SOX2 and NR2F1 coordinate the gene expression program of the early postnatal visual thalamus

Abstract

The thalamic dorsolateral geniculate nucleus (dLGN) receives visual input from the retina via the optic nerve, and projects to the cortical visual area, where eye-derived signals are elaborated. The transcription factors SOX2 and NR2F1 are directly involved in the differentiation of dLGN neurons, based on mouse work and patient mutations leading to vision defects. However, whether they regulate each other, or control common targets is still unclear. By RNA-seq analysis of neonatal dLGN from thalamo-specific Sox2 and Nr2f1 mouse mutants, we found a striking overlap of deregulated genes. Among them, Vgf, encoding a cytokine transported along thalamic-cortical axons is strongly downregulated in both mutants. Direct SOX2 binding to some of these genes was confirmed by CUT&RUN, which identified a SOX2 chromatin-binding pattern characteristic of the dLGN. Collectively, our genetic and molecular analyses on the SOX2 and NR2F1-coregulated genes contribute to our understanding of the gene regulatory network driving the differentiation and connectivity of thalamic neurons, and the vision impairments caused by mutations in these genes.

Keywords: CUT&RUN; DLGN; Nr2f1; Sox2; Thalamus; Vgf; Vision.

Fig. 1.

RNA-seq of the visual thalamus (dLGN) of Sox2 and Nr2f1 thalamic mutants identifies many genes differentially expressed in both mutants. (A) dLGN immunofluorescence with antibodies recognizing SOX2 (green) and NR2F1 (magenta) at P0 (wild type, con). Scale bar: 200 µm. (B) Sections comprising the dLGN at P0 used for RNA-seq analyses, before (left) and after (right) dissection. Scale bar: 200 µm. (C) Numbers of significantly dysregulated genes, identified in RNA-seq experiments, in Sox2 and Nr2f1 mutants, and in both mutants are shown. (D) Table showing Sox2 RNA levels (TPM) in dLGN of Nr2f1 mutants and controls (Nr2f1^flox/flox, Nr2f1^flox/+) and Nr2f1 RNA levels (TPM) in dLGN of Sox2 mutants and controls (Sox21^flox/flox, Sox2^flox/+) (TPM for each biological replicate for controls and mutants are shown). (E) GO analysis of genes differentially expressed in both Sox2 and Nr2f1 mutants (DEG; the 514 genes in B) reveals enrichment in categories involved in neuronal development. The GO Biological Processes, Cellular Component and Molecular Function categories that are significantly enriched within the indicated mutants are shown.

Fig. 2.

CUT&RUN of SOX2 binding in the P0 visual thalamus. (A) Schematic depiction of the CUT&RUN experimental design. Two independent biological replicates for SOX2 and an anti-HA negative control were performed from pools of six dorsal lateral geniculate nuclei (dLGN) from three brains. (B) Venn diagram overlap and signal intensity plots of the two SOX2 replicates, showing enrichment over the control in all peak regions. (C) Peak region annotation by HOMER, showing that SOX2 binds primarily promoter, intronic and intergenic regions in dLGN. (D) HOMER known motifs for dLGN SOX2 peaks. The SOX2 motif is the most highly ranked, followed by other SOX factors. (E) GO enrichment of biological processes for genes associated by GREAT to SOX2 dLGN peaks. Dot size shows number of peak associated genes, dot color represents −log10 FDR (FDR<0.05), and the x axis represents fold enrichment. The top 20 terms are shown. Enrichment terms include neuron development related processes. (F) Pie chart depicting the distribution of SOX2 dLGN peaks in unique peaks and those shared with NSC SOX2 datasets. (G) HOMER known motifs for dLGN and NS shared peaks. NFY and SOX2 are the top motifs. (H) HOMER known motifs for dLGN unique peaks. Top motifs include ZIC3 and RORα. Sox2 is ranked 10th. (I) CUT&RUN tracks as visualized in Integrative Genome Viewer (IGV), showing both dLGN and NS shared peaks (left) and dLGN unique peaks (right). (J) Schematic depiction of CUT&RUN and RNA-seq overlap, showing Sox2 peak associated genes that are transcribed (>5 TPM, 784/1102) and those that are differentially expressed (DEG) in Sox2 mutant dLGN (FDR<0.01, 327/784), and those that are up- (92) or downregulated (145) in the mutant. (K) STRING map of interactions between SOX2, NR2F1, and the 79 SOX2 dLGN direct targets that are dysregulated in both Sox2 and Nr2f1 mutant mice. Confidence was set on default (0.4), disconnected nodes were removed, and line thickness represents confidence. Red nodes are those known to be involved in neurogenesis, while blue nodes are not included in the neurogenesis set and could represent novel genes important to the generation of neurons in the visual thalamus. (L) CUT&RUN tracks showing SOX2 dLGN peaks near the important targets Vgf (left), Sox5 (center, intronic region), and Sox13 (right). While the peaks near Vgf and Sox5 are unique to the dLGN datasets, the Sox13 peak is also bound by SOX2 in NS.

