Introduction

The primary cilium is a tiny microtubule-based organelle present on the surface of nearly all mammalian cell types including neurons, which functions as a signaling hub and transduces several signaling pathways comprising Sonic Hedghog (SHH), Wnt, Delta/Notch and mTOR pathways ¹. Primary cilium dysfunction causes pleiotropic diseases named ciliopathies. KIF7 is a ciliary kinesin responsible for the trafficking and the positive and negative regulation of the GLI transcription factors in the primary cilium of mammals ². Loss of function of KIF7 in the mouse has shown that KIF7 regulates SHH signalling by acting downstream of Smoothened (Smo) and upstream of GLI2 and GLI3 ^3–7. Studies in mice established that KIF7 activity is dependent on the expression level of SHH. In the absence of SHH, KIF7 localizes to the base of the primary cilium ⁵. GLI factors are phosphorylated and addressed to the proteasome at the base of the primary cilium for degradation, leading to the formation of a cleaved and stable transcriptional repressor (GLI3-R). In the presence of SHH, KIF7 accumulates at the distal tip of the primary cilium ^4,5 and associates with full length GLI2/3 that become transcriptional activators (GLI-A) ⁸.

Since 2011, ten studies have identified patients carrying mutations in the KIF7 gene responsible of ciliopathies classified according to clinical features as hydrolethalus, acrocallosal, Joubert and Greig cephalopolysyndactyly syndromes ^9–18. MRI investigations revealed macrocephaly, ventricles enlargement and corpus callosum alterations. They also showed the characteristic hindbrain abnormalities observed in most ciliopathies such as molar tooth sign (MTS) and cerebellar atrophy. Patients also presented with mild but frequent cortical malformations as well as neurodevelopmental delay, intellectual disability and seizures ^11,12,15,19 which indicate cortical abnormalities.

In physiological conditions, the proper activity of the cortex relies on the excitatory/inhibitory balance i.e. on the ratio, positioning and connectivity of excitatory glutamatergic neurons (principal neurons, PN) and inhibitory GABAergic interneurons (cIN), generated from progenitors in the dorsal and ventral telencephalon, respectively. The cortical neurons of dorsal and ventral origin each migrate towards the developing cortex, the so-called cortical plate (CP), according to a specific timeline allowing the establishment of proper connections in the CP.

SHH plays a central role in the forebrain patterning and differentiation. At early embryonic stage, the ventral expression of SHH orchestrates the ventro-dorsal ^20,21 and medio-lateral ²² regionalization of the mouse forebrain and the differentiation of ventral cell types ²⁰. Shh ablation performed after forebrain patterning alters the specification of distinct subgroups of GABAergic interneurons ^23–26. Interestingly, conditional ablation of Shh and Smo in the embryonic cortex reduces the proliferation of dorsal progenitors ²⁷, demonstrating a minimal SHH expression in the dorsal telencephalon, even before birth ²⁸. Beside its role on proliferation and specification, SHH also controls the migration of interneurons to the cortical plate ^29–33. In the embryonic telencephalon, GLI transcription factors mediate SHH signals in complex and specific ways. The three GLI factors identified in mammals are expressed in the mouse forebrain: GLI1 along the source of SHH in the ganglionic sulcus, and GLI2 and GLI3 dorsally to the SHH source ^26,34,35. GLI1 acts uniquely as a pathway activator whereas GLI2 and GLI3 can be processed in transcriptional activators or inhibitors in the primary cilium. However, both GLI1 and GLI2 function primarily as transcriptional activators in response to SHH activity in the ventral forebrain ³⁵. Nevertheless, GLI1 and GLI2 mutants show mild phenotypes ^36,37 and GLI2 seems required to transduce high level SHH signals in mice ³⁸. In contrast, the development of the dorsal cortex, where principal excitatory neurons differentiate, depends mainly on the expression of the GLI3-R repressor, in agreement with the low cortical expression of SHH ^39,40. Following patterning, Gli3R/A ratio remains critical for specifying the fate of cortical progenitors and regulating cell cycle kinetics ^41–43.

Previous studies investigated the mechanisms underlying the corpus callosum agenesis in patients with KIF7 mutation using Kif7 knock-out (Kif7 -/-) mice ⁴⁴ and the consequence of KIF7 knock down on cortical neurogenesis in principal cells electroporated with Kif7 shRNA⁴⁵. It remains that the influence of developmental abnormalities associated with ciliopathies on the cortical cytoarchitecture is poorly understood. Given the crucial role of KIF7 in regulating the GLI pathways and the important role of Gli3-R and GLI2/3-A in regulating the early developmental stages of the dorsal and ventral telencephalon, respectively, we examined here the cortical development from E12.5, the establishment of long distance projections with the thalamus and the migration of GABAergic interneurons (cIN) in Kif7 -/- mice. The migratory behavior of Kif7 -/- cINs was investigated by time-lapse videomicroscopy in co-cultures and in organotypic cortical slices. We also compared the migration of cINs in Kif7 -/- cortical slices and in control slices treated with pharmacological activator and inhibitor of the SHH pathway. We moreover determined the local distribution of the SHH protein in the embryonic cortex.

Our results show developmental defects leading to permanently displaced neurons, abnormal formation of cortical layers and defective cortical circuits that could be responsible for epilepsy and/or intellectual deficiency in patients carrying KIF7 mutation ^9,11–18,44.

Results

Kif7 knock-out mice as a model to investigate the structural cortical defects that could lead to clinical feature in patients carrying KIF7 mutation

The KIF7 gene is located on chromosome 15 in human and encodes a 1343 aa protein containing, a kinesin motor domain and a GLI-binding domain in the N-terminal part followed by a Coiled-coil region and a Cargo domain able to bind a diverse set of cargos in the C-terminal part ⁴⁶. Table 1 summarizes the clinical features associated with mutations targeting either the kinesin or GLI binding domain, the Coiled-coil domain, or the N-terminal Cargo domain. Interestingly, various mutations are associated with the same clinical picture, suggesting that KIF7 mutations, whatever their nature, could lead to protein loss of function, for example by altering the protein structure and solubility as proposed by Klejnot et Kozielski ⁴⁷. All patients carrying mutation in the KIF7 gene have developmental delay (DD) and intellectual deficit (ID) associated with classical defects of ciliopathies (ventricle enlargement, macrocephaly, corpus callosum agenesis and MTS). Some patients have additional anatomical cerebral cortex defects such as poor frontal development, atrophy, pachygyria and heterotopia ^9,10,13,16, which could participate not only to DD and ID but moreover to seizure as observed in a quarter of patients. In this context, we used a murine model in which the Kif7 gene had been deleted (Kif7 -/-) ³ to investigate the consequence of KIF7 loss of function on the cortex development. Kif7 -/- mice have been previously characterized as dying at birth with severe malformations, skeletal abnormalities (digits and ribs), neural tube patterning defects including exencephaly in a third of the mutants, microphtalmia, lack of olfactory bulbs and CC agenesis ^3,4,44. In the present study, the KIF7 loss of function was analyzed in a mouse strain with a mixed genetic background in which exencephaly was observed at the same frequency in Kif7 -/- and control embryos. As previously reported, Kif7 -/- embryos displayed microphtalmia [illustrated at embryonic stage 14.5 (E14.5), Fig. 1A, black arrow] and polydactily (not illustrated). Interestingly, skin laxity previously reported in ciliopathic patients ^48,49 was observed in all mutants (Fig. 1A, white arrow). Comparison of brains from Kif7 -/- embryos and control littermates revealed the absence of olfactory bulbs in Kif7 -/- brains (illustrated at E14.5, Fig. 1B, white arrows on ventral face view).

