Williams syndrome (WS) is a neurodevelopmental disorder caused by a genomic deletion of ∼28 genes that results in a cognitive and behavioral profile marked by overall intellectual impairment with relative strength in expressive language and hypersocial behavior. Advancements in protocols for neuron differentiation from induced pluripotent stem cells allowed us to elucidate the molecular circuitry underpinning the ontogeny of WS. In patient-derived stem cells and neurons, we determined the expression profile of the Williams–Beuren syndrome critical region-deleted genes and the genome-wide transcriptional consequences of the hemizygous genomic microdeletion at chromosome 7q11.23. Derived neurons displayed disease-relevant hallmarks and indicated novel aberrant pathways in WS neurons including over-activated Wnt signaling accompanying an incomplete neurogenic commitment. We show that haploinsufficiency of the ATP-dependent chromatin remodeler, BAZ1B, which is deleted in WS, significantly contributes to this differentiation defect. Chromatin-immunoprecipitation (ChIP-seq) revealed BAZ1B target gene functions are enriched for neurogenesis, neuron differentiation and disease-relevant phenotypes. BAZ1B haploinsufficiency caused widespread gene expression changes in neural progenitor cells, and together with BAZ1B ChIP-seq target genes, explained 42% of the transcriptional dysregulation in WS neurons. BAZ1B contributes to regulating the balance between neural precursor self-renewal and differentiation and the differentiation defect caused by BAZ1B haploinsufficiency can be rescued by mitigating over-active Wnt signaling in neural stem cells. Altogether, these results reveal a pivotal role for BAZ1B in neurodevelopment and implicate its haploinsufficiency as a likely contributor to the neurological phenotypes in WS.
Williams syndrome (WS [MIM 194050]) is a neurodevelopmental disorder genetically defined by a typical hemizygous 7q11.23 microdeletion of 1.55 million base pairs (Mb) that encompasses approximately 28 genes (1). Individuals with this deletion exhibit a specific physical, cognitive and behavioral profile characterized by hypersociality and a relative preservation of language function despite overall intellectual disability (ID). Although the phenotypes of WS have been described for half a century and the exact genotype known for nearly two decades, the molecular circuitry linking specific genetic changes with the neurological features remains largely undetermined (2). Hemizygosity of elastin (ELN), located in the Williams–Beuren critical region deletion (WSΔ), causing supravalvular aortic stenosis (SVAS [MIM 185500]) is the only firmly established genotype–phenotype correlation in WS (3). Despite important efforts to build a cognitive map from the genome architecture of patients with atypical patient deletions (4) and the use of animal models to dissect the genetic networks underlying the phenotypes, these systems are faced with limitations that hamper discovery of clear-cut phenotype–genotype correlations (5) and the failure of mouse models to capture certain aspects of human physiology (6).
The technical breakthrough of reprogramming mature cells into self-renewing, induced pluripotent stem cells (iPSCs) enabled the generation of disease-specific human iPSC lines (7–9). WS iPSCs have been recently used to model the vascular phenotype and profile the transcriptional landscape of WS derivatives (10, 11); however, neurons were not investigated in these studies. Building upon these studies, we have generated iPSCs from fibroblasts of two individuals carrying the typical WS 7q11.23 microdeletion and directly induced these cells into neurons. Genome-wide transcriptional profiling allowed us to measure the expression of WSΔ genes and assess the transcriptional consequences of the microdeletion specifically in neurons. Thus, the goal of these studies is to link the WS microdeletion to a transcriptional profile.
During neural induction of WS iPSCs, we identified a differentiation defect characterized by aberrant cell cycle and activated Wnt signaling. We investigated the contributions of BAZ1B ((bromodomain adjacent to zinc finger domain, 1B), also known as Williams syndrome transcription factor, WSTF), an evolutionarily conserved protein tyrosine kinase and member of the ISWI ATP-dependent chromatin remodeler complex subfamily encoded in the WSΔ and expressed throughout neurodevelopment, to these defects. Genome-wide identification of BAZ1B binding sites implicated this gene in neurodevelopment and several WS phenotypes. Knocking down BAZ1B to haploinsufficient levels in neural stem cells recapitulated the transcriptional dysregulation we observed in WS and caused a neural differentiation defect. This defect was reversed by antagonizing Wnt signaling, thus implicating this gene and pathway in WS pathology. WS iPSCs serve as a powerful platform to study the neurodevelopmental pathology of the disease and to gain insights into the relationship between gene deletions and uniquely human neurocognitive functions.