Fig. 3.

Deconvolution analysis documents the loss of projection neurons transcriptional identities in Sox2 and Nr2f1 mutants. Cell type abundance variation in Sox2 mutant and control (Sox2^flox/flox or Sox2^flox/+) samples (A) and Nr2f1 mutant and control (Nr2f1^flox/flox or Nr2f1^flox/+) samples (B) expressed as fraction of cells of a given type gained or lost with respect to the average abundance across all samples considered. The cell types considered are those defined by their scRNA-seq transcriptional identity in (Kalish et al., 2018). Data (estimated fraction of each cell type) for all three mutant and control samples are shown numerically in Table S4 and graphically in A, B.

Fig. 4.

ISH and immunofluorescence document Vgf and Sox5 downregulation in Sox2 and Nr2f1 thalamic mutants. (A) Top, ISH with Nr2f1, Vgf and Sox5 probes on coronal sections of E18.5 brains of Sox2 mutants (Sox2 cKO) and control littermates (Sox2^flox/flox or Sox2^flox/+). Bottom, ISH with Sox2, Vgf and Sox5 on coronal sections of E18.5 brains of Nr2f1 mutants (Nr2f1 cKO) and control littermates (Nr2f1^flox/flox or Nr2f1^flox/+). At least four mutants and five controls were analyzed with each probe. Arrows indicate the dLGN. Note reduced expression of Vgf and Sox5 in the dLGN of both mutants compared to the respective controls. Nr2f1 and Sox2 appear very slightly affected in the Sox2 mutant and Nr2f1 mutant, respectively. Scale bar: 200 µm. (B) IF with antibodies recognizing SOX5 (red) on coronal sections of E18.5 brains of controls (Sox2^flox/flox, Sox2^flox/+; Nr2f1^flox/flox, Nr2f1^flox/+), Sox2 CKO and Nr2f1 CKO. DAPI (blue) marks nuclei. At this stage, the dLGN is only modestly reduced in the mutants (Chou et al., 2013; Armentano et al., 2007). Arrows point to SOX5-positive cells within the dLGN region. Note the reduction of the SOX5-positive area in the Sox2 and Nr2f1 mutant dLGN. SOX5 positivity is not altered in the adjacent hippocampal region in mutants. Scale bar: 200 µm. (C) Graphs show the quantification of SOX5 pixel intensity within the dashed rectangles in B, in controls, Sox2 cKO and Nr2f1 cKO dLGNs. Large dots represent the mean intensity for each brain analyzed and the small dots represent each individual measurement. Mean±s.e.m. is indicated. *P<0.05, **P<0.01, ***P<0.001; unpaired t-test. The results shown are representative of n=4 mutants and n=4 control brains analyzed.