Table 1.

Figure 1.

Kif7 invalidation alters the development of the cortex

The thinning of the dorsal cortex at E14.5 allowing the lateral ventricles to be seen through (Fig. 1B, black arrow on dorsal face view) proned us to investigate the structural organization of the forebrain over time (Fig. 1C1-3, E12.5-E16.5). Analyses performed on rostro-caudal series of frontal sections identified strong abnormalities in the cortex development. E12.5 Kif7 -/- embryos exhibited cortical wall folding and thinning in rostro-median sections (Fig. 1C1, arrowheads) and prominent diencephalon (Fig. 1C1, stars). At E14.5, cortex folding was not longer observed in Kif7 -/- embryos which exhibited enlarged lateral ventricles (compare left and right panels in Fig.1C2). The cortex remained thinned in the dorsal region and the latero-dorsal extension of the telencephalic vesicle was strongly reduced caudally. These defects were maintained in E16.5 embryos (Fig. 1C3) and the hippocampus did not develop properly (Fig. 1C3, caudal section). A quantitative analysis performed in E14.5 brains (Fig.1C4) confirmed the ventriculomegaly (upper left graph) and the cortex thinning (upper right graph) in the rostral and median sections despite unchanged total width and height of the telencephalon (Fig. 1C4, lowers graphs). KIF7 depletion in E10.5-E11.5 mouse embryos was reported to alter the dorso-ventral patterning of the medial forebrain ⁴⁴. The frontier of expression of GSH2 (Fig. 1D1,D2) and TBR2 (Fig. 1D3,D4), two transcription factors expressed in the ventral and dorsal telencephalon respectively, was slightly abnormal at E13-E14 in Kif7 -/- embryos with a small shift of GSH2(+) cells toward the ventricular angle (Fig. 1D2) and a dispersion of some TBR2(+) cells at the pallium-subpallium boundary (PSB) (Fig. 1D4). This suggests a near-normal dorso-ventral patterning of the telencephalic vesicles, however with a discrete blurring of the dorsal and ventral markers at the PSB.

In Kif7 -/- embryos, the cleavage of full length GLI3 (GLI3-FL) into the transcriptional repressor GLI3 (GLI3-R) is decreased ^3,4. Here, we analyzed the expression of GLI3-FL and GLI3-R in the cortex and in the medial ganglionic eminence (MGE) where cortical interneurons (cIN) are born. We showed minimal expression of GLI3-R and GLI3-FL in the MGE compared to the cortex (Fig. 2A) and a strong decrease in the cleavage of GLI3-FL in GLI3-R only in the cortex of Kif7 -/- embryos (Fig. 2A,B).

Figure 2.

The structural defects described above convinced us to examine the cellular organization of the cortex of Kif7 -/- embryos. At E12.5, the whole cortex is proliferative. From stage E14.5, the mouse cortex can be described as a stack of three specialized domains: i) the upper/superficial cortical layers containing post-mitotic cells generated in the proliferative zones of the cortex, with TBR1(+) post-mitotic neurons located in the cortical plate (CP) in-between the subplate (SP) and marginal zone (MZ) MAP2-expressing cells; ii) the deep proliferative layers, ventricular and subventricular zones (VZ and SVZ, respectively), located along the lateral ventricle; and iii) the intermediate zone (IZ) located between the CP and the proliferative layers, which hosts the radially and tangentially migrating neurons and growing axons. At the end of the embryonic development, the proliferative layers are strongly reduced due to neurogenesis decrease, whereas the cortical thickness had increased due to CP delamination. At E14.5, the TBR1(+) CP cells appeared more clustered in Kif7 -/- embryos than in wild type (WT) embryos where they formed a rather regular, radially organized CP (compare left and right panels in Fig. 3A1,3A2). The thickness of the upper MAP2(+) MZ layer was extremly irregular in Kif7 -/- embryos and MAP2(+) SP cells were missing in the dorsal cortex (Fig. 3A1,A2, white arrows indicate the limit of distribution of MAP2(+) SP cells). In E16.5 and E18.5 embryos, the CP was obviously thinner in the mutant as compared to the WT (Fig. 3B,C), even though CTIP(+) cells differentiated normally in V-VI layers (Fig. 3C). The CP thinning appeared even more pronounced in the caudal than in the median and rostral cortex. We then analyzed the proliferative layers focusing on TBR2(+) intermediate progenitors (IP) which normally form a dense layer in the cortical SVZ (Fig. 3D1, left). The TBR2(+) layer constantly showed an abnormal positioning in the dorsal cortex of Kif7 -/- embryos. TBR2(+) cells indeed reached the brain surface (Fig. 3D1, arrow) where they mixed up with post-mitotic TBR1(+) cells instead of keeping a minimal distance with TBR1 cells as observed in control brains (Fig. 3D2, arrow). As a consequence intermediate progenitors and post-mitotic neurons no longer segregated in the dorsal cortex of Kif7 -/- embryos where no IZ was observed (Fig. 3D2, arrow). Accordingly to the thinning of the cortex starting at E12.5, the TBR2(+) layer was thinner at all development stages analyzed (Fig. 3E,F). In contrast, the abnormal positioning of TBR2(+) cells at the brain surface in the dorsal cortex of E14.5 embryos was no longer observed at E16.5 (Fig. 3F) suggesting either a transient structural disorganization and/or developmental delay of the dorsal cortex in Kif7-/- embryos at E14.5. The TBR2 staining moreover revealed focal heterotopia in the dorsal or lateral cortex of E14.5 Kif7 -/- embryos, either at the ventricular or pial surface (Fig. S1).

Figure 3.

The loss of Kif7 alters the connectivity between the cortex and the thalamus

The intermediate zone (IZ) hosts migrating neurons and the growing cortical and thalamic projections ⁵⁰. After reaching the CP by radial migration, post-mitotic neurons extended pioneer axons oriented tangentially in the IZ and directed to the PSB ^51,52. We thus examined whether the structural defaults of the CP and SP and the lack of IZ in the dorsal cortex of Kif7-/- embryos did associate with developmental abnormalities of corticofugal projections. We labeled corticofugal axons by inserting small crystals of DiI in the CP of E14.5 paraformaldehyde fixed brains. After DiI had diffused along corticofugal axons, we analyzed labeled axons on coronal sections. In both control and Kif7 -/- brains, corticofugal axons extended below the CP toward the PSB (Fig. 4A). According to the latero-medial gradient of cortical development, DiI injections in the dorsal cortex of control and Kif7 -/- brains (Fig. 4A1) labeled much less axons than injections in the lateral cortex (Fig. 4A2,A3). Remarkably, the cortical bundles labeled from similar cortical regions were always much smaller and shorter in Kif7 -/- than in WT brains (compare right and left columns in Fig. 4A). Dorsal injections labeled large bundles crossing the PSB in WT brains, but only a few axons reached the PSB in Kif7 -/- brains (Fig. 4A1). The growth of corticofugal axons appeared severely hampered in the dorsal cortex of Kif7 -/- embryos. Lateral injections in WT brains labeled anterogradely corticofugal axons that reached the embryonic striatum and continued medially in the internal capsule (IC), and labeled retrogradely thalamic projections (Fig. 4A3, left panel). Lateral injections in Kif7 -/- embryos labeled anterogradely cortical axons that spread in the striatum (Fig. 4A2, right panel) and labeled a large bundle directed to the ventral pre-optic area (Fig. 4A3, right panel, white arrow), which was never observed in WT brains. Moreover, no thalamic axons were retrogradely labeled from the lateral cortex in Kif7 -/- embryos.

