The corticothalamic and thalamocortical tracts play essential roles in the communication between the cortex and thalamus. During development, axons forming these tracts have to follow a complex path to reach their target areas. While much attention has been paid to the mechanisms regulating their passage through the ventral telencephalon, very little is known about how the developing cortex contributes to corticothalamic/thalamocortical tract formation. Gli3 encodes a zinc finger transcription factor widely expressed in telencephalic progenitors which has important roles in corticothalamic and thalamocortical pathfinding. Here, we conditionally inactivated Gli3 in dorsal telencephalic progenitors to determine its role in corticothalamic tract formation. In Emx1Cre;Gli3fl/fl mutants, only a few corticothalamic axons enter the striatum in a restricted dorsal domain. This restricted entry correlates with a medial expansion of the piriform cortex. Transplantation experiments showed that the expanded piriform cortex repels corticofugal axons. Moreover, expression of Sema5B, a chemorepellent for corticofugal axons produced by the piriform cortex, is similarly expanded. Finally, time course analysis revealed an expansion of the ventral pallial progenitor domain which gives rise to the piriform cortex. Hence, control of lateral cortical development by Gli3 at the progenitor level is crucial for corticothalamic pathfinding.
Cortex and thalamus together represent a highly integrated processing unit for sensory information from the external environment. Both structures are interconnected by the thalamocortical tract which conveys sensory information to the cortex and by the corticothalamic tract which sends processed sensory information back to the thalamus thereby providing the feedforward and feedback mechanisms essential in this processing unit (Grant et al. 2012). Establishing these reciprocal pathways during embryonic development requires thalamocortical (TCAs) and corticothalamic axons (CTAs) to cover large distances, change direction several times and pass through a variety of brain territories. CTAs initially grow laterally through the cortical intermediate zone (IZ) to reach the pallial/subpallial boundary (PSPB). After a waiting period (Molnar and Cordery 1999; Jacobs et al. 2007), they enter the ventral telencephalon (VT) which acts as an important intermediate target by providing guide post cells and by secreting several axon guidance molecules (Metin et al. 1997; Richards et al. 1997; Braisted et al. 1999; Tuttle et al. 1999; Hevner et al. 2002; Tissir et al. 2005; Uemura et al. 2007; Zhou et al. 2008; Magnani et al. 2010; Molnar et al. 2012). In the VT, CTAs also interact with TCAs which migrate in the opposite direction and according to the handshake hypothesis (Molnar and Blakemore 1991) guide each other into their target area in the thalamus and cortex, respectively.
While the mechanisms underlying the guidance of CTAs through the VT have been extensively studied, much less is known how their entry into the subpallium is regulated. After reaching the PSPB, corticofugal axons exhibit a waiting period (Molnar and Cordery 1999; Jacobs et al. 2007), which is regulated by Sema3E/PlexinD1 signaling and which is required to coordinate their further progression through the internal capsule with the arrival of TCAs (Deck et al. 2013). Notably, CTAs enter the VT in a wide domain along the whole PSPB boundary. In contrast to the temporal regulation of CTA entry into the subpallium, the spatial requirements allowing CTAs to enter the subpallium in this wide domain remain unexplored.
The zinc finger transcription factor Gli3 plays crucial roles in corticothalamic/thalamocortical tract formation. We recently showed that in the Gli3 hypomorphic mouse mutant Polydactyly Nagoya (Gli3Pdn) the entry of the CTAs into the VT is delayed and thalamacortical axons form abnormal projections in the VT (Magnani et al. 2010). Moreover, the Gli3 compound mutant Gli3Xt/Pdn, which carries the Gli3 null allele Extra-toes (Gli3Xt) over the hypomorphic Gli3Pdn allele, shows dramatic defects in corticothalamic/thalamocortical tract development (Magnani et al. 2012). TCAs do not enter the cortex but are deflected ventrally at the PSPB while CTAs show a severe axon outgrowth defect in the cortex. This latter phenotype is caused by the nearly complete absence of subplate neurons which pioneer the corticothalamic tract. Although this phenotype emphasizes the importance of Gli3 for corticothalamic tract formation, it might have obscured additional roles of Gli3 in the developing cortex. To address this possibility, we specifically inactivated Gli3 in cortical progenitor cells using an Emx1Cre driver line. Here, we show that, in Emx1Cre;Gli3fl/fl embryos, CTAs do grow but leave the cortex in a narrow, restricted domain. Analyses of cortical lamination revealed that corticofugal projection neurons are formed in the conditional mutants but, surprisingly, the rhinal fissure separating the lateral neocortex from the piriform cortex is shifted medially leading to a medial expansion of the piriform cortex. Using transplantation experiments, we demonstrate that this expanded piriform cortex is repulsive for the growth of CTAs thereby restricting their entry zone into the VT. Moreover, the repulsive activity of the piriform cortex is at least partially mediated by Sema5B whose expression is also shifted medially and which acts as a chemorepellent on CTAs (Lett et al. 2009). Finally, we show that the expansion of the piriform cortex arises from an expansion of the ventral pallial progenitor domain. Taken together, these findings demonstrate that Gli3 controls the entry of CTAs into the VT by regulating the size of the ventral pallial progenitor domain and provide a novel role for the piriform cortex in regulating CTA pathfinding.
