Sox2 Acts in Thalamic Neurons to Control the Development of Retina-Thalamus-Cortex Connectivity.

Visual system development involves the formation of neuronal projections connecting the retina to the thalamic dorso-lateral geniculate nucleus (dLGN) and the thalamus to the visual cerebral cortex. Patients carrying mutations in the SOX2 transcription factor gene present severe visual defects, thought to be linked to SOX2 functions in the retina. We show that Sox2 is strongly expressed in mouse postmitotic thalamic projection neurons. Cre-mediated deletion of Sox2 in these neurons causes reduction of the dLGN, abnormal distribution of retino-thalamic and thalamo-cortical projections, and secondary defects in cortical patterning. Reduced expression, in mutants, of Sox2 target genes encoding ephrin-A5 and the serotonin transport molecules SERT and vMAT2 (important for establishment of thalamic connectivity) likely provides a molecular contribution to these defects. These findings unveil thalamic SOX2 function as a novel regulator of visual system development and a plausible additional cause of brain-linked genetic blindness in humans.

Introduction

The development of specific neuron-to-neuron connections is central to the formation and function of the nervous system. In the mammalian visual system, the retina sends visual information to the thalamic dorso-lateral geniculate nucleus (dLGN); the dLGN neurons in turn connect to the visual cerebral cortex, which elaborates visual information. In development, retinal axons, forming the optic nerve, outgrow toward the dLGN, and then generate connections with a precise spatial distribution, linking retinal projections from each specific eye to specific dLGN subregions, in response to region-specific signals produced by the dLGN. In turn, dLGN neurons develop precise connections to the visual cerebral cortex (Garel and Lopez-Bendito, 2014, Gezelius and Lopez-Bendito, 2017). The gene regulatory networks active in dLGN neurons, specifying their connectivity programs, are still poorly understood, and are being investigated (Gezelius and Lopez-Bendito, 2017, Horng et al., 2009).

In humans, mutations in the gene encoding the Sox2 transcription factor cause severe visual disease (Fantes et al., 2003, Williamson and FitzPatrick, 2014). Conditional eye-specific knockout (KO) in mouse showed that, in the visual system, Sox2 plays important functions in retinal and lens progenitor cells' development (Pevny and Nicolis, 2010, Taranova et al., 2006, Smith et al., 2009, Kamachi et al., 2001). However, roles for Sox2 in other components of the visual system, in particular the dLGN, are still unexplored. So far, Sox2 function has been prominently demonstrated in stem cells (embryonic, neural) (Arnold et al., 2011, Avilion et al., 2003, Favaro et al., 2009, Pevny and Nicolis, 2010, Bertolini et al., 2019); instead, neuronal differentiation implies Sox2 extinction, with rare exceptions (Graham et al., 2003, Lee et al., 2014).

Here, we report that SOX2 is highly expressed in postmitotic, fully differentiated projection neurons of the dLGN. We find that its thalamic-specific ablation in these neurons impairs the development of dLGN, in particular its connectivity with the retina and the visual cortex, and we identify specific SOX2 downstream target genes that likely contribute to these defects.

Results

Sox2 Is Expressed in Thalamic Projection Neurons

We first investigated Sox2 expression in the thalamus by in situ hybridization (ISH) and immunofluorescence (IF) (Figure 1). ISH detects high Sox2 expression in the dorsal thalamus at perinatal stages (embryonic day [E] 17.5), and in the sensory thalamic nuclei, including the dLGN, at postnatal stages (P8) (Figure 1A). At these late stages, thalamic cells consist of differentiated neurons and glia (Gezelius and Lopez-Bendito, 2017). Indeed, IF on the postnatal dLGN shows that the vast majority of cells positive for NEUN (a general marker of differentiated neurons) are strongly positive for SOX2 (Figures 1B and 1D). A proportion of glial cells (about 50%), marked by S100β, are also weakly SOX2 positive (Figures 1C and 1D). Overall, most (89%) SOX2-positive dLGN cells are represented by neurons (Figure 1E). On the other hand, interneurons, marked by GAD67, are SOX2 negative (Figure S1A), indicating that, within neurons, SOX2 activity is mainly restricted to glutamatergic (projection) neurons. The majority of oligodendrocytes, marked by OLIG2, are SOX2 negative with some minor exceptions (Figure S1B).

Figure 1

Sox2 Thalamic Expression and Its Thalamic Ablation

(A) In situ hybridization with a Sox2 probe on sagittal sections of mouse brain at E17.5 and P8. Sox2 thalamic expression in the dorsal thalamus (dTh) at E17.5 and in the thalamic nuclei MG, dLG, and VP at P8 can be observed.

(B–C″) Immunofluorescence on coronal sections of dLG of mouse brain at P8 with anti-SOX2 (green) and anti-NEUN (red), a neuronal marker, antibodies in (B–B″) and with anti-SOX2 (green) and anti-S100β (red), an astroglial marker, antibodies in (C–C″). Arrows indicate cells co-expressing SOX2 and NEUN (B–B″) or SOX2 and S100β (C–C″) (n = 3).

(D) Quantification of the number of cells in the dLG at P8 positive for both SOX2 and NEUN out of the total number of NEUN-positive cells (blue) and of the number of cells positive for both SOX2 and S100β out of the total number of S100β-positive cells (orange). Around 80% of NEUN-positive neurons express SOX2. Data are represented as mean ± standard deviation.

(E) Quantification of the number of SOX2-and NEUN-positive cells in the dLG at P8 out of the total number of SOX2-positive cells (blue) and of the number of SOX2-and S100β-positive cells out of the total number of SOX2-positive cells (orange). Around 90% of SOX2-positive cells are neurons (n = 3). Data are represented as mean ± standard deviation.

