Srsf1 and Elavl1 act antagonistically on neuronal fate choice in the developing neocortex by controlling TrkC receptor isoform expression

Abstract

The seat of higher-order cognitive abilities in mammals, the neocortex, is a complex structure, organized in several layers. The different subtypes of principal neurons are distributed in precise ratios and at specific positions in these layers and are generated by the same neural progenitor cells (NPCs), steered by a spatially and temporally specified combination of molecular cues that are incompletely understood. Recently, we discovered that an alternatively spliced isoform of the TrkC receptor lacking the kinase domain, TrkC-T1, is a determinant of the corticofugal projection neuron (CFuPN) fate. Here, we show that the finely tuned balance between TrkC-T1 and the better known, kinase domain-containing isoform, TrkC-TK+, is cell type-specific in the developing cortex and established through the antagonistic actions of two RNA-binding proteins, Srsf1 and Elavl1. Moreover, our data show that Srsf1 promotes the CFuPN fate and Elavl1 promotes the callosal projection neuron (CPN) fate in vivo via regulating the distinct ratios of TrkC-T1 to TrkC-TK+. Taken together, we connect spatio-temporal expression of Srsf1 and Elavl1 in the developing neocortex with the regulation of TrkC alternative splicing and transcript stability and neuronal fate choice, thus adding to the mechanistic and functional understanding of alternative splicing in vivo.

Graphical Abstract

Figure 1.

The balance between the two TrkC alternative splicing isoforms TK+ and T1 is regulated in a developmentally dynamic and cell type-specific manner. (A) Alternative splicing of the TrkC (Ntrk3) pre-mRNA produces the T1 and TK+ receptor variants. Two groups of mutually exclusive exons (13A-14A and 13–17) give rise to the distinct 3′ termini of the TrkC-TK+ and TrkC-T1 transcript variants. Correspondingly, these translate to distinct intracellular domains at the C-termini of the protein isoforms, giving rise to either a kinase domain (TK+) or a catalytically inactive domain (T1). Stop codons are indicated and demarcate the start of variant-specific 3′ UTRs. Binding sites for the probes used in RT-qPCR are indicated at the respective exon junctions. (B) The balance between TrkC-TK+ and TrkC-T1 changes during cortex development. In RNA prepared from cortices of increasing embryonic age, TaqMan quantitative real-time PCR for the two TrkC isoforms shows that the balance between TK+ and T1 shifts in favor of TK+ as cortex development progresses from embryonic day E 11.5 to E 18.5. N = 4. Bars: mean percentage of isoform from total TrkC expression (T1 plus TK+) ± SD. P value derived from unpaired, two-tailed Student's t test with Welch's correction. (C–E). The balance between TrkC-TK+ and TrkC-T1 in the developing cortex is cell-type specific. (C) Primary cortical cells from whole E 12.5 embryo litters were sorted into neuronal and stem cell populations by FACS after labelling with an anti-prominin-1 antibody (Prom-1) and collected for further analysis. (D) Example APC versus count plots used to distinguish viable Prom-1 positive and negative cells. Gating strategy according to signal from isotype control-stained cells. Complete gating strategy is presented in Supplementary Figure 1B. (E) RT-qPCR was performed on mRNA from the sorted neocortical cells, as described in (B). N = 3. P value derived from paired, two-tailed Student's t test. Pairing efficiency between Prom-1 + and Prom-1- results: r = 0.9962.

Figure 2.

