Celsr3 and Fzd3 regulate the development of reciprocal thalamocortical projections independently of their expression in cortical or thalamic neurons. To understand this cell non autonomous mechanism further, we tested whether Celsr3 and Fzd3 could act via Isl1-positive guidepost cells. Isl1-positive cells appear in the forebrain at embryonic day (E) 9.5-E10.5 and, from E12.5, they form 2 contingents in ventral telencephalon and prethalamus. In control mice, corticothalamic axons run in the ventral telencephalic corridor in close contact with Isl1-positive cells. When Celsr3 or Fzd3 is inactivated in Isl1-expressing cells, corticofugal fibers stall and loop in the ventral telencephalic corridor of high Isl1 expression, and thalamic axons fail to cross the diencephalon–telencephalon junction (DTJ). At E12.5, before thalamic and cortical axons emerge, pioneer projections from Isl1-positive cells cross the DTJ from both sides in control but not mutant embryos. These early projections appear to act like a bridge to guide later growing thalamic axons through the DTJ. Our data suggest that Celsr3 and Fzd3 orchestrate the formation of a scaffold of pioneer neurons and their axons. This scaffold extends from prethalamus to ventral telencephalon and subcortex, and steers reciprocal corticothalamic fibers.
During brain wiring, integrated and protracted mechanisms control axonal growth, fasciculation and branching, target selection, and the formation of appropriate synapses. Guidance of axonal growth cones requires attractive and repellant signals (Tessier-Lavigne 2002) generated by diffusible factors, extracellular matrix molecules and membrane proteins expressed by so-called guidepost cells (Tissir and Goffinet 2013).
The finely organized reciprocal wiring between thalamus and cortex is critical to cortical function (Sherman and Guillery 2011) and results from a tight regulation during development (Lopez-Bendito and Molnar 2003; Molnar et al. 2012). Growing thalamic and cortical axons cross, respectively, the diencephalic–telencephalic junction (DTJ) and the pallial–subpallial boundary (PSPB) before entering a permissive “corridor” in the internal capsule (IC), paved with guidepost cells that assist the progression of axonal growth cones (Lopez-Bendito et al. 2006; Bielle et al. 2011). In addition to guidance by corridor cells, fibers from cortex and thalamus may help each other reciprocally, a mechanism referred to as “handshake” or “rendez-vous” (Molnar and Blakemore 1991; Hevner et al. 2002; Deck et al. 2013). In the corridor, subcerebral cortical projections segregate from corticothalamic axons and are further directed to their targets by additional mechanisms, which include Ryk/Wnt, IGF1, L1-CAM, and Eph/ephrin signaling (Cohen et al. 1998; Dottori et al. 1998; Liu et al. 2005; Ozdinler and Macklis 2006; Greig et al. 2013).
Previous work identified a crucial role for the atypical cadherin Celsr3 and the 7 transmembrane domain receptor Fzd3 in steering cortical and thalamic axons (Wang et al. 2002; Tissir et al. 2005). Conditional inactivation experiments showed that Celsr3 and Fzd3 act via expression in guidepost cells along the pathway, but not in cortical or thalamic neurons themselves (Zhou et al. 2008; Tissir and Goffinet 2013; Hua et al. 2014; Qu et al. 2014). As some cells in the IC corridor express Isl1 transcription factor (Lopez-Bendito et al. 2006), we studied the expression of Isl1, and the wiring phenotype of mice with conditional inactivation of Celsr3 and Fzd3 upon Isl1-Cre expression. We show that 2 different pools of Isl1-positive cells are present in the early ventral telencephalon and prethalamus, and that cortical and thalamic axons in the IC corridor run in close contact with them. In Isl1-Cre induced conditional mutants, cortical axons stall and loop after entering the IC corridor, in a more medial position than in Dlx5/6-Cre;Celsr3f/− mice, whereas thalamic axons do not cross the DTJ. In control mice, early pioneer axons from ventral telencephalic and prethalamic Isl1-positive cells cross the DTJ in opposite directions prior to the arrival of thalamic fibers. These reciprocal projections and a few intermingled migrating cells form a bridge that provides a substrate for growing thalamic axons. That bridge does not form in Celsr3 and Fzd3 mutant mice, accounting for the thalamus nonautonomous derailing of thalamic axons (Qu et al. 2014).