(F–H′) In situ hybridization with a Sox2 probe on coronal sections of mouse brains at E15.5 (F and F′), E18.5 (G and G′), and P8 (H and H′) of Sox2 thalamic mutants Sox2ThΔ/Δ (F′, G′, and H′) and control littermates (Sox2ThΔ/+ or Sox2Th+/+) (F, G, and H). A clear ablation of Sox2 expression in the dorsal thalamus of Sox2 thalamic mutants (arrows) is observed at all stages compared with controls (E15.5 control n = 2, mutant n = 2; E18.5 control n = 3, mutant n = 2; P8 control n = 2, mutant n = 2). Scale bars, 600 μm in (A and F–H′) and 50 μm in (B–C″).

dTh, dorsal thalamus; MG, medial geniculate nucleus; dLG, dorso-lateral geniculate nucleus; VP, ventro-posterior nucleus.

Sox2 Deletion in the Developing Thalamus Leads to Reduced dLGN Size and Reduced Retinal Afferents to the dLGN

To delete Sox2 in the thalamus, we used a Sox2flox allele that we had previously generated (Favaro et al., 2009), in combination with a RORα-Cre transgene, active from E14.5, when thalamic neurons are already postmitotic (Chou et al., 2013). Sox2-efficient ablation was observed by ISH, already at E15.5 in the dorsal thalamus (Figures 1F and 1F′), and, subsequently (E18.5, P8; Figures 1G–1H′), in the dLGN and in the adjacent somatosensory thalamic nucleus (ventro-posterior nucleus [VPN]) (Chou et al., 2013). Sox2 thalamic mutants are termed Sox2ThΔ/Δ hereafter. IF shows that SOX2 is efficiently ablated from neurons, although it persists in at least some glia (Figures S1C and S1D).

To determine if Sox2 thalamic ablation results in defects in normal development of thalamic nuclei, in particular of the visual thalamic nucleus (dLGN), we analyzed by ISH the expression of the transcription factor Lef1, present in the dLGN, but almost absent in the adjacent ventro-lateral geniculate nucleus (vLGN) (Figures 2A–2D′). At the end of gestation (E18.5), the mutant dLGN did not overtly differ in size from controls (Figures 2A and 2A′); however, in postnatal development, the mutant dLGN failed to increase in size, contrary to controls, as seen by Lef1 ISH (Figures 2B–2D′) and Nissl staining (Figures 2E and 2E′).

Figure 2

The dLG Nucleus Is Reduced in Size in Sox2 Thalamic Mutants

(A–D′) In situ hybridization with a Lef1 probe on coronal sections of mouse brains at E18.5 (A and A′), P0 (B and B′), P7 (C and C′), and adults (D and D′) of Sox2 thalamic mutants (A′, B′, C′, and D′) and control littermates (A, B, C, and D). The dLG, marked by Lef1 expression, is reduced in size in the mutants starting at P0 (E18.5 control n = 6, mutant n = 6; P0 control n = 4, mutant n = 4; P7 control n = 5, mutant n = 3; adult control n = 2, mutant n = 2).

(E and E′) Nissl staining of coronal sections of mouse brains at P8 of Sox2 thalamic mutants (E′) and control littermates (E). The dLG is reduced in size in mutants compared with controls (control n = 2, mutant n = 2).

(F and F′) Immunofluorescence with an anti-vGLUT2 antibody (in red) on coronal sections of dorsal thalamus at P8 of Sox2 thalamic mutants (F′) and control littermates (F). Nuclei are marked by DAPI (blue). The mutant dLG marked by VGlut2-expressing retinal fibers appears reduced in size compared with controls (control n = 4, mutant n = 2). Dotted lines outline the dLG.

(G–H′) Immunofluorescence on coronal sections of mouse dLG at P8 of Sox2 thalamic mutants (G′ and H′) and control littermates (G and H) with antibodies anti-NEUN (a neuronal marker, in red) (G and G′) and anti-S100β (an astroglial marker, in red) (H and H′). Nuclei are marked by DAPI (blue) (control n = 3, mutant n = 3).

(I) Quantification of the percentage of NEUN-positive cells out of DAPI-positive cells in control and mutant dLG nuclei.

(J) Quantification of the percentage of S100β-positive cells out of total DAPI-positive cells in control and mutant dLG nuclei. A slight reduction of NEUN-positive neurons in the mutant dLG is observed. Error bars represent the standard deviation, * denotes a statistically significant difference p < 0.05 (p = 0.01308). Scale bars, 200 μm in (A–F′) and 50 μm in (G–H′). dLG, dorso-lateral geniculate nucleus; vLG, ventro-lateral geniculate nucleus.

In normal development of visual system connectivity, retinal axons outgrow toward the dLGN, and establish neuron-to-neuron connections with the cell bodies of dLGN neurons. The dLGN provides molecular signals that guide and precisely pattern the outgrowth of incoming retinal axon terminals to form appropriately localized connections to the dLGN during the perinatal and early postnatal periods; retinal afferents are, in turn, the source of important trophic signals, that allow the dLGN to complete its development and growth in early postnatal life (El-Danaf et al., 2015, Guido, 2018). IF with antibodies against the vesicular glutamate transporter 2 (VGLUT2) detects glutamatergic neuronal afferents from the retina, ensheathing and defining the dLGN. VGLUT2 IF showed an overall reduced area of the dLGN signal in the mutant at postnatal stages, confirming a reduced size (Figures 2F and 2F′), a result that is also observed at P16 and in adults (Figures S2A–S2B′). Interestingly, the area covered by the incoming VGlut2-positive fibers reaching the vLGN appeared slightly expanded (Figures 2F and S2A–S2B′). In addition, in both P16 and adult mutants, the signal was less uniformly distributed than in controls and more concentrated on the dorsal side of the dLGN (Figures S2A–S2B′).

