CHD7 controls cerebellar development via Reelin
JCI title: The chromatin remodeling factor CHD7 controls cerebellar development by regulatingreelin expression
Danielle E. Whittaker1,2, Kimberley L. H. Riegman1, Sahrunizam Kasah1, Conor Mohan1,Tian Yu1, Blanca Pijuan Sala1, Husam Hebaishi3, Angela Caruso4,5, Ana Claudia Marques6$, Caterina Michetti4,7, María Eugenia Sanz Smachetti1, Apar Shah1,Mara Sabbioni4, Omer Kulhanci8, Wee-Wei Tee9*, Danny Reinberg9, Maria Luisa Scattoni4, Holger Volk2, Imelda McGonnell2, Fiona C. Wardle3, Cathy Fernandes8,9, & M. Albert Basson1,9^.
1King’s College London, Department of Craniofacial Development and Stem Cell Biology, Floor 27, Guy’s Hospital Tower Wing, London, SE1 9RT, UK
2Department of Comparative Biomedical Sciences, Royal Veterinary College, 4 Royal College Street, London NW1 0TU, UK
3King’s College London, Randall Division, 3rd floor, New Hunt’s House, London, SE1 1UL, UK
4Neurotoxicology and Neuroendocrinology Section,Department of Cell Biology and Neuroscience, Istituto Superiore di Sanità, Viale Regina Elena 299,00161 Rome, Italy
5School of Behavioural Neuroscience,Department of Psychology, Sapienza University of Rome, via dei Marsi 78, 00185, Rome, Italy
6Department of Physiology, Anatomy and Genetics, University of Oxford, South Parks Road, Oxford OX1 3PT, UK
7Department of Physiology and Pharmacology "V. Erspamer", Sapienza University of Rome , Rome , Italy
8MRC Social, Genetic & Developmental Psychiatry Centre, PO82, Institute of Psychiatry, Psychology & Neuroscience, King's College London, De Crespigny Park, London SE5 8AF, UK
9Howard Hughes Medical Institute, Department of Molecular Pharmacology and Biochemistry, NYU School of Medicine, New York, USA
9King’s College London, MRC Centre for Neurodevelopmental Disorders, 4th floor, New Hunt’s House, London, SE1 1UL, UK
$Current address: Department of Physiology, University of Lausanne, Rue du Bugnon 7, 1005 Lausanne, Switzerland.
Current address: Institute of Molecular and Cell Biology, A*Star, Singapore.
^Corresponding author: Dr. M. Albert Basson, Department of Craniofacial Development and Stem Cell Biology, Floor 27, Guy’s Hospital Tower Wing, London, SE1 9RT, UK. Tel: +44(0)207 188 1804. E-mail:
To be published open access: Creative Commons CC-BY license
The authors have declared that no conflict of interest exists.
ABSTRACT
Mutation of the gene encoding the ATP-dependent chromatin remodelerCHD7 causes CHARGE syndrome.The mechanisms underlying the neurodevelopmental deficits associated with the syndrome, which includecerebellar hypoplasia, developmental delay, coordination problems and autistic features, are not known. CHD7 is expressed in neural stem and progenitor cells, but its role in neurogenesis during brain development remains unknown. Here we show that deletion ofChd7 from cerebellar granule cell progenitors (GCps) in the mouse results in reduced GCp proliferation, cerebellar hypoplasia, developmental delayand motor deficits. Genome-wide expression profiling revealed downregulatedReln geneexpression in Chd7-deficient GCps. RecessiveRELNmutations is associated with severe cerebellar hypoplasia in humans.We provide molecular and genetic evidence that reducedReln expression contributes to theGCp proliferative defectand cerebellar hypoplasia in GCp-specific Chd7 mouse mutants.Finally, we show that CHD7 is necessary for the maintenance of an open, accessible chromatin state at the Relnlocus.Taken together, this study shows that Reln gene expression is regulatedby chromatin remodeling, identifies CHD7 as a previously unrecognizedupstream regulator of Reln and provides direct, in vivoevidence ofa mammalian CHD protein controlling brain development by modulating chromatin accessibility in neuronal progenitors.
