Analysis of human samples reveals impaired SHH-dependent cerebellar development in Joubert syndrome/Meckel syndrome

Abstract

Joubert syndrome (JS) and Meckel syndrome (MKS) are pleiotropic ciliopathies characterized by severe defects of the cerebellar vermis, ranging from hypoplasia to aplasia. Interestingly, ciliary conditional mutant mice have a hypoplastic cerebellum in which the proliferation of cerebellar granule cell progenitors (GCPs) in response to Sonic hedgehog (SHH) is severely reduced. This suggests that Shh signaling defects could contribute to the vermis hypoplasia observed in the human syndromes. As existing JS/MKS mutant mouse models suggest apparently contradictory hypotheses on JS/MKS etiology, we investigated Shh signaling directly on human fetal samples. First, in an examination of human cerebellar development, we linked the rates of GCP proliferation to the different levels and localizations of active Shh signaling and showed that the GCP possessed a primary cilium with CEP290 at its base. Second, we found that the proliferation of GCPs and their response to SHH were severely impaired in the cerebellum of subjects with JS/MKS and Jeune syndrome. Finally, we showed that the defect in GCP proliferation was similar in the cerebellar vermis and hemispheres in all patients with ciliopathy analyzed, suggesting that the specific cause of vermal hypo-/aplasia precedes this defect. Our results, obtained from the analysis of human samples, show that the hemispheres and the vermis are affected in JS/MKS and provide evidence of a defective cellular mechanism in these pathologic processes.

Joubert syndrome (JS; Online Mendelian Inheritance in Man no. 213300) and Meckel syndrome (MKS; Online Mendelian Inheritance in Man no. 249000) are believed to represent the two extremes of the same multisystemic disorder. JS is characterized by a complex malformation of the cerebellum/brainstem, the most striking feature of which is hypoplasia or complete aplasia of the cerebellar vermis (1, 2). In MKS, which is lethal in utero, this defect is associated with occipital encephalocele or anencephaly, severely cystic kidneys, and abnormal liver and skeleton. To date, 16 genes responsible for JS (NPHP1, AHI1, ARL13B, INPP5E, KIF7, OFD1, TCTN1, and CEP41), MKS (MKS1, B9D1, and B9D2), or both (RPGRIP1L, CEP290, CC2D2A, TMEM216, TMEM67/MKS3, and TCTN2) have been identified (3–6). Although all JS/MKS genes encode a ciliary/basal body (BB) protein, which is compelling evidence that the primary cilium is involved in these pathologic conditions, the cellular mechanisms by which cilium dysfunction leads to severe brain malformations remain a mystery.

Because the use of human subjects to study human pathologies is problematic, insight into diseases usually comes from the analysis of murine models. However, none of the currently available murine models of JS/MKS faithfully recapitulate the cerebellar phenotype seen in JS/MKS. Interestingly, in ciliary conditional mutant mice, the cerebellum is hypoplastic and Sonic hedgehog (Shh)-dependent proliferation of granule cell progenitors (GCPs) is severely reduced, as in mice in which the JS-associated gene RPGRIP1L has been abrogated, suggesting that cilia-related defects in Shh-induced GCP expansion might explain the cerebellar abnormalities observed in JS (7–9). In contrast, Shh-dependent GCP proliferation and cerebellar structure were only mildly affected in Ahi1 or Cep290 KO mice, in which cilia formation is not modified (10). Thus, the analysis of ciliary mutants and JS/MKS mice models yields antagonistic hypotheses on the involvement of Shh-driven GCP proliferation in the etiology of the human forms of the syndromes.