Materials and Methods
Animals were handled in accordance with local guidelines for the ethical use of animals in research at the University of Edinburgh which conform with international guidelines. Emx1Cre (Gorski et al. 2002), Gli3fl/fl (Blaess et al. 2008), and Golli-τGFP (Jacobs et al. 2007) mice were kept on a mixed background, and were interbred. Emx1Cre;Gli3fl/+ mice were mated with Gli3fl/fl; Golli-τGFP mice to obtain Emx1Cre;Gli3fl/fl; Golli-τGFP conditional mutant embryos. Emx1Cre;Gli3fl/+ and Emx1Cre;Gli3fl/+; Golli-τGFP embryos were used as controls. Embryonic (E) day 0.5 was assumed to start at midday of the day of vaginal plug discovery. For each marker and each stage, 3–5 embryos were analyzed.
In Situ Hybridization and Immunohistochemistry
Antisense RNA probes for Dbx1 (Yun et al. 2001), Dlx2 (Bulfone et al. 1993), Dmrt5 (Saulnier et al. 2012), Gbx2 (Wassarman et al. 1997), Gli1 (Hui et al. 1994), Gli3 (NM_008130, Genbank, 132–5113 bp), Lhx2 (Liem et al. 1997), Liprin-β1 (Kriajevska et al. 2002), Nrp2 (Kolodkin et al. 1997), Ngn2 (Gradwohl et al. 1996), Patched1 (Goodrich et al. 1996), Pax6 (Walther and Gruss 1991), Pls3 (Oeschger et al. 2012), Sema5B (Skaliora et al. 1998), Sfrp2 (Kim et al. 2001), Shh (Echelard et al. 1993), Slc6α7 (Hoglund et al. 2005), Sox5 (Lefebvre et al. 1998), and Tgfα (Assimacopoulos et al. 2003) were labeled with digoxigenin. In situ hybridization on 10-μm serial paraffin sections of mouse brains were performed as described (Theil 2005).
Immunohistochemical analysis was performed as described previously (Theil 2005) using antibodies against the following antigens: caspase-3 (1:50, Cell Signaling); CS56 (1:1000, Sigma); Ctip2 (1:1000, Abcam); GFP (1:500, Abcam); Gsh2 (1:5000, provided by Kenneth Campbell); Satb2 (1:50, Abcam); Hippocalcin (1:3000; Abcam); Neurofilament (2H3) (DSHB, 1:5); Pax6 (1:200, DSHB); TAG-1 (1:100, DSHB); Tbr1 (1:2500, CHEMICON). Primary antibodies for immunohistochemistry were detected with Alexa- or Cy2/3-conjugated fluorescent secondary antibodies. For counter-staining, TOPRO-3 (1:2000, Invitrogen) was used.
Carbocyanine Dye Placement and Analysis
DiI crystals (D282, Molecular Probes) were placed in target areas of PFA fixed, E18.5 brains using glass capillaries. Brains were incubated in the dark at 37 °C in 4% PFA for 4–6 weeks. Brains were rinsed in PBS, embedded in agarose and sectioned coronally on a vibratome at 100 μm. Sections were cleared in 9:1 glycerol:PBS solution containing the nuclear counter-stain TOPRO3 (0.2 µM) overnight at 4 °C. After mounting in 9:1 glycerol:PBS, sections were examined under epifluorescence microscopy with a rhodamine filter using a Leica confocal microscope.
Organotypic slice cultures of E14.5 mouse telencephalon were prepared as previously described (Magnani et al. 2010). Brain slices were cultured on polycarbonate culture membranes (8 μm pore size; Corning Costar) in organ tissue dishes containing 1 mL of medium (Neurobasal/B-27 [Gibco] supplemented with glutamine, glucose, penicillin, and streptomycin). Slices were cultured for 72 h, fixed with 4% PFA and processed for anti-GFP immunofluorescence as described above.
To quantify the density of hippocalcin+ subplate cells, 160 µm2 wide sampling boxes were placed over the medial neocortex of E12.5 control and Gli3cKO embryos (3 embryos for each genotype). To compare cell densities, a Mann–Whitney test was used.
Analysis of the size of neocortex and piriform cortex was performed on data collected from P7 brains of 3 embryos of each genotype. Mann–Whitney test was used to compare the absolute sizes and the relative length of neocortex and piriform cortex. For statistical analyses, SPSS software was used. Asterisks indicate P < 0.05.