A postnatally reduced size of the mutant dLGN (but not vLGN) was further confirmed by area measurements on DAPI-stained sections, in which retinal afferents had been stained with cholera toxin (Figures S2C–S2E; cholera toxin labeling will be described in Figure 3).

Figure 3

Defects in Retinal Projections to the dLG Are Found in Sox2 Thalamic Mutants

(A–H″ and K) Choleratoxin (CT) subunit B -Alexa 488 (green) and CT subunit B-Alexa 594 (red) were injected, respectively, in the right and left eyes, and contralateral (green) and ipsilateral (red) axons to the left dLG and vLG are shown (see scheme of injection in K modified from Seabrook et al., 2017). Series of coronal sections of the dLG from anterior (left) to posterior (right) of representative Sox2 thalamic mutants (B, D, G, and H) and littermate controls (A, C, E, and F) are shown. (A and B) CT was injected at P0, and brains were analyzed at P1. Less retinal axons are observed in the mutant dLG. Arrows mark the medial border of contralateral projections. (C and D) CT was injected at P7, and brains were analyzed at P9. Retinal axons arrive to a smaller dLG in mutants, and projections are often fragmented (arrowheads). (E–H″) CT was injected at P21, and brains were analyzed at P24. Retinal projections in mutants arrive to a smaller dLG and are sometimes fragmented (arrowheads) compared with controls (P1 control n = 4, mutant n = 4; P9 control n = 3, mutant n = 4; P24 control n = 3, mutant n = 3).

(I) Quantification of the percentage of contralateral (CL) retinal axons that reach the dLG or vLG in mutants and controls at P1 and P9 (***p < 0.005, unpaired Student's t test).

(J) Quantification of the percentage of retinal axons, contralateral (CL) or ipsilateral (IL), that reach the dLG or vLG in mutants and controls at P24 (P1 control n = 4, mutant n = 3; P9 control n = 2, mutant n = 4; P24 control n = 3, mutant n = 3). Error bars represent standard deviation; * denotes a statistically significant difference with unpaired Student's t test, **p < 0.01, ***p < 0.005. Scale bars, 200 μm. dLG, dorso-lateral geniculate nucleus; iLG, intermediate lateral geniculate nucleus; vLG, ventro-lateral geniculate nucleus.

In the mutant dLGN (P8), the frequency of neurons (NEUN-positive cells) was reduced (by 18%), whereas that of glia (S100β-positive) was not significantly altered (Figures 2G–2J).

Taken together, these results highlight that the mutant dLGN does not grow, postnatally, contrary to the control dLGN, and that, concomitantly, a reduction of retinal afferents reaching the mutant dLGN is observed.

Retino-Thalamic Projections Are Abnormally Distributed in Sox2 Thalamic Mutants

In normal development, naso-temporal retinal projections from each eye cross on the brain midline (optic chiasm) to reach a specific region of the contralateral dLGN, whereas a minority of the fibers from the ventro-temporal crescent of the retina do not cross and project onto the ipsilateral dLGN. dLGN regions receiving the contra- or ipsilateral fibers have well-defined, complementary, mutually exclusive shapes, that can be visualized by separately marking the retinal fibers originating from the right and left eyes with cholera toxin subunit B labeled with different fluorochromes (Figure 3; see drawing in Figure 3K). We labeled retinal afferents at three time points (P0, P7, P21), defining successive steps of afferents arrival to the dLGN, and segregation within it, and analyzed the projections to the thalamus 1–3 days later (P1, P9 and P24, Figures 3A–3H″). In mutants, retinal afferents to the dLGN are reduced already at the earliest time point (P1, Figures 3A and 3B), particularly in the medial dLGN region (arrows). Of note, the size of the dLGN is still comparable in mutants and control at this early stage. At later stages (P9, P24), in controls, contra (green)- and ipsilateral (red) afferents complete their segregation to different regions of the dLGN, as expected (Figures 3C and 3E). In mutants, segregation occurs, but the pattern of ipsi- and contralateral afferent fibers appears abnormal: ipsilateral (red) fibers show a different distribution (compare the mutants in Figure 3D, and particularly Figures 3G and 3H, versus controls in Figures 3C, 3E, and 3F, respectively); contralateral (green) fibers also look abnormal, with a somewhat fragmented, “clumpy” fiber distribution, in particular the most anterior (left) sections (Figure 3D compared with Figure 3C, Figure 3G compared with Figure 3E; arrowheads in Figures 3D and 3G point to “clumps,” i.e., local dishomogeneities, in mutant). In addition, at both early (P1) and later (P9, P24) stages, in mutants, the fraction of retinal afferents reaching the vLGN (measured as the green area in the vLGN) is proportionally higher than that reaching the dLGN when compared with controls, suggesting afferents misrouting (Figure 3D versus Figure 3C, and Figures 3G and 3H versus Figures 3E and 3F; see Figures 3I and 3J for quantifications). Finally, in mutants, an abnormally increased fraction of contralateral fibers was also apparent in the intermediate lateral geniculate nucleus (iLGN) (Figure 3G versus 3E), and, within the vLGN, some overlap between contra- and ipsilateral projections was visible, possibly as a result of the excess fibers projecting to the vLGN in mutants (see above; Figure 3G compared with Figure 3E).