To investigate the relation between human ciliopathies and Shh-dependent GCP proliferation, we first analyzed Cep290-associated JS in eight human fetal samples and Cep290-RNAi–induced down-regulation mouse cell culture models, and demonstrate the localization of the CEP290 protein at the base of cilia and its requirement for cilium formation. We then analyzed GCP proliferation and Shh signaling in these fetal human JS/MKS samples and age-matched controls, and show that the high GCP proliferation rate is correlated to detectable levels of Shh mRNA expression by Purkinje cells (PCs), and is strongly reduced in Cep290-associated JS/MKS samples. Surprisingly, however, in JS/MKS, the reduction in the rate of proliferation of GCP is similar in the vermis and the cerebellar hemispheres, suggesting that it may be responsible for a global cerebellar phenotype in the disease. The vermal hypoplasia/aplasia would most probably result from a pathological event occurring earlier in development. Remarkably, similar GCP proliferation defects and impaired Shh signaling were also found in three other JS/MKS cases (TMEM67/MKS3, CC2D2A, unidentified mutation), as well as in a Jeune asphyxiating thoracic dystrophy (JATD) case. These findings show that impaired Shh signaling and defective GCP proliferation underlie JS/MKS syndromes and suggest that these defects might be common to a number of ciliopathies.

Results

Human GCP Possess Primary Cilia.

In mice, GCPs, located in the external granular layer (EGL), actively proliferate in response to Shh, leading to massive cerebellar growth and generation of the largest neuronal population in the brain, the granule cells (11–15). We and others have previously shown that murine GCP have a primary cilium, which is required for their proliferation in response to Shh (8, 9). In JS/MKS, ciliary genes mutations lead to a drastic hypoplasia of the cerebellar vermis that could be caused by defects in cilia function in the GCP.

To determine whether control human GCP possess a primary cilium, we immunostained paraffin or fresh frozen cerebellar sections from human fetuses aged 12 to 36 gestational weeks (gw) with cilia-specific antibodies (Arl13B or Adenylate cyclase 3). Because no specific staining was observed in cilia, probably as a result of antigen destruction by late postmortem fixation of the tissue and paraffin embedding, we examined a human gw 21 cerebellar sample, post mortem, by transmission EM. At this developmental stage, human GCP had primary cilia with the usual features: alar sheets, a basal foot, and a rootlet (Fig. 1A, arrows). Interestingly, every cilium analyzed (n = 18) grew from a ciliary pocket (16) and its base was coated with electron-dense material (Fig. 1A, arrowhead). These results thus show that control human GCPs possess a primary cilium. Unfortunately, we could not determine whether human GCPs possess a primary cilium in JS/MKS cases because all JS/MKS samples used in this study (as described later) are paraffin-embedded for clinical diagnosis and therefore not suitable for transmission EM analysis.

Fig. 1.

CEP290 is enriched at the base of murine and human GCPs; shRNA-mediated down-regulation of Cep290 reduces the number of ciliated cells. (A) Transmission EM images of a gw 21 sample of human EGL. Left: Three GCPs have a primary cilium (yellow arrow) that extends from the BB (black asterisk) near the centriole (blue asterisk). The primary cilium has a basal foot (black arrow) and a ciliary pocket, the base of which is coated by electron-dense material (yellow arrowhead). (B) Immunohistofluorescence imaging of CEP290 (green) and a BB rootlet component rootletin (red) on cryosections of mouse (Left) and human cerebellum (Right). In murine EGL, dots of CEP290 surround the rootlet; this pattern is observed from embryonic day 16 to postnatal day 14. In human EGL at gw 21, CEP290 is also localized in dots surrounding the BB rootlet. (C) Immunostaining of CEP290 (white), GFP (shRNA transfected cells, green), and the ciliary and BB marker polyglutamylated tubulin (GT335, red) in adherent astroglial stem cells, transfected with GFP-expressing plasmids driving the expression of scramble or CEP290 shRNA (green). Most control transfected cells express CEP290 and have a primary cilium. When CEP290 is no longer detectable in CEP290 shRNA-expressing cells, they do not possess a primary cilium. (D) Quantification of primary cilia in scramble or CEP290 shRNA-transfected cells. For CEP290 shRNA-transfected cells, only those in which CEP290 was no longer immunodetectable were quantified. CEP290 down-regulation leads to a 69% reduction in cells presenting a primary cilium. The result shown is the mean of three replicates. Error bars indicate SD. (Scale bars: A, left to right, top to bottom, 10 μm, 2.5 μm, and 0.5 μm; B, 5 μm; C, left to right, 2 μm and 1 μm.)

