The ciliogenic transcription factor RFX3 regulates early midline distribution of guidepost neurons required for corpus callosum development

Abstract

The corpus callosum (CC) is the major commissure that bridges the cerebral hemispheres. Agenesis of the CC is associated with human ciliopathies, but the origin of this default is unclear. Regulatory Factor X3 (RFX3) is a transcription factor involved in the control of ciliogenesis, and Rfx3-deficient mice show several hallmarks of ciliopathies including left-right asymmetry defects and hydrocephalus. Here we show that Rfx3-deficient mice suffer from CC agenesis associated with a marked disorganisation of guidepost neurons required for axon pathfinding across the midline. Using transplantation assays, we demonstrate that abnormalities of the mutant midline region are primarily responsible for the CC malformation. Conditional genetic inactivation shows that RFX3 is not required in guidepost cells for proper CC formation, but is required before E12.5 for proper patterning of the cortical septal boundary and hence accurate distribution of guidepost neurons at later stages. We observe focused but consistent ectopic expression of Fibroblast growth factor 8 (Fgf8) at the rostro commissural plate associated with a reduced ratio of GLIoma-associated oncogene family zinc finger 3 (GLI3) repressor to activator forms. We demonstrate on brain explant cultures that ectopic FGF8 reproduces the guidepost neuronal defects observed in Rfx3 mutants. This study unravels a crucial role of RFX3 during early brain development by indirectly regulating GLI3 activity, which leads to FGF8 upregulation and ultimately to disturbed distribution of guidepost neurons required for CC morphogenesis. Hence, the RFX3 mutant mouse model brings novel understandings of the mechanisms that underlie CC agenesis in ciliopathies.

Figure 1. Abnormal callosal axon pathfinding in Rfx3−/− mice.

(A–F) Immunohistochemistry for calretinin and Npn-1 (A1–A3 and B1–B3), for calbindin and Npn-1 (C1–C3 and D1–D3), and for GFAP and L1 (E1–E3 and F1–F3) in coronal CC sections from E18.5 WT (A1–A3, C1–C3 and E1–E3) or Rfx3−/− (B1–B3, D1–D3 and F1–F3) mice. A2, B2, C2, D2, E2 and F2 are higher magnifications of the lateral CC seen in A1, B1, C1, D1, E1 and F1, respectively. A3, B3, C3, D3, E3 and F3 are higher magnifications of the medial CC seen in A1, B1, C1, D1, E1 and F1, respectively. (A1–A3, C1–C3 and E1–E3) At E18.5, the hemispheres of WT brains have fused. Callosal fibres (in red) cross the midline and project into the contralateral cortex. (B1–B3, D1–D3 and F1–F3) Aberrant callosal axon bundles are observed in Rfx3−/− embryos (arrowheads). (B1–B3 and F1–F3) While the hemispheres have fused, most of callosal fibres do not cross the midline and form large ectopic bundles on the CC border, reminiscent of Probst bundles (PB). (D1–D3) Some Rfx3−/− embryos exhibit a more severe phenotype with an absence of midline hemispheric fusion and absolutely no callosal axons crossing the midline. In this case, a large bulge is observed along the inter-hemispheric fissure at the location where the callosal axons approach the midline (O). In all mutants, axonal defects are accompanied by cellular mis-positioning through the CC and the IG (open arrowheads). While calretinin+ or calbindin+ neurons, as well as, GFAP+ glia, are still present in the CC and the IG of Rfx3−/− mice, there is a midline disorganization and a lateral shift of these cell populations. Bar = 435 µm in A1, B1, C1, D1, E1, F1; 220 µm in A3, B3, C3, D2, D3, E2, F2 and 110 µm in A2, B2, C2, E3, F3.

Figure 2. Expression pattern of Rfx3 in the developing mouse telencephalon from E11.5 to E14.5.