WSΔ gene expression and transcriptional rewiring in WS iPS cells and induced neurons
We generated and characterized iPSCs from two individuals carrying the typical WS 7q11.23 microdeletion and two euploid controls. We derived one of the control lines from an individual with Angelman syndrome (AS [MIM 105830]), a neurological disorder characterized by autistic features, ID and the absence of speech that served as a control with another form developmental cognitive dysfunction (12). In addition to euploidy at the WSΔ locus, the controls were matched for age, gender and ethnicity to the WS individuals. We induced neurons directly from the iPSC lines through viral transduction of the neurogenic transcription factor, NEUROD1, allowing us to consistently and reproducibly obtain highly pure human-induced neurons (iNs) suitable for gene expression profiling (13). Gene expression profiles obtained from this neuron induction protocol were highly reproducible (r2 ≥ 0.98) for all pairs of biological replicates (Supplementary Material, Fig. S1A and S1B). Hierarchical clustering of gene expression profiles by genes with high variation or the expression profile of the WS locus showed that the two WS lines behaved similarly, as did the two control lines (Supplementary Material, Fig. S1C and S1D).
To determine the local and global consequences of the WS microdeletion on gene expression, we performed transcriptional profiling of WS cells and euploid controls of iPS cells and iNs (Fig. 1A). WS iPSCs exhibited typical embryonic stem cell morphology and stained positively for typical pluripotent stem cell markers including OCT4, SOX2 and NANOG (Fig. 1B and Supplementary Material, Fig. S2) and were validated for other stem cell characteristics (Supplementary Material, Fig. S3). Six of the WS deleted genes were significantly downregulated in WS iPSCs (Fig. 1D), despite few overall transcriptional differences between the different iPS cell lines (Fig. 1E and Supplementary Material, Table S1). Direct induction of neurons yielded highly pure populations of neurons staining positively for MAP2 and TUJ1 (Fig. 1C). High purity neuronal induction was equivalent between all lines, and no gross morphological differences distinguished WS neurons from controls (Supplementary Material, Fig. S4A,B), although further characterization of these neurons may reveal differences. When plated on a multielectrode array, neurons induced by this method were capable of firing action potentials that were blocked in the presence of tetrodotoxin (Supplementary Material, Fig. S4C,D). In contrast to the iPSCs, neurons induced from WS lines showed widespread transcriptional dysregulation compared with controls with 2272 differentially expressed probes corresponding to 778 genes downregulated in WS and 1008 upregulated genes (Fig. 1F and Supplementary Material, Table S2). Fourteen of the WS-deleted genes were expressed in neurons (log2 normalized expression > 8), and all of these were statistically downregulated to about half of control levels (average fold decrease = −1.8 ± 0.1; Fig. 1D). The expression profile of the WSΔ genes is therefore consistent with a lack of dosage compensation for the WS microdeletion, in line with findings reported in other human microdeletions (14).
Enrichment analysis of genes downregulated in WS neurons revealed widespread transcriptional dysregulation that included markedly downregulated genes implicated in cognition, synaptic transmission and ID (CACNA1C, GABRG2, GRIN3A, NLGN3; Fig. 1G and Supplementary Material, Table S3). Also among these downregulated terms was an enriched interaction network centered on the WSΔ gene STX1A (Supplementary Material, Fig. S5). All of the genes presented in this figure were significantly downregulated in the WS neurons. STX1A encodes syntaxin-1a, a protein highly abundant in the synaptic bouton (15) whose haploinsufficiency may contribute to the WS phenotypes (16), but in itself is not a satisfactory explanation for the complex cognitive phenotype. Based on many transcriptional profiling experiments in neurological diseases, downregulated synaptic transcripts are expected and may be a general feature of neurological conditions that affect cognition. On the other hand, the enriched upregulated pathways in WS neurons suggested novel upstream pathogenic mechanisms. The upregulated terms included cell cycle terms (cell cycle, mitosis) and Wnt-activated receptor genes, and suggested an underlying basis for a neurodevelopmental facet of the pathology that precedes synaptogenesis and may more directly underlie defects in neural circuitry.
Interestingly, we found that canonical neural stem cell markers (ASCL1, HES5, PAX6, SOX2) were upregulated in WS iNs, suggesting the transcriptional state of WS neurons was more immature compared with controls. Neuroepigenetic chromatin remodelers were also dysregulated in WS neurons, with three neuron-specific chromatin remodeling subunits downregulated in WS iNs and a neural progenitor-specific chromatin remodeler upregulated in these cells (17) (Supplementary Material, Fig. S6). Together this suggested a possible neuroepigenetic constriction point in the path to post-mitotic commitment of WS-induced neurons. Consistent with this observation, a recent mouse model recapitulating key features of WS revealed an increased proportion of immature neurons, further supporting this as a relevant cellular phenotype of WS (18). Moreover, as this neuroepigenetic switch is essential for proper synaptic development and neuronal plasticity (19, 20), an incomplete neurogenic commitment of WS iNs is a strong candidate event contributing to the widespread transcriptional deficits of synaptic genes we observed.