Materials and Methods
Animal procedures were approved by the Laboratory Animal Ethics Committee at Jinan University (Permit Number: 20111008001), and adequate measures were taken to minimize pain or discomfort for animals. To produce mice with regional inactivation of Celsr3, we crossed Isl1-Cre;Celsr3+/− males with homozygous ﬂoxed Celsr3f/f females to generate Isl1-Cre;Celsr3f/− animals; littermates with Isl1-Cre;Celsr3f/+ or Celsr3f/− genotype were used as controls. A similar strategy was used for regional inactivation of Fzd3. Isl1-Cre;Celsr3f/− and Isl1-Cre;Fzd3f/− mice are weak and have frequent hindlimb paresis, and most die around P21 (Chai et al. 2014). Rosa26Tomato and Rosa26YFP mice were crossed with Isl1-Cre mice to trace Cre-expression. The Thy1-YFP transgene was used to trace subcerebral tracts (Feng et al. 2000). Dlx5/6-Cre mice (Stenman et al. 2003) were used to compare distribution of Dlx5/6- and Isl1-positive cells.
Immunohistochemistry was carried out using the following antibodies: mouse anti-Calbindin (CB, 1:3000, Sigma), rabbit anti-Calretinin (CR, 1:400, Invitrogen), mouse anti-Reelin (Reln, G10, 1:2000, provided by AMG), rabbit anti-Pax6 (1:200, Covance), mouse anti-neurofilament (NF; 2H3,1:500, Hybridoma bank), mouse anti-Isl1 (39.4D5, 1:500, Hybridoma bank), rabbit anti-Isl1 (1:2000, Abcam), goat-anti cleaved Caspase3 (1:500, bcam), goat-anti-choline acetyltransferase (ChAT, 1:100, Millipore) and rat-anti Fzd3 (1:400, R&D). Cortical barrels were studied in vibratome sections of flattened cortex, stained with guinea pig anti-Vglut2 (1:2000, Millipore). Signal was detected with an anti-mouse/rabbit universal ABC kit (PK-6200, Universal, Vector) or the following fluorescent secondary antibodies: donkey anti-mouse Alexa fluor 546 (1:1000, Invitrogen), donkey anti-mouse Alexa fluor 488 (1:1000, Invitrogen), donkey anti-rabbit Alexa fluor 488 (1:1000, Invitrogen), donkey anti-rabbit Alexa fluor 546 (1:1000, Invitrogen), goat anti-guinea pig Alexa fluor488(1:1000, Invitrogen).
DiI and NeuroVue Tracing
E13.5 and P0 brains were fixed with 4% paraformaldehyde (PFA) overnight at 4°C. Small DiI crystals (D3911, Molecular Probes) were inserted in the thalamus or cortex using a tungsten needle. To trace pioneer axon projections, we prepared vibratome sections of E12.5 brains and implanted NeuroVue (24 837, NeuroVue Jade, Polysciences) or DiI in prethalamus or ventral telencephalon. To trace early thalamocortical axons, we prepared 600-µm thick oblique vibratome sections of E13.5 brains and implanted NeuroVue in dorsal thalamus. After incubation in phosphate buffer saline (PBS) containing 0.08% NaN3 at 37°C for 1–3 weeks, samples were cut (vibratome, 80 µm thickness) and observed under fluorescence microscopy.
Slice Culture and Neuron Tracing
Rosa26Tomato females were crossed with Isl1-Cre males and plugs were checked in the morning, scored as E0.5. Slice culture was carried out as reported (Zhou et al. 2007). Briefly, brains from E12.5 embryos were rapidly removed on ice and Isl1-Cre;Rosa26Tomato brains were selected under a fluorescence dissection microscope. Brains were embedded in 4% low melting agarose (1273c458, Amresce), and sectioned into 400-µm thick, oblique sections with a vibratome (MA752, Campden, USA). Selected sections containing prethalamus and ventral telencephalon were cultured in 12-well plates, in Dulbecco's modified Eagle's medium/F12 (DMEM-F12, 1 747 275, Gibco by Life Technologies) supplemented with B27 (1/50), G5 (1/100), and penicillin (1/1000) (all from Invitrogen). After recovering for 1 h at 37°C in the incubator, slices were positioned under a dissection microscope and small pieces of filter paper immersed in 10 µM CellTracker™ Green (CMFDA, c7025, Life Technologies) were inserted into the prethalamus or ventral telencephalon using a tungsten needle. Slices were returned to the incubator and cultured for 2 days. They were fixed with 4% PFA for 2 h at 4°C in the dark and mounted on slides after several washings in PBS. Images were captured with a confocal microscope (Zeiss LSM 780, Germany).