Corticothalamic Pathfinding is Impaired Following Gli3 Inactivation in Dorsal Telencephalic Progenitors
We previously reported severe outgrowth defects of cortical axons in the Gli3 compound mutant Gli3Xt/Pdn (Magnani et al. 2012). This severe phenotype made it impossible to dissect further roles of Gli3 in corticothalamic pathfinding. To investigate such potential role(s), we conditionally inactivated Gli3 specifically in dorsal telencephalic progenitors using an Emx1Cre strain (Gorski et al. 2002; Amaniti et al. 2013). In these conditional mutants, Gli3 inactivation occurs in the cortex in a medial to lateral spatial and temporal gradient so that Gli3 expression is undetectable in the lateralmost neocortex by E12.5 (Amaniti et al. 2013).
To analyze formation of the corticothalamic/thalamocortical tracts, we initially performed neurofilament immunohistochemistry on E18.5 Emx1Cre;Gli3fl/+ (control) and Emx1Cre;Gli3fl/fl (Gli3cKO) mutant brains. In control embryos, this revealed a strongly labeled cortical IZ and many separate axon bundles throughout the striatum (Fig. 1A). In Gli3cKO mutants, the IZ appeared thinner and we only detected a few bundles of axons in the dorsalmost part of the striatum (Fig. 1D). Interestingly, a pronounced axon bundle projected ventrally along the PSPB (Fig. 1D). These data indicate defects in the formation of the corticothalamic/thalamocortical tracts. However, it remained unclear whether corticothalamic, thalamocortical pathfinding or both are affected.
To address this question, we used Golli-τauGFP transgenic mice and also performed DiI axon tracing experiments. The Golli-τauGFP transgene is expressed in a subset of SP cells and layer VI neurons and labels their axonal trajectory as they leave the cortex, enter the striatum along most of the PSPB (Fig. 1B) and reach the thalamus in E18.5 control embryos (Fig. 1C) (Jacobs et al. 2007). In contrast, we only observed a few axon bundles in the dorsalmost part of the striatum in rostral sections of Gli3cKO mutants (Fig. 1E) while a separate axon bundle descended ventrally along the PSPB (Fig. 1E). In caudal sections, few GFP+ axons reached the thalamus (Fig. 1F). To further examine the CTA trajectory, we placed DiI crystals in the E18.5 cortex leading to the anterograde labeling of corticofugal projections and to retrograde labeling of TCAs in control embryos as revealed by the labeling of neuronal cell bodies in the thalamus (Fig. 1G). In Gli3cKO mutant brains, DiI placements revealed the trajectory of CTAs and TCAs. Their path through the striatum was restricted to its dorsalmost part and only a few thalamocortical neurons were labeled retrogradely (Fig. 1J). In addition, a prominent axon bundle projected along the PSPB (Fig. 1J). Finally, we placed DiI crystals in the thalamus to label TCAs specifically. To expose the thalamus for DiI placements, we initially used a coronal section through the thalamus thereby removing the caudal thalamus from the cortex. In this way, DiI labeling showed TCAs passing through the VT and entering the cortex in control brains (Fig. 1H) but all TCAs were deflected from the PSPB and none entered the cortex in Gli3cKO mutant brains (Fig. 1K). As DiI placement in the cortex retrogradely labeled neuronal cell bodies in the thalamus of the mutant, we repeated the thalamic DiI labeling experiments by exposing the thalamus by a midsagittal section which leaves all the thalamus connected to the telencephalon (Fig. 1I). Under these conditions, many TCAs still projected ventrally along the PSPB but some axons crossed the PSPB in its dorsalmost area (Fig. 1L). Overall, this analysis revealed severe CTA and TCA pathfinding defects in Gli3cKO embryos. Many axons abnormally project along the PSPB and only a few CTAs and TCAs, which are derived from neurons in the caudal thalamus, leave and enter the cortex, respectively. Strikingly, exit from and entry into the cortex is restricted to a small area in the dorsalmost part of the PSPB.
Subplate Neurons do not form a Distinct Layer in the Lateral Neocortex of Gli3cKO Mutants
Given the pathfinding defects described above, we became interested in analyzing the underlying cellular defects. As CTAs and TCAs form reciprocal pathways, the defects in the corticothalamic/thalamocortical tract described above could be caused by defects in either axonal population although the use of the Emx1Cre driver line for cortex-specific inactivation of Gli3 renders thalamic defects unlikely. Nevertheless, to test for the specificity of Gli3 recombination and to rule out potential defects in thalamic development due to inappropriate Cre activity, we performed in situ hybridization for Gli3 and several thalamic markers (Supplementary Fig. 1), but we did not detect any obvious changes in their expression patterns in the thalamus. Moreover, the VT serves as an important intermediate target for CTAs and TCAs. Whereas we had recently shown that Gli3 expression is not affected in the medial and lateral ganglionic eminences of Gli3cKO mutants (Amaniti et al. 2013), we here investigated Shh expression and signaling in the VT, since Gli3 has a key role in regulating Shh signaling in the VT and since abnormal patterning of the VT causes TCA pathfinding defects in Gli3Pdn/Pdn embryos (Magnani et al. 2010). However, this analysis did not reveal any obvious difference between control and Gli3cKO mutant embryos (Supplementary Fig. 2). We therefore focused our further analyses on the development of the cortex in Gli3cKO mutants.