Prioritization of BAZ1B as a candidate contributor to WS transcriptional dysregulation
After finding all of the expressed WSΔ gene levels at roughly half the control levels in neurons, we next sought to identify which of these may contribute to the observed widespread transcriptional dysregulation and neurogenic defect in WS iNs. Six genes within the WSΔ locus are predicted to function in transcriptional regulation or epigenetics (21), and three of these are predicted to be completely intolerant to loss of function mutations in healthy humans (22): GTF2I, GTF2IRD1 and BAZ1B. Extensive studies have linked the general transcription factors (GTFs) to neurocognitive aspects of WS (23–29), but the contributions of BAZ1B are not well studied.
BAZ1B is a component of the ISWI chromatin remodeling complex (30), which predominantly labels proliferating cell populations during early mouse brain development (31), and helps to establish the gene expression programs controlling neural maturation (32). BAZ1B regulates gene expression by modulating chromatin and BAZ1B knockout clones of human cells have large numbers of genes both up- and downregulated (33). In an expression dataset of cortical neurons derived from human pluripotent stem cells, BAZ1B levels are associated with cortical specification, and genes with similar co-expression patterns were enriched for cell-cycle roles (34). Thus BAZ1B is a compelling candidate contributor to the neuroepigenetic phenotype of WS neurons.
Genome-wide identification of BAZ1B binding sites implicates BAZ1B in neurogenesis and WS phenotypes
We used a human neural progenitor cell line as a platform to ascertain the role of BAZ1B in neural stem cell differentiation. Given its DNA and chromatin binding capabilities, we first sought to identify the genomic binding targets of BAZ1B in neural stem cell self-renewal and in neuronal differentiation. To this end, we performed chromatin-immunoprecipitation followed by next-generation sequencing (ChIP-seq) in these two conditions. In self-renewal conditions, we identified 1211 BAZ1B peaks and mapped these to the nearest transcription start sites to identify 1068 genes putatively regulated by BAZ1B (Fig. 2A and Supplementary Material, Table S4). Enriched gene ontology (GO) terms of BAZ1B target genes were related to neurogenesis, implicating BAZ1B as a regulator of neurodevelopment (Fig. 2B). BAZ1B target genes in self-renewal conditions were enriched for 161 target genes involved in nervous system development and 47 synaptic genes (Fig. 2B left and Supplementary Material, Table S5). Wnt signaling was enriched among BAZ1B target genes.
ChIP-seq in differentiated neurons further supported a role for BAZ1B in neurodevelopment. We identified 3043 BAZ1B peaks mapping to the nearest transcription start sites of 2339 genes in differentiated neurons (Fig. 2A and Supplementary Material, Table S6). This 2-fold expansion of target genes in differentiation over the precursor state suggests an expanding role for BAZ1B in neuronal differentiation. Important pathways in neuron differentiation, such as axon guidance and the formation of neuronal projections, were enriched among these targets (Fig. 2B right). Compared with targets in self-renewal conditions, BAZ1B target genes in differentiation conditions were further enriched for synaptic genes and less enriched in genes in nervous system development. This observation suggests a progressive shift in BAZ1B function from a role in neural progenitor cell cycle regulation to neuronal differentiation and synaptic gene activation. Of the 390 overlapping genes between precursor cells and differentiated cells, 59 genes were involved in nervous system development and 20 genes were synaptic. Taken together, BAZ1B appears to coordinate a set of neurogenic genes with cell cycle genes to control the final population of terminally differentiated neurons.
Disease enrichment of BAZ1B-regulated genes predicted several characteristic WS phenotypes, including anxiety and attention deficit hyperactivity disorder (ADHD [MIM 143465]) (Fig. 2C). BAZ1B target genes were also enriched for Simons Foundation Autism Research Initiative (SFARI) curated human genes (35). Of the 385 manually curated and scored genes in the human gene module of SFARI, 44 and 103 BAZ1B target genes in self-renewal and differentiation conditions respectively overlapped these autism-associated genes (P = 2.6 × 10−7, P = 1.9 × 10−18, hypergeometric test). The predicted functionality of BAZ1B target genes extended beyond the neural lineage to the musculoskeletal and cardiovascular phenotypes of WS. In particular, our results predict BAZ1B target genes also contribute to craniofacial dysmorphology in WS, consistent with mouse studies (36). Specific phenotypes regulated by BAZ1B were further predicted by the Human Phenotype Ontology (HPO) (Fig. 2D) (37). These predicted phenotypes, including short nose, small sella turcica and optic nerve dysplasia have been observed in WS (38,39). Strikingly, BAZ1B target genes delta-catenin (CTNND2) and KANSL1 are associated with the HPO term ‘conspicuously happy disposition’ implicating BAZ1B as a possible regulator of the WS behavioral profile. Altogether, BAZ1B target genes predict physical, behavioral and disease-relevant phenotypes, and allow us to link novel and specific genotype–phenotype correlations in WS.