Elavl1 and Srsf1 regulate TrkC alternative splicing in N2A cells and in primary cortical neurons. (A) Selection of splicing factors (SFs) and other RNA-binding proteins (RBPs) with potential involvement in TrkC alternative splicing. B-E. Elavl1 and Srsf1 control TrkC alternative isoform expression. (B) Strategy for radioactive splicing-sensitive RT-PCR for evaluating the TrkC-T1 and TrkC-TK+ splicing event. TrkC AS was assessed in N2A cell samples where RBPs defined in (A) were knocked down using siRNAs. (C) Exemplary result of a radioactive splicing-sensitive RT-PCR for TrkC-T1 and TrkC-TK+ on RNA from N2A cells treated with the indicated siRNAs. Percentage of TrkC-T1, as represented in (D) and (E), was quantified using a Phosphorimager and the ImageQuant TL software. Ctrl – siCtrl. (D) Summary plot for all tested RBPs and their effect on the proportion of the TrkC-T1 transcript variant normalized to TrkC-T1 percentage in the control siRNA samples. Gray dotted circles graduate the plot, indicating increases (positive values, outside the zero circle) or decreases (negative values, inside the zero circle) in TrkC-T1 percentage as compared to the siCtrl samples. Error bars were omitted for clarity. Statistically significant changes (siSrsf1 and siElavl1 samples) are represented separately with the corresponding descriptive and analytical statistical information in (E). (E) siRNA-mediated knockdown of Elavl1 or Srsf1 in N2A cells changes the ratio of TrkC-T1 to TrkC-TK+ significantly. N = 3. Bars: mean percentage of isoform from total TrkC expression (T1 plus TK+) ± SD. P values derived from Brown-Forsythe and Welch ANOVA with Dunnett's T3 multiple comparison post-hoc test. Overall P value: 0.0002. (F) Strategy for modulating SF levels in cortical neurons. Cortices from full litters of E 13.5 embryos were microdissected, dissociated into primary cortical cells and nucleofected with Srsf1 or Elavl1 expression plasmids, or empty expression constructs (pCAGIG = EV). Nucleofected cells were cultured for two days in vitro (DIV), after which total RNA or protein were extracted. (G) Srsf1 and Elavl1 alter transcript variant ratio of TrkC-T1 and TrkC-TK+ in cortical neurons. RT-qPCR on material from the nucleofected, cultured primary cortical cells, percentage of TrkC-T1 from total TrkC transcripts (T1 plus TK+) is shown. Lines represent paired replicates from the same experiment (cortical cells from one full litter split into the three nucleofection conditions). N = 4. P values from one-way ANOVA; overall P value: 0.0046. (H) Western blot of samples from (F), probed with a pan-TrkC antibody, which detects TrkC-TK+ (130 kDa) and TrkC-T1 (100 kDa). GAPDH was detected as a loading control. N = 4; P values from one-way ANOVA with Dunnett's T3 multiple comparison post-hoc test; overall P value: 0.0278.

Figure 3.

TrkC transcript levels are regulated by an Srsf1-dependent exonic splicing enhancer element in the first TrkC-T1-specific exon, exon 13A. (A) Splice site strength prediction of the Ntrk3 primary transcript. (B) Bioinformatic analysis of exon 13A suggests its subdivision in three major splicing-regulatory regions. The arrow points to a nucleotide predicted by HEXplorer to be of particular importance for conferring the splicing enhancer properties to fragment 13A-3. The dotted box marks the region most strongly impacted by this nucleotide and is shown magnified in Supplementary Figure 3B, along with the predicted effects of mutating this nucleotide. C-E. Minigene analysis of exon 13A splicing regulatory regions. (C) Splicing reporter used to assess enhancing or silencing properties of exon 13A fragments. The skipping and inclusion control vectors are described in (47). (D) The TrkC-T1 exon 13A reporter vectors were transfected into N2A cells and the splicing outcome assessed by RT-PCR. The mutation in 13A-3 predicted to disrupt Srsf1 binding impedes the splicing enhancing ability of this element, leading to a significant reduction of exon inclusion, as quantified in E. N = 3; P values derived from Brown-Forsythe and Welch ANOVA test with Šidak's post-hoc multiple comparisons test. Overall P value: <0.0001. The inclusion product of the 13A-3 reporter is larger due to the larger insert size (see B). (E) Quantification of (D). (F) To assess the involvement of Srsf1 in TrkC alternative splicing, TrkC-T1 exon 13A reporter vectors were transfected into N2A cells together with siRNAs as indicated, and the splicing outcome assessed by RT-PCR. N = 3. P values derived from ordinary ANOVA test with Šidak's post-hoc multiple comparisons test. Overall P value: <0.0001. (G) Quantification of (F). (H, I) Blocking the putative Srsf1 binding site in exon 13A of the TrkC pre-mRNA leads to a decrease in TrkC-T1 formation. A 2′-MOE antisense oligonucleotide complementary to the putative Srsf1 binding site in exon 13A (H) was transfected into N2a cells and its effect on TrkC alternative splicing assessed by RT-qPCR (I). N = 3; P values derived from ordinary one-way ANOVA with Šidak's post-hoc multiple comparisons test. Overall P value: <0.0001. (J, K). Radioactively labelled, in vitro transcribed RNA probes of the exon 13 A-3 fragment (E 13A-3) or of the same fragment with the mutation described in Supplementary Figure 4B (E 13A-3 mut.) were crosslinked by UV irradiation to nuclear extract proteins from N2a cells transfected either with an Srsf1 overexpression construct (pCAG-Srsf1) or with the empty vector (pCAGIG). Arrow: Srsf1 band (see also Supplementary Figures 3A and 4G).

