Abstract
Down syndrome (DS), caused by triplication of human chromosome 21 (HSA21), is the most common genetic origin of intellectual disability. Despite the limited success of current pharmacological interventions, little has been achieved to reverse the abnormal brain development in DS. Here, using human induced pluripotent stem cell (hiPSC)-based brain organoid and in vivo human neuronal chimeric mouse brain models, we demonstrate that the HSA21 genes OLIG1 and OLIG2 exhibit distinct temporal expression patterns during neuronal differentiation. The population of OLIG2-expressing ventral forebrain neural progenitors is overabundant in the cells derived from DS hiPSCs, which results in excessive production of calretinin- and somatostatin-expressing GABAergic interneurons in DS organoids and causes impaired recognition memory in DS chimeric mice. Furthermore, we find that overexpression of OLIG2 in DS alters the expression of GABAergic neuron lineage-determining transcription factors such as DLX1 and LHX8. We further show that OLIG2 can directly bind to promoter regions of DLX1 and LHX8 to increase their expression, leading to lineage specification of interneurons. Importantly, knockdown of OLIG2 largely reverses the abnormal global gene expression profile of early stage DS neural progenitors, reduces inhibitory neuronal population in DS organoids and chimeric mouse brains, and improves behavioral performance of DS chimeric mice. Therefore, OLIG2 is a potential target for developing personalized prenatal therapeutics for intellectual disability in subjects with DS.
Introduction
Down syndrome (DS), resulting from an extra copy of human chromosome 21 (HSA21), is the leading genetic cause of intellectual disability and affects one in every 700-800 live births (Parker et al., 2010). No therapies currently exist for the rescue of neurocognitive impairment in DS, mainly due to the lack of knowledge on the underlying molecular mechanisms (Letourneau and Antonarakis, 2012; Sturgeon and Gardiner, 2011). Previous studies have suggested that an imbalance in excitatory and inhibitory neurotransmission is one of the underlying causes of cognitive deficit of DS (Fernandez et al., 2007; Haydar and Reeves, 2012; Rissman and Mobley, 2011). The inhibitory GABAergic interneurons in the cerebral cortex are derived from the neuroepithelium of the embryonic ventral forebrain (Butt et al., 2005; Kessaris et al., 2006; Marin, 2012; Wonders et al., 2008). Many of these neuroepithelial cells express the HSA21 genes OLIG1 and OLIG2, members of the basic helix-loop-helix family of transcription factors. In mice, both Olig1 and Olig2 are expressed in the embryonic neuroepithelium of the ventral forebrain (Lu et al., 2000; Petryniak et al., 2007). In humans, OLIG2, but not OLIG1, is abundantly expressed in the embryonic ventral forebrain (Jakovcevski and Zecevic, 2005a), as opposed to their overlapping expression pattern in mouse embryonic brain. Thus, establishing the role of human OLIG genes, in regulating interneuron production is critical for understanding the mechanisms underlying cognitive deficit in DS and may help devise novel therapeutic strategies.
It is highly debatable how the production of GABAergic neurons is altered in DS and how OLIG genes are involved as regulators of GABAergic neuron production under normal and DS disease conditions. First, using mouse models, studies examining the functions of Olig genes in GABAergic neuron production remain inconclusive. Loss-of-function studies showed that only Olig1 repressed the production of GABAergic interneurons (Furusho et al., 2006; Ono et al., 2008; Petryniak et al., 2007; Silbereis et al., 2014). Gain-of-function studies showed that overexpression of Olig2 promoted the production of GABAergic neurons (Liu et al., 2015). However, this finding is confounded by the fact that the overexpression and mis-expression of Olig2 in inappropriate cells and developmental stages caused massive cell death in the mouse brain (Liu et al., 2015). Second, DS mouse models often show discrepancies in modeling DS-related genotype-phenotype relationships. Genotypically, Olig genes are only triplicated in a subset of DS mouse models, because different DS mouse models are trisomic for different human chromosome 21 orthologs. Phenotypically, Olig genes are not always overexpressed even if they are triplicated in the genome, due to different epigenetic modifications and chromatin states in mice (Aziz et al., 2018; Belichenko et al., 2015; Goodliffe et al., 2016). The discrepant findings in genotype and phenotypic expression of olig genes, and changes in the number of GABAergic neurons from different DS mouse models are summarized in supplementary table 1. Third, while studies in the Ts65Dn mouse model of DS indicated that GABAergic neurons were overproduced (Chakrabarti et al., 2010) and inhibiting the GABAergic transmission could alleviate cognitive deficit (Fernandez et al., 2007), studies using postmodern brain tissues from elderly DS patients (Kobayashi et al., 1990; Ross et al., 1984) and 2-dimensional (2D) cultures of DS human induced pluripotent stem cells (hiPSCs) (Huo et al., 2018) contradictorily showed reduced production of GABAergic neurons.
The lack of availability of functional human brain tissue from normal or DS patients is preventive for a detailed mechanistic understanding of the pathophysiology of DS. The well-accepted non-invasive approach for prenatal DS screening is usually performed around the end of the first trimester (Malone et al., 2005). Postmortem prenatal DS human brain tissues are scarcely available and are mostly in fetal stages ranging from gestation week (GW) 13-21. Analysis of these tissues yields a limited understanding of early prenatal and embryonic DS brain development (Bhattacharyya et al., 2009; Busciglio et al., 2002; Busciglio and Yankner, 1995; Esposito et al., 2008; Lu et al., 2012; Mao et al., 2003; Olmos-Serrano et al., 2016). Recent studies have demonstrated the utility of hiPSCs derived from individuals with DS as a human cellular model of DS brain development (Briggs et al., 2013; Chen et al., 2014; Jiang et al., 2013b; Shi et al., 2012; Weick et al., 2013). More importantly, as opposed to a 2D culture system, the hiPSC-based 3D brain organoid model is likely more appropriate, yielding more reliable results that recapitulate actual brain development in humans (Brawner et al., 2017; Centeno et al., 2018; Pasca, 2018; Simao et al., 2018).
In this study, we employed brain organoid and in vivo chimeric mouse brain models (Chen et al., 2016) to investigate the functions of OLIG genes in human interneuron development and pathogenesis. Using ventral forebrain organoids generated from OLIG2-GFP human pluripotent stem cell (hPSC) reporter lines, we demonstrated that human OLIG2+/GFP+ ventral forebrain NPCs generated various subclasses of GABAergic neurons and cholinergic neurons. We found that OLIG2 was overexpressed in DS hiPSC-derived organoids, compared with control hiPSC-derived organoids. Moreover, subclass-specific GABAergic neurons were overproduced in DS brain organoids, as well as in DS chimeric mouse brains. Significantly, DS chimeric mice exhibited impaired recognition memory. Importantly, we showed that inhibiting OLIG2 expression by shRNA normalized the overproduction of GABAergic neurons in both organoids and chimeric mouse brains and improved behavioral performance of the DS chimeric mice. Mechanistically, overexpression of OLIG2 resulted in a disrupted gene expression profile in DS organoids. Through direct interactions with GABAergic neuron lineage-determining transcription factors, OLIG2 promoted the production of subclass-specific GABAergic neurons. Development of prenatal therapy for DS has been put forward as a part of ‘fetal personalized medicine’ strategy (Bianchi, 2012; de Wert et al., 2017; Guedj et al., 2014). Our findings suggest OLIG2 as an excellent potential target for the personalized prenatal therapy for DS.
Results
Human PSC-derived OLIG2+ ventral forebrain NPCs give rise to GABAergic neurons
To test the hypothesis that human OLIG2 is involved in interneuron development, we utilized OLIG2-GFP hPSC (human embryonic stem cell, hESC and hiPSC) reporter lines generated in our previous studies (Liu et al., 2011; Xue et al., 2009). Neural differentiation of the hPSCs was induced by dual inhibition of SMAD signaling (Chambers et al., 2009). At week 2, neural rosettes were manually isolated from surrounding cells and expanded as primitive neural progenitor cells (pNPCs), as described in our recent study (Chen et al., 2016). At week 3, we further cultured these pNPCs using an established brain organoid culture protocol (Monzel et al., 2017; Pasca et al., 2015) (Fig. 1A). To obtain ventralized brain organoids, we treated the organoids with Sonic Hedgehog (SHH) and purmorphamine (Pur), a small-molecule agonist of SHH signaling pathway (Liu et al., 2013a; Liu et al., 2013b; Maroof et al., 2013; Nicholas et al., 2013), from week 3 to 5. After the treatment of SHH and Pur, many cells in the organoids started to show GFP fluorescence (Fig. 1A and supplementary Fig. 1A). At week 5, most of the cells in the organoids expressed the NPC marker Nestin (97.0 ± 1.4%), forebrain marker FOXG1 (96.6 ± 2.6%), as well as NKX2.1 (80.0 ± 5.6%), a marker of ventral prosencephalic progenitors (Sussel et al., 1999; Xu et al., 2004), but not dorsal brain marker PAX6 (Ma et al., 2012) (n = 4, Fig. 1B, C), indicating that the vast majority of the cells were ventral forebrain NPCs. In these organoids, we observed distinct rosette-like structures that resembled the proliferative regions of human ventricular zone containing SOX2+ progenitors, and β-III-Tubulin (βIIIT+) immature neurons (supplementary Fig. 1B). Consistent with our previous studies (Jiang et al., 2013b; Liu et al., 2011), the GFP fluorescence faithfully mirrored the expression of OLIG2 in the NPCs (Fig. 1B, C; about 38% of the total cells expressed both GFP and OLIG2). Interestingly, only a very small fraction of the NPCs expressed OLIG1 (1.1 ± 0.2%, Fig. 1B, C). OLIG1 expression did not overlap with OLIG2/GFP and appeared to be localized in the cytoplasm (Fig. 1B). This is consistent with a previous report that in human fetal brain tissue, although OLIG2 is expressed in the ventral forebrain throughout the ventricular zone of the ganglionic eminence, OLIG1 is expressed in very few cells (Jakovcevski and Zecevic, 2005a). These observations were confirmed by immunoblot analysis. OLIG2 was abundantly expressed and present mostly in the nuclear fraction, whereas OLIG1 was detected at a low level in the cytosolic fraction (Fig. 1D). Unexpectedly, qPCR results demonstrated that OLIG1 and OLIG2 mRNAs were both abundantly expressed but with different temporal profiles: during the patterning stage, the levels of both OLIG mRNAs were significantly increased from week 3 to 5. The discrepancy between OLIG1 transcript level and its protein level may suggest the involvement of precise spatiotemporal control of mRNA translation observed in cortical development (Kraushar et al., 2014; Kwan et al., 2012). After initiating neuronal differentiation, OLIG2 expression started to decrease over time and GFP fluorescence became dim at week 8 (Fig. 1A and Supplementary Fig. 1A). Nevertheless, at 8 weeks, some of the βIIIT+ neurons (25.8 ± 2.3%, n = 4), GABA+ neurons (24.8 ± 1.5%, n = 4), and S100β+ astrocytes (2.6 ± 0.5%, n = 4) were labeled by GFP staining, indicating these cells were likely derived from OLIG2+/GFP+ NPCs. However, we did not find any oligodendroglial lineage cells in these organoids cultured under the neuronal differentiation condition, which is consistent with previous reports (Lancaster and Knoblich, 2014; Pasca et al., 2015; Quadrato et al., 2016). The expression of both OLIG gene transcripts also significantly decreased at week 8 (Fig. 1E). Organoids had been known to lack oligodendrocytes (Pasca, 2018; Quadrato et al., 2016). It was not until recently that the induction of oligodendrogenesis and myelination was achieved and culture conditions appear to be critical for oligodendroglial differentiation in organoids (Madhavan et al., 2018). In a separate set of experiments, we cultured the 5-week-old organoids under glial differentiation condition and found MBP+ mature oligodendrocytes in the 8-week-old organoids (supplementary Fig. 1C), demonstrating that the OLIG2+/GFP+ NPCs in the organoids had the potential to differentiate to oligodendroglial lineage cells if supplied with the appropriate signals.