CEP290 Is Localized at Centrosome of Human GCP and Is Involved in Murine Neural Cilium Assembly.

So far, the largest number of ciliopathies is caused by mutations in CEP290/NPHP6 (17), and genetic abrogation of Cep290 in mice leads to mild cerebellar hypoplasia (10). The Cep290 transcript is expressed in GCPs and their progeny (18), but the subcellular location of the protein in mouse and human cerebellum is still unknown. By immunohistochemistry, we detected CEP290 in granular structures scattered around BB rootlets in mouse and human GCPs (Fig. 1B), as previously described in hTERT-RPE cells (5, 19). CEP290 was also found at the ciliary transition zone of cultured primary mouse neural progenitors, as reported in hTERT-RPE1 cells and IMCD3 cells (3, 5), as well as at the centrosome, as in IMCD3 cells (Fig. 1C and Fig. S1) (3, 20, 21).

CEP290 is involved in the assembly of primary cilium in several established cell lines (3, 19, 22–24), but no data are available regarding the brain. Therefore, we down-regulated Cep290 by RNAi in cultured primary mouse neural progenitors (the shRNAs were previously tested for efficiency in HEK cells; Fig. S1). Cells were transfected with shRNA and plated at high density to rapidly reach confluence. Three days later, almost 70% of the control cells, but only 20% of the CEP290-depleted cells, had primary cilia (Fig. 1 C and D). Thus, our results show that CEP290 is involved in the assembly of primary cilia in neural cells. In patients with CEP290 mutations, JS might therefore result from ciliary defects.

GCP Proliferation Is Impaired in Cerebellar Vermis and Hemispheres in JS/MKS.

We and others have previously shown that, in mice, selective genetic ablation of genes required for cilia formation (Kif3A, Ift88, or Ftm) in GCP leads to ataxia and cerebellar hypoplasia caused by impaired Shh-dependent GCP proliferation (8, 9, 25). Cep290 KO mice, however, have only a mild cerebellar phenotype that mainly results from Shh-independent mechanisms (10).

Given the prominent role of cilia in Shh signaling in most organs analyzed so far (26, 27), we quantified GCP proliferation in the cerebellum of 12 cases of JS/MKS caused by mutations in the ciliary genes CEP290, CC2D2A, or TMEM67/MKS3 or by unidentified mutations (Table S1) and 11 age-matched controls selected for their lack of cerebral involvement. Fetal cerebellar sections were stained with anti-Ki67 to label proliferating GCPs, which were quantified in the EGL of the vermis and the cerebellar hemispheres, and normalized to the EGL surface (Fig. 2 A and B). In control cases, the rate of GCP proliferation decreased greatly from age 12 gw to age 14 to 15 gw (P < 0.05). It then increased greatly from age 16 gw to age 21 gw (P < 0.005), after which it stabilized (Fig. 2B). GCP proliferation in the pathological case sampled before gw 16 was similar to GCP proliferation in controls. In contrast, in 10 of 11 cases sampled after gw 16, the rate of GCP proliferation was significantly reduced compared with the control cases (Fig. 2B). These results show that defective GCP proliferation after gw 16 is common to almost all cases of JS/MKS.

Fig. 2.