E13.5 brains were fixed in a solution containing 0.3% glutaraldehyde, 15% picronitric acid, and 4% PFA in 0.1 M phosphate buffer (pH 7.4), overnight at 4°C, and cut into 40–50 µm coronal sections using a vibratome. After rinses in PBS (pH 7.4), sections were treated with 1% NaBH4 and 0.3% H2O2 for 30 min, blocked with 0.5% bovine serum albumin and incubated with primary antibodies (mouse anti-NF and rabbit anti-Isl1) at 4°C overnight. Signal was detected with an ABC kit (PK-6200, Universal, Vector). Regions of interest were cut into ∼1-mm blocks which were treated with 1% osmium tetroxide and 1% uranylacetate, and embedded in EMbed 812 (Electron Microscope Sciences). Thin sections were stained with 2% uranyl acetate for 30 min and 1% lead citrate for15 min. Images were captured with a Philips 400 electron microscope.
In Situ Hybridization
Celsr3 digoxigenin-labeled riboprobes were prepared and used as described previously (Tissir et al. 2004). Briefly, E12.5 and E13.5 embryos were fixed in 4% PFA in PBS, cryoprotected in sucrose, and 20-µm thick cryostat sections were prepared under RNAse-free conditions. Sections were treated with 1 µg/mL Proteinase K (P-6556, Sigma) for 10 min, acetylated with 0.1 M triethanolamine/0.25% acetic anhydride for 15 min, washed, placed in prehybridization solution at 60°C for 4 h, and hybridized with 1.2 µg/mL probe at 60°C for 16–20 h. Signal was detected with anti-digoxigenin antibody (1:800, Roche) and visualized by NBT/BCIP staining.
Isl1 and Fzd3 RNAscope
E12.5 brains were dissected out in PBS, embedded with cryo-embedding medium (OCT) and frozen immediately on dry ice. Sixteen micrometer thick frozen sections were fixed in 4% PFA for 15 min at 4°C, dehydrated in serial ethanol (50%–70%–100%–100%–100%), each for 5 min, and processed according to the protocol provided in the RNAscope® kit (Advanced Cell Diagnostics. INC, Cat. 320 293). Briefly, after pretreatment with the prehybridization solution for 20 min at room temperature, the sections were hybridized with mixed probes of Isl1 and Fzd3 mRNA, for 2 h at 40°C. After 4-step amplification, the sections were counterstained with DAPI. All solution and probes are provided in the RNAscope® kit. The images were captured with a confocal microscope (Zeiss LSM 780, Germany).
Confocal and 2-Photon Microscopy
Confocal sections were made using a Zeiss LSM 700 or 780 and deconvolved using AutoQuantX (Media Cybernetics, USA) to improve the signal-to-noise ratio. A 2-photon microscope (Zeiss LSM 780) was used to prepare stacks of optical sections at 1 µm intervals in the region of the DTJ, and movies were generated using the Imaris software.
Isl1-Positive Cells Develop Early in Ventral Telencephalon and Prethalamus
Thalamocortical axons are guided by corridor cells in the IC, some of which express Isl1 and originate from the lateral ganglionic eminence (LGE) (Lopez-Bendito et al. 2006). However, when Celsr3 is inactivated in Gsh2-positive LGE progenitors (Corbin et al. 2003), no abnormal axonal bundles are found in the IC (Zhou et al. 2008), indicating that Celsr3 is not crucial for the function of LGE derived corridor cells or that some Isl1 guidepost cells originate from other sources than the LGE. To define better the localization and developmental history of Isl1-positive cells in the forebrain, we used Isl1 immunohistochemistry and Isl1-Cre;Rosa26Tomato mice. The first Isl1-positive cells were seen at the ventral aspect of the early diencephalon at E9.5-E10.5 (Fig. 1A1–B2,I). At E11.5, additional cells were found rostrally in the ventral telencephalon, and more caudally in the diencephalic region (Fig. 1C1–D2,I). This pattern was better defined at E12.5 and E13.5, with increasing numbers of Isl1-positive cells in ventral telencephalon, and segregation of 2 contingents in the diencephalon, a dorsal one in prethalamic areas (see below), and another one in the hypothalamic anlage, which will not be mentioned further (Fig. 1E1–H2,I). Unlike those in telencephalon, all Isl1-positive cells in the prethalamic region were Pax6 positive. They were negative for Reelin (Reln), Calretinin (CR), and Calbindin (CB) (Fig. 2), showing that they did not correspond to the thalamic eminentia, which is CR positive (Abbott and Jacobowitz 1999), nor to the zona limitans intrathalamica (also called “central diencephalic organizer”), which is Pax6-negative (Caballero et al. 2014). They also differed from the Reln-positive “reticular” thalamus, including zona incerta and ventral lateral geniculate nucleus (Schiffmann et al. 1997). The localization and marker expression of those Isl1-positive cells fitted best with their belonging to the prethalamus, a derivative of prosomere 3 (p3) (Vue et al. 2007; Scholpp and Lumsden 2010; Inamura et al. 2011).