Subplate (SP) neurons are amongst the earliest born cortical neurons and play a crucial role in pioneering the corticothalamic tract (McConnell et al. 1989; Jacobs et al. 2007). Recently, we reported that Gli3Xt/Pdn mutants lack subplate (SP) neurons (Friedrichs et al. 2008; Magnani et al. 2012) raising the possibility that their development is also affected in the Gli3cKO mutants. To explore this, we performed an expression analysis of SP-specific markers. In E12.5 control and Gli3cKO embryos, hippocalcin labels SP neurons in the preplate (Osheroff and Hatten 2009) suggesting that their formation is not affected in Gli3cKO mutants (Fig. 2A–C). Next, we investigated whether SP neurons occupy their normal position underneath the cortical plate in E14.5 embryos. In control embryos, expression analyses of CS56 (Fig. 2D,E), hippocalcin (Osheroff and Hatten 2009) (Fig. 2H,I) and Pls3 (Oeschger et al. 2012) (Fig. 2L,M) revealed the presence of a distinct subplate layer which is particularly obvious in the lateral neocortex where cortical development is more advanced. In Gli3cKO mutants, a distinct SP was only evident in medial neocortex but not in lateral neocortex (Fig. 2f,J,N). Thus, the formation of the subplate layer is affected in the lateral cortex of Gli3cKO mutants.
Corticofugal Projection Neurons are Formed Correctly but the Piriform Cortex is Expanded in Gli3cKO Mutants
Next, we examined the possibility that the axon pathfinding defects could be caused by a misspecification of corticofugal projection neurons. To analyze their formation, we performed in situ hybridization for the corticofugal determinant Sox5 (Kwan et al. 2008; Lai et al. 2008) which is expressed in SP and layer V/VI neurons (Fig. 3A). In Gli3cKO mutants, this expression was maintained. Interestingly, however, the Sox5 expression domain did not extend as far ventrally as in control embryos but stopped at the level of the dorsalmost striatum (Fig. 3B). We next investigated the expression of Tbr1 which labels subplate and layer VI corticothalamic projection neurons (Hevner et al. 2001). In control embryos, double labeling with the callosal determinant Satb2 (Alcamo et al. 2008; Britanova et al. 2008) revealed the position of Tbr1+ neurons in the deeper cortical layers underneath the Satb2+ callosal projection neurons of layers II/III and IV (Fig. 3C). This analysis also revealed the position of the rhinal fissure, a sulcus that is conserved across mammalian species and separates neocortex from the paleocortical piriform cortex (Ariens-Kapers et al. 1936) (Fig. 3C,D). In the piriform cortex, Tbr1 is expressed in layer II neurons and in the olfactory tubercle whereas Satb2 expression is largely absent (Fig. 3D). In Gli3cKO mutants, Tbr1+ and Satb2+ cells occupy their correct laminar position in the neocortex and in the piriform cortex but the rhinal fissure was shifted medially (Fig. 3E,F). This shift became also evident when we stained for Ctip2 which labels corticospinal motor neurons (Arlotta et al. 2005) in layer V of the neocortex and layer II neurons in the piriform cortex (Fig. 3G,H). Both these expression domains are present in Gli3cKO mutants, but the neocortical expression of Ctip2 was diminished laterally and its piriform expression increased medially (Fig. 3I,J). We also examined whether this expansion of the piriform cortex persists in postnatal brains (P7) when laminar organization of the piriform cortex is mature. Analyzing Slc6a7 and Liprinβ1 expression which marks the piriform cortex (Kriajevska et al. 2002; Hoglund et al. 2005) and measuring the length of piriform and neocortex confirmed that in the Gli3cKO mutants, the piriform cortex is significantly expanded medially (Supplementary Fig. 3). Taken together with our findings on subplate development, these data suggest that corticofugal projection neurons form and occupy their correct laminar position in medial neocortex but that the lateral neocortex is affected by a medial expansion of the piriform cortex.
Early Corticofugal Pathfinding Defects Coincide with a Medial Expansion of the Piriform Cortex
Based on these findings, we next investigated a potential correlation between the CTA pathfinding phenotype and the defects in the development of the lateral cortex. To compare the expansion of the piriform cortex with the extent and size of the corticofugal exit zone, we analyzed Gli3cKO;Golli-τauGFP+ mutants by double immunofluorescence analyses for Ctip2 and GFP which revealed the position of the rhinal fissure and corticofugal axons, respectively. In both control and Gli3cKO embryos, Golli-τauGFP+ axons turn medially and enter the striatum whereby the rhinal fissure provides a good landmark for the ventral most axonal exit point (Fig. 3K–N). Whereas this point is located in the lateral cortex of control embryos, it is shifted medially close to the angle region in Gli3cKO mutants (Fig. 3M). Thus, the expansion of the piriform cortex spatially correlates with the restricted region through which corticofugal axons leave the cortex.