BAZ1B knock-down to haploinsufficient levels recapitulates WS transcriptional dysregulation and a Wnt-dependent neurogenic commitment defect
Having shown the direct binding of the haploinsufficient WS gene, BAZ1B, to neurogenic and neuron differentiation genes, we next wanted to determine the transcriptional consequences of BAZ1B haploinsufficiency on neural stem cell differentiation. We selected two independent short-hairpin RNAs (shRNA) targeting BAZ1B to haploinsufficient levels. These shRNAs reduced mRNA expression levels to 62 ± 1 and 57 ± 3% of control levels (Fig. 3A, P < 0.001), resembling the levels we observed in WS iNs. We infected neural progenitor cells with either BAZ1B shRNAs or scrambled shRNA, and grew these cells under differentiation conditions (Fig. 3B). Transcriptional profiling of these cells revealed widespread expression changes, in agreement with previous BAZ1B knock-down studies (33). Compared with scrambled shRNA differentiated neurons, thousands of genes were affected by BAZ1B haploinsufficiency, including 1697 unique genes upregulated and 1199 downregulated (Supplementary Material, Fig. S7 and Table S7). Hierarchical clustering of these cells by the most variably expressed genes clustered BAZ1B knock-down cells as an outgroup to neural progenitors cells and differentiated neurons, demonstrating widespread transcriptional consequences attributable to dosage reduction of BAZ1B (Fig. 3C). As anticipated, the set of genes perturbed by BAZ1B haploinsufficiency was enriched for direct target genes identified in our ChIP-seq analysis. Remarkably, GO analysis of the transcriptional consequences of dosage reduction of BAZ1B to haploinsufficient levels recapitulated two transcriptional hallmarks we observed in WS neurons, namely upregulated mitotic genes and downregulated genes enriched for roles in nervous system development and disease (Fig. 3D). This concordance of GO terms was driven by a strong overlap of dysregulated genes in both WS neurons and BAZ1B knock-down neurons. In fact, direct and indirect BAZ1B targets identified by ChIP-seq and knock-down explained a combined 42% of all transcriptional dysregulation in WS iNs (P = 3.4 × 10−79, hypergeometric test; Fig. 3E). Included in this overlap were an enrichment of Wnt-activated receptor genes and canonical neural stem cell markers (Supplementary Material, Table S8), consistent with a role for BAZ1B in regulating these pathways.
Brightfield microscopy showed that both shRNAs targeting BAZ1B dramatically blocked the differentiation of neural progenitors, whereas scrambled control shRNA had no effect (Supplementary Material, Fig. S8). This profound delay of differentiation was confirmed by staining for proliferation marker Ki67 (Fig. 4A), showing a significantly elevated proportion of BAZ1B haploinsufficient neural progenitors continued to proliferate. Quantifying these results, 43.8 ± 2% of BAZ1B shRNA-treated cells remained proliferative even after 6 days of differentiation conditions compared with <1% of cells treated with a scrambled control (P = 1 × 10−05). Expression of MAP2 and other neuronal marker genes was also blocked in these cells (Fig. 4B and Supplementary Material, Fig. S8B). Cell cycle analysis of these cells in differentiation conditions showed a 4-fold increase of cells in S phase (P < 0.001) and a 2.25-fold increase in G2/M cells (P < 5 × 10−5) upon BAZ1B knock-down, indicative of an increased proliferative capacity in the population versus differentiation (Fig. 4C). These results implicate haploinsufficiency of BAZ1B as altering the balance between progenitor self-renewal and neuronal differentiation.
Given the well-established role of Wnt signaling in regulating neurogenesis (40, 41), and the enrichment of Wnt signaling terms in both WS neuron upregulated genes and BAZ1B ChIP-seq target genes, we asked whether we could rescue the differentiation defect induced by BAZ1B haploinsufficiency by inhibiting this pathway. We antagonized Wnt/β-catenin signaling in BAZ1B knock-down neural progenitor cells during differentiation using the small molecule XAV939 which stimulates β-catenin degradation by stabilizing axin (42). Wnt/β-catenin pathway inhibition was able to partially rescue the differentiation defect induced by BAZ1B haploinsufficiency, visibly increasing neurite outgrowth in the cells (Fig. 4D) and partially restoring the expression of the neuron marker MAP2 (Fig. 4E). As shown by Ki67 staining, a dose-dependent decrease in proliferating cells could be achieved by increasing Wnt-antagonist concentration (Fig. 4F). A modest dose of XAV939 (0.5 μm) lowered the percentage of proliferative cells from around 28% of the total population to under 5% (P < 0.01). A higher dose (1 μm) further reduced the proliferative proportion to near wild-type levels. These effects were independent of apoptosis, which did not increase upon BAZ1B knock-down or treatment with XAV939 (Supplementary Material, Fig. S8C). Overall, these results confirm that the BAZ1B haploinsufficiency-mediated imbalance in neural progenitor cell cycle exit depends on Wnt/β-catenin signaling and can be rescued by inhibiting this pathway. An overactivation of this pathway, as observed in the WS iNs transcriptomes, may be a novel and targetable pathway in WS.