As Celsr3 is required in guidepost cells that express Dlx5/6 (Zhou et al. 2008), we compared the distribution of Isl1- and Dlx5/6-positive cells, using Isl1 immunohistochemistry and EGFP expression from the Dlx5/6-Cre-EGFP transgene (Stenman et al. 2003). At E10.5, Isl1 was expressed at the midline, and EGFP laterally to the Isl1-positive region, with little overlap; some EGFP-positive cells were located in the ventral telencephalon where no Isl1-positive cell could be seen (see Supplementary Fig. 1A–C). At E13.5, Isl1-positive cells were abundant in the central tier of the ventral telencephalon, whereas the peak of EGFP expression was in the dorsolateral tier (see Supplementary Fig. 1D–F). At that stage, some Isl1-positive cells in the IC region were also EGFP-positive (see Supplementary Fig. 1G–I). In the diencephalon, the distribution of EGFP-positive cells was wider than that of Isl1-positive cells (see Supplementary Fig. 1F). Most Isl1-positive cells in prethalamus were EGFP-positive, whereas dorsally located EGFP-positive cells did not express Isl1 (see Supplementary Fig. 1J–L). Previous in situ hybridization (ISH) studies showed that Celsr3 and Fzd3 mRNA are widely expressed in ventral telencephalon and prethalamus. To confirm coexpression of Fzd3 and Isl1 proteins in the prethalamus and ventral telencephalon, we used IHC. As shown in Supplementary Figure 2A–D, at E12.5 most Isl1-positive cells were positive for Fzd3. The cellular colocalization of Fzd3 and Isl1 mRNAswas also shown using RNAscope (see Supplementary Fig. 2E–H).
In mice, the first thalamocortical and corticothalamic axons respectively reach the medial and lateral aspects of the ventral telencephalon around E13.5 and then progress rapidly in the corridor. To see whether Isl1-positive cells are closely apposed to growing axons and therefore qualify as candidate guidepost cells, we examined the region of the IC corridor using DiI tracing to label corticofugal axons, and Isl1-Cre;Rosa26YFP mice to label Isl1-positive cells. DiI-labeled axons ran almost parallel to one another and came in close contact with Isl1-positive cells (Fig. 3A), which was confirmed by NF and Isl1 double immunostaining (Fig. 3B). We then immunostained vibratome sections with mouse anti-NF and rat anti-Isl1 antibodies, detected signal with diaminobenzidine, and observed the distribution of the precipitate by electron microscopy. As shown in Figure 3C, NF-positive axons were found in direct contact with the surface of Isl1-positive cells, identified by their immunopositive nuclei. These results suggest that Isl1-positive cells in the IC are in a position to interact directly and serve as guidepost to axons of cortical and presumably thalamic origin.
What could be the fate of early Isl1-positive guidepost cells? To answer this question, we combined lineage tracing using Isl1-Cre and the Rosa26Tomato transgene, with IHC for the ventral telencephalic marker ChAT, and the apoptosis marker activated (cleaved) Caspase3. We examined ventral telencephalic areas at E15.5, P0, P5, and P10. ChAT immunoreactivity was detected in almost all cells from the Isl1 lineage from P5 onward, whereas activated Caspase3 immunoreactivity was very low, almost undetectable at all stages studied (see Supplementary Fig. 3). Those results indicate that most early born Isl1 cells, presumably including guidepost cells, do not undergo apoptosis and differentiate in cholinergic and possibly some other types of neurons in the postnatal period.