We also traced the retinal axons reaching the dLGN through 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI) labeling in early postnatal life (at P0, and P5) (Figures S2F–S2G′). In thalamic Sox2 mutants, an overall reduction of retinal afferents reaching the dLGN is seen already at P0, and, more markedly, at P5 (Figures S2F–S2G′), matching the abnormality observed in the mutant by cholera toxin labeling.

Of note, the thalamic defects did not result in gross alterations of the adult retina or in changes of the number and distribution of the retinal ganglion cells (RGC; Figures S2H and S2I), indicating that, in spite of abnormal targeting, RGC viability is preserved.

Overall, these data indicate a role for Sox2 within dLGN cells, in the generation of appropriate connections of retinal afferents to the thalamus, suggesting a possible role for Sox2 in regulating guidance cues in the thalamus.

Thalamo-Cortical Connections and Normal Postnatal Patterning of the Cerebral Cortex Are Perturbed in Sox2 Thalamic Mutants

To directly investigate thalamo-cortical connections in mutants, we performed immunohistochemistry (on cortical flat mounts, scheme in Figure 4A, and telencephalon coronal sections) with antibodies against VGlut2 (Figures 4B and 4D), the serotonin transporter (SERT) (Figures 4C and 4E), and serotonin (5-hydroxytryptamine [5-HT[) (Figure 4F), all marking the developing thalamo-cortical axons (TCA) in the young postnatal brain (Chou et al., 2013, Lebrand et al., 1996). With all three antibodies, a reduction of the staining was observed in the visual cortex (V1) in Sox2ThΔ/Δ mutants (Figures 4B′–4F′, compared with Figures 4B–4F), indicating that TCA originating from mutant dLGN are strongly abnormal.

Figure 4

Thalamo-Cortical Connectivity and Cortical Patterning Are Abnormal in Sox2 Thalamic Mutants

(A) Representation of a cortical flat mount.

(B–C′) Immunohistochemistry on tangential sections of cortical flat mounts at P8 of Sox2 thalamic mutants (B′ and C′) and controls (B and C) with VGlut2 (B and B′) and serotonin transporter (SERT) (C and C′) both expressed by thalamic projections to the cortex. Staining in the mutant visual cortex is reduced (*) with both markers compared with controls (VGlut2 control n = 16, mutant n = 7; SERT control n = 6, mutant n = 3).

(D and D′) Immunofluorescence with VGlut2 on coronal sections of mouse brains at P8. VGlut2-positive thalamo-cortical projections to V1 are reduced in Sox2 mutants (D′) compared with controls (D) (white arrow) (control n = 4, mutant n = 2).

(E–F′) Immunohistochemistry on coronal sections of mouse brains at P8 of Sox2 thalamic mutants (E′ and F′) and controls (E and F) with SERT (E and E′) and serotonin (5HT) (F and F′). Staining in the mutant V1 is reduced with both markers (white arrows) (SERT control n = 4, mutant n = 2; 5-HT control n = 5, mutant n = 3).

(G–H″) (G–H′) Whole-mount in situ hybridization for the cortical markers Lmo4 (G and G′) and Bhlhb5 (H and H′) (schematic expression pattern in G″ and H″) on Sox2 thalamic mutants and littermate controls. The mutant visual area (V1) is not clearly distinguished from the neighboring higher-order visual areas (VHO) compared with control V1 (see *) (Lmo4 control n = 6, mutant n = 3; Bhlhb5 control n = 11, mutant n = 5).

(I–I″) (I and I′) Coronal sections at P8 after insertion of DiI crystals (scheme in I″ modified from Chou et al., 2013) in the primary visual cortex (V1) (red) and DiA in the somatosensory cortex (S1) (green) showing axons projecting to the dLG in controls and not in Sox2 thalamic mutants (control hemispheres n = 28, mutant hemispheres n = 15). Scale bars, 600 μm. V1, primary visual area; VHO, higher-order visual area; S1, primary somatosensory area; A1, primary auditory area; dLG, dorso-lateral geniculate nucleus; VP, ventro-posterior nucleus; MG, medial geniculate nucleus; F/M, motor cortex.

The correct development of TCA, projecting from the dLGN to the visual cortex, is essential for the development of the cortical visual areas, in particular for the diversification of the primary visual area (V1) and the adjacent higher-order visual areas (VHO) (Chou et al., 2013). In the normal postnatal brain (P7), the Lmo4 gene is expressed in VHO, but not in the adjacent V1, with a clear boundary between positive and negative regions; in the mutant, however, this boundary is not well defined, and expression in V1 is more pronounced and more similar to that in VHO (Figures 4G and 4G′). We also obtained similar results with two different markers, Bhlhb5, normally expressed in the V1, but less so in the VH0 area (Figures 4H and 4H′), and Rorβ (not shown). The observed cortical defect is reminiscent of alterations previously found in Nr2f1 (COUP-TF1) thalamic mutants obtained by RORα-Cre-mediated deletion of the Nr2f1 gene, encoding a transcription factor important for dLGN neurons development. In these mutants, the cortical defects are thought to be secondary to defective TCA connections of the dLGN to the visual cortex (Chou et al., 2013).

Overall, the detection of a cortical patterning defect in our thalamic mutant, together with the abnormalities of the incoming thalamic connections (as indicated by the reduced VGlut2 staining), indicates a defect in the development of TCA, affecting their ability to correctly pattern the postnatal cortex.

The thalamus is also the target of cortico-thalamic axon afferents (CTA) (Garel and Lopez-Bendito, 2014). We injected DiI (red) and 4-(4-(dihexadecylamino)styryl)-N-methylpyridinium iodide (DiA) (green) in the visual and somatosensory cortex, respectively (scheme in Figure 4I″); these compounds diffuse along the neuronal membranes, allowing visualization of neuronal projections. Labeling of visual cortex with DiI showed that development of CTA to the dLGN was compromised in Sox2ThΔ/Δ mutants, whereas the connections between the somatosensory cortex and the VPN were comparatively less affected (Figures 4I–4I″).