WS, with a well-defined deletion and characteristic neurocognitive and behavioral profile, offers one of the most powerful systems to connect specific gene deletions to the molecular mechanisms underlying human neurocognitive phenotypes. Patients with atypical deletions have been used to try to link specific genes to WS phenotypes (5, 43, 44), and in fact, partial deletions of just an ∼950 kb interval that spans ABHD11 on the centromeric side to GTF21 on the telomeric side have been reported (43, 45). However, these rare partial deletions cannot exclude a position effect on the neighboring chromosome region (46) or mutations not detectable with conventional fluorescence in situ hybridization (FISH) analysis.
WS iPSCs now provide an unprecedented vantage to understand the effects of the 7q11.23 microdeletion and serve as a novel model to identify disease-relevant cellular phenotypes. Recent work has examined the vascular phenotype and transcriptional dysregulation caused by this microdeletion in WS iPSC derivatives (10, 11), but neither study generated neurons, the most relevant cell type to investigate neurological phenotypes of WS. The latter study implicated a dosage imbalance in GTF2I, a GTF deleted at the WS locus. However, this gene was not studied in the context of WS neurons, is encoded in a highly redundant manner with many closely related genes and pseudogenes on chromosome 7, and only accounted for 10–20% of the widespread transcriptional dysregulation observed in the WS IPSCs (11). A lack of major transcriptional consequences for the duplication or deletion of Gtf2i was reported in mouse primary cortical neurons (47) suggesting that other WS-deleted genes may contribute to transcriptional dysregulation. Our patient-derived neurons enabled us to discover novel WS pathogenic molecular mechanisms connected to the neurological phenotypes of WS and exhibited disease-relevant hallmarks such as the downregulation of key genes involved in synaptic transmission and cognition. With all detected WSΔ genes expressed at only around half the control levels, a combination of these genes, including STX1A, likely contributes to the neurological aspects of WS. Upregulated pathways in WS neurons, including cell cycle terms and Wnt signaling, informed novel mechanistic insights into WS pathology. These pathways indicated an incomplete developmental transition from neural progenitors to differentiated neurons in WS, the same cellular phenotype recently observed in a mouse model of WS (18).
A limitation to our study is that we generated iPS cells from only two individuals with WS and two controls. However, the reproducibility between differentiations and the similarity of gene expression between the two samples suggest the transcriptional consequences of the WS microdeletion are profound and robustly detectible. While previous iPS cell-derived neurons of AS individuals have reported thousands of transcriptional differences in these cells (48), the use of AS iPSCs and neurons as controls in our experiment may enhance our ability to detect transcriptional changes specific to WS versus other neurodevelopmental disorders. Comparing the WS iNs with either AS or healthy control iNs, we found largely overlapping gene sets, including cell cycle, Wnt signaling and neural stem cell genes (Supplementary Material, Fig. S9), further supporting these pathways as transcriptional signatures of WS iNs. The full datasets generated in this work are openly available, allowing further exploration of the transcriptional dysregulation we observed in WS neurons.
We have identified BAZ1B, a component of the ISWI subfamily of ATP-dependent chromatin remodelers, as a likely critical contributor to the unique neurocognitive phenotypic features of WS. Systematic overexpression (49) or deletion experiments (50) have shown the vast majority of eukaryotic genes are not dosage sensitive, and gene regulatory networks are typically robust to the failure of random individual nodes (51). Yet, certain genes within the WSΔ locus at 7q11.23 are dosage sensitive, and their haploinsufficiency manifests as multiple systems-level phenotypes. Critical determinants of haploinsufficiency include high and developmentally regulated expression, evolutionary conservation, and high connectivity in protein–protein interaction networks (52). BAZ1B fits these criteria. Confirming the essentiality of BAZ1B to normal human development is a significant depletion of loss of function mutations in this gene in over 60 000 human exomes from the Exome Aggregation Consortium (ExAC) (22). Consistent with a pivotal role in neurodevelopment and WS, ChIP-seq identified thousands of genomic loci bound by BAZ1B proximal to genes enriched for functions in neurogenesis, neuron differentiation and specific WS phenotypes. Among these were several cell-cycle-related genes including cyclin D1 (CCND1) that can increase neurogenesis and reduce basal progenitor population when inhibited, resulting in a reduced surface area of the cerebral cortex (53). BAZ1B dosage reduction to haploinsufficient levels in neural progenitor cells elicited widespread transcriptional dysregulation and induced a differentiation defect resembling our observations in WS-derived neurons. Remarkably, when knock-down experiments brought BAZ1B to haploinsufficient levels in a neural progenitor cell line, the resultant dysregulation overlapped a plurality (42%) of all the transcriptional changes we observed in WS neurons. These results implicate BAZ1B as a novel regulator of neurogenesis and support an etiological role for BAZ1B haploinsufficiency as contributory to the WS neurocognitive phenotypes. The size of BAZ1B and its exquisite dosage sensitivity when deleted or duplicated precluded facile rescue experiments, which could definitively establish the causality of BAZ1B haploinsufficiency underlying the transcriptional changes we observed in WS neurons. Despite this omission, our knock-down experiment in neuronal progenitors and our ChIP-seq data independently support a pivotal role for BAZ1B in neurodevelopment and prioritize this gene for future studies.