Corticothalamic and Corticospinal Projections are Defective in Isl1-Cre;Celsr3f/− Mice
Inactivation of Celsr3 upon Isl1-Cre expression was validated previously in E12.5 spinal cord (Chai et al. 2014). We confirmed that gene inactivation proceeded as predicted upon Isl1-Cre expression, by comparing Celsr3 mRNA expression by in situ hybridization in control Isl1-Cre;Celsr3f/+ and Isl1-Cre;Celsr3f/− mutant embryos at E13.5. We noted that the global signal was consistently weaker in mutant than in control samples, even in regions such as the cortex or ganglionic eminences that do not express Cre, because only one active allele (Celsr3f) is present in mutants, versus 2 (Celsr3f/+) in controls. In control embryos, Celsr3 mRNA signal was distributed in postmitotic cells in all cortical fields, the dorsal tier of the ventral telencephalon and the middle region of the IC, as well as in the diencephalon, including the hypothalamic anlage, the prethalamus and thalamus (see Supplementary Fig. 4A–C). In contrast, in mutant embryos (see Supplementary Fig. 4D–F), Celsr3 mRNA was preserved and detected in all cortical fields, ventral telencephalon, hypothalamic anlage and thalamus, though less intense for reasons explained above. Signal was completely absent, however, in the intermediate region of the IC, where the Isl1-positive signal is intense, and in the prethalamus where Isl1 expression is also high (see Fig. 1G). Thus, Celsr3 mRNA expression was inactivated in the central area of the ventral telencephalon and in prethalamus, consistent with the distribution of Isl1-positive cells described above.
The distribution of axonal bundles was compared in control and Isl1-Cre;Celsr3f/− mutant mice using anti-NF immunohistochemistry at E18.5–P0 (Fig. 4). Analysis of 10 mutant animals disclosed some variation in phenotype expressivity. In all of them, a prominent whorl of strongly NF-positive axons was evident at the center of the ventral telencephalon (Fig. 4B, arrow). In 4 out of 10, all thalamic axons were derailed in lateral hypothalamus and subpial telencephalic location (Fig. 4B, arrowhead). The phenotype was slightly less pronounced in the other 6 samples, in which most thalamic axons followed an aberrant ventral path in the telencephalon (see Supplementary Fig. 5G, arrow), whereas others followed a more normal path in the IC (asterisks in Supplementary Fig. 5B,G). This observation of incomplete phenotype expressivity in about half Isl1-Cre;Celsr3f/− mutants suggested that Celsr3 may act redundantly with Celsr2, as it does in other contexts (Qu et al. 2010, 2014; Chai et al. 2014). This was assessed by examining Isl1-Cre;Celsr2f/−;Celsr3f/− and Celsr2−/−;Isl1-Cre;Celsr3f/− mice with joint inactivation of Celsr2 and Celsr3. In all those samples, a drastic phenotype was found (see Supplementary Fig. 5C,D,H,I), comparable to the strongest phenotype generated by Celsr3 inactivation alone, confirming redundancy of Celsr3 and Celsr2. We examined Isl1-Cre;Fzd3f/− mice and saw that their IC phenotype was consistently very strong, similar to inactivation of both Celsr2 and 3 (see Supplementary Fig. 5E,J), in accord with previous observations (Chai et al. 2014; Hua et al. 2014; Qu et al. 2014).
Intriguingly, looping bundles in the middle tier of the IC corridor in Isl1-Cre;Celsr3f/− and Isl1-Cre;Fzd3f/− were located more medially than those in Dlx5/6-Cre;Celsr3f/− mice, where axons stall at the entry of the corridor, at a level facing the LGE (Zhou et al. 2008). To study thalamocortical projections, we implanted DiI crystals in the thalamus and cortex at P0. Upon DiI placement in thalamus in control brains, thalamocortical axons were followed through the IC and reached the cortex (Fig. 4C). By contrast, in mutant brains, whereas some thalamocortical axons followed the same path as in control brains (Fig. 4D, arrow), others were derailed ventrally (Fig. 4D, arrowhead). DiI injection in cortex in control animals labeled fibers along the IC and resulted in retrograde filling of thalamic neurons (Fig. 4E). Similar injections in the mutant cortex filled corticothalamic axons which looped in the IC (Fig. 4F, arrowhead); only a minority progressed to the thalamus where scarce cells were retrogradely labeled (Fig. 4F′). Subcerebral projection axons were traced at P20 using the Thy1-YFP transgene (Feng et al. 2000). In contrast to control axons, mutant subcerebral projection axons looped in the IC and a few were misrouted to the thalamus (Fig. 4G,H). Information from vibrissae is relayed by thalamocortical axons to the somatosensory cortex, where they end in cortical barrels (Waite and Cragg 1979). In P7 tangential sections, well-organized barrels were visualized in control brains, whereas barrels were absent or barely defined in mutant samples (Fig. 4I,J). Of note, when atrophic barrels were present in mutants with incomplete phenotype, they were still located at the expected location and layer (Fig. 4K,L). Thus, inactivation of Celsr3 in Isl1-positive cells results in failure of thalamocortical, corticothalamic, and subcerebral axons to progress through the IC.