Corticofugal axons reach the PSPB by E14.5 where they leave the cortex after a short waiting period. Therefore, we investigated whether the expansion of the piriform cortex is already present in conditional mutant embryos at E14.5. Nrp2 expression marks the continuous olfactory cortical structures, including the presumptive piriform cortex and the olfactory tubercle but is absent from neocortical neurons (Chen et al. 1997) (Fig. 4A). In Gli3cKO embryos, we observed Nrp2 expression expanding medially into the region normally occupied by lateral neocortex (Fig. 4D). In addition to its expression in cortical plate neurons, Tbr1 is also expressed in neurons of the presumptive piriform cortex and in the endopiriform nucleus (Ceci et al. 2012) (Fig. 4B,C). In Gli3cKO mutants, Tbr1+ cells acquired their correct position in the cortical plate, in the presumptive piriform cortex, and in the endopiriform nucleus but the latter 2 structures are expanded medially (Fig. 4E,F). Therefore, our analysis indicates an expansion of the piriform cortex in E14.5 Gli3cKO embryos.
To investigate the navigation of corticofugal projection at this stage, we used Golli-τauGFP transgenic mice and immunofluorescence analysis for TAG-1 that labels early corticofugal axons (Denaxa et al. 2001). In control brains, Golli-τauGFP+ and TAG-1+ axons projected ventrally and turned medially at the PSPB (Fig. 4G,H,I). In Gli3cKO mutants, Golli-τauGFP+ axons were present but did not a medial turn at the PSPB (Fig. 4J) whereas TAG-1+ axons stopped their descent at the level of the angle region (Fig. 4J,K,L). Overall, these data suggest that an early medial expansion of the presumptive piriform cortex coincides with a defect in the descent of corticofugal efferents.
The Expanded Piriform Cortex Inhibits the Growth of Corticofugal Axons
These findings indicate that an expansion of the piriform cortex temporally and spatially coincides with corticofugal pathfinding defects raising the possibility that the expanded piriform cortex restricts the exit zone of corticofugal axons. To test this hypothesis, we performed in vitro transplantation experiments using E14.5 Golli-τauGFP+ embryos to reveal the migration of corticofugal axons. We first cultured coronal sections of control;Golli-τauGFP+ and Gli3cKO;Golli-τauGFP+ embryos to verify that the pathfinding defects of corticofugal axons are recapitulated under our culture conditions. After 3 days in culture, control explants showed axons growing into the striatum (n = 11/11) (Fig. 5A,B) but, in mutant sections, no GFP+ axons entered the VT (n = 11/11) (Fig. 5C,D). Next, we went on to homotopically transplant control;Golli-τauGFP− lateral cortex including lateral neocortex and piriform cortex into the corresponding region of a host control;Golli-τauGFP+ section. This transplantation led to the growth of GFP+ axons into the striatum indicating that corticofugal axons maintain their normal projection pattern under these transplant conditions (n = 8/11) (Fig. 5E–G). To test whether the expanded piriform cortex of Gli3cKO mutants interferes with the growth of corticofugal axons, we transplanted the expanded Gli3cKO;GFP− piriform cortex homotopically into the lateral cortex of control;Golli-τauGFP+ embryos thereby mimicking the expansion of the piriform cortex in an otherwise wild-type environment. Under these conditions, no GFP+ axons projected to the striatum (n = 8/10) suggesting that the expanded piriform cortex is not permissive to corticofugal growth (Fig. 5H–J). Finally, we tested whether replacement of the expanded piriform cortex of Gli3cKO;Golli-τauGFP+ mutants by control;Golli-τauGFP− lateral cortex would allow corticofugal axons to enter the VT. This transplantation led to a partial rescue since GFP+ axons did project but they followed the PSPB ventrally and failed to enter the striatum (n = 13/16) (Fig. 5K–M). Taken together these data indicate that the expanded piriform cortex of Gli3cKO mutants inhibits the growth of corticofugal axons.