The identification of dosage sensitivity of the chromatin-remodeling protein BAZ1B in widespread transcriptional dysregulation of neurodevelopmental pathways arrives at an emerging theme in the literature implicating chromatin remodeling genes underlying autism spectrum disorder (ASD) and ID (54, 55). Recent genome-wide association and exome-sequencing studies have identified causative mutations for a range of neurodevelopmental disorders in several ATP-dependent chromatin remodeling proteins (56–64). These complexes overlap at target sites to coordinate gene expression (65) and epigenetically orchestrate the complexity and dynamism of the neurodevelopmental transcription program characterized by the activation and precise spatio-temporal regulation of an estimated 86% of the total coding transcriptome (66). In a comprehensive functional genomics approach integrating ChIP-seq, knock-down experiments and transcriptomics, we have elucidated the effects of BAZ1B dosage on neurodevelopmental pathways. BAZ1B target genes are enriched for Wnt signaling genes and this pathway is activated in WS neurons. Wnt signaling regulates the balance between neuronal progenitor proliferation and differentiation, with loss-of-function and overexpression causing reciprocal effects of depleting or expanding the progenitor pool (67, 68). By antagonizing this pathway, we can rescue the switch defect caused by BAZ1B haploinsufficiency in neural progenitor differentiation. These results suggest that BAZ1B haploinsufficiency in neural progenitors is sufficient to tip the balance away from neuron differentiation toward prolonged self-renewal. Precise control of cell cycle exit in neural precursors regulates cortical histogenesis (67), and even slight alterations of this control result in profound downstream consequences on brain size, regionalization and micro-circuitry (69). An approximate 10–15% reduction in brain volume has been documented in WS (70), with regions of increased cortical thickness and complexity (71, 72). These findings support a Wnt-mediated cell-cycle exit defect caused by BAZ1B haploinsufficiency as a plausible mechanism contributing to the abnormalities in brain volume, thickness and complexity in WS. Notably, FZD9 and BCL7B, two additional genes deleted in WS, are contiguous with BAZ1B and also function in Wnt signaling (73, 74). Both these genes are expressed at roughly half the levels of control neurons in the WS iNs in this study and may contribute to the Wnt-mediated cell-cycle abnormalities we observed.
The genetic pathways identified are particularly insightful, because symmetrical copy number variations of 7q11.23 display symmetrically opposite phenotypes with regard to the social facets of brain function. Whereas WS is often called the opposite of autism, the WS-locus duplication (MIM 609757) is associated with ASD (75, 76), confirming that this region contains one or more genes whose dosage contributes to these specific cognitive phenotypes. A set of brain functions mediated by a genetically based brain circuit may bundle seemingly diverse behaviors, which in the case of 7q11.23-mediated social behavior include excessive friendliness, unreserved behavior around strangers, an engaging personality, striking verbal ability and an affinity for music. All these could be construed as facets of social behavior under the control of a genetic program that precisely times developmental progression of neural progenitors and their integration into mature neural circuits. We hypothesize that BAZ1B and the regulatory node identified here follow the corresponding symmetrically divergent pattern in which delayed cell cycle exit and concomitant effects on chromatin remodeling mediates the WS hypersocial features, and early cell cycle exit mediates the impaired social behaviors of ASD. This hypothesis dovetails with orthogonal findings that neurological disorders affecting language and intelligence exhibit premature differentiation of neural progenitor cells (77).
Materials and Methods
Generation and characterization of iPSCs
Skin biopsies were collected from two females diagnosed with WS at the Alexandru Obregia Clinical Hospital of Psychiatry in Bucharest, Romania, pursuant to approval by institutional review boards at Alexandru Obregia Clinical Hospital of Psychiatry and Victor Babes National Institute of Pathology. Both lines carried the typical 1.55 Mb hemizygous WS deletion, confirmed by FISH probes spanning the locus. A clinical description of these individuals is provided in Supplementary Material. Epidermal fibroblasts were reprogrammed to iPSCs using integrating retroviral ‘Yamanaka factors’ (7). Apparently, healthy, young female donor dermal fibroblasts were purchased from Cell Applications Inc. (San Diego, CA) and reprogrammed to iPSC as a control. Fibroblasts obtained from an age, gender and ethnically matched control with AS (MIM 105830) were also collected and reprogrammed.