GCP proliferation is severely impaired in the cerebellar vermis and hemispheres of most JS/MKS cases. (A) Immunolabeling of cycling cells expressing the marker Ki67 (red) in the EGL in the gw 23 JS/MKS case 070151 and a gw 21 control. In all panels, a dashed line outlines the EGL. Note that there are fewer labeled cells in the EGL of the JS/MKS case (Right) compared with control (Left), although thickness of the EGL is comparable in both cases. (B) Quantification of Ki67-positive cells per 100 μm² in the EGL of 12 cases of JS/MKS (CEP290, MKS3, CC2D2A, and mutations still under investigation), one case of JATD, one case of HLS, and 11 age-matched controls. In the controls, GCP proliferation is high at 12 gw, greatly decreases at approximately 14 to 15 gw (P < 0.05), increases again from gw 15 to gw 21 (P < 0.005), and then stabilizes at a significantly higher plateau. The number of Ki67-positive cells in the cerebellum (vermis and hemispheres) is dramatically decreased in the majority of JS/MKS cases (10 of 12) and in the JATD case, compared with controls. (C) Ki67-positive cells per 100 μm² in the cerebellar vermis and hemispheres in the control and JS cases in B. As in all control cases (green), in five of seven of the JS/MKS cases (blue) studied, the proliferation defect was similar in the vermis (light green/blue) and the hemispheres (dark green/blue). (Scale bar: 10 μm.) ***P < 0.001, **P < 0.01, and *P < 0.05.

GCP Proliferation Is Also Disrupted in JATD.

Ataxia, which frequently results from cerebellar dysfunction, and cerebellar abnormalities have been reported in several ciliopathies [Bardet-Biedl syndrome (BBS), orofaciodigital syndrome (OFD), nephronophtisis (NPHP), JATD] (28–31). We have quantified GCP expansion in a gw 17 JATD sample, in which postaxial polydactyly, short lower limbs, a pear-shaped thorax with short ribs, and retrognathism were observed, but development of the cerebellar hemispheres and vermis appeared normal on neuropathological examination. GCP proliferation was severely impaired in this subject (Fig. 2B), supporting the hypothesis that defective GCP expansion might be common to a number of ciliopathies.

GCP Proliferation Is Similar in Cerebellar Vermis and Hemispheres in JS/MKS and Controls.

To test whether the vermis was more affected than the hemispheres in JS/MKS, given the particularly severe hypo-/aplasia in this territory in the disease, proliferation rates were evaluated separately in each structure and compared with controls. Although the vermis can be completely lost in JS/MKS, the structure was present in all 12 of our JS/MKS cases. Nevertheless, only seven of the 12 cases could be analyzed quantitatively because the structure was damaged or lost as a result of sectioning in the sagittal plane (Fig. S2). Surprisingly, in five of seven JS cases, proliferation rates in the vermis and in the hemispheres were similar, as seen in all the control cases analyzed (Fig. 2C). As the cerebellar hemispheres are affected to the same extent as the vermis in JS/MKS, vermal hypo-/aplasia most probably results from a pathogenic event occurring earlier in cerebellar development, not from impaired GCP proliferation.

Shh Pathway Is Active in Early Fetal Human Cerebellum.

In the mouse cerebellum, massive proliferation of GCPs and consequent cerebellar growth are driven by a potent mitogen, SHH, which is secreted by PCs starting at embryonic day 17.5 and induces Gli1 expression in the GCPs (11–15). Until very recently, the precise timing of Shh pathway activation and its functional relevance to GCP proliferation and cerebellar growth had not been described in humans.

To establish the cellular and temporal patterns of SHH expression in the human cerebellum, we assessed SHH mRNA expression by in situ hybridization from 12 to 34 gw. Before gw 17, SHH mRNA levels were undetectable in CaBP+ cells and in the EGL (Fig. 3A). Unlike mouse PCs that start expressing SHH when they integrate the PC cell layer (14, 15), at gw 17, staining was detected in a subset of PCs that were still migrating (30) (Fig. 3A). As PCs aligned into a multicellular layer, SHH expression intensified and was observed in all PCs at 33 gw (Fig. 3A).

Fig. 3.