Sema5B Expression is Expanded in the Lateral Cortex of Gli3cKO Mutants
To address the possible molecular basis for the inhibitory activity of the piriform cortex on the navigation of corticofugal axons, we screened the expression of several guidance molecules in E14.5 Gli3cKO mutants. While the expression of Sema3C, Sema5A, and Netrin1 which have previously been implicated in thalamocortical and corticothalamic pathfinding (Metin et al. 1997; Bagnard et al. 1998) is not altered (Supplementary Fig. 4), we identified severe changes in Sema5B expression. Sema5B is normally expressed in the ventricular zones of the neocortex and of the caudal ganglionic eminence and separately in the piriform cortex and the endopiriform nucleus (Skaliora et al. 1998) (Fig. 6A,B). Moreover, Sema5B guides CTAs through the internal capsule by specifically repelling CTAs and not TCAs (Lett et al. 2009) raising the possibility that Sema5B might mediate the chemorepellent activity of the expanded piriform cortex in Gli3cKO mutants. Indeed, Gli3cKO mutants showed normal Sema5B expression in the ventricular zones but its piriform expression was expanded laterally (Fig. 6C,D). Notably, its expression in the endopiriform nucleus extended nearly as far as the expression in the cortical VZ at all rostrocaudal levels (Supplementary Fig. 5) leaving only a small gap between the 2 expression domains.
As we showed above, TAG-1+ axons stopped their descent in the lateral neocortex (Fig. 4) suggesting that the expanded Sema5B expression may restrict the corticofugal descent. To address this question, we performed immunofluorescence for TAG-1 and in situ hybridization analysis for Sema5B on adjacent sections. Overlaying these sections indicated that in control embryos, TAG-1+ axons were capable of projecting ventrally till they reached the Sema5B expression domain in the EN where they turned medially toward the LGE (Fig. 6E,F,I,J). In conditional mutants, TAG-1+ axons already encountered the expanded Sema5B expression domain at the level of the angle region where only a small Sema5B-negative area remained for their exit into the striatum (Fig. 6G,H,K,L). Taken together with the known chemorepellent activity of Sema5B on CTAs (Lett et al. 2009), these findings suggest that the expanded expression of Sema5B might be responsible for the chemorepellent activity of the piriform cortex.
Regionalization Defects in Gli3cKO Mutants
Finally, we investigated how Gli3 inactivation in the cortex leads to the expansion of the piriform cortex in Gli3cKO mutants. As Gli3 mRNA expression is confined to cortical progenitor cells and as the progenitor domain of the ventral pallium (VP) gives rise to a subpopulation of cells in the piriform cortex (Kawasaki and Hirata 2002), we hypothesized that defective VP development in Gli3cKO mutants might underlie the expansion of the piriform cortex. The VP is the ventral most part of the cortex and is delineated on its ventral side by the PSPB, which plays a key role in corticothalamic pathfinding (Lopez-Bendito and Molnar 2003). We therefore analyzed the expression of the cortical markers Pax6 and Ngn2 and of the subpallial markers Gsh2 and Dlx2 which show sharp expression boundaries at the PSPB in E12.5 control embryos. In Gli3cKO mutants, these expression patterns are largely maintained although there was some intermingling of cortical and LGE progenitor cells (Supplementary Fig. 6). A similar defect in PSPB formation was also observed in E14.5 Gli3cKO embryos (Supplementary Figs 7 and 8) when TCAs start crossing the PSPB. However, the extent of this defect is similar to that found in Gli3Pdn/Pdn embryos, which do not show any change in the exit domain of corticofugal axons (Magnani et al. 2010), suggesting that these PSPB defects are unlikely to underlie the pathfinding defects in Gli3cKO mutants.
To analyze the extent and size of the VP domain, we determined the expression pattern of the VP-specific markers Sfrp2, Tgfα, and Dbx1. Their expression is restricted to the ventral pallium of control brains (Fig. 7A–C), but Gli3cKO mutants showed a moderate dorsal expansion of Sfrp2 and Tgfα expression (Fig. 7F,G). Moreover, Dbx1 expression was already upregulated in the lateral cortex in E11.5 Gli3cKO embryos. This upregulation gradually extended medially to cover the complete cortical ventricular zone by E14.5 (Fig. 7H and Supplementary Fig. 9). We also investigated whether other transcription factors, known to control the development of the piriform cortex, are affected in Gli3cKO mutants. Conditional inactivation of Lhx2 in cortical progenitors leads to an ectopic piriform cortex (Chou et al. 2009) while Dmrt5 mutants show an expansion of the piriform cortex similar to Gli3cKO mutants (Saulnier et al. 2012). In control brains, Lhx2 and Dmrt5 expression were restricted to the ventricular zone with a medial to lateral gradient (Fig. 7D,E). In Gli3cKO mutants, Lhx2 and Dmrt5 expression was downregulated in the lateral neocortex whereas, in more medial cortical areas, there were no obvious expression defects (Fig. 7I,J). Taken together, these data suggest that Gli3 is required in cortical progenitors early in development to repress VP gene expression by regulating Lhx2 and Dmrt5 expression in the lateral neocortex.