(A) A schematic procedure for deriving ventral forebrain organoids from OLIG2-GFP knockin hiPSCs or hESCs by the treatment of sonic hedgehog (SHH) and purmorphamine (Pur). Insets: representative bright-field and fluorescence images at different stages. Scale bars: 100 μm.
(B) Representatives of Nestin-, FOXG1-, NKX2.1-, PAX6-, OLIG2-, OLIG1- and GFP-expressing cells in 5 to 6-week-old ventral forebrain organoids. Scale bars: 20 μm or 10 μm in the original or enlarged images, respectively.
(C) Quantification of pooled data from OLIG2-GFP hiPSCs and hESCs showing the percentage of Nestin+, FOXG1+, NKX2.1+, PAX6+, GFP+, OLIG2+, OLIG1+, Nestin+/GFP+, FOXG1+/GFP+, NKX2.1+/ GFP+, PAX6+/GFP+, OLIG2+/GFP+, and OLIG1+/GFP+ cells in week 5-week-old organoids. The experiments are repeated for four times (n = 4) and for each experiment, 4 to 6 organoids from each cell line (total about 50 organoids) are used. Data are presented as mean ± s.e.m.
(D) 5-week-old ventral forebrain organoids were subjected to subcellular fractionation assay. Western blotting analysis shows that OLIG2 is localized in the nuclear fraction (nuc), whereas OLIG1 localized in the cytosolic fraction (cyto). β-tubulin is used as a cytosolic marker and Histone 3 is used as a nuclear fraction marker.
(E) qPCR analysis of OLIG2 and OLIG1 mRNA expression in the organoids at different stages. The experiments are repeated for four times (n = 4) and for each experiment, 20 to 30 organoids derived from OLIG2-GFP hESCs and hiPSCs are used. One-way ANOVA test. **P < 0.01, ***P < 0.001, and NS represents no significance. Data are presented as mean ± s.e.m.
(F) Representatives of βIIIT-, GABA-, S100β-, and GFP-expressing cells in 8-week-old ventral forebrain organoids. Scale bars: 20 μm or 10 μm in the original or enlarged images, respectively.
(G) Quantification of pooled data from OLIG2-GFP hiPSCs and hESCs showing the percentage of βIIIT+, GABA+, S100β+, βIIIT+/GFP+, GABA+/GFP+, and S100 β+/GFP+ cells in 8-week-old organoids (n = 4, and for each experiment, 4 to 6 organoids from each cell line are used). Data are presented as mean ± s.e.m.
(H) A schematic showing that 5-week-old ventral forebrain organoids derived from OLIG2-GFP hPSC-derived are disaggregated to single cells and then subjected to FACS. The purified GFP+ cells were further cultured under 3D conditions to form organoids.
(I) FACS analysis showing that in 5-week-old organoids, the percentage of GFP+ cells is much higher in the group treated with SHH and Pur than that in the control group treated with DMSO. The experiments are repeated for four times (n = 4) and for each experiment, 40 to 60 organoids derived from OLIG2-GFP hiPSCs and hESCs are used to collect GFP+ cells.
(J) Representatives of GABA-, calbindin (CB)-, calretinin (CR)-, parvalbumin (PV)-, somatostatin (SST), and neuropeptide Y (NPY)-expressing GABAergic neurons in 8-week-old GFP+ cell-derived organoids. Scale bars: 20 μm.
(K) Quantification of pooled data from OLIG2-GFP hiPSCs and hESCs showing the percentage of GABA+, CR+, CB+, SST+, PV+, and NPY+ neurons in 8-week-old GFP+ cell-derived organoids (n = 4 and for each experiment, 4 to 6 organoids from each cell line are used). Data are presented as mean ± s.e.m.
To further delineate neuronal fate of these OLIG2+ NPCs, we purified the GFP+ cells using fluorescence-activated cell sorting (FACS). At week 5, the SHH and Pur-treated organoids were disaggregated to single cells and subjected to FACS. About 40% of the total cells were collected as GFP+ cells (Fig. 1H, I). The purity of these cells was verified by immunostaining, showing that all of the sorted cells expressed GFP and OLIG2 (supplementary Fig. 2A). These cells also expressed Nestin, NKX2.1, and the mid/forebrain marker OTX2, confirming that these GFP+ cells were OLIG2-expressing ventral telencephalic NPCs (supplementary Fig. 2A). Then we cultured those purified OLIG2+/GFP+ cells in 3D to generate organoids. GFP fluorescence was strong at week 5 but became dim at week 8, indicating the decreased expression of OLIG2 (Supplementary Fig. 2B). We then assessed cell composition at week 8. We found NeuN+ (89.4 ± 3.7%, n = 4) and S100β+ (7.0 ± 0.4%, n = 4) cells in the organoids, but not NG2+ or PDGFRα+ cells (supplementary Fig. 2C, D), demonstrating that the majority of OLIG2+/GFP+ NPCs differentiated into neurons with a small fraction giving rise to astrocytes, but no differentiation to oligodendroglia under neuronal differentiation condition. Most strikingly, the majority of OLIG2+/GFP+ NPCs in the organoids efficiently differentiated into GABAergic neurons (57.3 ± 4.9%, n = 4; Fig. 1J, K). GABAergic interneurons are conventionally categorized based on neuropeptide and Ca2+ binding protein expression (Petilla Interneuron Nomenclature et al., 2008). We thus examined the subclass composition of the GABAergic neurons derived from OLIG2+ NPCs, by staining the major subclass markers calretinin (CR), calbindin (CB), parvalbumin (PV), somatostatin (SST), and neuropeptide Y (NPY). As shown in Fig. 1J, K, of these markers, CR was most highly expressed (24.5 ± 3.5%, n = 4), and CB (16.0 ± 1.9%, n = 4), SST (4.9 ± 0.3%, n = 4), and PV (5.2 ± 0.9%, n = 4), were robustly detected. A small percentage of NPY+ neurons (1.5 ± 0.4%, n = 4) was also detected. As opposed to 3D cultures, in the 2D cultures maintained for the same period of time, we found that the OLIG2+/GFP+ NPCs only differentiated into CR+ neurons and other subclasses of GABAergic neuron were not detected (supplementary Fig. 2E, F), suggesting that the 3D organoid models might be more effective in recapitulating a broader range of GABAergic neuron differentiation as seen in the developing brain. In addition, we interestingly found that a small fraction of OLIG2+/GFP+ NPCs (2.4 ± 0.7%, n = 4) were able to generate cholinergic neurons identified by the expression of choline acetyltransferase (ChAT) in the organoids (Fig. 1J, K). Taken together, these results show that the ventral forebrain organoid model in this study recapitulates the expression pattern of OLIG1 and 2 observed in human fetal brain tissue. Moreover, we provide direct evidence demonstrating that human OLIG2+/GFP+ NPCs give rise to different subclasses of GABAergic neurons and cholinergic neurons in the 3D cultures, in addition to having the potential to differentiate into oligodendrocytes, astrocytes, and motoneurons as reported in our previous studies (Jiang et al., 2013b; Liu et al., 2011).