(A–D) In situ hybridization for Rfx3 mRNAs on coronal brain sections of wild type (B1–B3) and Rfx3−/− (A1–A3) embryos at E11.5. In situ hybridizations for Rfx3 (in red) combined with immunohistochemical staining for calretinin (C1–C3 and D1–D3) (in green) on coronal brain sections of wild type embryos at E12.5 (C1–C3) and E14.5 (D1–D3). A1, B1, C1 and D1 are coronal sections at the corticoseptal boundary (CSB, *) level, while A3, B3, C3 and D3 are caudal coronal sections at the level of the cortical hem (CH). A2, B2 and C2 are higher power views of the CSB seen in A1, B1 and C1 respectively. D2 is a lateral view of the telencephalon. (A and B) At E11.5, Rfx3 is strongly expressed in wild type mice throughout the entire neuroepithelium of the CSB, and at more caudal levels in the cortical hem (B1–B3). The Rfx3 hybridization signal is specific since no signal is visible in the same brain area of Rfx3−/− (A1–A3). (C–D) From E12.5 to E14.5, Rfx3 expression is restricted to the cingulate cortex (CCi) that contains pioneer callosally projecting neurons, and throughout the CSB at the midline where the CC will form (C1–C2 to D1–D2). In addition, Rfx3 is detected within the glial wedge (GW, open arrows) and the septum. (C3 to D3) On more caudal sections, Rfx3 mRNAs are expressed in the cortical hem (CH), choroid plexus, ventral pallium (VP) and preoptic area (POA) (open arrows). Bar = 435 µm in A1, A3, B1, B3, C1, C3, D1, D2, D3 and 60 µm in A2, B2, C2.

Figure 3. Aberrant localization of midline neurons before CC formation at E14.5.

In situ hybridization for reelin mRNAs (A1–A3 and B1–B3) and DAB staining for calretinin (C1–C3 and D1–D3) and calbindin (E1–E3 and F1–F3) on coronal rostromedial slices from E14.5 WT (A1–A3, C1–C3 and E1–E3) and Rfx3−/− (B1–B3, D1–D3 and F1–F3) mice. A2, A3, B2, B3, C2 and D2 are higher power views of the corticoseptal boundary (CSB,*) seen in A1, B1, C1 and D1 respectively. E2, E3, F2 and F3 are higher power views of the induseum griseum (IG) region seen in E1 and F1, respectively. C3 and D3 are higher power views of the cortical marginal zone (MZ) seen in C1 and D1 respectively. (A–F) As early as E14.5, before inter-hemispheric midline fusion occurs, the organization of reelin+ (B1–B3), calretinin+ (D1–D2) and calbindin+ (F1–F2) midline neurons is severely affected in the CSB and the IG regions of the Rfx3−/− mutant (arrows and open arrowheads). In addition, the neurons lose their tangential organization through the Rfx3−/− cortical MZ (B3, D3 and F2–F3; arrows). Bar = 300 µm in A1, B1, E1, F1; 150 µm in A2, B2, C1, C3, D1, D3, E2, F2 and 60 µm in A3, B3, C2, D2, E3, F3.

Figure 4. Abnormal neuron localization and aberrant callosal axon pathfinding at the onset of CC formation.