Figure 4.

To identify the trajectories of thalamic axons in Kif7 -/- embryos, we immunostained coronal sections of E14.5 brains with antibodies against the Netrin G1a (NG1a), an early marker of thalamocortical axons (TCA, Fig. 4B) ⁵³. At E14.5, NG1a antibodies labeled a larger population of neurons in the dorsal thalamus of Kif7 -/- than in WT brains (Fig. 4B, stars). In control brains, TCA made a right angle turn to join the IC in the striatum. Then they extended to the PSB and some axons entered the lateral cortex (Fig. 4B2,B3, left panel). In Kif7 -/- embryos, most thalamic axons formed a thick bundle oriented ventrally in the basal forebrain (Fig. 4B4, white arrow in right panel). A minor structural defect was evidenced at the telo-diencephalic junction by Pax6 staining in Kif7 -/- brains (Fig. S2), recalling the prethalamus defect described in the ciliary Rfx3 mutant ⁵⁴. Given the complex trajectory of NG1a thalamic axons, we immunostained thalamic axons in whole brains and imaged them after transparisation using a light-sheet microscope. The three-dimensional reconstruction of labeled projections confirmed that thalamic axons made a sharp turn to reach the IC and then navigated straight to the PBS in WT brains (Fig. 4C, left panels and Movie S1). In Kif7 -/- embryos, most thalamic axons stopped their course after leaving the diencephalon, and formed two short bundles: a large one oriented to the IC (Fig. 4C, right panels, arrowheads and Movie S2) and another one oriented to the amygdala, more caudally and ventrally (Fig.4C, white arrows). This last projection was never observed in WT brains. Similar analyses performed at E16.5 showed that two days later, most TCA had reached the cortex in both WT and Kif7 -/- embryos (Fig. 4D, arrowheads) despite a small delay in Kif7 -/- embryos and abnormal trajectories in the PSB region (Fig. 4D1,D2). The ectopic ventral caudal projection to the amygdala was still present at E16.5 in Kif7 -/- embryos (Fig.4D1, arrow). At E18.5, the TCA projections in the Kif7 -/- embryos appeared almost normal in the rostral cortex (Fig. 4E1) but drastically reduced in the median and caudal cortex (Fig. 4E2, E3, E4, arrowheads). In addition, a thick bundle of thalamic axons followed an abnormal trajectory at the PSB (Fig. 4E2,E3, arrow). Neurofilament staining with Smi32 antibodies confirmed the strong reduction of projections in the cortex and IC of Kif7 -/- embryos (Fig. 4F).

Kif7 invalidation alters the cortical distribution of cIN at E14.5

Because cIN are born in the basal forebrain and likely depend on contact/functional interactions with pioneer corticofugal projections for the first stages of their migration ^55,56, we examined their distribution in the developing cortex of Kif7 -/- animals. cIN are generated outside of the cortex, in the medial and caudal ganglonic eminences (MGE, CGE) and preoptic area (POA) of the ventral forebrain. They enter the lateral cortex at E12.5 - E13.5, depending on mouse strains, and colonize the whole cortex by organizing two main tangential migratory streams, a superficial one in the MZ, and a deep and large one in the lower TBR2(+) IZ/SVZ ^57,58. We thus analyzed the cortical distribution of tdTomato(+) MGE-derived cIN in WT and Kif7 -/- Nkx2.1-Cre;Rosa26-tdTomato transgenic embryos. An abnormal MGE cell distribution was observed in the developing cortex of Kif7 -/- embryos from E14.5 (Fig. 5 and Fig.S3). In mutant brains, both the latero-medial extent of the superficial and deep tangential migratory streams and the space separating them were significantly shortened compared to WT brains (Fig. 5A2,B2, B3). The closeness of the two streams in mutants was consistent with the reduced thickness of the corticofugal and TCA projections that navigate tangentially in the IZ (illustrated in Fig. 4). The superficial migratory stream was thinner and denser in Kif7 -/- embryos compared to WT (Fig. 5A1,5B1, right panels). Remarkably, cIN no longer colonized the dorsal cortex of Kif7 -/- embryos and the deep tangential stream stopped in the region where the layer of TBR2(+) SVZ cells switched to the cortical surface (Fig. 5A1, white arrow). The tangential progression of cIN in the developing cortex is controlled by CXCL12, a chemokine transiently expressed in TBR2(+) cells and in meninges ^60,61. CXCL12 prevents the premature occurrence of the tangential to radial migration switch required for cIN to colonize the CP ^59–62. In situ hybridization experiments revealed that CxCl12 mRNA was not expressed in the dorsal cortex of E14.5 Kif7 -/- embryos (Fig. 5C, black arrow). To characterize the consequences of the abnormal expression pattern of CXCL12 at E14.5 on the migratory behavior of cIN, we sliced coronally E14.5 Kif7 -/- and WT forebrains and imaged by time-lapse video-microscopy the trajectories of MGE-derived tdTomato-expressing cIN in the dorsal cortex of organotypic slices (Fig.5D). While most cIN presented tangential and oblique trajectories in WT slices, a majority of cIN exhibited radially oriented trajectories in Kif7 -/- slices (Fig.5D). CXCL12 expression in SVZ strongly decreases at E16.5 ^59,63 and analyses performed at E16.5 in fixed Kif 7 -/- and WT embryos showed that cIN had pursued their migration and reached the dorsal cortex (Fig.S3A) even though the tangential migration of cIN in Kif7 -/- embryos remained delayed as compared to WT embryos. Analyses performed in a surviving P0 animal (Fig. S3B,C) revealed an abnormal distribution of cIN in the CP characterized by a strong decrease of cIN density in the infragranular layers that were moreover much thinner than in control brains (see CTIP2 staining in Fig. S3C).

Figure 5.

The abnormal distribution of cIN in the cortex of Kif7 -/- embryos reflected the structural and molecular abnormalities that we had identified in their developing cortex. We next examined if migratory defaults proper to Kif7 -/- cIN could moreover contribute to their abnormal cortical distribution.

Kif7 invalidation affects the migratory behavior of cIN in co-cultures and organotypic slices

We thus compared the migratory behavior of Kif7 -/- and WT cIN using an in vitro model previously established in the lab ⁶⁴ to compare the dynamics of mutant and WT cIN migrating on a substrate of WT cortical cells (Fig.6A). WT and Kif7 -/- E14.5 MGE explants from tdTomato expressing embryos were cultured on WT dissociated cortical cells (Fig. 6A1). Fluorescent migrating MGE cells were imaged using time-lapse video-microscopy and tracked over time (Fig.6A2). Dynamic parameters (speed during moves, stops, Fig. 6A3-A6) and trajectories (Fig. 6A7) were analyzed on movies. Kif7 -/- cIN migrated slightly more rapidly than WT cIN (Fig. 6A3) mainly because the duration of their stops was significantly reduced (Fig. 6A4). The saltatory behavior of cIN was preserved (same frequency of fast nuclear movements, Fig. 6A5) despite a slight decrease of the speed of fast nuclear movements (Fig. 6A6). WT cIN migrated along pretty straight trajectories (Fig.6A2, left panel and Fig. 6A7, black curve) whereas Kif7 -/- cIN showed reduced directionality persistence over time (Fig. 6A2, right panel and Fig. 6A7, blue curve). Kif7 ablation thus affected the migratory behavior of cIN in a cell-autonomous manner and the migratory defaults of Kif7 -/- cIN evoked those of Kif3a -/- cIN whose primary cilium is non-functional (See Figure S6 in ²⁹).