Abnormal OLIG2 protein expression in DS hiPSC-derived ventral forebrain organoids
Since both OLIG1 and OLIG2 are encoded by HSA21, we thus hypothesized that they would be overexpressed in DS hiPSC-derived organoids because of the increased gene dosage in DS and that this would impact neuronal differentiation. Using the method established with OLIG2-GFP reporter lines (Fig. 1A), we derived ventral forebrain organoids from the three pairs of control and hiPSCs generated in our previous study (Chen et al., 2014). As shown in supplementary Table 2, the hiPSC lines derived from DS patients include two DS hiPSC lines (DS1, female; and DS2, male) and isogenic disomic (Di)-DS3 and trisomic (Tri)-DS3 hiPSCs that were generated from a single female patient. As reported in our previous study (Chen et al., 2014), the same age-matched hiPSC lines (control1 and control2 hiPSCs) generated from healthy individuals were used as controls. As shown in Fig. 2A, control and DS pNPCs stably exhibited disomy and trisomy of HAS21, respectively. In addition, at the DNA level, all DS pNPCs consistently had ~ 1.5-fold higher OLIG1 and OLIG2 gene copy numbers than the corresponding control pNPCs (supplementary Fig. 3A). After two weeks of SHH and Pur treatment (week 5), control and DS pNPCs similarly formed organoids with ventricular zone like regions containing Ki67+ and SOX2+ progenitors, and βIIIT+ immature neurons (Fig. 2B). At week 5, similar expression of brain identity markers was observed in control vs. DS organoids (Fig. 2C, D; 99.2 ± 0.2%, vs. 99.3 ± 0.2% for Nestin+ cells; 96.6 ± 1.4% vs. 95.4 ± 1.7% for FOXG1+ cells; 77.1 ± 4.3% vs. 74.9 ± 5.8% for NKX2.1+ cells; and 0.09 ± 0.2% vs. 0.15 ± 0.2% for PAX6+ cells), indicating the efficient patterning of DS and control pNPCs to ventral forebrain identify. The pooled data were generated from experiments repeated for four times (n = 4) using the three pairs of control and DS hiPSCs and for each experiment, 4 to 6 organoids from each cell line (total around 70 control and 70 DS organoids) were used.
(A) Fluorescence in situ hybridization (FISH) analysis confirms that three pairs of control (Cont1, Cont2, and Di-DS3) and DS (DS1, DS2, and Tri-DS3) pNPCs stably maintain disomy and trisomy of HAS21, respectively. Scale bar represents 10 μm.
(B) Representatives of Ki67, SOX2, and βIIIT immunostaining, showing ventricular zone (VZ)-like areas found in both control and DS hiPSC-derived 6-week-old organoids after SHH and Pur treatment. Scale bars: 100 μm.
(C) Representatives of Nestin-, NKX2.1-, FOXG1-, PAX6-, and OLIG1-expressing cells in control and DS iPSC-derived 5-week-old ventral forebrain organoids after SHH and Pur treatment. Scale bars: 20 μm.
(D) Quantification of pooled data showing the percentage of Nestin-, NKX2.1-, FOXG1-, PAX6-, and OLIG1-expressing cells from three pairs of control and DS hiPSCs derived 5-week-old ventral forebrain organoids. The experiments are repeated for four times (n = 4) and for each experiment, 4 to 6 organoids from each cell line (total around 70 control and 70 DS organoids) are used. Student’s t test. NS represents no significance. Data are presented as mean ± s.e.m.
(E) Representatives of OLIG2-expressing cells from three pairs of control (Cont1, Cont2, and Di-DS3) and DS (DS1, DS2, and Tri-DS3) iPSC-derived 5-week-old ventral forebrain organoids after SHH and Pur treatment. Scale bars: 20 μm.
(F) Quantification of the percentage of OLIG2+ cells in control or DS hiPSC-derived 5-week-old ventral forebrain organoids. The experiments are repeated for four times (n = 4) and for each experiment, 4 to 6 organoids from each cell line were used. Student’s t test. ***P < 0.001, comparison between DS and corresponding control organoids. Data are presented as mean ± s.e.m.
(G) qPCR analysis of OLIG2 and OLIG1 mRNA expression in 5 to 6-week-old control and DS organoids. The experiments are repeated for four times (n = 4) and for each experiment, 20 to 30 organoids respectively derived from control and DS hiPSC lines are used. Student’s t test, ***P < 0.001, comparison between DS and corresponding control organoids. Data are presented as mean ± s.e.m.
(H) Western blotting analysis of OLIG2 and OLIG1 expression in 5 to 6-week-old control and DS organoids.
(I) Quantification of pooled data showing OLIG1 and OLIG2 expression by western blot in the organoids. The experiments are repeated for three times (n = 3) and for each experiment, 30 to 40 organoids respectively derived from three pairs of control and DS hiPSCs are used. Student’s t test. **P < 0.01. Data are presented as mean ± s.e.m.
(J-L) Transcriptome correlation between 5-week-old organoids and post-mortem human brain samples from the BrainSpan project. The x axis shows the post-conception age in weeks or the brain region of the BrainSpan post-mortem brain samples. The y axis shows the mean spearman correlation. AMY, amygdala; CBC, cerebellar cortex; DFC, dorsolateral frontal cortex; HIP, hippocampus; ITC, inferolateral temporal cortex; MFC, medial prefrontal cortex; OFC, orbital frontal cortex; STC, superior temporal cortex; STR, striatum. In (H) x axis shows BrainSpan agglomerated brain regions (i.e., more brain regions were merged into a single ‘‘larger’’ region): VF (i.e., VF, CGE, LGE, MGE, STR), ventral forebrain; NCX (i.e., DFC, ITC, MFC, OFC, STC), neocortex; HIP, hippocampus; AMY, amygdala; URL (i.e., CBC, CB, URL), upper rhombic lip.
As predicted, a significantly higher percentage of OLIG2+ cells was found in DS organoids, compared to control organoids (~70% in control vs. ~40% in DS, Fig. 2E, F; n = 4). Consistent with this, we observed a significantly higher expression of OLIG2 mRNA (6 to14 fold), compared to control organoids (Fig. 2G). This finding was verified by western blot analysis (Fig. 2E, F). In addition, we also observed a higher expression of OLIG1 transcripts in DS organoids (6 to 17 fold, Fig. 2G). However, very few OLIG1+ cells were identified in both DS and control organoids by immunostaining and no significant differences were noted between the two groups (Fig. 2C, D). Western blot analysis confirmed the similarly low expression of OLIG1 protein in both control and DS organoids (Fig. 2H, I). These results indicate that our organoid model recapitulates the temporal expression of OLIG genes observed in human brain tissue, further implicating the potential role of OLIG2 in neuronal differentiation in early stage NPCs.
To assess the maturity and region specificity of the DS and control organoids, we obtained global transcriptome profiles of organoids by RNA-seq and compared them with BrainSpan, the largest dataset of post-mortem human brain transcriptomes from embryonic ages to adulthood (Kang et al., 2011). Remarkably, the transcriptome analyses suggested that the hiPSC-derived 5-week-old organoids were most similar to the human brain in 8 to 9 weeks post-conception (pcw) (Fig. 2J). With regards to regional specificity, our preparation best reflected human ventral forebrain, including caudal ganglion eminence (CGE), lateral ganglion eminence (LGE), medial ganglion eminence (MGE), and striatum (STR), with smaller homology to the neocortex (NCX), hippocampus (HIP), amygdala (AMY), and upper rhombic lip (URL) (Fig. 2K). We further comprehensively analyzed the maturity and ventral forebrain specificity of our organoids. Consistent results demonstrated that our preparation best reflected the 8 to 9 pcw of MGE, CGE, and LGE, particularly the MGE. However, much less homology was found to STR that only appears after 9 pcw (Fig. 2L).
Overabundance of GABAergic neurons in DS ventral forebrain organoids and early postnatal DS human brain tissue
To investigate the generation of GABAergic neurons in DS human brain, particularly in early brain development stages, we cultured the control and DS organoids under the neuronal differentiation condition shown in Fig. 1A. Both DS and control organoids similarly increased in size over time in culture (supplementary Fig. 3B, C). The NPCs in DS and control organoids also efficiently differentiated to neurons. At week 8, we saw robust expression of MAP2 (Fig. 3A). We next compared the cell compositions in the DS and control organoids (Fig. 3B, C). Similar percentages of NeuN+ neurons (82.3 ± 3.3% vs. 84.4 ± 3.3% in control and DS organoids, respectively; n = 4) and S100β+ astrocytes (7.7 ± 0.6% vs. 7.9 ± 0.9% in control and DS organoids, respectively; n = 4) were found, indicating similar neural differentiation in the two groups of organoids. Notably, GABA staining robustly labeled neuronal processes in addition to cell bodies in the organoids (Fig. 3D and large-area images of organoids in supplementary Fig. 3D). The GABA fluorescence intensity (FI) in the DS organoids was also significantly higher than that in control organoids (Fig. 3E, n = 4). More than half of the cells (~ 60%) were GABA+ neurons in DS organoids, compared with only ~35% in control organoids (Fig. 3D, G; n = 4).
(A) Representatives of MAP2 expression in 8-week-old control (Cont) and DS ventral forebrain organoids. Scale bar: 50 μm. (B) Representatives of NeuN- and S100β-expressing cells in 8-week-old control and DS organoids. Scale bar: 20 μm. (C) Quantification of pooled data from three pairs of control and DS hiPSCs showing the percentage of NeuN+ and S100β + cells in 8-week-old control and DS organoids. The experiments are repeated for four times (n = 4) and for each experiment, 4 to 6 organoids from each cell line (total around 70 control and 70 DS organoids) are used. Student’s t test. NS represents no significance. Data are presented as mean ± s.e.m. (D) Representatives of GABA-expressing neurons from three pairs of control (Cont1, Cont2, and Di-DS3) and DS (DS1, DS2, and Tri-DS3) hiPSC-derived 8-week-old organoids. Scale bar: 20 μm. (E) Quantification data showing the fluorescence intensity (FI) of GABA in 8-week-old control and DS organoids. The experiments are repeated for four times (n = 4) and for each experiment, 4 to 6 organoids from each cell line are used. Student’s t test. *P < 0.05, ** < 0.01, comparison between DS and corresponding control organoids. Data are presented as mean ± s.e.m. (F) Representatives of CR-, CB-, SST-, PV-, NPY-, and ChAT-expressing neurons in 8-week-old control and DS organoids. Scale bar: 20 μm. (G) Quantification data showing the percentage of GABA+, CR+, SST+, CB+, PV+, NPY+, and ChAT+ neurons in control and DS hiPSC-derived 8-week-old ventral forebrain organoids. The experiments are repeated for four times (n = 4) and for each experiment, 4 to 6 organoids from each cell line are used. Student’s t test. **P < 0.01, ***P < 0.001 and NS represents no significance, comparison between DS and corresponding control organoids. Data are presented as mean ± s.e.m. (H) Western blotting analysis of GAD65/67, CR, and SST expression in postmortem cerebral cortex tissue from control and DS patients at ages less than one-year old. (I) Quantification of GAD65/67, CR, and SST expression in postmortem cerebral cortex tissue from control and DS patients. Student’s t test. *P < 0.05 and NS represents no significance. Data are presented as mean ± s.e.m.