SHH signaling is active in control cases in which the exponential increase in the EGL surface begins. (A) Combined in situ hybridization of SHH or GLI1 mRNA and calbindin immunostaining at gw 12, 17, and 33. At gw 12, neither the PCs (arrows) nor the EGL express SHH, whereas the EGL expresses GLI1. At 17 gw, around the observed peak of GCP proliferation, some calbindin-positive PCs still migrating toward the pial surface express detectable levels of SHH mRNA (arrows) whereas others do not (arrowheads). At the same time, GLI1 mRNA is expressed in the EGL. At gw 33, PCs, aligned in a single-cell row beneath the EGL, all express high levels of SHH mRNA (arrows). (B) Quantification of the EGL surface on horizontal H&E-stained sections (green dots). From the onset of detectable SHH expression (i.e., gw 17, blue arrow), the exponential curve fitting all the data points starts to diverge from the linear regression calculated from the data points at gw 10 to gw 15. The divergence corresponds with the timing of the GCP proliferation peak (Fig. 2, purple arrows), suggesting that SHH-dependent increase in the rate of GCP proliferation is responsible for the exponential increase in the surface of the EGL. (Scale bars: 20 μm.)

To gain insight into the timing of the GCP response to SHH signaling, we performed in situ hybridization to evaluate GLI1 mRNA expression in human GCPs on the same samples previously assessed for SHH mRNA expression. Although we were not able to detect the expression of SHH in PCs before 17 gw, we observed that GCPs expressed GLI1 at all stages analyzed (Fig. 3B), suggesting that an additional source of SHH might exist before 17 gw (32, 33). However, SHH mRNA expression detected in the PCs from 17 gw coincides with a peak in GCP proliferation (Figs. 2B and 3); thus, these results strongly suggest that massive GCP proliferation correlates with active Shh signaling in the developing human cerebellum.

SHH Expression Is Correlated with Cerebellar Growth.

In mice, it has been proposed that the potent mitogenic effect of SHH is not required for the basal proliferation of GCPs but rather for the massive postnatal proliferation of these cells that is necessary for lobe growth (14). To investigate whether SHH expression is also correlated with the surface growth necessary for lobulation in the human cerebellum, we measured the surface area of the EGL throughout normal fetal development on histologically stained horizontal sections. The surface of the EGL increased exponentially with time. The curve diverged from the linear at approximately gw 16, indicating an important increase in the rate of expansion of the EGL surface at this point in development (Fig. 3B). These results suggest that strong levels of PC-derived SHH drive the massive GCP proliferation necessary for lobe growth.

To assess the contribution of the early activation of Shh in the EGL (Fig. 3A), we evaluated GCP proliferation at 15 gw in a case of JS/hydrolethalus syndrome (HLS) caused by an inactivating mutation in KIF7, the human orthologue of Drosophila Costal2, a key component of the Hedgehog signaling pathway. Although Shh signaling was compromised (Fam 1, Fetus 3) (34), GCP proliferation was not impaired at gw 15 in this case (Fig. 2B). This supports the possibility that SHH pathway activation before 16 gw only mildly contributes to the massive GCP proliferation driving lobe growth.

Response of GCP to Shh Signaling Is Impaired in JS/MKS.

Shh pathway activation can be impaired as a result of defects in SHH expression or in the GCP response to this signal. Faulty migration of PCs might contribute to a deficit in Shh signaling, as suggested by the heterotopic PC and local interruptions in the PC layer observed in all our JS/MKS cases (Fig. S2). However, combined in situ hybridization of SHH and calbindin labeling of PCs on sections from eight JS/MKS and four control cases showed that, when PCs reached their destination, they expressed normal levels of SHH (Fig. 4 A and B). On the contrary, the levels of GLI1 mRNA and PTC in the GCP were reduced in most analyzed JS/MKS cases in which GCP expansion was defective, indicating an impaired response of GCPs to SHH (Fig. 4).

Fig. 4.