Overall, our observations show that Sox2 ablation affects TCA development, and, secondarily, cortical patterning and CTA development.

Specific Genes Important for dLGN Development Are Downregulated in the Mutant dLGN

To identify Sox2 target genes whose deregulation in Sox2 mutants might be responsible for the observed defects of connectivity, we studied the expression of candidate genes known to play key roles in dLGN development, by ISH and immunohistochemistry (Figures 5 and S3). In the dLGN, (and VPN), Sox2 was co-expressed with Nr2f1 (data not shown); however, no important reduction was detected, in mutants, in the expression of Nr2f1 by ISH (Figures 5B and 5B′) or immunohistochemistry (at E18.5 and P8, not shown). Also, no change was found in the expression of other genes normally active, and important, in or for the dLGN, tested at E18.5 or at E15.5: Zic4 (Horng et al., 2009), NtnG1, Sema6A (Gezelius and Lopez-Bendito, 2017, Little et al., 2009), Gbx2 (Miyashita-Lin et al., 1999, Sur and Rubenstein, 2005), Sox11, Klf6, Zic1, and Ntn1 (Braisted et al., 2000) (Figures 5C–5E′, S3B–S3D′, and S3F–S3H′). This indicates that a “general” dLGN gene expression program was retained in mutant dLGN.

Figure 5

Efna5 Expression Is Reduced in the dLG of Sox2 Thalamic Mutants, and a DNA Region within the Efna5 Locus Is Bound and Activated by SOX2

(A–E′) In situ hybridization on coronal sections of mutant (A′, B′, C′, D′, and E′) and control (A, B, C, D, and E) mouse forebrain at E18.5 with Sox2 (A and A′), Nr2f1 (B and B′), Zic4 (C and C′), Netrin G1 (NtnG1; D and D′), and Semaphorin 6A (Sema6A; E and E′) probes. Sox2 expression is clearly ablated in the mutant thalamus (A′ compared to A), whereas the expression of the other markers appears unchanged (at least three controls and three mutants were analyzed for each probe).

(F–H″) Time course of Ephrin-A5 (Efna5) expression by in situ hybridization on coronal sections of controls, heterozygotes, and homozygote Sox2 thalamic mutant forebrains at E15.5 (F–F″), E18.5 (G–G″), and P7 (H–H″). Efna5 is specifically downregulated in the dLG and VP of Sox2 homozygote thalamic mutants (white arrows), whereas it remains unchanged in other domains of expression both at E18.5 (G–G″) and P7 (H–H″). A mild Efna5 downregulation is also observed in Sox2 heterozygotes (G′ and H′) (E15.5 control n = 3, heterozygote n = 3, mutant n = 4; E18.5 control n = 8, heterozygote n = 7, mutant n = 8; P7 control n = 3, heterozygote n = 3, mutant n = 5).

(I) Sox2 ChIP-seq profile across the Efna5 locus from neurosphere cultures derived from postnatal mouse telencephalon showing two Sox2 peaks (5′, red circle; 3′, blue circle) within the first intron of the Efna5 gene.

(J) Schematic representation of plasmids containing the 3′ (blue) or 5′ (red) Sox2-bound regions upstream of a minimal promoter (grey) and luciferase reporter gene (yellow).

(K) Co-transfection in Neuro-2a cells of the constructs in (J) with increasing amount of SOX2-expressing vector or with the corresponding empty vector (+, 1:0.06; ++, 1:0.125; +++, 1:0.187; ++++, 1:0.25; +++++, 1:0.5 reporter:transactivator molar ratio). The previously known activation of the Nkx2.1 promoter by the SOX2-expressing vector was used as a positive control of transactivation (Ferri et al., 2013). Results are the mean of three independent transfections, each performed in triplicate. Error bars represent standard deviation. Scale bars, 200 μm. hip, hippocampus; dLG, dorso-lateral geniculate nucleus; VP, ventro-posterior nucleus; dTh, dorsal thalamus.

However, in the mutant dLGN we observed a strong reduction in the expression of ephrin-A5 (encoded by the Efna5 gene) by ISH E18.5, i.e., before the visible patterning defect, and postnatally (P7) (Figures 5F–5H″). Of note, some reduction was consistently observed also in heterozygotes (Figures 5G′ and 5H′, compared with Figures 5G and 5H).

Ephrin-A5 is a signaling molecule involved in axon guidance in the developing brain (Kania and Klein, 2016) and is expressed, in the normal dLGN, in a gradient. Germline KO studies showed that, in the absence of Efna5, 2, and 3, the establishment of the correct pattern of contra- and ipsilateral retina-dLGN projections is severely disrupted (Huberman et al., 2005, Pfeiffenberger et al., 2005, Vanderhaeghen et al., 2000); these previous results suggest that Efna5 downregulation in Sox2 mutants may significantly contribute to the observed defective patterning of retinal afferents. Interestingly, SOX2 chromatin immunoprecipitation (ChIP-seq) detects two SOX2-bound regions in the Efna5 gene in neural stem or progenitor cells, located in the first intron (Figure 5I). We cloned each of the two DNA regions encompassing the SOX2 peaks upstream of a minimal promoter and a luciferase reporter gene (Figure 5J) and co-transfected the constructs with increasing amounts of a SOX2 expression vector (Ferri et al., 2013, Mariani et al., 2012, Panaliappan et al., 2018) into neural (Neuro-2a) cells (Figure 5K). The most upstream peak (5′ peak) did not show any response over the SOX2 amounts tested; the 3′ peak, however, showed a strong dose-dependent transactivation of the luciferase reporter, to an extent similar to that previously observed with the promoter of the Nkx2.1 gene, a previously identified SOX2 target (Ferri et al., 2013) (Figure 5K). These observations indicate the presence of a SOX2-responsive regulatory region (putative enhancer) within the Efna5 gene.