We next examined the differentiation to various subclasses of GABAergic neuron in week 8 organoids. There were significantly more CR+ (~35% vs. ~15% in DS and control organoids, respectively; n = 4) and SST+ (~8% vs. ~2% in DS and control organoids, respectively; n = 4) GABAergic neurons in DS organoids than in control organoids (Fig. 3F, G and supplementary Fig. 3D). There were no significant differences in the percentage of NPY+, CB+, and PV+ GABAergic neurons (Fig. 3F, G). In addition, the percentage of ChAT+ cholinergic neurons were also similar in the two groups of organoids (n = 4). To validate these findings, we examined the expression of CR, SST and GAD65/67, an enzyme required for GABA synthesis and a marker of GABAergic neurons, in postmortem cerebral cortex tissue from DS and control subjects at ages of less than one year old (supplementary Table 3). These human brain tissues were applicable for western blot analysis, since the denatured SDS-PAGE and Coomassie blue staining showed that the human brain sample preparations exhibited the absence of smear (little or no degradation) and sharp resolution of protein bands (supplementary Fig. 3E). Strikingly, we observed that protein levels of CR and GAD65/67 in DS patients, as compared to healthy controls, were significantly higher (Fig. 3H, I). There was a trend of increased expression of SST in DS patients but with large variations, so no statistical significance was detected. This could be due to the relatively much smaller number of SST+ neurons in cerebral cortex in neonatal human brain, compared to CR+ neurons (Paredes et al., 2016). Taken together, these results demonstrate that the organoid model effectively recapitulates the overabundance of GABAergic neurons seen in the developing DS human brain. We further reveal that among various subclasses of GABAergic neurons, CR+ and SST+ GABAergic neurons are preferentially overproduced in the early developing DS brain.
DS neuronal chimeric mice show overproduction of GABAergic neurons
The in vitro data we obtained from the brain organoids was exciting in that the overexpressed OLIG2 in DS appears to bias neuronal differentiation towards inhibitory neurons. However, these neurons represented early developmental stages (Nicholas et al., 2013; Pasca et al., 2015; Patterson et al., 2012). We thus sought to use a human neuronal chimeric mouse brain model (Chen et al., 2016) to examine the stability of the inhibitory neuronal fate of engrafted DS or control NPCs in vivo. We disassociated the 5-week-old, patterned early ventral forebrain DS and control organoids into single progenitor cells and transplanted them into the brains of neonatal rag1-/- immunodeficient mice at regions overlying lateral ventricles (Fig. 4A). For the transplantation experiments, two pairs of control and DS hiPSC lines, including control1 and DS1, and isogenic Di-DS3 and Tri-DS3 hiPSC lines, were used, because based on our results in organoids, all DS lines consistently showed the similar trends of changes with statistical significance when compared with corresponding control lines (Fig. 2G, 3E, 3G, 4J, and 4K). Moreover, inclusion of isogenic hiPSC lines can potentially limit the need for multiple iPSC lines to control for genetic variation (Chen et al., 2014; Maclean et al., 2012; Weick et al., 2013). The donor-derived cells were tracked by staining human nuclei (hN) with specific anti-human nuclei antibody. We found that both engrafted DS and control cells were similarly distributed in mouse brains. At two months post-injection, the engrafted cells were mainly found near the injection sites (Fig. 4B1), with some of the cells having migrated to the deep layers of cerebral cortex (Fig. 4B2). Occasionally, a small cluster of cells was found in the superficial layers of cerebral cortex (Fig. 4B2), which might have been deposited when the injection needle was retracted. The majority of hN+ donor-derived cells expressed NKX2.1, indicating that engrafted cells retained their ventral brain identity in vivo (supplementary Fig. 4A, B). At six months post-transplantation, we observed that around 60% of donor-derived cells were LHX6+ in the cerebral cortex in both control and DS groups (supplementary Fig. 4C, D), suggesting that the majority of the cells were post-mitotic interneuron precursors derived from the NKX2.1+ progenitors (Zhao et al., 2008). At six months post-injection, a large population of human cells dispersed widely in the cerebral cortex (Fig. 4C). Notably, there was a high density of donor-derived cells in the superficial layers and a low density in the deep layers (Fig. 4C1), suggesting the migration of donor-derived cells from injection sites to the cerebral cortex. Doublecortin (DCX) is a marker of young migrating neurons (Gleeson et al., 1999). At 4 months post-injection, many hN+ cells expressed DCX and had bipolar processes, further suggesting migration of the engrafted cells (supplementary Fig. 4E). We also tested injecting cells into deeper sites near the subventricular zone (SVZ) and interestingly found that most of the engrafted cells preferentially migrated along the rostral migratory stream to the olfactory bulb (supplementary Fig. 5). However, in contrast to our findings in the cerebral cortex, very few of these human cells in the olfactory bulb differentiated into CR+ cells (supplementary Fig. 5C, F), further suggesting that that the vast majority of the neural progenitors generated using our current protocol were NKX2.1 lineage and mainly gave rise to cortical or striatal interneurons, rather than olfactory bulb interneurons.
(A) A schematic diagram showing that hiPSC-derived ventral forebrain organoids are dissociated into single cells and then engrafted into the brains of P0 rag1-/- mice.
(B) Representatives of engrafted human cells at two months post-transplantation. Anti-human nuclei antibody (hN) staining selectively labels engrafted human cells. Engrafted cells are found in the brain regions overlying the lateral ventricles (LV; B1) and the cerebral cortex (B2). Scale bars: 100 μm.
(C) Representatives of hN+ donor-derived cells at 6 months post-transplantation, showing wide distribution of donor-derived cells in the cerebral cortex. Areas in C1 and C2 are enlarged to show the different cell density in the cerebral cortex and the distribution of donor-derived cells in the hippocampus (HIP), respectively. Scale bars: 100 μm.
(D) A representative image of engrafted lenti-GFP-labeled hiPSC-derived ventral forebrain NPCs. Notably, the GFP+ neural processes are widely dispersed in the hippocampus. Scale bar: 50 μm.
(E) Representatives of NeuN-, GFAP-, MBP-, and OLIG2-expressing cells differentiated from engrafted control and DS ventral forebrain NPCs in the cerebral cortex at 6 months post-transplantation. Scale bars: 50 μm.
(F) A representative image showing high fluorescence intensity of GABA staining in areas enriched of donor-derived hN+ cells. Scale bar: 50 μm.
(G) Representatives of GABA staining in the cerebral cortex at 6 months after transplantation of control or DS NPCs. Scale bars: 50 μm.
(H) Representatives of CR-, SST-, CB-, PV-, NPY-, and ChAT-expressing donor-derived hN+ cells in the cerebral cortex at 6 months post-transplantation. Scale bars, 50 μm.
(I) Quantification of pooled data from control (Cont1 and Di-DS3) and DS (DS1 and Tri-DS3) hiPSCs showing the percentage of NeuN+, GFAP+, MBP+, and OLIG2+ cells among the total hN+ cells in the cerebral cortex at 6 months post-transplantation (n = 7 mice for each group). Student’s t test. **P < 0.01, ***P < 0.001, and NS represents no significance. Data are presented as mean ± s.e.m.
(J) Quantification data from control and DS groups showing the fluorescence intensity (FI) of GABA staining in the cerebral cortex at 6 months post-transplantation (n = 7 mice for each group). Student’s t test. **P < 0.01, ***P < 0.001, comparison between DS and corresponding control groups. Data are presented as mean ± s.e.m.
(K) Quantification data from control and DS groups showing percentage of CR+, SST+, CB+, PV+, NPY+, and ChAT+ cells among the total hN+ cells in the cerebral cortex at 6 months post-transplantation (n = 7 mice for each group). Student’s t test. ***P < 0.001, comparison between DS and corresponding control groups. Data are presented as mean ± s.e.m.