Figure 4.

The ratio of Srsf1 to Elavl1 steers the alternative splicing choice between TrkCT1 and TrkC-TK+. (A) Srsf1 to Elavl1 ratios were modulated in N2a cells by transfection with either expression constructs, siRNAs against the two transcripts, or combinations thereof. Combinations are indicated above. The resulting ratio of Srsf1 to Elavl1 transcripts was determined by RT-qPCR and plotted below. The percentage of each transcript was calculated based on the C_T values by assuming one cycle difference in C_T to indicate a twofold difference. The bars are divided at the mean percentage from N = 3 replicates. Error bars represent the standard deviation of the three results. The ΔΔC_T values for Srsf1 and Elavl1 transcript levels relative to Hprt transcript levels and matching control sample are summarized in Supplementary Figure 5. (B) The effect of the Srsf1 and Elavl1 modulations in (A) was assessed by radioactive, splicing-sensitive PCR specific to the TrkC-T1/TK+ alternative splicing event, as described in Figure 2B. Each lane corresponds to the Srsf1/Elavl1 ratio and transfection conditions indicated above it in panel (A). The quantification was performed by normalizing the intensity of the TrkC-T1 band to the total signal from the TrkC-T1 and TrkC-TK+ bands. The bars are divided at the mean percentage from N = 3 replicates. Error bars represent the standard deviation from the mean of the three results. P values derived from ordinary one-way ANOVA with Šidak's post-hoc multiple comparisons test. (C) Correlation analysis of the percentage of Srsf1 in Srsf1 + Elavl1 transcripts with the percentage of TrkC-T1 in the TrkC-T1 + TrkC-TK+ transcripts for the experiment presented in (A) and (B). Dots represent individual biological replicates. pCAGIG – pCAG-IRES-GFP, empty vector; pCAG-Srsf1 – overexpression vector containing the Srsf1 CDS; pCAG-Elavl1 – overexpression vector containing the Elavl1 CDS; siCtrl – control siRNA (siAllstar); siSrsf1 – pool of siRNAs against mouse Srsf1; siElavl1 – pool of siRNAs against mouse Elavl1.

Figure 5.

Srsf1 and Elavl1 are differentially expressed in the developing neocortex. (A) Srsf1 mRNA levels are high in the ventricular zone and low in the intermediate zone and cortical plate, while Elavl1 is uniformly expressed. RNA in situ hybridization with probes against Srsf1 and Elavl1 showed different expression patterns in the developing cortex. Exemplary coronal sections of brains at the indicated embryonic stages show strong expression of Srsf1 in the ventricular zone, with much weaker expression outside this compartment. VZ – ventricular zone, CP – cortical plate, Pia m. – Pia mater, LV – lateral ventricle. Dashed boxes represent areas shown magnified in the last or last two magnification insets, respectively. (B) Srsf1 and Elavl1 are strongly co-expressed in neural progenitor cells in the developing neocortex. Expression plots generated using Seurat v4 and FindAllMarkers package for cells known to be part of the pyramidal neuron lineage. Expression levels generated by the SCTransform package. L – layer. NPCs – neural progenitor cells. Immat. neurons - immature neurons. CA1, CA3 - Cornu ammonis areas of the hippocampus; DG – dentate gyrus of the hippocampus. (C) The ratio of Srsf1 and Elavl1 mRNA levels in neural progenitor cells flips during corticogenesis. Data obtained from the same analysis as in (B), depicting the stage-specific expression of Srsf1 and Elavl1 mRNAs in different neural progenitor subsets. Asterisks denote significant differences between Elavl1 and Srsf1 expression values when testing the fit with a negative binomial model with stage and cell type as interaction terms. Underlying coefficients and further analyses are detailed in Supplementary Figure 6D–F. Pax6+ cells: apical NPCs. Tbr2+ cells: basal NPCs. (D) Elavl1 protein distribution differs from the distribution of its mRNA in the developing cortex. Immunofluorescent micrographs of E 11.5 to E 16.5 cortex sections stained with antibodies against Elavl1 and the neuronal marker MAP2 show an increased signal intensity for Elavl1 in the nascent and formed cortical plate as compared to the VZ/nascent SVZ as in the CP. VZ - ventricular zone, SVZ – subventricular zone, IZ – intermediate zone, CP - cortical plate, MZ – marginal zone. Asterisk denotes putative interneurons, which migrate into the neocortex beginning with E 15.5. Scale bars: 50 μm.