(A–H) Immunohistochemistry for calretinin and reelin (A1–A2 and B1–B2), for calretinin and neuropilin-1 receptor (Npn-1) (C1–C2 and D1–D2), for calbindin and L1 receptor (E1–E2 and F1–F2) and for GFAP and L1 receptor (G1–G2 and H1–H2), in coronal CC sections from WT (A1–A2, C1–C2, E1–E2 and G1–G2) and Rfx3−/− (B1–B2, D1–D2, F1–F2 and H1–H2) mice. A2, B2, C2, D2, E2, F2, G2 and H2 are higher magnifications of the midline seen in A1, B1, C1, D1, E1, F1, G1 and H1. (A1–A2 to D1–D2) From E15.5 to E16.5, calretinin+ guidepost neurons fail to form a well organized band of neurons at the CSB (*) and are dispersed in the septum of Rfx3−/− mice (B2 and D2, white open arrowheads). Reelin+ and calretinin+ neurons loose their tangential organization through the cortical marginal zone (MZ) (compare B2 to A2, red open arrowheads). (E1–E2 and F1–F2) At E16.5, calbindin+ neurons (green) do not organize appropriately within the indusium griseum (IG) and accumulate at the CC midline in Rfx3−/− mice (compare F2 to E2, open arrowheads). (G1–G2 and H1–H2) At E16.5, the organization of GFAP+ glial cell populations within the CC is indistinguishable between WT and Rfx3−/− mice. (A to H) Axonal misrouting of pioneer callosal axons from E15.5 to E16.5. (A1–A2, C1–C2, and E1–E2) In WT brains, pioneer callosal fibres grow within the CSB and reach the midline. (B1–B2, D1–D2 and F1–F2) In Rfx3−/− brains, most callosal fibres form ectopic bundles of axons in the septum (B2 and D2) and the IG (F2) on either side of the midline (white arrowheads). Bar = 435 µm in C1, D1, E1, F1, G1, H1; 220 µm in A1, B1, C2, D2, E2, F2; 110 µm in G2, H2; 60 µm in A2, B2.

Figure 5. Midline integrity is necessary for pathfinding by callosal axons.

(A1) Experimental paradigm used to confirm the growth of E16.5 Rfx3+/+ control callosal axons in midline structure transplants from Rfx3+/+ control mice. (A2–A3) DiI labeling showing that WT callosal axons grow normally and cross the midline when they are confronted to a WT environment. (B1) Experimental paradigm used to confirm the growth defects of E16.5 Rfx3−/− callosal axons in midline transplants from Rfx3−/− mice. (B2–B3) DiI labeling showing that Rfx3−/− callosal axons are misrouted and do not cross the midline of Rfx3 mutants. (C1) Experimental paradigm used to study the growth defects of E16.5 control Rfx3+/+ callosal axons in transplants of midline structures from Rfx3−/− mice. (C2–C3) DiI labeling showing that WT callosal axons are misrouted and do not cross the midline of Rfx3 mutants. (D1) Experimental paradigm used to test whether the midline integrity is necessary and sufficient to direct the growth of callosal axons. To this end, control Rfx3+/+ midline regions are transplanted in Rfx3−/− slices. (D2–D3) DiI labeling showing the complete restoration of Rfx3−/− callosal axon pathfinding. Dashed lines outline the CC transplant localizations. Brain slices in A2–A3, B2–B3, C2–C3 and D2–D3 were counterstained with Hoechst. Bar = 435 µm in A2, B2, C2, D2 and 220 µm in A3, B3, C3 and D3.

Figure 6. Mice carrying a conditional deletion of Rfx3 in guidepost cells have a normal corpus callosum.