To specifically study the human GABAergic neurons developed in the context of cerebral cortex, we focused on analyzing the animals that received cell transplantation to the sites overlying the lateral ventricles and had wide distribution of donor-derived cell in the cerebral cortex at the age of 6 months. In a separate group of animals, prior to transplantation, we labeled the cells with lentiviral-CMV-GFP and observed wide distribution of GFP+ processes in the hippocampus, despite the fact that a small population of hN+ cells were seen in the hippocampus (Fig. 4D). In addition, the control and DS cells were highly proliferative before transplantation, as demonstrated by high percentages of Ki67+ cells in both groups (supplementary Fig. 6A, C). We also found that there were more OLIG2+/Ki67+ cells among total OLIG2+ cells in DS organoids than in control organoids, indicating that DS OLIG2+ ventral forebrain NPCs are more proliferative than control OLIG2+ ventral forebrain NPCs (supplementary Fig. 6A, C). After transplantation, the Ki67+/hN+ cells dramatically decreased from 2 months (supplementary Fig. 6B, D; 10.2 ± 1.7% and 14.3 ± 3.3% for control and DS group, respectively; n = 7) to 6 months (0.9 ± 0.4% and 1.4 ± 0.6% for control and DS group, respectively; n = 7). We did not observe any tumor formation of the transplanted cells in all the animals examined. Next, we analyzed the differentiation of the engrafted cells in the cerebral cortex of 6-month-old chimeric mice. As shown in Fig. 4E, I, we found that a large proportion of the engrafted DS and control cells gave rise to NeuN+ neurons. Consistent with previous studies using human samples (Chen et al., 2014; Esposito et al., 2008; Guidi et al., 2008), the DS NPCs exhibited impaired neurogenesis, as indicated by a slightly reduced percentage of NeuN+/hN+ neurons among total hN+ cells in the DS cell transplantation group (84.2 ± 3.2%, n = 7), as compared to the control cell transplantation group (76.7 ± 3.3%, n = 7). A small fraction of the engrafted cells gave rise to GFAP+ astrocytes and MBP+ mature oligodendrocytes. No significant differences were noted in the percentages of astrocytes and mature oligodendrocytes in the two groups. In addition, some of the donor-derived cells also expressed OLIG2, which at this stage were likely oligodendroglia progenitor cells, because no Nestin+ human NPCs were found in these 6-month-old chimeric mouse brains. We observed a higher percentage of OLIG2+/hN+ cells among total hN+ cells in the DS cell transplantation group (18.9 ± 3.40%, n = 7) than that in control cell transplantation group (12.5 ± 1.9%, n = 7), which might reflect the altered oligodendroglial differentiation seen in Ts65Dn mice and DS human brain tissue (Olmos-Serrano et al., 2016). To examine GABAergic neuron differentiation from the engrafted cells, we double-stained the brain sections with GABA and hN. As shown in Fig. 4F and supplementary Fig. 7A, B, GABA staining labeled not only the cell body, but also the processes. As such, the areas that were strongly immunopositive for GABA staining were enriched for hN+ donor-derived cells, indicating the efficient differentiation of the engrafted cells to GABAergic neurons. These interneurons also expressed vesicular GABA transporter (VGAT) (supplementary Fig.7C). We then quantified the FI of GABA staining in the areas that were enriched for hN+ donor-derived cells to compare the production of GABAergic neurons from transplanted DS and control cells. Consistent with the observations in organoids, engrafted DS cells overproduced GABAergic neurons, indicated by the much higher FI of GABA staining in DS cell transplantation group, as compared to control cell transplantation group (Fig. 4G, J). Similar results were also obtained by examining the expression of GAD65/67 (supplementary Fig. 7D, E). Furthermore, we examined the different subclasses GABAergic neuron. Significantly higher percentages of CR+/hN+ and SST+/hN+ neurons among total hN+ cells were found in DS cell transplantation group than in control cell transplantation group (Fig. 4H, K), which was similar to the observations in organoids. No significant differences were noted in the percentage of CB+/hN+, PV+/hN+, and NPY+/hN+ GABAergic neurons, as well as ChAT+/hN+ cholinergic neurons (Fig. 4H, K). Previous studies in mice show that a subpopulation of neocortical interneurons co-express CR and SST (Xu et al., 2006). To further examine whether this subpopulation of interneurons was overly produced from transplanted DS cells, we triple-stained control and DS chimeric mouse brains with CR, SST, and hN. At 6 months post-transplantation, none of the human interneurons in control chimeric mice co-expressed CR and SST, whereas in DS chimeric mice, CR+/SST+ human interneurons were occasionally seen (supplementary Fig. 7F; less than 1% of total hN+ cells). These results suggest that the subclass of CR+/SST+ interneurons is increased in the DS chimeric mice, but they only represent a very small population. The overly produced interneurons from the engrafted DS cells are mainly CR+/SST- and CR-/SST+ interneurons.
In order to examine whether the donor cell-derived neurons integrated into the brain, we immunostained synaptic markers. As shown in supplementary Fig. 8A, many of the hN+ cells in the cerebral cortex were surrounded by the presynaptic marker synapsin I, suggesting the formation of synapses. Then, we double-stained human-specific MAP2 (hMAP2), which selectively labels dendrites of donor-derived human neurons, and postsynaptic density protein 95 (PSD-95), a postsynaptic marker. The PSD-95+ puncta were found to distribute along the hMAP2+ dendrites (supplementary Fig. 8B), further demonstrating that the human neurons formed synapses in the mouse brain. Furthermore, we examined the expression of c-Fos, an activity-dependent immediate early gene that is expressed in neurons following depolarization and often used as a marker for mapping neuronal activity (Loebrich and Nedivi, 2009). As shown in supplementary Fig. 8C, D, about 20% of the hN+ neurons at 6 months were positive for c-Fos, indicating that they received synaptic neurotransmission and were active in the mouse brain. Moreover, in a separate group of animals, we labeled donor cells with lentiviral-CMV-GFP before transplantation. At 6 months post-transplantation, we performed electron microscopic analysis. We observed synaptic terminals formed between human neurons labeled by diaminobenzidine (DAB) staining against GFP or a human-specific cytoplasmic marker STEM121 and mouse neurons that were not labeled by the DAB staining (supplemental Fig. 8E). Numerous synaptic contacts from mouse neurons to human neurons were found in the sections and those were asymmetric synapses, which are typically excitatory synapses (Harris and Weinberg, 2012). On the other hand, the synaptic contacts from human neurons to mouse neurons were symmetric synapses, indicating that those synapses are inhibitory synapses (Harris and Weinberg, 2012). Taken together, these results indicate that human neuronal chimeric mouse model in which ventral forebrain NPCs are engrafted effectively recapitulates the overproduction of GABAergic neurons from DS NPCs in brain organoids. Moreover, the donor cell-derived GABAergic neurons can integrate into the mouse brain and are functionally active.
OLIG2 overexpression in DS biases gene expression attributed to GABAergic neurons
To gain mechanistic insight into how DS ventral brain NPCs were biased to generation of GABAergic neurons, we performed next-generation whole-genome deep sequencing and analyzed the transcriptomic profiles of 5-week-old DS and control organoids, including one pair of control and DS organoids (control1 and DS1), and the pair of isogenic Di-DS3 and Tri-DS3 organoids. A dendrogram demonstrated that two control organoids (control1 and Di-DS3 organoids) and two DS organoids (DS1 and Tri-DS3 organoids) clustered closer to each other, respectively, indicating that they had similar biological properties within the same group, while the gene expression of DS organoids was distinct from that of control organoids (Fig. 5A). Then we did two-step analyses to narrow down to the most prominent differentially expressed genes (DEGs) that were identified in common between control1 vs. DS1 and Di-DS3 vs. Tri-DS3 organoids. Previous studies have shown that comparisons of RNA-seq data between isogenic pairs are often distinct from those between cell lines with different genetic backgrounds (Lin et al., 2015). Thus, choosing only transcripts overlapping between the two pairs of cultures provides an added level of confidence in their reproducibility. As shown in Fig. 5A, B, we began with comparing expression of transcripts between control1 and DS1 organoids. We identified a total of 2,920 DEGs, including 1,796 up-regulated genes and 1,124 down-regulated genes. To reduce the potential interference of different genetic background between control1 and DS1 cells, we then examined the 2,920 DEGs in the organoids generated from isogenic Di-DS3 and Tri-DS3 hiPSCs. We identified that a total of 610 overlapping DEGs, including 398 up-regulated genes and 212 down-regulated genes. Notably, there were 943 DEGs (506 up-regulated genes and 437 down-regulated genes) in the comparison between Di-DS3 and Tri-DS3 organoids that were not seen in the comparison between control1 and DS1 organoids. Next, we analyzed the distribution of the overlapping 610 DEGs on each human chromosome. As shown in Fig. 5C, we found that the majority of the DEGs were non-HSA21 genes (577 of 610 DEGs), similar to previous studies using human samples (Olmos-Serrano et al., 2016; Weick et al., 2013). However, HSA21 had the highest percentage of DEGs among all the chromosomes. We also observed that all of the DEGs on HSA21 were upregulated in DS samples and the overexpression of these genes on HSA21 resulted in both positive and negative regulation of the genes on other chromosomes. To further investigate how inhibiting the expression of OLIG2 would change the global gene expression profile in DS during early embryonic brain development, we generated Tri-DS3 hiPSCs expressing OLIG2 shRNA (Tri-DS3 + OLIG2shRNA hiPSCs) by infecting these hiPSCs with lentivirus carrying the OLIG2 shRNA. Then we generated organoids from the hiPSCs (Tri-DS3 + OLIG2shRNA organoids). RNA-seq results revealed that many DEGs were reversed after OLIG2 knockdown, as compared to the organoids derived from the Tri-DS3 hiPSCs received control shRNA with scrambled and non-targeting sequences (Tri-DS3 + ContshRNA organoids) and control1 organoids (Fig. 5D). A Venn diagram showed that the expression of 255 of 398 upregulated overlapping DEGs and 166 of 212 downregulated overlapping DEGs was effectively reversed (Fig. 5E). We also analyzed the reversed DEGs after OLIG2 knockdown compared between Di-DS3 and Tri-DS3 and their distribution on chromosomes (supplementary Fig. 9). To explore the possible biological processes altered by the overlapping DEGs, gene ontology (GO) analyses of the upregulated and downregulated genes were performed. Biological processes such as myelin sheath, transcription, and cell proliferation were significantly enriched, consistent with previous findings (Lockstone et al., 2007; Lu et al., 2012; Olmos-Serrano et al., 2016) (Fig. 5F, G). Interestingly, GO and top 20 annotated pathways revealed that an enrichment of downregulated genes in DS organoids was particularly related to neurodevelopmental terms such as nervous system development, synaptic vesicle recycling, axon guidance, synaptic vesicle, axon extension, dendrite and neuron projection guidance, indicating that the embryonic neuronal development as early as 8 to 9 pcw is altered in DS (Fig. 5F). In addition, we found that the upregulated genes in DS organoids were enriched in extracellular matrix terms, such as extracellular matrix organization, extracellular matrix structural constituent, extracellular matrix disassembly, indicating that the altered cells’ dynamic behavior in DS might affect cell adhesion, communication, and differentiation and fate specification (Fig. 5G). Moreover, the GO analyses demonstrated that most of the changed pathways in both DS1 and Tri-DS3 groups were also significantly rescued, such as nervous system development, axon guidance, and extracellular matrix organization, suggesting that inhibiting OLIG2 expression in DS effectively reversed the tendency for abnormal neurodevelopment (Fig. 5F, G).