JCI-edited abstract:
The mechanisms underlying the neurodevelopmental deficits associated with CHARGE syndrome, which include cerebellar hypoplasia, developmental delay, coordination problems and autistic features, have not been identified. CHARGE syndrome has been associated with mutations in the gene encoding the ATP-dependent chromatin remodeler CHD7. CHD7 is expressed in neural stem and progenitor cells, but its role in neurogenesis during brain development remains unknown. Here we have shown that deletion of Chd7 from cerebellar granule cell progenitors (GCps) results in reduced GCp proliferation, cerebellar hypoplasia, developmental delay, and motor deficits in mice. Genome-wide expression profiling revealed downregulated expression of the gene encoding the glycoprotein reelin (Reln) in Chd7-deficient GCps. RecessiveRELN mutations have been associated with severe cerebellar hypoplasia in humans. We found molecular and genetic evidence that reductions in Reln expression contribute to GCp proliferative defects and cerebellar hypoplasia in GCp-specific Chd7 mouse mutants. Finally, we showed that CHD7 is necessary for maintaining an open, accessible chromatin state at the Reln locus. Taken together, this study shows that Reln gene expression is regulated by chromatin remodeling, identifies CHD7 as a previously unrecognized upstream regulator of Reln, and provides direct in vivo evidence that a mammalian CHD protein can control brain development by modulating chromatin accessibility in neuronal progenitors.
INTRODUCTION
Mutations in genes encoding chromatin remodeling factors have emerged as a major cause of neurodevelopmental disorders(1). The mechanisms employed by these factors to ensure precise regulation of developmental gene expressionremain largely unexplored. Furthermore, the neuroanatomical abnormalities that result from these mutations and the behavioral and psychiatric features associated with specific brain defects remain unidentified for most of these conditions. Chromodomain-Helicase-DNA-binding (CHD) proteins are ATP-dependent chromatin remodeling factors that use ATP to catalyze nucleosome translocation along chromatin, presumably to modulate access of transcriptional regulators (2). Mutations in CHD7cause CHARGE syndrome, a complex developmental syndrome defined by a constellation of birth defects, which includecoloboma, heart defects, atresia of the choanae, retarded growth and development, genital and ear abnormalities (3-5). Although neurodevelopmental abnormalities are not considered for clinical diagnosis, 99% of patients exhibit developmental delay and74% suffer from intellectual disability(6). The alterations in chromatin and brain structure that underlie these deficits have not been identified.
ChIP-seq experiments in a variety of cell lines have identified widespread CHD7 recruitment across the genome (7-9). CHD7 appears to localize primarily to distal regulatory elements marked by lysine 4 monomethylation on histone 3 (H3K4me1). In vitro nucleosome remodeling assays have confirmed the ability of CHD7 to translocate nucleosomes along a chromatin template and demonstrated a loss or reduction in this activity in CHARGE syndrome-associated CHD7 mutants (10). These findings point to chromatin remodeling defects as a central pathogenic mechanism in CHARGE syndrome. However, whether this holds true in vivo and the effect of such changes on developmental gene expression, have not been investigated.
As a first step towards identifying the neuroanatomical alterations that may underlie the neurological deficits in CHARGE syndrome, we recently reported cerebellar hypoplasia in 35% of patients with CHD7 mutations (11). It remains unknown to what extent cerebellar defects contribute to the developmental delay and intellectual disability associated with CHARGE syndrome.
The cerebellum develops over several weeks in mice and several months in humans. The cerebellar territory is established during early embryogenesis by the action of secreted signaling molecules from a secondary organizer located at the mid-hindbrain boundary(reviewed in 12). We previously reported that bi-allelic expression of Chd7 was essential for maintaining appropriate levels of Fgf8 expression at this early stage of cerebellar development (11).As deregulatedFgf8 expression and signaling from this embryonic organizerselectively disrupts the formation of the cerebellar vermis, these findings identified a role for CHD7 in early embryonic cerebellar development(13, 14). Towards the end of embryogenesis, the cerebellum initiates a period of rapid growth, primarily driven by the proliferation of GCps in the external germinal layer (EGL) on the surface of the cerebellar anlage. The primary mitogen driving GCp proliferation is Sonic Hedgehog (SHH), which is produced by post-migratory Purkinje cells (PCs) that become organized in a layer beneath the EGL during late embryonic stages (15-17). This process is associated with the formation of cerebellar folia. Intriguingly, we have also detectedcerebellar foliation anomalies in CHARGE syndrome patients(11), implying a role for CHD7 in GCp development.