All iPSCs exhibited typical embryonic stem cell morphology and stained positively for typical pluripotent stem cell markers (Supplementary Material, Fig. S2). Derived iPSCs expressed pluripotency genes and formed three germ layers during embryoid body differentiation (Supplementary Material, Fig. S3A and S3B). iPSCs stably retained normal (46XX) karyotypes and underwent demethylation of the OCT4 promoter upon reprogramming to pluripotency (Supplementary Material, Fig. S3C and S3D). G-banded karyotyping was performed by WiCell (WiCell Research Institute Inc., Madison, WI). iPSCs were grown in feeder-free conditions on Matrigel (BD Biosciences, San Jose, CA) in mTeSR1 media (StemCell Technologies, Inc., Vancouver) and manually passaged every 5–7 days.
Direct neural induction of iPSCs
We induced neurons directly from the iPSC lines through viral transduction of a master neurogenic transcription factor, NEUROD1 (13, 78). In brief, iPSCs were passaged as single cells with Accutase (StemCell Technologies), and 500 000 cells were plated in one well of a 6-well plate on Matrigel in the presence of ROCK inhibitor Y-27632 (StemCell Technologies). On the next day, cells were transduced with lentiviral particles encoding NEUROD1-ires-eGFP-PuroR driven by a tetracycline inducible promoter and doxycycline (1 µg/ml, Sigma) was added to the media. Two days later, cells were re-plated at a dilution of 1:6 onto Matrigel and puromycin (1 µg/ml) was added to the media to remove non-transduced cells. This was designated as day 0. Cytosine beta-d-arabinofuranoside (Ara-C, Sigma) was added at a low concentration for days 2–6 to remove any proliferative cells that escaped selection to ensure pure non-dividing neuronal cultures. Purity of this protocol exceeded 90% by Tuj1 staining and did not differ across lines (Supplementary Material, Fig. S4B). Induced neurons were maintained in N2/B27 media (Gibco-Life Technologies, Carlsbad, CA) without additional neurotrophic factors. Cells were cultured for 14 days before RNA extraction. Doxycycline was withdrawn from the media after day 7 to minimize the effect of transgene overexpression on neuronal transcriptomes at the time of analysis.
Transcriptome profiling and differential gene expression analysis
RNA samples were collected using the mirVana kit (Life Technologies), and RNA expression was measured with Illumina HumanHT-12 Expression BeadChips (Illumina, Inc., San Diego, CA) by expert technicians at the microarray core facility at Sanford–Burnham Medical Research Institute. Data were analyzed using the beadarray (79) and limma (80) packages in R (v2.15.2) (81). Expression data were quantile normalized and log2 transformed for analysis. All samples were profiled in biological triplicate, i.e. three independent replicates per line. Gene expression levels across biological replicates were highly reproducible, with r2 ≥ 0.96 in all cases (Supplementary Material, Fig. S1); therefore, gene expression levels for the replicates were averaged. Differentially expressed genes were determined by fitting linear models to the expression data and performing empirical Bayes testing. P-values were adjusted for multiple testing according to Benjamini–Hochberg method (82). All differentially expressed gene probes (P < 0.01) in iPSCs and neurons are provided in Supplementary Material, Table S1 and Table S2.
Gene ontology and pathway enrichment analysis
For GO analysis, gene lists for enrichment analysis were obtained from PharmGKB (83), KEGG pathways from the Encyclopedia of Genes and Genomes (84) and GO terms from AmiGO (85). P-values were calculated from a cumulative hypergeometric distribution, calculated using the phyper command in R and adjusted for multiple testing. The total population size was set to 21 462, representing the number of genes detectable using the Illumina HumanHT-12 Expression BeadChips. Additional GO analysis was performed with WebGestalt (86).
Human neural progenitor cell culture and differentiation
We used human neural progenitor cells which were derived from the ventral mesencephalon of human fetal brain (ReNcells) purchased from EMD Millipore (Billerica, MA, SCC008). ReNcells were maintained on Matrigel in ReNcell NSC maintenance medium (EMD Millipore) with 20 ng/ml human FGF2 (Peprotech, Rocky Hill, NY) and 20 ng/ml human EGF (Peprotech). For differentiation, ReNcells were seeded on Matrigel-coated plates and induced to differentiate by withdrawing growth factors in ReNcell NSC media. To antagonize Wnt/β-catenin signaling, XAV939 (Selleckchem, 1 µm) was added to differentiating ReNcells.
Cells were fixed with 4% paraformaldehyde, permeabilized with 0.25% Triton X-100 and then blocked with 10% fetal bovine serum (FBS). Samples were stained with primary antibodies Ki67 (Cell Signaling 9129), Tau (T46, Life Technologies 13-6400), TuJ1 (Covance MMS-435P), MAP2 (Millipore AB5622) overnight at 4°C. Secondary antibodies used include Alexa Fluor 555-goat anti-mouse IgG, and Alexa Fluor 488 or 555-goat anti-rabbit IgG (Invitrogen). Images were acquired using an Olympus IX71 fluorescence microscope with MetaMorph software (Molecular Devices).
