within selected GO terms that account for the enrichment are shown, and any genes in more than one GO term indicated by the overlapping nature of the Venn diagram. Note that while selected pathways are shown in this figure, all significantly enriched pathways are shown in Source_Data.xlsx.

FIGURE 5

Study of human and mouse gene basal and activity-dependent alternative splicing in mouse Tc1 neurons. (A) PSI of hCh21 exons (with a 1:1 mouse/human ortholog) in Hum-ESC^CORT neurons vs. the PSI of the same human exon in mouse Tc1 neurons. (B) PSI of the mouse orthologs of the exons from Figure 5A in mouse Tc1 neurons (x-axis) vs. PSI of the same exons in DIV10 Mus-PRIM^CORT neurons. (C) A comparison of the difference in basal PSI in orthologous exons within mouse Tc1 neurons compared to the difference in the same exons between mouse (DIV10 Mus-PRIM^CORT) and human (hESC^CORT) neurons. Correlation coefficient r is shown. (D) A comparison of KCl-induced PSI changes in human Ch21 exons in Hum-ESC^CORT neurons vs. mouse Tc1 neurons. (E) For alternatively spliced exons (80 > PSI > 20) the effect of KCl stimulation on PSI as a percentage of the maximum possible PSI change was compared between Hum-ESC^CORT neurons vs. mouse Tc1 neurons. For data points relating to this figure see Source_Data.xlsx.