Figure 6.

We then analyzed the dynamic behavior of Kif7 -/- cIN in the mutant cortex (Fig. 6B). In WT Nkx2.1-Cre;Rosa26-tdTomato transgenic slices, fluorescent cIN migrated tangentially from the PSB to the dorsal cortex in two main streams, the MZ and the lower IZ/SVZ (Fig. 6B1, Supplementary Movie S3). All along these pathways, cIN sporadically operated a tangential to oblique/radial migration switch to colonize the CP. High cell density in the MZ prevented cell monitoring and MZ cells were discarded from analyses. In the deep stream, a large proportion of imaged cells (59.3 %) either maintained tangentially oriented trajectories (28.7 %, red trajectories in Fig. 6B1,B3, left column of Fig. 6B2) or reoriented to the CP along oblique or radial trajectories (30.6 % green trajectories in Fig. 6B1,B3, left column in Fig. 6B2). Most of these cIN reached the CP surface and remained there. A small proportion of cIN moved from the deep stream to the ventricular side of the slice (11.7%, blue trajectories in Fig. 6B1,B3, left column in Fig. 6B2). A significant proportion (15.6 %) migrated radially over the cortex thickness, inverting their polarity in the MZ or at the ventricular surface (pink trajectories in Fig. 6B1,B3, left column in Fig. 6B2). Rare cIN in the VZ, deep stream and CP moved very short distances or did not move (3.6%, dark colors in the left column of Fig. 6B2). Finally, 8 % cIN (orange) moved backward to the ventral brain and 1.7 % (grey) showed no specific directionality or layer specificity and were classified as chaotic (Fig. 6B1,B3, left column in Fig. 6B2). The majority of cIN presented a characteristic saltatory behavior, alternating stops and fast moves ⁶⁴. The dynamics of cells recorded in the deep stream and/or moving to the CP is shown in Fig. 6B4-6 (left column on histograms).

The trajectories followed by cIN in the lateral/latero-dorsal cortex of Kif7 -/- slices (where CxCl12 mRNA was detected as shown in Fig. 5C) dramatically differed from those described above (compare WT and Kif7 -/- columns in Fig. 6B2, schemes in Fig.6B3 and Supplementary Movies S3 and S4). The number of cIN migrating tangentially in the deep migratory stream and/or reorienting to the CP dropped drastically (17.8 % as compared to 59.3 % in WT slices) to the benefit of immobile cells (35.7 % as compared to 3.6 % in WT slices) especially in the SVZ/lower IZ. The proportions of cIN moving to the ventricle or to the CP were similar (10.2 % and 13.9 %, respectively, as compared to 11.7 % and 30.6 % in WT slices) suggesting that the radial asymmetry of cortical slices was lost for Kif7 -/- cIN. The same proportion of cIN migrated radially across the cortical layers as in WT slices (17.2 % versus 15.6 % in WT) but the proportion of cIN with chaotic trajectories strongly increased (7.8 % versus 1.7 % in WT slices). Another major change was a significant decrease of the migration speed (excluding immobile cIN) (Fig. 6B4) with less frequent but longer stops in Kif7 -/- than in WT slices (Fig. 6B5,B6), confirming the dynamic parameters measured on co-cultures.

Kif7 ablation has been shown to activate SHH transcriptional signaling by blocking the cleavage of GLI3 in GLI3-R repressor, or conversely to inhibit SHH transcriptional signaling by preventing the formation of GLI activators (GLI1/2)³. Our western blot analysis confirmed that the GLI3-R/GLI3-FL ratio dropped in the Kif7 -/- cortex according to the expected inhibition of the GLI3 processing, a result not observed in the MGE. Since Kif7 -/- cIN are born in the MGE and migrate in the cortex where SHH signaling is activated due to an abnormal GLI3 processing, we first examined whether the application of SHH on a WT forebrain slice could mimic the migratory defaults of Kif7 -/- cIN in the mutant cortical slices. SHH application on control slices minimaly affected the trajectories of cIN (Fig. 6B2 compare WT and SHH (WT) columns). An increased proportion of cIN migrated radially across the cortical thickness (29 % compared to 15.6 % in WT slices) at the expense of cIN migrating from the deep stream to the CP (20.8 % instead of 30.6 % in WT slices). Frequent radial movements directed to the VZ (Fig. 6B2,B3, Supplementary Movie S5) did not evoke the migratory defaults of Kif7 -/- cIN in the mutant slices. Moreover, the frequency of chaotic trajectories did not increase (grey, Fig. 6B2) and the main dynamic alterations, a decreased speed (Fig.6B4) associated with an increased duration of stops (Fig. 6B6), were milder than those observed in Kif7 -/- cIN. On the contrary, the application of the SHH pathway inhibitor cyclopamine on WT cortical slices altered cIN trajectories in a way that resembled alterations in Kif7-/- slices (compare WT and cyclo (WT) columns in Fig. 6B2 and scheme in Fig. 6B3). For example, the proportion of immobile cIN and of cIN moving short distance reached 33.7 % as compared to 3.6 % in WT slices and 35.7% in Kif7-/- slices (Fig. 6B2). Cyclopamine application mainly affected the ability of cIN to move tangentially in the deep stream and to switch to the CP (Fig. 6B2,B3, Supplementary Movie S6). Nevertheless, the migration speed defects of Kif7 -/- cIN and of cyclopamine treated cIN differed (Fig. 6B4).

In conclusion, Kif7 -/- cIN exhibited intrinsic migratory defaults resembling those of cIN with a primary cilium ablation and showed cortical trajectories defaults resembling those of cyclopamine treated WT cIN. Kif7 ablation in cIN had the same effect than SHH signaling inhibition on cIN migratory behavior.

The strong influence of cyclopamine on the migration of cIN in WT cortical slices led us to determine the distribution of the endogenous activator SHH in the developing cortex.

Distribution of Shh mRNA and SHH protein in the E14.5 cortex

Shh mRNA is strongly expressed in the mantle zone of the ganglionic eminences (MGE, CGE) and POA during the time cIN differenciate ^24,52 but its expression level in cIN after they reach the cortex is controversial ^29,31,65. We thus reexamined Shh mRNA expression in the forebrain and SHH protein distribution in cortical layers. In situ hybridization (ISH) experiments performed at E13 confirmed the strong expression of Shh mRNA in the ventral forebrain, especially in the SVZ and mantle zone of the MGE (Fig. 7A1) in agreement with the SHH-dependent Gli1 mRNA expression in a VZ subdomain of the MGE ⁶⁶. In contrast, Shh transcripts were bearly detectable in the cortex. Using RNAscope, which has higher sensitivity and spatial resolution than standard ISH, we re-examined the expression pattern of Shh mRNA at E14.5. In the ventral forebrain, RNAscope Shh transcript detection showed similar expression pattern as ISH (Fig. 7A1,A3,B1,B2). In addition, Shh mRNA and Lhx-6 mRNA, a cIN marker, were expressed in overlapping domains in the mantle zone of the MGE (Fig. 7A2-4). In the cortex, radially-oriented cells in the VZ/SVZ and tangentially oriented cells in the deepmigration stream expressed Shh mRNA at a very low level (one or two dots versus large cluster in the MGE) (Fig. 7C). Among the large number of cells expressing Lhx-6 transcripts (Fig. 7C1,C2), only 5% co-expressed Shh mRNA (Fig. 7C3).