Several genes associated with cerebellar hypoplasia, developmental delay, ataxia and intellectual disability in humans, are expressed in cerebellar GCps(reviewed in 18). Homozygous mutations in RELN, which encodes the secreted glycoprotein Reelin, or the gene encoding its receptor VLDLR, are responsible for severe cerebellar hypoplasia and intellectual disability in humans (19, 20). Studies in mouse models have localized Reln expression to GCps and have reported essential roles for Reelin signaling in PC migration, maturation and GCp proliferation (21-23). RELN has also been associated with psychiatric disease in several contexts. For example, Reelin expression has been reported to be reduced in postnatal cerebella from autism patients (24, 25).
In this manuscript,we report a role for CHD7 in controlling GCp proliferation and survival,and a striking downregulation ofRelngene expression in Chd7-deficient GCps. Increasing Reln expression partially rescued GCp proliferation and cerebellar hypoplasia, providing functional genetic evidence that Reln downregulation contributes to the cerebellar hypoplasia phenotype. Finally, we show that DNA accessibility is reduced at the Reln locus, and throughout the genomein CHD7-deficient GCps, consistent with a role for CHD7 in maintaining DNA accessibility through nucleosome remodeling.
RESULTS
Chd7 deletion from neuronal progenitors results in cerebellar hypoplasia
Chd7 is expressed in neural stem cells, where it has been reported to regulate the expression of developmental and disease-associated genes (7). Chd7 expression localizes to regions of ongoing neurogenesis in the developing brain: the ventricular zone (vz), hippocampus (Fig. 1A,C), as well as the rhombic lip stream (RLS), the ventricular zone (vz) and external granule cell layer (EGL) of the cerebellum (Suppl. Fig. 1A-E). Chd7 expression is maintained in post-mitotic, differentiating and migrating neural progenitors in the rostral migratory stream (RMS) of the forebrain, and upregulated in the differentiating inner EGL (iEGL) and internal granule cell layer of the cerebellum (Suppl. Fig. 1E; Fig. 1A,C), implicatingChd7 in neural differentiation. Severalrecent studies have identified roles for Chd7 in adult neural stem cell populations in the forebrain (26-28); yetChd7 appears to have a relatively minor role in regulating neural stem cell expansion in the embryonic and perinatal forebrain (26).To determine whether Chd7 has a function in embryonic neurogenesis, we inactivated Chd7 in neuronal progenitors using a Nestin-cre line (29). In agreement with previous demonstrations of robust gene recombination in the cerebellar anlage by E11 (14), Chd7 mRNA was not detectable in the Nestin-cre;Chd7f/f conditional knockout (cko) cerebellum at E12.5 (Suppl. Fig. 1F,G) and CHD7 protein, present in cell nuclei in the VZ and migratory cells emerging from the rhombic lip in control embryos, was absent from these cells (Suppl. Fig. 1H,I). The expression of Fgf8 from the mid-hindbrain organizer was maintained and the expression patterns of the Otx2 and Gbx2 homeobox genes were not altered in these mutants (Suppl. Fig. 1J-O), indicating that the deletion of Chd7 after the establishment of this embryonic organizer, had little effect on the expression of genes linked to organizer function (11).
Newborn ckoanimals lacked Chd7 expression throughout the entirebrain(Fig. 1B,D). These micewere born at the expected Mendelian ratios but postnatal survival rate was reduced (Suppl. Table 1). Two mutants survived to P21 and thesewere smaller than their littermates, overall brain size was reduced accordingly, but cerebellar size was disproportionally reduced (Compare cerebella outlined in Fig. 1E withF, and G with H). Histological examination confirmed pronounced hypoplasia of all cerebellar lobules in the vermis(Fig. 1I,J). The cerebellar hemispheres were also hypoplastic and displayedhighly irregular cerebellar foliationwith apparently mis-directed folia (Fig. 1K,L) and ectopic clusters of granule cellsaround clusters of PCs (Fig. 1N). Following specific folia in serial sections (Suppl. Fig. 2) revealed that vermis folia IV-V and IX extended abnormally into the lateral hemispheres (Suppl. Fig. 2I,J).