Pioneer Axons From Prethalamus and Ventral Telencephalon Cross the DTJ in Celsr3/Fzd3-Dependent Manner
We wondered whether Isl1-positive guidepost cells in the ventral telencephalon and prethalamus could send pioneer axons across the DTJ, which would act like a bridge and guide later arriving thalamic and cortical axons. To assess this, we used anti-NF immunostaining, and Isl1-Cre;Rosa26Tomato transgenic mice to label Isl1-positive cell bodies and their neurites, at E12.5, before any thalamic or cortical axons reach the DTJ. Fibers originating from Isl1-positive cells were identified by double labeling for NF and the red fluorescent protein. In control embryos, prominent early fibers originating from Isl1-positive cells were seen in the ventral telencephalon and prethalamus, and crossed the DTJ (see Supplementary Fig. 6A–D), forming a bridge. In contrast, in Isl1-Cre;Celsr3f/− mutants, many early fibers were tomato-positive and therefore originated from early Isl1-positive cells, yet none crossed the DTJ (see Supplementary Fig. 6E–H). To provide more evidence for such an early bridge, we prepared vibratome sections at E12.5 in an oblique orientation to include prethalamus and ventral telencephalon (Fig. 5A), and used 2-photon imaging and tridimensional reconstruction to analyze the region of the DTJ. Isl1-positive cells and their neurites were labeled using the Rosa26Tomato transgene. In control samples (n = 11), fluorescent fibers joining the prethalamus and ventral telencephalon across the DTJ were clearly visualized (Fig. 5B, arrows; see Supplementary Movie 1). At later stages, E15.5 and P0, fiber bundles in the region increased in size due to addition of thalamocortical and corticothalamic fibers, among which early axons from the bridge appeared to subsist (Fig. 5G,H). In Isl1-Cre;Celsr3f/− mutants (n = 19), the bridge was completely absent in 14 samples (Fig. 5C,F; see Supplementary Movie 2), and partial (defined by a few fibers crossing the DTJ) in 5 samples (Fig. 5D, arrowheads, F). No bridge was found in any Isl1-Cre;Fzd3f/− mutant (n = 5, Fig. 5E,F).
To confirm that Isl1-positive cells in ventral telencephalon and prethalamus sent pioneer axons across the DTJ, we implanted small NeuroVue filter fragments in E12.5 vibratome slices (600-µm thickness) prepared from mutant (Isl1-Cre;Celsr3f/−;Rosa26Tomato) and control (Isl1-Cre;Rosa26Tomato) embryos. When NeuroVue was inserted in the prethalamus in control slices (n = 11), Isl1-positive cell bodies (78.82 ± 11.16 cells/sample) were retrogradely labeled in the ventral telencephalon (Fig. 6A,B,B′,E). Among 5 mutant samples, no retrogradely labeled cell bodies were seen in 4, and 17 retrogradely labeled cells were seen in one (Fig. 6C,C′,D,E). Reciprocally, upon NeuroVue insertion in the control ventral telencephalon (n = 10), Isl1-positive cell bodies (59.20 ± 11.60 cells/sample) were retrogradely labeled in the prethalamus (Fig. 6F,G,G′,J). However, only one mutant embryo showed very few labeled cells (10 cells per sample) and no cell bodies were retrogradely labeled in 4 other mutant samples (Fig. 6H,H′,I,J). Those tracing experiments show that Isl1-positive cells in the ventral medial telencephalon and in the prethalamus contribute pioneer axons to the bridge in both directions, across the DTJ. We also carried out DiI labeling experiments by placing tracer in the prethalamus in E12.5 Dlx5/6-Cre;Celsr3f/+ (control) and Dlx5/6-Cre;Celsr3f/− (mutant) embryos, combined with visualization of EGFP and Isl1 IHC to identify retrogradely labeled Dlx and Isl1 guideposts, respectively. Among retrogradely labeled cells in the ventral telencephalon, some were double positive for Dlx5/6 and Isl1, and some were only Isl1-positive (see Supplementary Fig. 7A,B,F), in accord with the partial colocalization (see Supplementary Fig. 1). In nice Dlx5/6-Cre;Celsr3f/− mutants, no retrogradely cells were found in 6 and only a few cells were found in 3 embryos (see Supplementary Fig. 7C,D,E,F), confirming that Dlx5/6 and Isl1-positive guideposts in ventral telencephalon both contribute to the early scaffold across the DTJ.