Conserved SHH expression and reduced GLI1 mRNA and PTC levels in most JS/MKS cases with defective GCP proliferation. (A) GLI1, PTC, and SHH expression in JS samples compared with controls. SHH, GLI1, and PTC levels of expression were assessed by in situ hybridization (SHH, GLI1) or immunohistochemistry (PTC). SHH mRNA expression is conserved in the vast majority of JS/MKS cases with defective GCP proliferation, but GLI1 mRNA and PTC levels are reduced or undetectable, indicating a defective response to SHH. (∼, comparable to WT; ND, not determined; NE, not expressed; −, diminished.) (B) Combined SHH in situ hybridization and calbindin immunolabeling of JS case 080144 and an age-matched control. SHH mRNA levels in JS calbindin-positive PCs (green) are comparable to control levels. The EGL and PC layer (PCL) are delimited by a dashed line. (C) GLI1 in situ hybridization or PTC immunohistochemistry and DAPI staining in two JS cases and age-matched controls. GLI1 mRNA and PTC expression are reduced in the JS EGL. (Scale bar: 20 μm.)

Discussion

Our study shows that SHH-dependent GCP proliferation is impaired in fetuses with JS/MKS in whom ciliary genes are mutated. This cellular mechanism, which we demonstrate here in humans, is found in patients with both JS and MKS, as well as in cilia or RPGRIP1L mutant mice (8, 9), and is thus conserved through evolution. GCP expansion and the response to SHH were also defective in a case of JATD, raising the possibility that impaired Shh-dependent GCP proliferation might be common to a number of ciliopathies.

Previous studies in which GCP cilia were ablated in mouse models provided the first evidence that defects in Shh-dependent GCP proliferation may contribute to the cerebellar phenotype observed in JS and MKS (23, 24). It was unclear, however, whether these defects were responsible for hypo-/aplasia of the vermis or affected cerebellar development globally. In the present study, we evaluated GCP proliferation at different developmental stages in 12 cases of JS/MKS, one case of JS/HLS, and one case of JATD. Interestingly, although defective GCP proliferation and vermis hypoplasia were observed in most of our JS/MKS cases (11 of 12), vermis hypoplasia was also found to be present in case 18485, which does not present a GCP proliferation defect. Furthermore, in five of seven analyzed fetuses, defective GCP proliferation affected the vermis and the cerebellar hemispheres to the same extent (Fig. 2). These results suggest that defective GCP expansion is not responsible for vermal hypo-/aplasia in JS/MKS, but, most probably, for the global cerebellar phenotype. In a recent study in Cep290 and Ahi1 KO mice, as well as in three subjects with JS, cerebellar midline fusion was incomplete (10). According to this study, this cerebellar defect could explain vermian hypoplasia and would originate from defective Wnt signaling (10). Therefore, it would be interesting, in a future work, to assess Wnt pathway activation in JS/MKS fetal tissue samples by performing in situ hybridization of two direct Wnt targets, AXIN2 (35) and LEF1 (36, 37).

In humans, genotype–phenotype correlations have previously been established for some JS genes or pathological CEP290 alleles (35, 36). In our study, there were no obvious correlations between the different CEP290 alleles responsible for JS and the degree to which GCP proliferation was impaired: the two pairs of siblings analyzed (cases 16676 and 16290 and cases 18485 and 050158; Fig. 2B and Table S1) exhibit radically different degrees of GCP proliferation compared with their age-matched controls, even though they have the same CEP290 alleles. We did observe, however, that, when TMEM67/MKS3 was mutated (case 060247), proliferation was more severely impaired than in JS/MKS caused by other mutated genes (Fig. 2B), suggesting a possible phenotype–genotype correlation between the mutated gene and the proliferation defect. Additional cases of JS/MKS with TMEM67/MKS3 mutations need to be analyzed to confirm this observation. In mice, such a genotype–phenotype correlation is not observed. Rather, a correlation might exist between ciliary function and GCP proliferation: mouse models in which GCP cilia were abrogated (Kif3a, Ift88, or RPGRIP1L mutants) showed very severe defects in GCP proliferation (23, 24). In contrast, Cep290 and Ahi1 KO mice and mice with spontaneous deletion of the mouse orthologue Tmem67/Mks3, had no defects in either cilia formation or Shh-dependent GCP proliferation (10, 38). Such strong differences in phenotype between humans and mice further outline the necessity of studies addressing the cellular and molecular mechanisms of human pathologic conditions in actual human tissue samples.