To identify the developmental stage at which these phenotypes first emerge, sagittal sections from E12.5 to P14 cerebella were examined. Cerebellar size was normal at E12.5 (Suppl. Fig. 1F-I) and E14.5 (Fig. 1O,P). Clear cerebellar growth retardation became evident at E16.5 (Fig. 1Q,R),a time point shortly after the initiation of GCp proliferation.At birth (P0), cerebellar hypoplasia was most clearly present in the cerebellar vermis, where it was associated with a failure to initiate the formation ofmost of the cardinal cerebellar fissures (Fig. 1S,T). By P14, the cerebellum was reduced in size, with irregular foliation that included expansion of vermis lobules IV-V and IX into the hemispheres (Fig. 1U-X). Taken together, these experiments identified a critical role for Chd7 in cerebellar growth after E14.5, implying a functionfor CHD7 in the expansion of cerebellar GCps.
CHD7 regulates the proliferation, differentiation and survival of cerebellar GCps
To specifically determine the function of Chd7 within the granule cell lineage, we deleted Chd7 from these cells from the time of their specification in the rhombic lip using a Math1-cre transgene (30). Transgenic expression resulted in the recombination of the Chd7 conditional allele from the early stages of EGLformation at E14.5 (Suppl. Fig. 3C,D). By postnatal stages, efficient Chd7 deletion was evident in GCps in the anterior and central cerebellar vermis, with Chd7 expression reduced somewhat in anterior lobule VIII and spared in the most posterior (posterior IX and X) lobules, in agreement with previous reports on the activity of this transgene (Suppl. Fig. 3E-H)(31). Efficient Chd7 deletion was evident throughout the entire cerebellar hemispheres (Suppl. Fig. 3I,J). Immunostaining showed the samepattern of CHD7 protein depletion in the cerebellar vermis (Suppl. Fig. 3K-R). Together, these experiments identified the granule cell lineage as the predominant cell type in the postnatal cerebellum that express CHD7, with only faint expression remaining in cells in the white matter after GC-specific Chd7 deletion(Suppl. Fig. 3H,J). Efficient Chd7 deletion in GCps was further confirmed by qRT-PCR analysis of purified GCps from P7 animals (Suppl. Fig. 3S).
TheGCp-specific conditional mutants, from hereon referred to as cko animals, survived to adulthood. Examination of wholemount P21 cerebella revealed significant cerebellar hypoplasia (Fig. 2A,B). Histological sections confirmed hypoplasia of lobules I-VIII (Fig. 2C,D), with the most striking hypoplasia presenting in central lobules. All lobules in the hemispheres were smaller and vermis folia IV-V and IX were again found to extend abnormally into the hemispheres(Suppl. Fig. 4 and Fig. 2E,F) with associated foliation irregularities (Fig. 2G,H). Cerebellar development was followed from E13.5 to reveal the first signs of reduced cerebellar growth. Surface area measurements to estimate cerebellar size were performedon Cresyl violet-stained sagittal sections through the most medial aspect of the cerebellar vermis. Delayed initiation of cerebellar foliation in the vermis became evident at E17.5 (Suppl. Fig. 5) before cerebellar size was significantly altered (Mean area +/- SEM (cn) = 0.26 mm2 ± 0.068, Mean (cko) +/- SEM = 0.26 mm2 ± 0.026). Hypoplasia became evident at E18.5 (Mean (cn) +/- SEM = 0.50mm2 ± 0.014, Mean (cko) +/- SEM = 0.30mm2 ± 0.026).To visualize the trajectory of cerebellar growth, representative sections (vermis and hemisphere) from control and cko cerebella were traced and overlaid (Fig. 2I). This analysis revealed interesting differences: whereas vermis growth was clearly reduced from E18.5 onwards, hypoplasia of the hemispheres only became obvious from P7 (Fig. 2I).