(A) A heatmap showing the differentially expressed genes (DEGs) in 5-week-old organoids derived from control (Cont1 and Di-DS3) and DS (DS1 and Tri-DS3) hiPSCs. Notably, the two disomic control organoids and two trisomic DS organoids cluster closer to each other, respectively.
(B) Venn diagram showing the numbers of DEGs in 5-week-old control and DS organoids. Among these genes, 398 and 212 DEGs are commonly upregulated and downregulated, respectively in the two pairs of control and DS organoids.
(C) The total number of DEGs, percentage of DEGs, and number of up-regulated or down-regulated DEGs on each chromosome in 5-week-old control and DS organoids.
(D) A heatmap showing the DEGs in 5-week-old control (Cont1 and Di-DS3) organoids, DS (DS1 and Tri-DS3) organoids, and organoids derived from Tri-DS3 hiPSCs infected with lentivirus that carry control shRNA (Tri-DS3 + ContshRNA) or OLIG2 shRNA (Tri-DS3 + OLIG2shRNA).
(E) Venn diagram showing the numbers of rescued genes after OLIG2 knockdown (KD). Among these genes, 255 and 166 genes are the overlapping DEGs identified from the two pairs of control and DS organoids.
(F) GO analyses of the overlapping down-regulated DEGs and the rescue effects after OLIG2 knockdown.
(G) GO analyses of the overlapping up-regulated and the rescue effects after OLIG2 knockdown.
(H) A heatmap showing the expression of genes critical for GABAergic neuron differentiation and specification in the 5-week-old organoids.
(I) qPCR analysis of OLIG2, DLX1, DLX2, DLX5, DLX6, LHX6, LHX8, and SOX6 mRNA expression in 5-week-old organoids. The experiments are repeated for four times (n = 4) and for each experiment, 20 to 30 organoids are used in each group. One-way ANOVA test. ***P < 0.001 and NS represents no significance. Data are presented as mean ± s.e.m.
(J) ChIP assay showing that OLIG2 directly interacts with the promotor of DLX1 and LHX8. The regions highlighted in red are the only ones that are immunoprecipitated by OLIG2 antibody.
(K) Luciferase reporter assay showing that OLIG2 expression vector is capable of upregulating DLX1 or LHX8 luciferase activity, compared with control vector. The experiments are repeated for three times (n = 3). Student’s t test. * P < 0.01, ** P < 0.05. Data are presented as mean ± s.e.m.
In order to investigate the molecular mechanisms underlying the overproduction of GABAergic neurons in DS, we examined the expression of transcription factors that play critical roles in GABAergic neuron development and specification (Cobos et al., 2006; Kelsom and Lu, 2013). RNA-seq results indicated that expression of many of these transcription factors was altered in DS organoids, as compared to control organoids. Among this group, DLX1/2, LHX6/8, and SOX6 are reported to regulate specification of CR+ and SST+ interneurons(Azim et al., 2009; Cobos et al., 2005; Flandin et al., 2011). NPAS1/3, which also regulates interneuron fate(Stanco et al., 2014), was abnormally increased in the DS organoids. In addition, we also found that other genes that regulates interneuron neurogenesis and migration, such as ASCL1(Castro et al., 2011) (Casarosa et al., 1999), ARX (Colasante et al., 2008), and NR2F2 (Kanatani et al., 2008) were also altered in DS organoids. Furthermore, we analyzed expression of these transcription factors after OLIG2 knockdown. Strikingly, the majority of them was effectively reserved (Fig. 5H). We further performed qPCR to verify RNA-seq results using brain organoids from the isogenic pair of hiPSCs. As shown Fig. 5I, qPCR results showed that the expression of all the transcripts was increased in DS. Among all the genes we examined, only DLX 5 and 6 showed inconsistent results between RNA-seq and qRT-PCR verification. Notably, unlike other genes in the RNA-seq analysis (Fig. 5H), only DLX 5 and 6 exhibited different expression levels between Tri-DS3 organoids and Tri-DS3+ContshRNA organoids which were derived from the same Tri-DS3 hiPSCs but received infection of lentivirus carrying control shRNA. This result suggests that expression levels of DLX5 and 6 might not be able to be stably detected by RNA-sequ. This could be caused by the low expression levels of DLX5 and 6, because our qRT-PCR analysis indicated that DLX5 and 6 were present at very low abundance in all the cell preparations (< 0.015% and 0.004% of GAPDH for DLX5 and 6, respectively in control organoids, as compared to 0.48% and 0.73% of GAPDH for DLX1 and 2, respectively). Moreover, our qPCR results demonstrated that inhibiting OLIG2 expression reversed the increased expression of the majority of the transcription factors, such as DLX1/2, LHX6/8 and SOX6 (Fig. 5I).
As a transcription factor, OLIG2 was reported to regulate gene expression through activating the promoter and enhancer (Kuspert et al., 2011; Ligon et al., 2007). To further test our hypothesis that OLIG2 could regulate the expression of these transcriptional factors directly through interacting with the promoters of these genes, we performed a chromatin immunoprecipitation (ChIP) assay. We focused on examining the interactions between OLIG2 with LHX8 or DLX1 promoter sequences, because LHX8 had the highest fold-change in the DS (Fig. 5I) and DLX1 was reported to be crucial for the production and longevity of CR+ and SST+ GABAergic neuron (Cobos et al., 2005). We designed 6 to 7 pairs of primers to span the 2000bp promoter sequences of upstream of the transcriptional start sites (TSS) of LHX8 and DLX1. As shown in Fig. 5J, OLIG2 antibody immunoprecipitated the regions from -162 to - 345, -1149 to -1321, and -1756 to -1890 in LHX8 promoter region and from -320 to -452 in DLX1 promoter region, demonstrating direct interactions of OLIG2 with promoters of LHX8 and DLX1. We further performed dual luciferase reporter assay to examine the functionality of OLIG2 interactions with DLX1 and LHX8. Based on our ChIP-PCR assay results, we cloned the promoter sequences of DLX1 (−10 to −1250) and LHX8 (−1 to −2000) and inserted them into pGL3-Basic luciferase reporter vector. We then transfected these plasmids into control hiPSC-derived pNPCs. As shown in Fig. 5K, when an OLIG2 expression vector was co-transfected into these cells, compared to co-transfection with a control vector, nearly 2.6 or 2.1-fold upregulation of luciferase activity was induced for DLX1 or LHX8, respectively (Fig. 5K). Thus, these data provide direct evidence indicating that OLIG2 functions upstream of DLX1 and LHX8 and promotes their expression by regulating their promoter activity.
Collectively, all these results demonstrate that OLIG2 overexpression is likely associated with the abnormal gene expression profiles and that altered biological processes are critically involved in early DS embryonic brain development. More importantly, we find that OLIG2 directly binds and regulates the promoter regions of pro-inhibitory neuronal differentiation transcription factors, which might be the molecular underpinnings for a biased neural differentiation in DS.
Inhibiting OLIG2 expression rescues overproduction of GABAergic neurons in DS organoids and chimeric mouse brains
Given the mechanistic insight presented in this study that overexpressed OLIG2 was responsible for the abnormal neuronal differentiation, we hypothesized that inhibiting the expression of OLIG2 would reverse the overproduction phenotypes under trisomy 21 conditions. To test this hypothesis, again we took the RNAi knockdown approach and used two DS hiPSC lines expressing OLIG2 shRNA or control shRNA (In addition to the Tri-DS3 hiPSCs, we also generated DS1 hiPSCs expressing OLIG2 shRNA or control shRNA). We cultured the 5-week-old DS + OLIG2shRNA and DS + ContshRNA ventral forebrain organoids under neuronal differentiation conditions and also generated chimeric mouse brains by engrafting the NPCs dissociated from those organoids. After 3 weeks (8-week-old organoids), we examined the cell compositions and found that NeuN, S100β, PDGFRα, and NG2 expression were similar in disomic control, DS + OLIG2shRNA, and DS + ContshRNA organoids (supplementary Fig. 10A, C). Then we examined the population of GABAergic neurons. The percentages of GABA+ cells and FI of GABA staining were significantly decreased in DS + OLIG2shRNA organoids, as compared to DS + ContshRNA organoids (Fig. 6A, B). Moreover, the percentages of CR+ and SST+ GABAergic neurons were decreased after inhibition of OLIG2 expression (Fig. 6C, D). The percentages of other subclasses of GABAergic neurons and cholinergic neurons were not significantly changed. In 6-month-old chimeric mouse brains, we found that the percentage of NeuN+/hN+ among total hN+ cells increased in the mouse brains received DS + OLIG2shRNA NPC transplant, as compared to the mouse brains received DS + ContshRNA NPC transplant (Supplementary Fig. 10B, D). Consistent with the results in organoids, FI of GABA staining was decreased after inhibiting OLIG2 expression (Fig. 6E and F). The overproduction of CR+ and SST+ GABAergic neurons was rescued and no significant differences of the population of CB+, PV+, NPY+, and ChAT+ neurons were noted (Fig. 6E and F). Therefore, overexpression of OLIG2 in DS results in overproduction of subclass-specific GABAergic neurons, which can be rescued by inhibiting OLIG2 expression.