Figure 7.

Antibodies that recognize the N-ter part of SHH [SHH-Nter, e.g. “activated” SHH ⁶⁷] strongly labeled in C57/Bl6 mouse brain the same regions known to strongly express SHH mRNA ^65,68,69, such as the lateral wall of the third ventricle (Fig. S4A1) and the zona limitans intrathalamica at E12.5 (Fig. S4A2) and the choroid plexus and septum at E14.5 (Fig. S4B). SHH-Nter immunostaining was detected in the cortex at E12.5 and E14.5 (Fig. S4C,4D). To further investigate the expression of SHH in the cortical wall at E14.5, we performed immunostaining experiments in Nkx2.1-Cre/R26R-tdTomato embryos (Fig. 7D). SHH-Nter antibodies detected large bright dots that distributed radially away from the ventricle in the lateral LGE and at the ventricular angle (green dots in Fig. 7D1, upper panel). In the cortex, SHH-Nter antibodies immunostained bright tiny dots that show remarkable distribution (Fig. 7D2,D3). In the lateral (Fig. 7D2) and dorsal (Fig. 7D3) cortex, the bright SHH-ter immunopositive dots were aligned radially in the VZ and in the CP according to the radial arrangement of cortical progenitors and CP cells. In contrast, dots were aligned tangentially in the SVZ/IZ where cIN distribute among tangentially oriented growing axons. The intensity and density of positive dots appeared lower in the dorsal than in the lateral cortex suggesting a latero-dorsal gradient of distribution (compare Fig. 7D2 and 7D3). Some CP cells presented a diffuse cytoplasmic staining, which was not observed in the bottom layers. Together, our ISH and immunostaining experiments showed that the cortical SHH protein has probably an extrinsic origin as previously described at the early stage of cerebellum development ⁷⁰.

Discussion

Our aim in the present paper was to better understand the developmental origin of functional abnormalities (e.g. epilepsy, cognitive deficits, …) in cortical circuits of human patients with KIF7 mutations by characterizing the developmental abnormalities of the cerebral cortex in a Kif7 -/- mouse model. We show that Kif7 ablation affects the development of the two populations of cortical neurons, cortical plate neurons and GABAergic interneurons, in specific and distinct ways, accordingly with their dorsal and ventral origin in the telencephalon. The alterations of the CP development (lack of most subplate cells in the dorsal cortex, and delayed corticofugal projection) are remiscient of defaults previously described in GLI3 null mutants that lack most subplate neurons and pioneer cortical projections ⁷¹. The abnormal distribution of Kif7 -/- cIN in the developing cortex partly reflects the structural and functional alterations of cortical layers on which cIN migrate (e.g. abnormal CXCL12 expression in the SVZ). However, we identify additional migration defaults of Kif7 -/- cIN that recall those of cIN lacking a primary cilium or of wild type cIN treated with the SHH pathway inhibitor cyclopamine, suggesting that the SHH pathway activity is inhibited in Kif7 -/- cIN. Kif7 ablation thus alters differently the SHH pathway activity in CP and cIN subpopulations, in agreement with their dorsal or ventral origin.

Kif7 knock-out mice as a model to study the effect of KIF7 mutations in human

KIF7 is a ciliary kinesin whose mutations in patients and ablation in mice alter the length rather than the structure of primary cilia. Primary cilium length was increased in patient fibroblasts ¹⁰, in MEF isolated from Kif7 -/- and in the neural tube of Kif7 -/- embryos at E10.5 ⁷²³. Owing that KIF7 regulates the trafficking and processing of GLI factors, the cilium-dependent SHH signaling was altered in both patient fibroblasts ¹⁰ and in Kif7 -/- embryos ^3,73 with GLI3-R decrease and GLI1/GLI2 up-regulation. These common features observed in cells from patients and in Kif7 -/- mice models, as well as a previous study of the corpus callosum agenesis in Kif7 mutant mice, which helped to dissect the mechanism at work in patients ⁴⁴ allowed to propose that the Kif7 -/- mouse is a good model to study the alterations that lead to clinical features in patients. Most Kif7 -/- mice died at the end of gestation. However, the brain developed up to birth allowing to study its development.

All patients carrying mutations in the KIF7 gene on both alleles have developmental delay/intellectual disability (ID) and some have seizures. Investigations on cerebral malformation by MRI have shown that some of them have alterations in the cerebral cortex including cortical atrophy ^9,12,13 and pachygyria ¹⁶. Since detecting defects in the cerebral cortex is more demanding in MRI quality images than corpus callosum agenesis or molar tooth sign, we can speculate that the number of patients with cerebral cortex defects is underscored. Malformations of cortical development was recognized as causes of ID ⁷⁴ and epilepsy ⁷⁵ and the cellular defects underlying these pathologies could imply defects in cell proliferation, neuronal migration and differenciation that lead to improper excitatory/inhibitory balance, connectivity into the cortex or of the cortex with other brain structures. The mechanism of action of KIF7 to transduce SHH ciliary dependent signaling has been debated. It has been accepted that SHH activates KIF7 binding to microtubules and its accumulation to the tip of the cilium, making the kinesin motor domain the most essential domain for SHH transduction. A more recent study proposed that KIF7 is an immobile kinesin translocated to the cilium tip after SHH binding by a complex KIF3A/Kif3B/KAP ⁷⁶ suggesting that domains in the KIF7 protein essential to its function are more complex. This could explain why any mutation, even point mutation in the Coiled-coil domain or in C- terminal part of the KIF7 protein induces clinical feature as severe as when mutation led to a highly truncated protein. Another explanation is that any mutation causes unproper folding of the protein probably inducing its degradation by the proteasome.

KIF7 loss of function alters the cerebral cortex organization and the connectivity between the cortex and the thalamus

Our study demonstrates that in WT embryos, the levels of GLI3-FL and GLI3-R were much higher in the developing cortex than in the MGE and the cleavage of GLI3-FL in GLI3-R was more important in the cortex. Our observations are coherent with previous studies that identified a preponderant role of GLI3-R signaling to pattern the dorsal telencephalon and to regulate the proliferation and differenciation of neurons in the cortex ^41,42. Moreover, the strong decrease of the cleavage of GLI3-FL in GLI3-R that we observed only in the cortex of Kif7 -/- embryos suggests that the hyper SHH signaling resulting from this decreased clivage affects the cortex, but not the MGE. The abnormal structure of the dorsal telencephalon in Kif7 -/- embryos (enlarged lateral ventricles, cortical wall folding and thinning) resembled the abnormal structure observed in ciliary mutants ^54,77 and in GLI3 mutants ^78–80. As in Kif7-/- embryos, a mild PBS disorganization and an abnormal differentiation of the caudal telencephalic vesicle and telo-diencephalic boundary that is dependent on SHH and GLI3 for its formation ⁸¹ were also reported in these mutants ^77,78. The cellular disorganization observed in the developing cortex of Kif7 -/- embryos also recall defects of Gli3 mutants, especially the impaired preplate and SP neurons differentiation, the abnormal cortical layer formation and abnormal axonal growth ^78–80. Several of these traits have been observed in the cortex of the ciliary mutant Rfx3 -/- that exhibits also a decreased GLI3-R/GLI3-FL ratio in the dorsal telencephalon ⁵⁴. Interestingly, the knock down of KIF7 in E14.5 cortical cells altered several developmental stages of cortical neurons, leading to a decrease of apical radial progenitors, basal IP divisions, and to impaired neuronal polarity and migration ⁴⁵. Sporadic heterotopia as we observed in the cortex of Kif7 -/- embryos, have been reported in one patient ¹⁶. Cortical heterotopia is not a landmark of ciliopathy but has been also described at E11 in the forebrain of the Ift88 hypomorphic mutant cobblestone (cbs, ⁷⁷. And cellular alterations that should contribute to the formation of heterotopia, e.g. defective capacity to polarize and to migrate to the cortical plate, were described in post-mitotic neurons electroporated with shRNA libraries targeting ciliary proteins such as KIF7, BBS1, BBS10, NPHP8, ALMS1 ⁴⁵. Heterotopias were also detected in a Joubert syndrom patient carrying a mutation in the CELSR2 gene coding for a planar cell polarity protein ⁸² essential for neural progenitor cell fate decision, neural migration, axon guidance and neural maturation ^83,84.