Formation of the corticothalamic tract requires complex interactions between CTAs and the environment through which these axons migrate. Here, we analyzed the spatial requirements for CTAs to leave the cortex and enter the VT using Gli3 conditional mouse mutants. We show that Gli3 is required in cortical progenitor cells to regulate the relative size of neocortex and piriform cortex. In Gli3cKO embryos, the ventral pallial progenitor domain and the piriform cortex, which is derived from the VP, expand medially. This expansion coincides with an extended expression of Sema5B, a known chemorepellent for CTAs, thereby restricting the entry zone of corticofugal axons into the VT.
The Size of the CTA Entry Zone is Regulated by Chemorepulsion From the Piriform Cortex
Entry of CTAs into the VT represents a crucial step during the formation of the corticothalamic tract. Strikingly, CTAs enter the striatum in a broad domain along the PSPB forming separate fascicles thereby giving the striatum its typical striped appearance. How this broad entry into the striatum is regulated spatially remains largely unexplored. Here, we show that, in Gli3 conditional mutants, CTAs can only enter the striatum in a very restricted domain immediately adjacent to the angle region while many CTAs follow the PSPB. Similarly, only few TCAs enter the cortex through this narrow entry domain while many TCAs are deflected at the PSPB. However, Gli3 inactivation is confined to the dorsal telencephalon (Amaniti et al. 2013) and thalamic and ventral telencephalic patterning, the expression of several axon guidance molecules in the VT and the formation of the permissive corridor in the MGE (Supplementary Figs 1, 4, 7, and 8) appear not be affected in Gli3cKO mutants. These findings suggest that to the TCA pathfinding defects are secondary to CTA pathfinding errors consistent with the requirement of corticofugal axons to guide TCAs into the cortex (Chen et al. 2012). Therefore, we focused our further investigation on the development of the dorsal telencephalon. Interestingly, the restricted entry domain negatively correlated with an expansion of the piriform cortex. The piriform cortex has previously been implicated in the early development of the corticofugal pathway by extending the first corticofugal projections to cross the PSPB (Molnar et al. 1998). Here, we define an additional role by using a transplantation assay to show that the piriform cortex is repulsive to the growth of corticofugal axons. This analysis also revealed that the growth of CTAs is restored in Gli3cKO embryos after replacing the mutant expanded piriform cortex with control lateral cortex. However, under these conditions, CTAs project along the PSPB similar to the situation in Gli3cKO mutants suggesting that there may be additional cell-autonomous defects in cortical projection neurons. However, the molecular basis of such a cell-autonomous effect remains unclear since the expression of the Sox5, Tbr1, and Ctip2 transcription factors and of the Robo1 receptor which have important functions in corticofugal pathfinding (Hevner et al. 2001; Arlotta et al. 2005; Lopez-Bendito et al. 2007; Kwan et al. 2008; Lai et al. 2008) showed no obvious defects in Gli3cKO embryos (Fig. 3 and Amaniti et al. 2013). Moreover, the increased size of the piriform cortex leads to an expansion of Sema5B expression. In control embryos, the Sema5B expression domains in the germinal zones of the cortex and VT are well separated from its expression in the piriform cortex and endopiriform nucleus opening up a wide corridor for CTAs to reach the PSPB and enter the VT. In contrast, due to the expansion of the piriform cortex, only a small Sema5B-negative area remains between the ventricular and endopiriform nucleus in Gli3cKO embryos. Consequently, TAG1+ pioneer axons encounter Sema5B-expressing cells right at the angle region. Interestingly, recent Sema5B knock-down experiments and ectopic application of Sema5B-expressing cells on cultured sections of the VT have demonstrated a chemorepulsive activity for Sema5B specifically on CTAs but not TCAs (Lett et al. 2009). While we cannot exclude the possibility that other guidance molecules produced by the piriform cortex take part in mediating its chemorepellent activity, our findings suggest a mechanism by which the piriform cortex may regulate the size of the CTA entry zone into the VT at least in part by controlling Sema5B expression.
Our findings on the role of the piriform cortex add to recent findings on the mechanisms regulating the entry of CTAs into the VT. CTAs experience an extensive waiting period at the PSPB (Molnar and Cordery 1999; Jacobs et al. 2007) which is regulated by Sema3E signaling from the globus pallidus (Deck et al. 2013). In addition to Sema3E-mediated chemorepulsion, chemoattraction by Netrin-1 derived from the VT can induce turning and therefore appears responsible for corticofugal growth cone reorientation toward the VT (Metin et al. 1997; Richards et al. 1997). However, since these axon guidance molecules are produced by the VT, they are unlikely to play a role in the development of the CTA guidance defects after cortical specific inactivation of Gli3 in Gli3cKO mutants. Finally, the PSPB has a key role during corticothalamic and thalamocortical development. Several mouse mutants show severe pathfinding defects at the PSPB (Lopez-Bendito and Molnar 2003) which expresses several corticofugal guidance cues including Netrin-1, Sema3C, and Sema5A (Jones et al. 2002). The PSPB might also act as a temporal physical barrier to corticofugal axons (Jones et al. 2002; Pinon et al. 2008; Chen et al. 2012). Interestingly, the PSPB does not form properly in Gli3cKO mutants as exemplified by mixing of some pallial and subpallial progenitor cells. However, the extent of PSPB malformation in Gli3cKO embryos appears similar to that found in Gli3Pdn/Pdn embryos in which PSPB crossing of corticofugal axons occurs though with a delay. Taken together, a picture is emerging where multiple, complex interactions between CTAs and their environment at the PSPB spatially and temporally control the entry of CTAs into the VT.