The bridge was formed by pioneer fibers among which a few cells with features of migrating cells appeared dispersed, as indicated by NF immunostaining of E12.5 Isl1-Cre;Rosa26Tomato embryos (Fig. 7A,B). To understand this better, we injected cell tracker green in prethalamus and ventral telencephalon in oblique slices at E12.5, and cultured slices for 2 days to trace putative migrating cells. As shown in Figure 7C–F, tracer injection in the prethalamus and the ventral telencephalon resulted in labeling of sparse cells in the ventral telencephalon and the prethalamus, repectively.
Thalamocortical Axons Progress on the Early Bridge Constructed by Isl1 Cells
To study the relationship of thalamocortical axons and the bridge, we implanted NeuroVue into the dorsal thalamus to trace thalamocortical axons in mutant (Isl1-Cre;Celsr3f/−;Rosa26Tomato) and control (Isl1-Cre;Rosa26Tomato) embryos at E13.5, the stage when the first axons from the thalamus cross the DTJ. In control samples (n = 12), a well-defined bundle of thalamocortical axons (Fig. 8A,B, green arrowheads) was seen at the dorsal aspect and in close contact with the bridge (Fig. 8A,B, red arrowheads), crossing the DTJ and reaching the ventral telencephalon directly. In 2 out of 10 mutant embryos, a few fibers crossed the DTJ in contact with very thin bridges (Fig. 8C,F); furthermore those axons ran erratically in the ventral telencephalon after crossing the DTJ (Fig. 8D, arrowheads). No thalamocortical axons crossed the DTJ in 8 other samples in which the bridge was completely absent (Fig. 8E,F).
Altogether, our data demonstrate that Celsr3 and Fzd3 are required for establishment of early reciprocal projections between Isl1-positive cells in medial ventral telencephalon and prethalamus. Those pioneer projections form an early bridge that provides a substrate and assists the progression of thalamic axons that begin to cross the DTJ one day later.
This study shows that expression of Celsr3 and Fzd3 in 2 early populations of Isl1-positive neurons, one in the ventral telencephalon and another one in prethalamus, is required to steer cortical axons in the IC corridor, and thalamic axons through the DTJ (Schema in Fig. 9).
Together with previous observations using Dlx5/6-Cre inactivation (Zhou et al. 2008), the present data show that, depending on which guidepost cell is lacking Celsr3 or Fzd3, corticofugal axons stall at different levels in the IC corridor. Inasmuch as corticothalamic axons grow normally in mice with inactivation of Celsr3 or Fzd3 in the cortical anlage (Zhou et al. 2008), the role of Celsr3 and Fzd3 in that context is therefore noncortical cell autonomous, in that both proteins are required in guidepost cells but not in cortical neurons of origin (Hua et al. 2014; Qu et al. 2014) . This suggests a model in which Celsr3 and Fzd3, as well as probably Linx (Mandai et al. 2014) act on cortical axons indirectly, by instructing the formation of an early permissive corridor in the ventral telencephalon, composed of Isl1-positive and other guidepost cells, which at that early stage are local pioneer neurons and their axons (Fig. 9A). Regional gene inactivation upon expression of Cre in different segments of that early scaffold may result in stalling of cortical axons at the corresponding defective segments (Fig. 9B–D). Intriguingly, the axon balling in the IC corridor generated by inactivation of Celsr3 or Fzd3 upon Dlx5/6 or Isl1-Cre expression is never found when inactivation is driven by Foxg1-Cre, which is expressed more widely than Dlx5/6 or Isl1, in ventral as well as dorsal telencephalon (Zhou et al. 2008). This fits in nicely with the known role of early cortical preplate pioneer axons in steering corticothalamic projections (McConnell et al. 1989). Celsr3 and Fzd3 could mediate reciprocal interactions among early neurons in preplate and ventral telencephalon and their axons across the PSPB. This might be analogous to the reciprocal projections that form the bridge at the DTJ (discussed below), even though a clearly defined anatomical bridge is not found at the PSPB, presumably because projections that cross that boundary are more widely dispersed and do not form as clearly defined a bundle as those that cross the DTJ (Lopez-Bendito et al. 2006; Molnar et al. 2012; Deck et al. 2013).