Beside delayed growth of cortical and thalamic axons, Kif7 -/- embryos exhibited connectivity defects between the thalamus and cortex, with abnormal trajectories of axon bundles in the basal forebrain, at the telo-diencephalic junction and along the PSB. To our knowledge, the ectopic projection of cortical axons to the pre-optic area had not been observed in GLI3 mutants that exhibited instead an ectopic thalamic projection to the amygdala ^78,79. In the ciliary mutants Rfx3 -/-, ectopic cortical and thalamic axons were indeed observed in the basal telencephalon ⁵⁴. Cortical neurons electroporated in utero with shRNAs to deplete KIF7 or ciliary proteins belonging to the BBSome, extended axons that showed abnormal fasculation and trajectories ⁴⁵, demonstrating a cell-autonomous role of ciliary genes to control the axonal pathfinding. Finally, a large population of thalamic axons that had reached the cortex of Kif7 - /- embryos showed a massive elimination at E18.5 in the caudal telencephalon. This massive loss paralleled the loss of CP cells, as previously observed in Rfx3 -/- and lnpp5e -/- mutants⁵⁴. Although disorganized thalamo-cortical connectivity had never been reported in patients with KIF7 mutation or in patients with ciliopathy, thalamocortical connectivity is essential for cognitive performances in young infants ⁸⁵ and its alteration leads to neurodevelopemental delay and later cognitive deficits ⁸⁶. Moreover, a link beween ciliary function, cortex development and connectivity with the thalamus had been previously identified in mutant mice lacking the planar cell polarity proteins CELSR2 and FZD3. These mice display ciliary defects and an abnormal development of tracts between the thalamus and the cortex leading to an hypotrophy of the intermediate zone where thalamo-cortical and corticofugal axons navigate ^87,88.

Extra-cortical origin of SHH

Beside its major function in forebrain patterning ^20,40, SHH regulates the embryonic and perinatal cerebral cortex development ^27,28. The origin of the morphogen in this brain structure remains obscure. In agreement with a recent set of data from Moreau et al. ⁸⁹, we showed that cIN born in the MGE or POA no longer express SHH after they reach the developing cortex. In addition, the level of local Shh mRNA synthesis was extremely low in the cortex at E14.5. In contrast, SHH protein was detected in immunoposive dots distributed throughout the cortex. Dots aligned radially in the VZ and CP, and tangentially in the IZ. This cortical distribution is compatible with an extrinsic origin of SHH and a transventricular delivery as described in the cerebellar ventricular zone ⁷⁰. SHH is indeed present in the cerebrospinal fluid and likely secreted by the choroid plexus, since Shh mRNA is expressed in the choroid plexus of mice from E12 to E14.5 ⁹⁰. Accordingly, we showed here that the choroid plexus of the lateral ventricles was strongly immunopositive for SHH at E14.5. Interestingly, the CDO/BOC co-receptor - that favors SHH transport - was also detected along radially oriented structures by immunostaining in the mouse embryonic cortex (not illustrated). SHH could thus be transported away from the ventricular zone and/or meninges in vesicular structures able to diffuse accross the cortical thickness. Accordingly, it has been shown that SHH transport needs ESCRT-III member protein CHMP1A ¹⁵ and SHH related proteins are found in exosomes secreted from Caenorhabditis elegans epiderman cells ⁷², Drosophila wing imaginal discs ⁸⁸, chick notochord cells and human cell lines ⁹¹. How SHH is transported throughout the cortex should be further explored. It has already be shown that SHH is present in multivesicular bodies, in neurites and filipodia in the postnatal hippocampus and cerebellum, and in vesicles located inside and outside of hippocampal neurons in culture ⁹².

Kif7 deletion leads to defects in cortical interneuron migration

The defects induced by gene mutations responsible for primary cilium dysfunction, including Kif7 deletion, have been examined in principal neurons born in the dorsal telencephalon, which organize the cortical layers ^93,94. However, the cIN are born ventrally in the MGE where the expression levels of GLI3-R and GLI3-FL are extremely low in both WT and Kif 7-/- embryos. cIN leave the MGE and then migrate to the cortex during the embryonic development. Their primary cilium is functional and can transduce local SHH signals to influence their directionality, especially the tangential to radial reorientation of their trajectories toward the cortical plate ²⁹. In the present study, we showed that the abnormal distribution of cIN in the mutant cortex of Kif7 -/- embryos resulted from cell autonomous defects and from extrinsic cortical defects, especially the transient lack of CXCL12 expression in the dorsal cortex at E14.5. Interestingly, Kif7 ablation affected the migratory behavior of cIN cultured on a WT cortical substrate in a similar way as did the ablation of Kif3a ²⁹, another ciliary gene required for the transduction of SHH signals ⁹⁵. Accordingly, cyclopamine application on WT cortical slices altered the migratory behavior of cIN in a way that mimicked the migratory behavior of Kif7 -/- cIN. We thus concluded from these observations that Kif7 ablation inhibited the Shh pathway activity in cIN. By consequence, Kif7 -/- cIN, which normally differentiate in a ventral telencephalic domain expressing high SHH level, present dynamic and trajectories alterations that mimics those of cIN lacking SHH signalling. Therefore Kif7 ablation in MGE-derived cIN does not activate SHH signalling by decreasing the GLI3-R/GLI3-FL ratio, as observed in the dorsal telencephalon. On the contrary, Kif7 ablation in MGE-derived cIN likely inhibits SHH signalling by a mechanism independent of GLI3 processing that was unchanged in the MGE of Kif7 -/-embryos (Fig.2). In addition to this cell autonomous defect that impaired cIN migration, we show that the lack of CXCL12 expression in the dorsal cortex transiently and locally modified the directionality of Kif7-/- cIN altering thereby the dorsal cortex colonization by cIN in Kif7 mutants ^59,96.

Interestingly, acute SHH application on WT cortical slices slightly increased the percentage of radially migrating cIN. In contrast, cyclopamine that inhibited the SHH pathway blocked the migration of a large population of cIN in the deep cortical layers demonstrating that endogeneous cortical SHH is likely to activate the SHH pathway in migrating cIN to control their trajectories.