RNA isolation and quantitative real-time PCR
Total RNA was extracted using the mirVana kit (Life Technologies) and reverse transcribed with the SuperScript III First-Strand Synthesis System (Life Technologies). Real-time quantitative PCR was performed using a QuantStudio 12K Real-Time PCR System (Life Technologies) with Power SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA). HPRT or GAPDH mRNA levels were used as a normalization control.
Lentiviral vectors (pLKO.1) containing shRNA targeting BAZ1B were purchased from Sigma (Sigma-Aldrich, St. Louis, MO). Lentiviral vector plasmids were transfected into HEK293T cells along with psPAX2 (Addgene #12260) and pMD2.G (Addgene #12259). After 48 h, supernatants were harvested, filtered through a 0.45 μm filter and concentrated by ultracentrifugation. Human neural progenitor cells were infected with virus, and stably integrated shRNA lines were selected using puromycin.
Cell cycle analysis
Cells were dissociated as single cells with Accutase (Life Technologies) and fixed with ethanol for 1 h on ice, followed by RNase treatment. Propidium iodide (Sigma, 10 μg/ml) was added to the samples and data were acquired by BD Accuri C6 flow cytometry (BD Biosciences). Cell cycle analyses were performed in biological triplicate.
Annexin V staining was performed with Dead Cell Apoptosis Kit (Life Technologies) according to the manufacturer's instruction. Briefly, differentiated ReNcells were washed with annexin-binding buffer and incubated with Alexa 488-anti-annexin V antibody at room temperature for 15 min. After additional washing with annexin-binding buffer, images were acquired using an Olympus IX71 fluorescence microscope with MetaMorph software (Molecular Devices).
Chromatin-immunoprecipitation and sequencing
ChIP-seq was performed in accordance with ENCODE guidelines (87). ChIP was performed using the EZ-ChIP kit (Millipore) according to manufacturer's instructions. Briefly, ReNcells grown under either self-renewal conditions or differentiation for 6 days were crosslinked with 1% paraformaldehyde for 10 min. Chromatin was isolated and sonicated using a Covaris S2 (Covaris, Woburn, MA). Sheared chromatin was incubated with anti-BAZ1B antibody (Abcam ab50850) overnight at 4°C followed by Protein G pulldown.
Ion Xpress Plus Fragment Libraries were prepared and sequenced on an Ion Proton PI chip on the Ion Proton System (Life Technologies, Carlsbad, CA) according to company supported protocols. DNA reads were aligned to the human genome (hg19) using bowtie2 (v 2.1.0) (88). ChIP-seq peaks were called using MACS (v 1.4.0) (89) and annotated using custom R scripts. BAZ1B pulldown was performed in duplicate, and replicates were combined for analysis. RNA-polymerase II (POL2) was included as a positive control. Input chromatin was used as a negative control.
Multi-electrode array recording
A glial bed was prepared from C57BL/6 mice. Briefly, cortices were dissected from postnatal day 1 mice, dissociated using papain and trituration, then plated at a density of 700 cells/mm2. Cells were incubated at 37°C with 5% CO2 and maintained in MEM supplemented with 5% heat-inactivated fetal calf serum and Mito + serum extender. At 7 days after plating, primary neurons were killed using a solution composed of 200 uM glutamate. NEUROD1-transduced cells were plated on glial bed 2 days after initiating doxycycline treatment and grown for 2 weeks. Extracellular action potentials were recorded using a Multi Channel Systems MEA 2100 system (MCS, Reutlingen, Germany) with a 120 count multi-electrode array with 100 μm interelectrode spacing and 30 um electrode diameter, at a sampling frequency of 20 kHz. 1 uM tetrodotoxin was added to verify biological origin of action potentials. All animal usage was performed following University of California, Santa Barbara Institutional Animal Care and Use Committee approved protocols.
Sequencing data have been deposited to GEO with accession number GSE71664.
ExAC Exome Aggregation Consortium Browser, http://exac.broadinstitute.org/.
Human Phenotype Ontology, http://www.human-phenotype-ontology.org/.
Above all, we gratefully acknowledge the fibroblast donors who enabled this study. We also acknowledge Dr Hyung-Seok Kim for the generous help with deriving some of the iPSC lines in this study. We gratefully acknowledge Dr Kang Liu and the Sanford Burnham Medical Research Institute's Microarray Core facility. We acknowledge the use of the Laboratory for Stem Cell Biology and Engineering at UC Santa Barbara, which is supported by the California Institute for Regenerative Medicine (CIRM) Grant CL1-00521-1. J.J. was supported by CIRM fellowship TG2-01151. We gratefully acknowledge support from the Dr Miriam and Sheldon Adelson Medical Research Foundation.
Conflict of Interest statement: The authors declare no conflicts of interest.
- williams syndrome
- signal transduction
- gene expression
- transcription, genetic
- stem cells
- stem cells, pluripotent
- nervous system development
- neural development
- candidate disease gene
- neural stem cell
- neuron differentiation