Figure 6.

Srsf1 and Elavl1 act antagonistically on TrkC-T1 levels to control the numbers of corticofugal neurons (CFuPNs) and callosal projection neurons (CPNs) in the developing cortex. (A) Srsf1 overexpression increases the number of CFuPNs in vivo,while its downregulation decreases it, and elicits the opposite effect on CPNs. Srsf1 overexpression constructs (pCAG-Srsf1-IRES-GFP) or empty vector constructs (pCAG-IRES-GFP, abbreviated pCAGIG) were electroporated into the lateral ventricles of E 12.5 embryos. Cells co-expressing GFP and one of the neuronal fate markers Ctip2 (CFuPN subtype) or Satb2 (CPN subtype) were quantified at E 16.5. N = 5. Similarly, plasmids expressing either a scrambled shRNA or an shRNA directed against Srsf1 were electroporated into the lateral ventricles of E 12.5 embryos. Brains were analyzed at E 16.5 as described in (A). shScrambled: N = 5. shSrsf1: N = 6. P values derived from unpaired Student's t test with Welch's correction. Box plot whiskers: minima and maxima of the sample. Horizontal line: median. Plus sign: mean of the sample. Empty arrows: GFP + Satb2 double-positive cells. Full arrows: GFP + Ctip2 double-positive cells. Scale bars = 50 μm. (B) Elavl1 expression level manipulations have an effect opposite to that of Srsf1 on the proportions of CFuPN/CPN neurons in vivo. Experiment performed as described in (A), using pCAG-Elavl1-IRES-GFP expression constructs. pCAGIG: N = 10. pCAG-Elavl1: N = 7. Similar to (A), we also used an shRNA against Elavl1. shScrambled: N = 4. shElavl1: N = 5. Statistics and labeling as in (A). Scale bars = 50 μm. (C) The effects of Elavl1 and Srsf1 on CFuPN-CPN neuron production depend on TrkC-T1 levels. Constructs were electroporated as shown and the quantification was performed as described in (A). The effects of Srsf1 overexpression are abolished upon the knockdown of TrkC-T1. N = 4 for pCAGIG + shScrambled, N = 5 for pCAG-Srsf1 + shTrkC-T1. Mean ± SD: Ctip2 - 1 ± 0.2237 in control versus 1.141 ± 0.1787 in pCAG-Srsf1 + shTrkC-T1; Satb2 - 1 ± 0.1109 in control versus 0.9914 ± 0.06534 in pCAG-Srsf1 + shTrkC-T1. Similarly, Elavl1 overexpression has no effect on the proportion of Ctip2 or Satb2 cells when TrkC-T1 levels are increased. N = 3 for both pCAGIG and pCAG-Elavl1 + pCAG-TrkC-T1. Mean ± SD: Ctip2 - 1 ± 0.2146 in control versus 1.007 ± 0.1723 in pCAG-Elavl1 + pCAG-TrkC-T1; Satb2 - 1 ± 0.1159 in control versus 0.99 ± 0.14 in pCAG-Elavl1 + pCAG-TrkC-T1. Statistics and labeling as in A. Scale bars = 50 μm. (D) Proposed model of the neuronal subtype fate regulation by Srsf1 and Elavl1 and TrkC isoform expression in the developing cortex. Srsf1 is labelled with question marks as we have only shown a change in mRNA.