In conclusion, using the Kif7 -/- murine model, we show that the Kif7 deletion is responsible for a large range of developmental defects in the cerebral cortex, which include: i) alteration of the inhibitory interneurons migration by a combination of intrinsic and extrinsic factors, ii) abnormal formation of cortical layers by the cortical plate neurons and iii) major pathfinding defects in axonal projections between the cortex and the thalamus. Patients carrying mutations in the KIF7 gene are classified as ciliopathic patients and display developmental delay, ID and epilepsy. These clinical features could thus be caused by an abnormal distribution of excitatory and inhibitory cortical neurons leading to impaired excitatory/inhibitory balance, and by the development of abnormal connections in the cerebral cortex and with others brain structures. By demonstrating focal heterotopia and abnormal thalamo-cortical connectivity in the murine model, we hope to open a new field of clinical investigations using MRI and tractography to identify such defects in patients carrying mutations in the KIF7 gene.

Material and methods

Mice

E14.5 embryos of the swiss (for in situ hybridization and RNAscope) and C57/Bl6J (for sonic hedgehog immunostaining) strains were obtained from pregnant females purchased at JanvierLabs (France). Male and female wild type (WT) and Kif7 -/- animals (E12.5, E14.5, E16.5, E18.5 and P0) expressing tdTomato or not in the MGE-derived cIN (Nkx2.1- Cre;Rosa26-tdTomato) were generated in our animal facility by mating male Kif7 +/-; Nkx2.1-Cre;Rosa26-tdTomato and female Kif7 +/- mice to produce embryos and P0. Thirty four pregnant females were used. Kif7 +/- mice are a generous gift of Pr Chi-Chung Hui (the Provider, SickKids Hospital, Toronto, USA) and Prof Bénédicte Durand (the Transferor, CNRS, Lyon, France). Mice were genotyped as described previously ³. The day of the vaginal plug was noted E0.5. Experiments have been validated and approved by the Ethical committee Charles Darwin (C2EA-05, authorized project 02241.02) and mice were housed and mated according to European guidelines. Both male and female animals were used in this study. Sex is not considered as a biological variable in embryos.

Brain lysate preparation and immunoblotting

Cortex and MGE from E14.5 WT and Kif7 -/- embryos were dissected and frozen at −80°C until use. Samples were sonicated in 80µl Laemmli buffer and heated for 10 min at 37°C after adding 2.5% β-mercapto ethanol. Proteins were loaded on SDS-PAGE 7% gels (BIO-RAD) and transferred to 0.45 μm Nitrocellulose membranes (BIO-RAD). Membranes were cut to detect signals from proteins above and below 55 KDa and were blocked with 50 g/l non-fat dry milk in PBS/0.1% Tween 20 for 1h at RT, incubated with primary antibodies in the same solution O/N at 4°C, 1 h at RT with appropriate IRDye-conjugated secondary antibodies, and imaged and quantified using ChemiDocMP Imaging System (BIO-RAD). Commercial antibodies anti GLI3 (R&D Sytem-Biotechne, goat, #AF369, 1:200), anti-actin (Millipore, mice, #MAB1501, 1:10.000), Donkey-anti-goat-IR800 (Advansta, R-05781, 1:20.000), Goat-anti-mouse-StarBight Blue 700 (BIO-RA, 12004159, 1:2.500). Band intensities were quantified using gel analysis of Fiji software.

Immunohistochemistry and imaging

Heads of embryos and P0 animals were fixed by immersion in 0.1% picric acid /4% paraformaldehyde (PAF) in phosphate buffer (PB) for 4 hours and then in 4% PAF in PB overnight. After PBS washes, brains were dissected, included in 3% type VII agarose (Sigma, A0701) and sectionned at 70 μm with a vibratome. Immunostaining was performed on free-floating sections as explain previously ²⁹. Primary antibodies were goat anti SHH-Nter (1:100, R&D system AF464), goat anti Netrin G1a (NG1a) (1:100, R&D system AF1166), rabbit anti TBR1 (1:1000, Abcam ab31940), rabbit anti TBR2 (1:1000, Abcam ab23345), rabbit anti PAX6 (1:100, clone poly19013, Covance PRB-278P), rabbit anti GSH2 (1:2000, Millipore ABN162), chicken anti MAP2 (1:500, Novus, NB30213), mouse anti SMI-32 (1:500, Covance SMI-32R), and rat CTIP2 (1:1000, Abcam ab18465). Primary antibodies were incubated in PGT (PBS/gelatine 2g per L/TritonX- 100 0.25%) with 0.1% glycine and 0.1% lysine for anti Netrin G1a antibodies and 3% BSA for anti SHH-Nter antibodies. Primary antibodies were revealed by immunofluorescence with the appropriate Alexa dye (Molecular Probes) or Cy3 (Jackson laboratories) conjugated secondary antibodies diluted in PGT (1:400). Bisbenzimide (1/5000 in PBT, Sigma) was used for nuclear counterstaining. Sections were mounted in mowiol/DABCO (25mg/mL) and imaged on a macroscope (MVX10 olympus) or an epifluorescence microscope (LEICA DM6000) using X40 and X63 immersion objectives, or a confocal microscope (Leica TCS SP5) using a x40 objective. Double immunostaining were performed when possible and images were representative of what was observed in more than 4 animals per genotype except for heterotopia. Macroscop images were treated to remove background around sections using wand tool in Adobe Photoshop 7.0 software. On epifluorescence and confocal microscop images, acquisition parameters and levels of intensity in Adobe Photoshop 7.0 software were similar for WT and Kif7 -/-. Merged images were acquired on the same section except for TBR1 with TBR2 acquired on adjacent sections. The fluorescence intensity along the depth of the cortex was assessed using the plot profile function of ImageJ under a 500 pixels wide line starting in the ventricle to the surface of the brain.

In situ hybridization

Specific antisense RNA probe for Shh, Gli1 and CxCl12 genes (gift of Marie-Catherine Tiveron) were used for in situ hybridization analyses. DIG-probes were synthesized with a labeling kit according to the manufacturer’s instructions (Roche, France). In situ hybridization was performed according to Tiveron et al ⁹⁷ with few modifications (see Supplementary material). Sections were mounted on glass slides, dried, dehydrated in graded ethanol solutions, cleared in xylene and coverslipped with Eukitt.

RNAscope

Embryonic E14.5 heads were cuted in cold Leibovitz medium (Invitrogen) and immersion fixed in cold 4% (w/v) paraformaldehyde (PFA) in 0.12 m phosphate buffer, pH 7.4, overnight. Brains were then cryoprotected in PFA 4% /sucrose 10 %, embedded in gelatin 7.5% /sucrose 10% at 4°C and frozen. Biological samples were kept at −80°C until coronally sectioned at 20 μm with cryostat. RNAscope experiment was performed according manufacturer’s instructions (RNAScope^® Multiplex Fluorescent V2 Assay, ACDbio) after a pre-treatment to remove gelatin. Two different probes/channels were used (C1 for Shh, C2 for Lhx-6). Probes were diluted according manufacturer’s instructions (1 Vol of C2 for 50 vol of C1) and dilution of the fluorophores (Opal 520 and 570) was 1:1500. Sections were mounted in mowiol/DABCO (25mg/mL) and imaged on a confocal microscop (Leica TCS SP5) using an immersion X10 or X63 objectives on 10 µm stacks of images (1 image /µm).