Further to Efna5 downregulation, we had observed a reduction of serotonin (5-HT), and its transporter SERT in TCA at P8 (see above, Figure 4). As serotonin levels in the first two postnatal weeks have been shown to have a role in regulating thalamo-cortical projections (see below), we investigated in more detail the levels of serotonin, and of components of its pathway, during development, following Sox2 thalamic loss.

5-HT marks the TCA during the time window when they first establish their patterned connections with cortical neurons. 5-HT is not synthesized by thalamic neurons, but is transiently uptaken by TCA via the serotonin transporter SERT (Lebrand et al., 1996); pharmacological manipulation of 5-HT uptake, or knockout of the SERT-encoding gene, perturbs sensory TCA development, particularly to the somatosensory cortex, indicating a developmental function for this transient 5-HT uptake (Chen et al., 2015, Gaspar et al., 2003, Persico et al., 2001). This raised the possibility that an early reduction in 5-HT content within thalamic neurons might, as well, contribute to the observed dLGN TCA development defect in Sox2 mutants. Indeed, we observed a reduction of 5-HT levels in cell bodies and axons in the mutant dLGN (more pronounced in homozygous mutants but also detected in heterozygotes) already at P1 (when the dLGN size is still not importantly reduced in homozygous mutants) (Figures 6A–6A″) and at P8 (Figures 6B–6B″). Furthermore, IF for 5-HT indicated that intracellular 5-HT accumulation is not observed in the mutants at P8 (Figures 6C–6C‴; compare Figures 6C and 6C′, control, showing intracellular perinuclear red 5-HT staining, with Figures 6C″ and 6C‴, mutant, showing strongly reduced intracellular 5-HT). We then investigated the expression of the SERT-encoding gene by ISH at early stages (E18.5), preceding the overtly defective phenotype (Figures 6D–6D″). SERT mRNA signal intensity was reduced in Sox2ThΔ/Δ and, to a lesser extent, in heterozygotes, at E18.5 (Figures 6D–6D″) (when the mutant dLGN size is similar to control), and P7 (Figures S3I–S3I″). We also observed, at E18.5, a reduction in the expression of the mRNA encoding the vesicular monoamine transporter 2 (vMAT2) (Figures S3J–S3J″), the transporter that packages 5-HT into synaptic vesicles, protecting it from degradation (Gaspar et al., 2003). These findings indicate that 5-HT metabolism or transport is compromised in Sox2 thalamic mutants, which might plausibly contribute to the abnormal development of thalamo-cortical connectivity, in accordance with previous observations on the somatosensory TCA connections (Persico et al., 2001, Chen et al., 2015, Gaspar et al., 2003). SOX2 ChIP-seq in neural stem or progenitor cells detects low-level SOX2 binding to the promoter region of the SERT-encoding gene Slc6a4 (Figure S4A), whereas in the vMAT2-encoding gene (Slc18a2), three intragenic SOX2-binding peaks are detected (Figure S4B). This suggests that SOX2 might directly participate in the regulation of the vMAT2, and, possibly, the SERT-encoding gene.

Figure 6

Serotonin Levels in Mutant dLG Are Reduced

(A–B″) Immunohistochemistry for serotonin (5-HT) on coronal sections of Sox2 mutant (A″ and B″), Sox2 heterozygote (A′ and B′), and control (A and B) mouse brains at P1 (A–A″) and P8 (B–B″). The level of serotonin in the dLG is dependent on the number of copies of the Sox2 gene. White arrows indicate the ventral border of the dLG, and dotted lines outline the dLG.

(C–C‴) Immunofluorescence for 5-HT (red) on coronal sections of dLG of Sox2 thalamic mutants (C″ and C‴) and controls (C and C′) at P8. 5-HT levels in the mutant dLG are greatly reduced. Nuclei are marked by DAPI (blue). Note perinuclear 5-HT in controls. C′ and C‴ are magnifications of details in C and C″, respectively (P1 control n = 4, heterozygote n = 3, mutant n = 3; P8 control n = 3, heterozygote n = 2, mutant n = 3).

(D–D″) In situ hybridization on coronal sections of E18.5 forebrains of Sox2 homozygote mutant (D″), Sox2 heterozygote (D′), and control (D) with a SERT probe. SERT is downregulated in the Sox2 mutant thalamus compared with control (black arrows). A mild downregulation is also observed in Sox2 heterozygous thalami indicating a dose-dependent effect of Sox2 loss (control n = 5, heterozygote n = 3, mutant n = 4). Scale bars, 200 μm. dLG, dorso-lateral geniculate nucleus; vLG, ventro-lateral geniculate nucleus; VP, ventro-posterior nucleus.

Overall, our findings indicate that Sox2 controls, possibly through direct binding, downstream target genes important for dLGN development.