To determine the underlying cause of cerebellar hypoplasia, GCp proliferation was quantified by BrdUincorporation between E18.5 and P14. This analysis revealed a significant reduction in GCp proliferation in vermis lobules I-VIII at early postnatal stages (Suppl. Fig. 6A-B), most prominent at P0 (2J). No significant reduction in GCp proliferation was seen in hemispheres (Suppl. Fig. 6D). Theseresultsindicated that reduced GCpproliferation likely contributed to reduced growth of the vermis, but not the hemispheres. To assess the contribution of apoptosis, sections were stained with an antibody to activated cleaved caspase 3. The number of apoptotic GCpswasincreased in both vermis and hemispheres at P7 (Suppl. Fig 6E,F), but only reached statistical significance in the hemispheres (Fig. 2K). Thesedatasuggested that the reduced postnatal growth of the hemispheres was largely caused by increased cell death, whereas both proliferative and apoptotic changes contributed to vermis growth.
Reduced Purkinje cell numbers in Chd7 cko cerebella
To determine the impact of the diminished production of granule neurons in the cerebellum on PCs, PC number and distribution were examined. Total PC numbers were reduced in cko cerebella at P21 (Fig. 3A). A finer analysis of PC numbers across different cerebellar regions revealed that the reduction in PC number was due tofewer PCs in lobules I-VIIof ckoanimals (Fig. 3B), in agreement with the pronounced hypoplasia of anterior and central cerebellar lobules in the mutants (Fig. 2D). PC density was not altered, indicating that the reduction in PC numbers remainedproportional to the reduced cerebellar size (Fig. 3C-E). PCs in the cko cerebellar vermis were organized in monolayers for most of the cerebellum, with small regions of slightly disorganized cells observed at P7 (Fig. 3F-I) and P21 (Fig. 3N-Q). By contrast, large patches of disorganized PCs were present in the hemispheres at P7 (Fig. 3J-M) and P21 (Fig. 3R-U).
To understand the cause of these abnormalities, we investigated the distribution of Lhx1/5+ PC progenitors over time as they migrate from their site of origin, the vz towards the pial surface(23). This analysis revealed a relatively normal PC distribution, including formation of the PC plate under the pial surface at E16.5 (Suppl. Fig. 7A-H). Abnormal PC organization was evident by E18.5, with more dispersed PC progenitors in the central vermis (Suppl. Fig. 7Q) and apparently mis-localized progenitors in the hemispheres (Suppl. Fig. 7S,T), consistent with the altered PC distributions seen in these regions at later stages. As CHD7 is also expressed in cells in the vz, we asked whether Nestin-crecko mice where Chd7 has beendeleted from these progenitorsin addition to GCps,exhibited more pronouncedPCs developmental defects. Indeed, although a clear PC plate still formed in the medial cerebellum by E16.5 (Suppl. Fig. 7I), PC distribution was clearly irregular in the rest of thecerebellum (Suppl. Fig. 7J-L). Mis-localized cells were evident in the vermis and hemispheresby E18.5 (Suppl. Fig. 7U-X). Together, these findings suggested that the deletion of Chd7 from the early vz progenitors contributed to PC defects in Nestin-cko mice, although we cannot rule out the possibility that the deletion of Chd7 from the RLS and EGL earlier than inMath1-cre cko mutants also contributed to these defects.
GCp-specific Chd7 conditional mouse mutants exhibit motor delay and coordination deficits
Next, we asked whether the cerebellar defects were sufficient to cause behavioral abnormalities in these mice. Mutant mouse pups exhibited normal growth (Suppl. Fig. 8A). Pups were assessed for the acquisition of developmental milestones, which revealed a delay in acquiring the righting reflex (Fig. 4A), the ability to turn around when placed head facing downwards on a sloping platform (negative geotaxis, Fig. 4B) and the ability to reach out towards an object (Fig. 4C). To determine whether adult cko animals had any motor deficits, their motor coordination ona revolving rotarod was examined. Male cko mice performed significantly worse on this test than their sex-matched control littermates (Fig. 4D), whereas female cko mice showed no difference compared to controls (Fig. 4E).There were no significant differences in body weight or grip strength between control and cko animals (Suppl. Fig. 8B,C), excluding general muscle weakness as a potential cause of this motor phenotype. The performance of cko animals of both sexes improved to a similar extent over the three days of testing, indicating that these mutants had no deficit in motor learning.Mutant animals showed no signs of repetitive behaviors (Fig. 4F) or anxiety (Suppl. Fig. 8D,E).