DiI experiments

Tiny crystals of DiI (1,1′-dioctadecyl-3,3,3′3′-tetramethylindocarbocyanine perchlorate) were inserted in the dorsal or lateral cortex of WT and Kif7 -/- embryonic brains fixed at E14.5 by immersion in 4% PAF. Injected brains were stored in 4% PAF at room temperature in the dark for several weeks to allow DiI diffusion. Before sectioning, brains were included in 3% agar. Coronal sections 60 μm thick were prepared with a vibratome and collected individually in 4% PAF. Sections were imaged with a macroscope (MVX10 Olympus).

Whole brain clearing and imaging

E14 and E16 embryos (WT, n=5; Kif7 -/-, n=6) were collected in cold L15 medium, perfused transcardially with 4% PAF using a binocular microscope. Brains were dissected and postfixed 3 days in 4% PAF at 4°C, and stored in PB at 4°C until clearing. All buffer solutions were supplemented with 0.01% Sodium Azide (Sigma-Aldrich). Whole brain immunostaining and clearing was performed at RT under gentle shaking according to a modified version of the original protocol ⁹⁸. Briefly, perfused brains were dehydrated in graded methanol solutions (Fischer Scientific) (20, 50, 80% in ddH₂O, 2×100%, 1h30 each), bleached overnight in 3% hydrogen peroxide (Sigma-Aldrich) in 100% methanol and rehydrated progressively. After PBS washes, brains were blocked and permeabilized 2 days in PBGT (0.2% gelatin (Merck), 0.5% triton X-100 (Sigma-Aldrich) in PBS). Brains were incubated 3 days at 37°C in primary goat anti Netrin G1a (1:100, R&D system AF1166) and rabbit anti TBR1 (1:1000, Abcam ab31940) antibodies. After 1 day of PBGT washes, brains were incubated 1 day at 37°C in secondary 647 donkey anti-goat (1:1000, Molecular probes,) and cy3 donkey anti-rabbit (1:1000, Jackson laboratories) antibodies. Immunolabeled brains were washed 1 day in PBS, embedded into 1.5% low-melting agarose (type VII, Sigma) in 1% ultra-pure Tris-acetate-EDTA solution, placed in 5-ml tubes (Eppendorf, 0030119452) and dehydrated 1h in each methanol baths (50%, 80% and 2×100%). Samples were then incubated for 3h in 33% methanol / 66% dichloromethane (DCM, Sigma-Aldrich), washed in 100% DCM (2x30 min) and incubated overnight (without shaking) in dibenzyl ether (DBE). Brains were stored in DBE at RT. Cleared samples were imaged on a light-sheet microscope (LaVision Biotec, x6.3 zoom magnification) equipped with a sCMOS camera (Andor Neo) and Imspector Microscope controller software. Imaris (Bitplane) was used for 3D reconstructions, snapshots and movies.

Brain slices

Brains of E14.5 WT and Kif7 -/- transgenic Nkx2.1-cre, tdTomato embryos were dissected in cold L15 medium, embedded in 3% type VII agarose and sectioned coronally with a manual slicer as explained in ²⁹. Forebrain slices 250 µm thick were transferred in Millicell chambers and cultured for 4 hours in F12/DMEM medium supplemented with CFS 10% in a CO2 incubator (Merck Millipore) prior to pharmacological treatment and recording. Slices were incubated in drug for 2 hours before transfer in culture boxes equipped with a bottom glass coverslip for time-lapse imaging.

Pharmacological treatment

Either recombinant Mouse SHH (C25II), N-terminus (464-SH-025, R&D systems) at 0.5µg/mL final or cyclopamine (C4116, Sigma) at 2.5 µM final were added to the culture medium of slices and renewed after 12 hours. Recombinant SHH was reconstituted at 100 µg/mL in sterile PBS with BSA 0.1%. The stock solution of cyclopamine was 10 mM in DMSO and the culture medium of control experiments contained 1/4000 DMSO (vehicle). Control experiments in culture medium with and without vehicle did not differ and were analyzed together.

Co-cultures

Co-cultures were performed on polylysine/laminin-coated glass coverslips fixed to the bottom of perforated Petri dishes in order to image migrating MGE cells. Brains were collected in cold PBS at embryonic day E14.5. Cortices and MGE explants were then dissected in cold Leibovitz medium (Invitrogen). Cortices were mechanically dissociated. Dissociated cortical cells were cultured on the coated glass coverslips and left in the incubator (37L°C, 5% CO2) for 1 hours. A wound divided the substrate in two-halves on which WT and Kif7 -/- MGE explants from littermate embryos were placed (Fig. 6A). Co-cultures were cultured for 24h and imaged.

Videomicroscopy and image processing

Slices and co-cultures were imaged on an inverted microscope (Leica DMI4000) equipped with a spinning disk (Roper Scientific, USA) and a temperature-controlled chamber. Multi-position acquisition was performed with a Coolsnap HQ camera (Roper Scientific, USA) to allow the recording of the whole cortex. Images were acquired with a X20 objective (LX20, Fluotar, Leica, Germany) and a 561 nm laser (MAG Biosystems, Arizona). Z-stacks of 30 μm were acquired 50 µm away from the slice surface, with a step size of 2 μm and a time interval of 2 or 5 minutes for at least 21 hours. Acquisitions were controlled using the Metamorph software (Roper Scientific, USA). Cell trajectories were reconstructed on movies by tracking manually the cell rear with MTrakJ (ImageJ plugin, NIH, USA) and were clustered according to their position in cortical layers and to their orientation (see details in Fig. 6). The migration speed of cells, the frequency and duration of pauses were extracted from tracking data using excel macros. The directionality of cells was analysed along time using a macro from ⁹⁹.

Study design

WT group was compared to Kif7 -/- group and to pharmacological treatment groups for brain slice experiments. The experimental unit was a single animal for immunohistochemistry analysis and individual cell for videomicroscopy analysis.

Statistical analysis

All data were obtained from at least three independent experiments and are presented as mean ± SEM (standard error of mean). Statistical analyses were performed with the GraphPad Prism software or R. Statistical significance of the data was evaluated using the unpaired two-tailed t test, the Mann–Whitney test, the Chi2 test or the Two-way ANOVA test followed by a post-hoc test. Data distribution was tested for normality using the D’Agostino and Pearson omnibus normality test. Values of P<0.05 were considered significant. In figures, levels of significance were expressed by * for P<0.05, ** for P<0.01, *** for P<0.001 and **** for P<0.0001.

Acknowledgements

Marie-Christine Tiveron is acknowledged for the gift of the CXCL12 expression vector. We thank all members of the Métin’s Team for constructive discussions. We gratefully acknowledge the Imaging plateform facility of the Fer à Moulin Institute for the use of their microscopes, and the animal facility of the Fer à Moulin Institute for animal care and breeding.

Additional information

Data availabitily

The data presented in this study are available from the corresponding author upon request without undue reservation.

Funding

This work was supported by Institut National de la Santé et de la Recherche Médicale (INSERM), Centre National de la Recherche Scientifique (CNRS), Sorbonne University. Agence Nationale de la Recherche (ANR, grant MIGRACIL to C.M.), Fondation pour la Recherche sur le Cerveau (grant R11080DD to C.M.), Fondation J. Lejeune (grant R14108DD to C.M.).

Author contributions

CM and JM conceptualized the project and supervised the work. MP, VG, SSL, JP, AM, NR, SL, CM and JM perfomed experiments. MP, SSL, AM, NR, CM and JM analyzed the data. MP, VG, SSL, CM and JM wrote the paper.

Additional files

Supplemental Figures