Gli3 Controls the Medial/Lateral Extent of the Piriform Cortex
The piriform cortex makes up the largest subdivision of the olfactory cortex, receives monosynaptic input from the olfactory bulb (Price 1973) and odors are transformed into odor maps in distinct regions of the piriform cortex (Zou et al. 2005). This phylogenetically distinct cortex consists of only 3 layers, and it develops in an “inside-out” fashion similar to the 6 layered neocortex (Bayer 1986). The piriform cortex reveals a mosaic nature with at least 3 populations of cells deriving from progenitors located in the ventral and dorsal pallium and in the dorsalmost LGE (Kawasaki and Hirata 2002). Although the piriform cortex arises at the lateral extreme of the dorsal telencephalon, there has been very little progress in understanding its specification and its separation from the adjacent neocortex. Our findings of an expansion of the piriform cortex in Gli3cKO embryos and our subsequent molecular analyses might shed some light on these issues. An expansion of the piriform cortex has previously been observed in Gli3Xt/Xt null mutants (Vyas et al. 2003) and in the Gli3 compound mutant Gli3Xt/Pdn (Magnani et al. 2012). These findings further emphasize Gli3's importance in the development of the piriform cortex but the molecular mechanisms underlying this control were not investigated. Our detailed analyses of the Gli3cKO mutants provide important clues to the temporal and molecular requirements of Gli3 function in this process. As the piriform cortex at least partially derives from VP progenitors, we analyzed the expression of several VP markers and identified a moderate medial expansion of the Sfrp2 and TGFα expression domains and a striking upregulation of Dbx1 throughout the cortical germinal layer in Gli3cKO mutants. Dbx1 has previously been implicated in delineating the VP domain (Yun et al. 2001; Assimacopoulos et al. 2003; Vyas et al. 2003) but obviously this dramatic upregulation is not sufficient to induce the expression of other VP markers throughout the cortical VZ. These findings are therefore consistent with the idea that progenitors of the lateralmost neocortex but not those of more medial neocortex have been transformed to a VP fate. Interestingly, inactivation of Gli3 in the lateral cortex of Gli3cKO mutants only occurs between E11.5 and E12.5 (Amaniti et al. 2013) suggesting that, at this stage, cortical progenitors are still plastic with respect to the generation of neocortical or piriform neurons. This also assigns a role for Gli3 at a late stage in patterning whereas previously it has been mainly implicated in early forebrain patterning (Hebert and Fishell 2008). Moreover, this role appears to be independent of Gli3's function in Shh signaling as the cortex-specific inactivation does not affect Shh expression and signaling in the VT. Shh-independent functions of Gli3 were previously described in patterning of the dorsal telencephalon (Rash and Grove 2007) and during limb development (Hui and Angers 2011) but the genes regulated by Gli3 in a Shh-independent manner remain unknown. As the Gli3 repressor form predominates in the developing cortex (Fotaki et al. 2006), it is conceivable that Gli3 directly controls the expression of Dbx1, Sfrp2, and TGFα, however, the time course and the extent of their upregulation argues against this possibility especially as the Dbx1 upregulation occurs from lateral to medial opposite to the Gli3 inactivation gradient. Gli3 could also regulate VP expression indirectly by controlling the expression of Lhx2 and Dmrt5. Conditional inactivation of Lhx2 in cortical progenitor cells leads to a duplication of the piriform cortex (Chou et al. 2009) while loss of Dmrt5 function results in a similar expansion of the piriform cortex as in Gli3cKO embryos (Saulnier et al. 2012). The downregulation of both factors in the lateral cortex of Gli3cKO mutants provides an alternative explanation for the expansion of the piriform cortex in these animals. Therefore, our findings place Gli3 upstream of these transcription factors and imply a key role for Gli3 in controlling the relative size of neocortex and piriform cortex.
Taken together, our findings emphasize the importance of Gli3 in regulating the transition from neocortex to piriform cortex which has major implications for corticofugal axon pathfinding as this affects the size of the entry zone of corticofugal axons into the striatum. Future studies will aim to identify the mechanism(s) by which Gli3 controls this process at the molecular level.
This work was supported by a grant from the Medical Research Council to T.T. (G0801359).
We are grateful to David Price and Tom Pratt for helpful comments on the manuscript, Trudi Gillespie for help with confocal imaging and Alex Joyner and Sandra Blaess for the Gli3fl/fl mouse line. Conflict of Interest: None declared.