(A) Representatives of GABA-expressing cells in 8-week-old DS + ContshRNA and DS + OLIG2shRNA organoids. Scale bars: 20 μm.
(B) Quantification of the percentage of GABA+ cells and FI of GABA staining in 8-week-old control (Cont) organoids, DS + ContshRNA organoids, and DS + OLIG2shRNA organoids. The data are pooled from control (Cont1 and Di-DS3) hiPSCs, DS (DS1 and Tri-DS3) hiPSCs, and the two lines of DS hiPSCs infected with lentiviruses that carry control shRNA or OLIG2 shRNA. The experiments are repeated for four times (n = 4) and for each experiment, 4-6 organoids from each cell line are used. One-way ANOVA test, **P < 0.01 and ***P < 0.001. Data are presented as mean ± s.e.m.
(C) Representatives of CR-, SST-, CB-, PV-, NPY-, and ChAT-expressing cells in 8-week-old DS + ContshRNA and DS + OLIG2shRNA organoids. Scale bars: 20 μm.
(D) Quantification of pooled data showing the percentage of CR+, SST+, CB+, PV+, NPY+, and ChAT+ cells in 8-week-old disomic Cont organoids, DS + ContshRNA organoids, and DS + OLIG2shRNA organoids (n = 4 and for each experiment, 4 to 6 organoids from each cell line are used). One-way ANOVA test, ***P < 0.001, NS represents no significance. Data are presented as mean ± s.e.m.
(E) Representatives of GABA-expressing donor-derived hN+ cells in the cerebral cortex at 6 months after transplantation of DS + ContshRNA or DS + OLIG2shRNA ventral forebrain NPCs. Scale bars: 50 μm.
(F) Quantification of FI of GABA staining in the cerebral cortex at 6 months after transplantation of Cont, DS + ContshRNA or DS + OLIG2shRNA ventral forebrain NPCs (n = 6-8 mice for each group). The data are pooled from mice received transplantation of the NPCs derived from control (Cont1 and Di-DS3) hiPSCs, DS (DS1 and Tri-DS3), and the two lines of DS hiPSCs infected with lentiviruses that carry control shRNA or OLIG2 shRNA. One-way ANOVA test, **P < 0.01 and ***P < 0.001. Data are presented as mean ± s.e.m.
(G) Representatives of CR-, SST-, CB-, PV-, NPY-, and ChAT-expressing donor-derived hN+ cells in the cerebral cortex at 6 months after transplantation of DS + ContshRNA or DS + OLIG2shRNA ventral forebrain NPCs. Scale bars: 50 μm.
(H) Quantification of percentage of CR+, SST+, CB+, PV+, NPY+, and ChAT+ cells among total hN+ donor-derived cells in the cerebral cortex at 6 months after transplantation of Cont, DS + ContshRNA or DS + OLIG2shRNA ventral forebrain NPCs (n = 6-8 mice for each group). The data are pooled from mice received transplantation of the NPCs derived from control (Cont1 and Di-DS3) hiPSCs, DS (DS1 and Tri-DS3), and the two lines of DS hiPSCs infected with lentiviruses that carry control shRNA or OLIG2 shRNA. One-way ANOVA test, **P < 0.01 and ***P < 0.001. Data are presented as mean ± s.e.m.
DS chimeric mice exhibit impaired recognition memory that can be improved by inhibiting OLIG2 expression
Cognitive impairment, characterized by learning and memory deficits, is a primary phenotypic manifestation of DS patients and mouse models of DS (Contestabile et al., 2010). Based on the findings that the donor cell-derived neurons integrated into the mouse brain (supplementary Fig. 8B), we hypothesized that engrafted control and DS cells might differently impact the behavioral performance of the animals and inhibiting the expression of OLIG2 in DS cells would reverse the different behavior phenotypes. To test our hypothesis, we examined five groups of mice: (1) vehicle control group that received injection of phosphate-buffered saline (PBS); (2) control chimeric mice that received transplantation of control1 or Di-DS1 cells; (3) DS chimeric mice that received transplantation of DS1 or Tri-DS3 cells; (4) DS + ContshRNA chimeric mice that received transplantation of DS1+ContshRNA or Tri-DS3+ContshRNA cells; and (5) DS + OLIG2shRNA chimeric mice that received transplantation of DS1+OLIG2shRNA or Tri-DS3+OLG2shRNA cells. We performed histological assessment after behavioral testing and assessed the engraftment efficiency and degree of chimerization by quantifying the percentage of hN+ cells among total DAPI+ cells in sagittal brain sections covering regions from 0.3 to 2.4 mm lateral to midline, as reported in the previous studies (Chen et al., 2015; Han et al., 2013). We found that the degree of chimerization was similar among the four groups of animals that received cell transplant (supplementary Fig. 10E).
We performed open field test to evaluate basal global activity in the five groups of mice. Mice were placed into a chamber under dim ambient light conditions and activity was monitored for 5 minutes. We found no significant difference in the total traveling distance and time in the center area (Fig. 7A; n = 10 - 15), suggesting no significant difference in locomotor activity and exploratory behavior among the four groups. Next, we performed novel object recognition test to assess the learning and memory that is associated with the prefrontal cortex and hippocampus (Antunes and Biala, 2012; Wan et al., 1999). Mice were placed in the center of the acquisition box and allowed for 10 min to investigate two identical objects during the training day. 24 hours later, one old object was replaced with a novel object and the time spent on each object was recorded on the testing day (Fig. 7B). Similarly, no difference in the total traveling distance was observed (Fig. 7B). We did not observe any significant behavioral changes in the preference ratio to the novel object between the PBS control group and control chimeric mouse group (new Fig. 7). Interestingly, DS chimeric mice spent less time with the novel object, compared to control chimeric mice and PBS control mice (Fig. 7B; 0.62 ± 0.024, 0.68 ± 0.028, 0.50± 0.017 for PBS control mice, control chimeric mice, and DS chimeric mice, respectively; n = 10-15), suggesting that DS chimeric mice had impaired recognition memory. Moreover, this impairment was effectively rescued by inhibiting OLIG2 expression (Fig. 7B; 0.49 ± 0.020 and 0.60 ± 0.016 for DS + ContshRNA and DS + OLIG2shRNA chimeric mice, respectively; n = 10-15). Next, we performed the elevated plus maze test to assess anxiety of the chimeric mice to further rule out the effects of anxiety on learning performance. Based on rodents’ tendency toward dark, enclosed spaces, and an unconditioned fear of heights/open spaces, mice were placed in the center of the apparatus with two open and two enclosed elevated arms and allowed to explore the apparatus for 5 min. We did not observe any significant difference in the total traveling distance, number of entries, and time spent in the open arms (Fig. 7C; n = 10-15), suggesting the anxiety was similar among the five groups.
(A) Open filed test showing the traveling distance and the percentage of time that PBS control mice, Cont chimeric mice, DS chimeric mice, DS + ContshRNA chimeric mice, and DS + OLIG2shRNA chimeric mice spend in the center area of (n = 10-15 mice for each group). One-way ANOVA test. NS represents no significance. Data are presented as mean ± s.e.m.
(B) Left, a schematic diagram showing novel object recognition test. Right, quantification of the traveling distance and the preference ratio to the novel object of the five groups of chimeric mice (n = 10-15 mice for each group). One-way ANOVA test. ***P < 0.001, **P < 0.01. NS represents no significance. Data are presented as mean ± s.e.m.
(C) Elevated plus-maze test showing the traveling distance, the number of entries, and the time that the chimeric mice spend on open arms (n = 10-15 mice for each group). One-way ANOVA test. NS represents no significance. Data are presented as mean ± s.e.m.
Discussion
In this study, we demonstrate that human OLIG2-expressing ventral forebrain NPCs give rise to multiple subclasses of GABAergic neurons and cholinergic neurons. This population of OLIG2+ NPCs is overabundant in DS, which leads to overproduction of subclass-specific GABAergic neurons likely via acting on GABAergic neuron lineage-determining transcription factors. When engrafted into mouse brains, the DS chimeric mice exhibit impaired recognition memory, compared to control chimeric mice. Importantly, inhibiting OLIG2 expression not only rescues the overproduction of GABAergic neurons, but also corrects many altered biological processes, particularly those involved in neuronal development, and improves behavioral performance of DS chimeric mice.