Alterations of the Thalamo-Cortical Connections and of the Patterning of the Cerebral Cortex Are Also Observed in the Somatosensory Axis in Sox2 Thalamic Mutants

In examining the components of the visual system, the focus of the present article (dLGN, visual cortex), we noticed that the ventro-posterior (VP) somatosensory thalamic nucleus and VP thalamo-cortical connectivity also showed abnormal features, suggestive of a more general role for Sox2 in the development of sensory organs-thalamo-cortical connectivity. Within the VP thalamic nucleus, that receives somatosensory input from the periphery and, in turn, projects to the somatosensory cortex, Sox2 is highly expressed in wild-type mice, similarly to the dLGN, but is absent in the Sox2 mutant (Figure 1). We found, by VGlut2 immunohistochemistry, that the map organization of barreloids in the mutant VPN appears perturbed (Figures 7A–7A″). We then visualized TCA from the VP to the somatosensory primary cortex (S1) at P8 by immunohistochemistry for VGlut2 on tangential sections of flattened cortices; we observed that the general topographic organization of the TCA projecting to the S1 area was affected in the mutant cortex. Not only was VGlut2 staining less intense but also the number of barrel fields, as outlined by VGlut2 staining, was reduced compared with controls (Figures 7B–7B″).

Figure 7

Thalamo-Cortical Somatosensory Afferents and the Somatosensory Thalamic Nucleus (VP) Are Affected

(A–A″) Immunofluorescence for VGlut2 at P8 on coronal section of the VP thalamic nucleus of controls (A′) and Sox2 thalamic mutants (A″). (A) Drawing of a typical coronal section through the thalamus (control n = 4, mutant n = 2).

(B–B″) Immunohistochemistry for VGlut2 of tangential sections of cortical flat mounts (drawing in B) of controls (B′) and Sox2 thalamic mutants (B″) (control n = 16, mutant n = 7).

(C–E′) Immunohistochemistry on coronal sections of mutant and control forebrain at P8 with SERT (C and C′), serotonin (5-HT; D and D′), and VGlut2 (E and E′) antibodies. Barrel fields are not detected in mutant cortex (see *) (SERT control n = 4, mutant n = 2; 5-HT control n = 5, mutant n = 3; VGlut2 control n = 4, mutant n = 2).

Scale bars, 200 μm in (A′–B″) and 600 μm in (C–E′).

VP, ventro-posterior nucleus; S1, primary somatosensory cortex; V1, primary visual cortex; A1, primary auditory cortex; dLG, dorso-lateral geniculate nucleus; hip, hippocampus.

A similar phenotype was observed by immunohistochemistry for SERT, serotonin, and VGlut2 on coronal sections at P8 (Figures 7C–7E′); in particular, cortical barrel fields were less defined and weakly stained (Figures 7C–7E′).

Interestingly, the expression of the Efna5, SERT, and vMAT2 genes, which we had found downregulated in the mutant dLGN, is concomitantly downregulated also in the mutant VP (Figures 5F–5H″, 6, and S3I–S3J″). These findings suggest that at least some of the Sox2-dependent gene regulatory network in the thalamus is shared between different thalamic nuclei.

Discussion

Severe defects of vision are a hallmark of Sox2 deficiency in humans. Known mechanisms underlying the severe visual defects in SOX2 deficiency included, so far, the recognition of the functional importance of SOX2 in the development of retina and crystallin, as revealed by eye-specific mouse conditional KOs (see Introduction). The importance of SOX2 for the development of dLGN and for its retinal or cortical connections demonstrated in this article provide a new, important site of Sox2 function in the visual system, and an additional potential explanation for the visual impairment in SOX2-mutant patients.

The strong expression of SOX2 in postmitotic dLGN projection neurons (Figure 1) was unexpected. In fact, Sox2 functions are critical in stem cells of various types, in particular embryonic, neural, and others, in which it is necessary to preserve stemness (Avilion et al., 2003, Bertolini et al., 2016, Favaro et al., 2009, Kondoh and Lovell-Badge, 2016, Bertolini et al., 2019), and in the reprogramming of differentiated cells to stem cells (Takahashi and Yamanaka, 2016). In contrast, in several neural cell types such as differentiated neurons and glia, Sox2 is either not expressed, or important in very specific cell types, such as retinal Müller glia (Taranova et al., 2006) or cerebellar Bergmann glia (Cerrato et al., 2018).

The development of connectivity between retinal, dLGN, and visual cortex neurons is the result of complex interactions between developing neurons and their environment, involving signaling molecules and their receptors, which are often co-expressed in the growth cones of outgrowing neurons as well as on the target neurons. The possibility to selectively knock out Sox2 in the dLGN (and not the cortex or the retina) with RORα-Cre allowed us to unambiguously attribute the defects observed in mutants in the development of visual connectivity to defects arising, at least primarily, in thalamic neurons, as a result of Sox2 loss.

One important defect consists in a reduction, and a mispatterning, of neuronal afferents from the retina that reach the dLGN (Figures 3 and S2). As Sox2 was not deleted in the developing eye, these defects must primarily depend on defective Sox2 function in the thalamus itself. It is known that retinal innervation plays an important trophic role in dLGN development, particularly in the early postnatal phase (El-Danaf et al., 2015, Guido, 2018), at a time when the Sox2 mutant thalamus fails to continue growing in size as seen in controls (Figure 2 and S2). Thus, the reduced retino-thalamic innervation may contribute to the reduced dLGN size seen postnatally in thalamic Sox2 mutants. KO experiments identified signaling molecules, and their receptors, important for the directional development of axons, and their appropriate targeting. Among these, we find ephrin-A5 to be importantly downregulated in the mutant dLGN (Figure 5). Of note, Efna5 is expressed in both thalamus and retina (Pfeiffenberger et al., 2005), and the KOs that demonstrated the importance of ephrin-A5 for the correct patterning of ipsi- and contralateral eye fibers (Pfeiffenberger et al., 2005) ablated the Efna5 gene throughout the brain (including thalamus and retina); our results are in agreement with the hypothesis that the Efna5 gene, regulated by Sox2, plays at least part of its function in retinal axon guidance by acting specifically in the thalamus.