(A–B) Rfx3 mRNA (red) and GLAST protein (green) expression on coronal brain sections of control Rfx3 ^f/+ ;hGfap-Cre^−/− (A1–A2) and Rfx3 ^f/− ;hGfap-cre^+/− (B1–B2) embryos at E15.5. A2 and B2 are higher magnifications of the glial wedge (GW) seen in A1 and B1, respectively. (A1–A2) In control Rfx3 ^f/+ ;hGfap-Cre^−/− mice, Rfx3 is strongly expressed through the cortex, the induseum griseum (IG) and the CSB, as well as, the GW. (B1–B2) No more Rfx3 mRNA is detected in guidepost glia and neurons of Rfx3 ^f/− ;hGfap-Cre^+/− CC. (C–F) Immunohistochemistry for GFAP and L1 receptor (C and D) and for calretinin and Npn-1 receptor (E and F) in coronal CC sections from E18.5 control Rfx3 ^f/+ ;hGfap-Cre^−/− (C and E) and Rfx3 ^f/− ;hGfap-Cre^+/− (D and F) mice. In mice where Rfx3 is inactivated after E14.5 in midline neurons and glia, the CC and callosal axons develop normally. (G–H) Rfx3 mRNA (red) and calretinin protein (green) expression on coronal brain sections of control Rfx3 ^f/+ ;Emx1-Cre^−/− (G1-G2) and Rfx3 ^f/− ;Emx1-cre^+/− (H1–H2) embryos at E14.5. G2 and H2 are higher magnifications of the corticoseptal boundary (CSB, *) seen in G1 and H1, respectively. (G1–G2) In control Rfx3 ^f/+ ;Emx1-Cre^−/− mice, Rfx3 is strongly expressed through the calretinin+ glutamatergic neurons of the cortex and of the CSB. (H1–H2) In Rfx3 ^f/− ;Emx1-Cre^+/− brains, no more Rfx3 mRNA is detected in midline glutamatergic guidepost neurons of the CSB. (I–L) Immunohistochemistry for GFAP and L1 (I and J) and for calretinin and Npn-1 (K and L) in coronal CC sections from E18.5 control Rfx3 ^f/+ ;Emx1-Cre^−/− (I and K) and Rfx3 ^f/− ;Emx1-Cre^+/− (J and L) mice. Rfx3 inactivation after E12.5 in guidepost glutamatergic neurons of the CSB does not affect callosal axon navigation. Bar = 435 µm in A1, B1, G1, H1; 220 µm in C, D, E, F, I, J, K, L; 110 µm in G2, H2 and 40 µm in A2, B2.

Figure 7. Disturbed expression of Fgf8 and of the ratio of GLI3 repressor/GLI3 activator forms in Rfx3−/− CSB.

In situ hybridization for Fgf8 (A and B), Sprouty2 (C and D) mRNAs on coronal sections from E12.5 WT (A and C) and Rfx3−/− (B and D) mice at the CSB. Fgf8 expression domains is expanded into the rostromedial pallium in Rfx3−/− embryos. Interestingly, the frontier region between the septum and the cortex is reduced in the mutant compare to WT (arrowhead). In addition, Sprouty2 expression is slightly increased in the Rfx3 mutant. Bar = 500 µm in all figures. (E) Western blot analysis of E13.5 individual forebrains from Rfx3^+/+, Rfx3^+/− and Rfx3^−/− embryos from a same litter. As control, extracts from bodies of Gli3^+/+ or Gli3^Xt/Xt embryos were included. No GLI3 protein is produced in Gli3^Xt mutants allowing the identification of GLI3 specific bands. (F) Quantification of the Western blot shows that the ratio of GLI3 repressor form to the full-length form is reduced in Rfx3 deficient mice compared to heterozygotes and WT mice.

Figure 8. Fgf8 ectopic expression causes severe guidepost neurons mislocalization at the midline.

Immunohistochemical staining for calretinin in control (A1–A3, C1–C3) and FGF8-treated (B1–B3, D1–D3) coronal CC organotypic sections. (A2–A3, B2–B3, C2–C3, D1–D3) are higher power views of the CSB region (*) seen in (A1, B1, C1, D1). In control conditions, after BSA bath application (A1–A3) or after implanting BSA-coated beads in the rostromedial pallium (white arrowheads) (C1–C3), neurons labelled with calretinin are properly positioned in cortical layers and at the CSB (*) as it is observed in vivo. By contrast, FGF8 bath application (B1–B3) or FGF8-coated beads implantation in the rostromedial pallium (white arrowheads) (D1–D3) results in the complete disorganization of calretinin-positive neurons that are dispersed in the entire rostromedial pallium. They failed to form a proper structure at the CSB, in the IG and to organize in layers within the cortex. They disappear at the MZ or form aggregates (arrowheads). Bar = 435 µm in A1, B1, C1, D1; 220 µm in A2, B2, C2, D2; 110 µm in A3, B3, C3, D3.