Materials and Methods

Tissue Samples.

The human fetal material was obtained from the tissue collections of the Histology, Embryology and Cytogenetics Department of the Necker Children’s Hospital, Paris, France. The tissue was obtained from spontaneous or medically induced abortions, with appropriately informed, written maternal consent for brain autopsy, the use of the brain tissue, and access to medical records for research purposes. The study was approved by the Ethics Boards of Hôpital Necker-Enfants Malades. Unstained paraffin-embedded JS and control tissues, as well as H&E-stained sections, came from the tissue archives. We also obtained unfixed control cerebellar samples. The mice were C57Bl6/J or mixed backgrounds purchased from Janvier.

Fluorescence Microscopy and in Situ Hybridization.

Unfixed human and mouse tissue was flash-frozen, embedded in Tissue-Tek optimal cutting temperature compound (1S-LB-4583-EA; Sakura Finetek), cut into 10-μm-thick sections, mounted, and dried.

For CEP290/rootletin immunostaining, sections were fixed for 10 min in −20 °C methanol. For the other antigens or in situ hybridization, tissue sections were dewaxed and the antigen retrieved by incubation for 30 min at 96 °C in preheated retrieval solution (S2367; Dako), pH 9, and then allowed to cool at room temperature for another 30 min in the same solution. The sections were then incubated for 12 to 72 h at 4 °C with the following primary antibodies: affinity-purified rabbit anti-human and anti-mouse CEP290 (1:500) (gift from S. Saunier, Hôpital Necker-Enfants Malades, Paris, France) (7), goat anti-rootletin (sc67-824, 1:100; Santa Cruz), mouse anti-detyrosinated α-tubulin (clone 1D5, 302011, 1:1,000; Synaptic Systems), rabbit anti-GFP (A11122, 1:1,000; Molecular Probes), mouse anti-Ki67 (MIB1 clone, M7240, 1:50; DAKO), rabbit anti-Pax6 (AB5409, 1:1,000; Chemicon), and rabbit anti-calbindin (CB38a, 1:20,000; Swant). Species-specific secondary antibodies from Sigma-Aldrich were then applied for 1 h to the slides. Finally, the sections were counterstained with DAPI (10 μg/mL; Sigma) and mounted in Fluoromount.

For in situ hybridization on human tissue, antisense riboprobes were labeled as described previously (8) by in vitro transcription of cDNAs encoding mouse Shh (gift from A. McMahon) or human GLI1 (gift from A. Ruiz i Altaba). In situ hybridization was then performed as described previously (8). Standard H&E histological staining was performed on paraffin sections.

Stained cilia, rootlet, and Ki67 were imaged with a Zeiss Observer Z1 microscope equipped with an Apotome device and a Hamamatsu ORCA R2 C10600 camera. Combined in situ hybridization and immunostaining were imaged with a Leica DM 5500B microscope and a Hamamatsu ORCA ER C4742-80 camera.

Transmission EM.

Postmortem fetal tissue was dissected in PBS solution into 1 × 2-mm blocks, which were then fixed in 2% PFA/1% glutaraldehyde solution in 0.1 M phosphate buffer for 3 h. After rinsing in PBS solution, the tissue was postfixed in 1% osmium tetroxide for 30 min on ice, protected from light, with shaking. The tissue blocks were then dehydrated in 50% and 70% ethanol baths in water for 7 min each, and then stained in 1% uranyl acetate in methanol. After final dehydration, samples were immersed for 40 min in a graded series of ethanol/Epon (2/1, 1/1, 1/2 ratios), then in pure Epon. Samples were then mounted in Epon blocks for 48 h at 60 °C to ensure polymerization. Ultrathin sections (70 nm) were cut sagittally on an ultramicrotome (Ultracut E; Reichert-Jung) and analyzed with a Jeol 10–11 transmission electron microscope.

Quantification of GCP Proliferation Rate and Cerebellar Surface.