Our finding that SOX2 can directly bind the Efna5 gene (Figure 5) yields a potential direct Sox2 target within dLGN neurons, and potential regulatory regions for the Efna5 gene, whose future molecular study could provide insight into dLGN neurons gene regulatory networks.

Sox2-mutant thalamic neurons of the sensory nuclei dLGN (and VP) fail to undergo normal post-natal development, and to generate normal axon outgrowths providing appropriate connections with the cortex (visual and somatosensory, respectively) (Figures 4, 5, and 7). Importantly, these two nuclei are the only ones (with the auditory nucleus) among the many thalamic nuclei that express high levels of Sox2 in neurons. In addition, in Sox2 mutants, both nuclei show decreased expression of genes known to control axon connectivity (ephrin-A5, and SERT) (Figures 5, 6, and S3). In particular, it is known that the knockout of SERT affects VP-originating fibers reaching the somatosensory cortex (Gaspar et al., 2003, Persico et al., 2001, Chen et al., 2015, Teissier et al., 2017); studies of the TCA connecting the dLGN to the visual cortex in SERT mutants were not reported. In Sox2 mutants, the connectivity defects between VP and somatosensory cortex mirror the defect seen in SERT KO mice (Figure 7) (Gaspar et al., 2003, Persico et al., 2001); hence the reduction of SERT and 5-HT expression in Sox2 mutants, together with the altered connections between dLGN with cortical visual areas, and of VP with somatosensory areas, is in agreement with the hypothesis that Sox2-dependent regulation of serotonin transport or metabolism may contribute to sensory TCA connectivity.

It should be noted that additional molecular or cellular defects may be relevant to the abnormal connectivity of thalamic nuclei to the cortex. The observed reduced size of the dLGN (Figure 2), implies a reduction of the number of projections to the cortex (Figure 4). However, the size of the VPN is not, or much less, reduced in mutants, when compared with the dLGN (see Figures 2A–2D′and 6A–6B″), yet thalamic afferents connecting the VPN to the somatosensory cortex (normally forming the barrel fields, see Figure 7, controls) are severely abnormal and disorganized in mutants (Figures 7B–7E′). In addition, the distribution of afferents from the periphery to the VP (forming the barreloids) also appears abnormal (Figures 7A–7A″). Hence, overall, factors other than thalamic nuclear size, specifically conditioning the patterning of neuronal afferents to and from the thalamus, appear to be involved in the observed defects.

Many other defects, such as altered expression of additional signaling molecules, or their receptors, in thalamic neurons, may likely contribute to the alterations observed in Sox2 mutants, and remain to be identified. Future studies of gene expression in mutant thalamic neurons at the genome-wide level (Kalish et al., 2018), will allow to investigate this point in more detail.

A cortical patterning defect in thalamic Sox2 mutants, i.e., the poor definition of V1 and VHO regions of the visual cortex, first pointed to the existence of abnormal TCA in mutants (Figure 4).

Very similar defects were described following thalamic ablation, via RORα-Cre, of Nr2f1, a gene whose mutation in humans leads to cerebral visual impairment (Bosch et al., 2014, Bosch et al., 2016), raising the question of whether they may functionally interact. We find wide co-expression of SOX2 and NR2F1 in dLGN neurons (not shown); however, we do not detect major alterations in Nr2f1 levels in the mutant dLGN (Figure 5), ruling out that the failure to activate Nr2f1 in dLGN is a key mechanism underlying the defects in Sox2 mutants. The similarities in the phenotype of the two mutants might be due to a general biological process, i.e., a reduction of TCA will always result in VHO patterning defects; due to deregulation of a set of common critical Sox2/Nr2f1 targets, controlling TCA development; or due to a combination of both mechanisms. Future combined studies of Sox2 and Nr2f1 targets will clarify this point.

The identification of Sox2 targets, such as Efna5, SERT, and vMAT2 that are downregulated not only in homozygous Sox2 mutants but also in heterozygotes (Figures 5, 6D, S3, and S4), points to potential candidates for roles in mediating Sox2 function in human visual defects; interestingly, in human patients, heterozygous Sox2 mutation is sufficient to cause blindness, pointing to dosage-sensitive gene regulation by SOX2 as an important phenomenon in the pathogenesis of these defects.

Genetic defects of vision are often polygenic in nature; importantly, some of these defects are dependent on brain, rather than eye, abnormalities (Williamson and FitzPatrick, 2014), but so far SOX2 has only been considered a gene responsible for anophthalmia and microphthalmia, i.e., congenital defects that primarily affect the eye. Our findings raise the possibility that genetically altered Sox2 levels in the thalamus (as obtained by regulatory mutations, or by mutations in Sox2-controlling transcription factors) may also play roles in brain-related visual defects. The importance of Sox2 for gene regulatory networks in the thalamus could allow to identify overlaps with the gene regulatory networks controlled, in the thalamus, by different genes, whose mutation also causes visual disease (as hypothesized for Nr2f1, see above). A better understanding of these networks has the potential to generate new, unifying hypotheses for therapeutic approaches.

Limitations of the Study

Sox2, being a transcription factor, potentially regulates hundreds of genes; it is thus likely that many factors, encoding a variety of molecules affecting the development of neuronal connectivity and dLGN growth in size, are misregulated in Sox2-mutant thalami. The present work identified candidate molecules, based on previous knowledge of their potential effects; it will be necessary to study, by genome-wide RNA sequencing, the alterations of gene expression in mutant versus wild-type dLGN (and VP), to obtain an unbiased catalog of potential effectors of the altered phenotypes. Functional rescue studies will have to be attempted, to confirm the roles in disease of candidate genes.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.