DTJ crossing requires Celsr3, Fzd3, and Linx (Wang et al. 2002; Tissir et al. 2005; Mandai et al. 2014). Celsr3 and Fzd3 are not required in thalamic axons themselves, but are crucial in guidepost cells (Qu et al. 2014). Our results identify those cells as p3-derived, Pax6- and Isl1-expressing cells, different than calretinin-positive cells from the thalamic eminence and Reln-expressing cells in reticular thalamus. Those data are in agreement with observations that: (1) Inactivation of Fzd3 upon Calretinin-Cre expression fails to generate any thalamic fiber misrouting (Hua et al. 2014); (2) The expression and function of Pax6 in prethalamus is required for thalamic axon navigation (Jones et al. 2002; Simpson et al. 2009); and (3) The Olig2 transcription factor, which is expressed in the ventricular zone of p3, regulates prethalamus formation and is crucial for thalamic fiber progression through prethalamus, and for DTJ crossing (Vue et al. 2007; Inamura et al. 2011; Ono et al. 2014).
Early pioneer axons emanating from Isl1-positive cells in medioventral basal forebrain and prethalamus cross the DTJ in both directions and a few Isl1-positive cells migrate among them from both directions. The early bridge formed by Isl1-positive pioneer cells and axons is followed by developing thalamic axons when they grow toward ventral telencephalon. This bridge cannot form in Celsr3 and Fzd3 mutant mice, and this presumably explains why thalamic fibers fail to turn and cross the DTJ in those mutants and possibly other mutants with similar phenotypes. At later developmental stages, some pioneer Isl1 ventral telencphalic guidepost cells differentiate into cholinergic neurons and their axons might be incorporated and diluted among GABAergic pallidothalamic fibers, whose development remains poorly understood (Metin and Godement 1996; Molnar and Cordery 1999; Ehrman et al. 2013; Jia et al. 2014). This is reminiscent of cortical subplate cells, which serve as guideposts and later become deep layer 6b neurons. Like for Isl1 guideposts, whether some subplate cells commit apoptosis after serving their guidepost function remains quite controversial (Hoerder-Suabedissen and Molnar 2013).
Our observations lead us to propose that Celsr3, Fzd3 and possibly Linx preside over the formation of an early scaffold made of pioneer neurons and their axons. This guidepost scaffold encompasses the bridge across the DTJ, the corridor in the ventral telencephalon, and early axons from ventral telencephalon and cortical preplate that cross the PSPB (McConnell et al. 1989; Hoerder-Suabedissen and Molnar 2015). Although molecular mechanism remain to be elucidated, it is tempting to consider a model inspired by Wnt-signaling complexes, which include, in addition to Frizzled receptors, 7 pass adhesion G-protein coupled receptors (GPR) such as GPR124 (Zhou and Nathans 2014), and Lrp5/6 transmembrane proteins (Tamai et al. 2000; Wehrli et al. 2000). Available data, particularly on conditional inactivation phenotypes, are compatible with the following hypothetical mechanism: 1) Fzd3 could impart directional information by reading gradients of diffusible ligands such as Wnt; (2) Homophilic interactions between Celsr3 cadherins, which belong to the family of adhesion GPR, would promote adhesion among guidepost neurons and/or their axons; and (3) Linx could act as an essential structural component of the Celsr3/Fzd3 complex, like Lrp5/6 in canonical Wnt-signaling complexes.
This work was supported by the following grants: National Natural Science Foundation of China (31070955, 31200826), National Basic Research Program of China (973 Program, 2014CB542205), Project of Internation as well as Hong Kong, Macao & Taiwan Science and Technology Cooperation Innovation Platform in Universities in Guangdong Province (2013gjhz0002), Programme of Introducing Talents of Discipline to Universities (B14036), Science & Technology Planning and Key Technology Innovation Projects of Guangdong (2014B050504006), Health and Medical Collaborative Innovation Major Projects of Guangzhou (201400000003-2), and by Belgian grants FNRS PDR T0002, Welbio CR-2012-007, IAP “Wibrain” P7/20 from Belspo to A.M.G. Funding to pay the Open Access publication charges for this article was provided by National Natural Science Foundation of China (31070955, 31200826).
We wish to thank K. Campbell for Dlx5/6-Cre mice and Fadel Tissir for critical comments. Conflict of Interest: None declared.