KI67-positive GCP quantification was accomplished on half horizontal sections of control and human JS fetal cerebella, comprising half the vermis and one hemisphere (routine sectioning for diagnosis done at the Histology, Embryology and Cytogenetics Department of the Necker Children’s Hospital).

For small samples (to 18 gw), the entire EGL was photographed with a 20× objective on a Zeiss Observer Z1 microscope equipped with an Apotome device and a Hamamatsu ORCA R2 C10600 camera. The EGL was then manually reconstituted and, based on EGL length, divided into 16 parts. The number of Ki67⁺ cells and the EGL surface were manually quantified for each part by using ImageJ (National Institutes of Health; on average, 3500 Ki67⁺/Pax6⁺ cells were counted per fetus). For bigger samples, this method became technically very challenging. We therefore sampled the EGL by taking 16 to 20 equidistant images at 20× throughout the EGL. In each picture, the number of Ki67⁺ cells and the EGL surface were manually quantified. In both cases, either when the full EGL was quantified or when it was sampled, the quantified regions are spatially comparable. The 16 to 20 measurements of the proliferation rate per sample were then averaged to obtain the final proliferation rate for each fetus.

The EGL surface area was measured on representative individual sections from each stage (n = 1 or 2 per fetus and n = 1–3 fetuses per stage), which were selected for being situated at the same anteroposterior level: they all encompass the dentate nucleus. The H&E-stained sections were scanned to obtain low-magnification and high-resolution images. On each image, the EGL surface was manually outlined and measured by using ImageJ.

Validation of RNAi-Mediated Down-Regulation of CEP290.

To test the efficiency of CEP290 down-regulation by the CEP290 shRNA, HEK 293T cells were cotransfected with a mCEP290-GFP plasmid (gift from J. Gleeson, University of California, San Diego, CA) and a pGIPZ CEP290shRNAmir-GFP plasmid or a pGIPZ-GFP control plasmid (Abgene). After 24 h, cells lysates were run on a NuPage 4% to 12% Bis-Tris gel. After transfer, the membrane was cut below the 110-kDa marker, and the upper half was blotted for GFP (homemade polyclonal anti-GFP antibody) (39) with the lower half blotted for α-tubulin (Hybridoma Bank clone 12G10, 1/5,000), following standard procedures. Protein bands were visualized with HRP-labeled protein A or anti-mouse antibody, followed by detection with chemiluminescence and quantification with ImageJ software.

Primary Cortical Neural Stem Cell Culture and shRNA Assay.

The cortex of postnatal day 0 to 2 mice was dissected (exclusion of the choroid plexus, hippocampus, corpus callosum, olfactory bulbs, and meninges) and mechanically dissociated, digested with papain. The cells were plated at high density in 10% FCS-containing medium, and allowed to grow to confluence. The primary cortical astroglial cells were then detached and nucleofected in suspension with the expression arrest pGIPZ CEP290shRNAmir-GFP plasmid or with a pGIPZ-GFP control plasmid (Abgene) using a Nucleofector device (Lonza) and the Basic Neuron Small Cell Number Nucleofector kit (VSPI-1003; Lonza). The cells were then plated at high density to form pure confluent monolayers of monociliated astroglia that were fixed for 1.5 min in 4% paraformaldehyde/0.33 M sucrose at 4 °C, then for 10 min in −20 °C MeOH (40). After triple staining for GFP, CEP290, and polyglutamylated tubulin (GT335; gift from C. Janke, Institut Curie, Orsay, France), as described earlier, cilia presence was assessed. In cells in which CEP290 was down-regulated, only GFP⁺/CEP290⁻ cells were quantified.

Statistical Analysis.

The results (i.e., average of the 16 to 20 measures of the proliferation rate obtained for each fetus) are presented as mean ± SD. A two-sided Wilcoxon test was used for all statistical analyses except for comparison of GCP proliferation in the cerebellar hemispheres vs. the vermis, for which a two-sided averages test was used.