Studies in mouse models show discrepant findings on the roles of Olig genes in GABAergic neuron development under normal (Furusho et al., 2006; Liu et al., 2015; Ono et al., 2008; Petryniak et al., 2007; Silbereis et al., 2014) and DS disease conditions (Belichenko et al., 2015; Das and Reeves, 2011; Goodliffe et al., 2016). Moreover, the expression domains of OLIG1 and OLIG2 are significantly different between the rodent and human embryonic brain (Jakovcevski and Zecevic, 2005a; Lu et al., 2000; Petryniak et al., 2007). These argue for the development of species-specific research tools to investigate their functions in human brain development and pathology. In this study, we unravel the contribution of OLIG2 overexpression in DS to developmental abnormalities in humans by using a novel hiPSC-based paradigm to precisely model DS. In this paradigm, we employ a complementary combination of human iPSC-derived cerebral organoid and developmentally-appropriate chimeric mouse brain model systems. These systems embrace the complexity of human neural cells developed in 3D conditions that retain cell-cell/cell-matrix interactions. First, we generate human ventral forebrain organoids from hiPSCs and find that compared to conventional 2D monolayer cultures, this 3D brain organoid model possesses significant advantages for modeling human embryonic ventral forebrain development and neuronal differentiation. The ventral forebrain organoids closely recapitulate the expression pattern of OLIG1 and OLIG2 in the human embryonic brain under healthy and DS disease states. As opposed to 2D cultures, the OLIG2+ NPCs in the 3D cultures are capable of efficiently generating distinct subclasses of neurons. Differentiation of human cholinergic neurons is rarely seen from ventral forebrain NPCs in 2D cultures (Ma et al., 2012; Maroof et al., 2013; Nicholas et al., 2013), unless nerve growth factor, a survival factor for cholinergic neurons, is added (Liu et al., 2013b). Interestingly, we find that without adding nerve growth factor, the purified OLIG2+ NPCs also generate cholinergic neurons in the organoids. Second, building upon previous studies (Chen et al., 2015; Han et al., 2013; Windrem et al., 2017; Windrem et al., 2008), we generate human neuronal chimeric mice by transplanting ventrally patterned pNPCs into the neonatal mouse brain. The development of this human chimeric mouse model permits the study of human interneuron development and disease pathogenesis in vivo, which allows specifying 1) whether the inhibitory neuron cell fate remains stable under long-term survival conditions; and 2) whether grafted “disordered” neurons cause any behavioral abnormalities. Recent studies in human brain tissue have demonstrated the widespread and extensive migration of immature neurons for several months in the neonatal brain (Paredes et al., 2016; Sanai et al., 2011). In particular, a study identifies that these immature neurons are mainly young GABAergic neurons, which form an arching or “Arc” structure adjacent to the anterior body of lateral ventricle and within the neighboring subcortical white matter. This “Arc” migratory stream takes the human GABAergic neurons to the frontal lobe of infants and young children (Paredes et al., 2016). In our study, at 2 months post-transplantation, most of the engrafted cells migrate away from injection sites and integrate into the brain regions overlying the lateral ventricles, which might mimic the “Arc” structure seen in the infant human brain. Very interestingly, at 6 months, the donor-derived cells, mostly GABAergic neurons, reside predominantly in the cerebral cortex. Previous studies have documented the transplantation of hPSC-derived progenitors of GABAergic neurons into rodent developing brains and demonstrated active migration of these GABAergic neurons to the cerebral cortex (Maroof et al., 2013; Nicholas et al., 2013). However, enrichment of human GABAergic neuron in the cerebral cortex, as shown in the present study, has not been reported previously. In our study, injection of the cells to the selected sites (anterior anlagen of the corpus callosum that overlies lateral ventricles, but not the deeper sites near the SVZ) of the neonatal mouse brain within a day of birth may facilitate the integration, migration, and neuronal differentiation of the engrafted cells. Moreover, these cells are also expected to divide for at least 6 months in the mouse host (Chen et al., 2016; Windrem et al., 2014), which may result in the large number of human donor cells in the mouse brain seen at 6 months. Therefore, the brain organoid and human neuronal chimeric mouse brain models provide powerfully tools to study human interneuron development both in vitro and in vivo and reveal the contribution of abnormal development of interneuron to cognitive behavioral deficits in DS as well as other neurodevelopmental disorders.
These human-centric systems provide unprecedented opportunities to yield new findings on DS disease pathogenesis. Previously, only DS human fetal brain tissues from dorsal brain regions (mostly frontal cortex) in ages ranging from GW 13-21 were used for pathological studies, isolating DS NPCs for in vitro characterization, and gene profiling studies (Bhattacharyya et al., 2009; Busciglio et al., 2002; Busciglio and Yankner, 1995; Esposito et al., 2008; Lu et al., 2012; Mao et al., 2003; Olmos-Serrano et al., 2016). However, as early as GW5, OLIG2 is present in regions of embryonic ventral neuroepithelium (Jakovcevski and Zecevic, 2005a). The biological events and consequences elicited by the overexpression of OLIG2 within the first trimester in DS human embryonic brain is largely unclear. Previous studies have demonstrated that hPSC-derived brain organoids are most effective in modeling neurodevelopmental disorders that emerge within the first trimester (Lancaster and Knoblich, 2014; Lancaster et al., 2013; Mariani et al., 2015). The comparison between our organoid samples and the BrainSpan dataset confirms that the gene expression patterns in the organoids exhibit similarities with human embryonic brain at 8 to 9 pcw (equivalent to GW 10 to 11) that precede fetal stages (Engle et al., 2004). Thus, these organoids provide unprecedented opportunities to investigate DS human brain development at embryonic stages. We identify a large number of DEGs across the DS and control organoids. The GO analyses indicate that biological processes, such as nervous system development, axon guidance, and extracellular matrix organization, are abnormally altered in DS. Importantly, we find that by directly interacting with GABAergic neuron lineage-determining transcription factors, overexpression of OLIG2 leads to overproduction of the subclass-specific GABAergic neurons in both DS organoids and DS chimeric mouse brains. We propose that the increased expression of DLX1 in DS organoids may be largely responsible for the overproduction of CR+ interneurons, because previous studies in mice have shown that Dlx1 is involved in production of CR+ interneurons (Cobos et al., 2005); and that increased expression of LHX6/8 and SOX6 may be largely responsible for the overproduction of SST+ interneurons, because Lhx6/8 and Sox6 regulate specification of SST+ interneurons (Azim et al., 2009; Flandin et al., 2011). In addition, these transcription factors may also work together to fine-tune the production of different subclasses of interneurons, since previous studies have also shown that Dlx5/6 are directly downstream of Dlx1/2 (Anderson et al., 1997a; Anderson et al., 1997b; Zerucha et al., 2000) and Sox6 functions downstream of Lhx6 (Batista-Brito et al., 2009). Interestingly, our RNA-seq analyses also show abnormal expression of genes involved in oligodendrocytes differentiation and myelination in DS, which is in consistence with a recent study that reported defective oligodendrocyte differentiation and myelination in Ts65Dn mice and human brain tissue (Olmos-Serrano et al., 2016). However, our organoid and chimeric mouse models in their present state are not optimal for studying oligodendrocyte development in DS, as few oligodendroglia cells are generated in our models. Further development of protocols to generate brain organoids with robust oligodendrocyte differentiation and formation of myelin cytoarchitecture (Madhavan et al., 2018; Quadrato et al., 2016; Rodrigues et al., 2017) as well as generation of DS human chimeric mouse brain with myelin-deficient Shiverer mice (Wang et al., 2013; Windrem et al., 2014) may help us better understand the roles of overexpressing OLIG genes in defective human oligodendrocyte differentiation and myelination in DS.
Notably, a recent study in human brain tissue (Paredes et al., 2016) demonstrated that from postnatal 1 day to 5 months, among various subclasses of GABAergic neurons, CR+ neurons were the predominant subclasses that migrated to the cingulate gyrus from the “Arc” structure, with the most drastic addition to the cortex (over 4-fold increase from 100,000 to over 400,000 CR+ cells/cingulate segment). This finding suggests that those CR+ neurons may play important roles associated with the critical period of brain plasticity in the early postnatal developing human brain. Intriguingly, our immunoblot results show an overabundance of CR in DS cortical brain tissue, collected from patients less than one-year old. We find that DS NPCs overproduce CR+ and SST+ GABAergic neurons in the chimeric mouse brain. Although a large number of human cells were seen in the mouse brains, we did not observe any significant behavioral changes between the PBS control group and control chimeric mouse. We further found that only around 20% of the total human cells were likely to be functionally active. Thus, the low percentage of active cells among the total control human interneurons may not be able to cause significant behavioral changes in control chimeric mouse group, as compared to the PBS control group. Furthermore, DS chimeric mice exhibited impaired recognition memory in novel object recognition test, compared to control chimeric mice and PBS vehicle control mice, despite the fact that similarly around 20% of total DS cells were c-Fos+ active interneurons. These data suggest that the impaired recognition memory in DS chimeric mice is very likely caused by the addition of these subclasses-specific human interneurons. This is further supported by our observations that inhibiting OLIG2 in DS cells corrected the abnormal production of the CR+ and SST+ interneurons and that DS+OLIG2shRNA chimeric mice had improved recognition memory, compared to the DS+controlshRNA chimeric mice. Collectively, these results suggest that cognitive impairments in DS may be associated with the abnormal production of subclass-specific GABAergic neurons from embryonic ventral forebrain NPCs.
We also discovered novel findings that are distinct from those identified previously in mice and in human samples. Studies using postmodern brain tissues from elderly DS patients (Kobayashi et al., 1990; Ross et al., 1984) and 2D cultures of DS iPSCs (Huo et al., 2018) showed reduced production of GABAergic neurons. The previous studies only used postmodern brain tissues from elderly patients with Alzheimer-like degeneration (Kobayashi et al., 1990; Ross et al., 1984), which may confound the evaluation of the cell counts of GABAergic neurons (Contestabile et al., 2017). Our results using DS hiPSCs and human brain tissues from DS patients less than one year old clearly demonstrate overproduction of SST+ and CR+ GABAergic neurons. A recent study (Huo et al., 2018) using DS hiPSCs reported discrepant findings that DS cells produced less CR+ neurons, although they similarly overproduced SST+ neurons. We propose that the discrepancy may result from the fact that traditional 2D cultures were used in that study, since both previous studies (Brawner et al., 2017; Centeno et al., 2018; Pasca, 2018; Simao et al., 2018) and our results have consistently demonstrated the advantages of modeling human embryonic brain development using the 3D brain organoid model. In addition, the different transplantation strategies may also result in the discrepant in vivo findings. Our transplantation strategy was designed to more appropriately mimic the GABAergic neuron development in an immature brain in the context of cerebral cortex, as opposed to the strategy of transplantation of young GABAergic neurons into a mature brain (8 to 10 weeks old mice) at sites very deep in the brain (medial septum), used in the recent study (Huo et al., 2018). These differences demonstrate the importance of using a developmentally-appropriate paradigm. In addition, we find that human OLIG2, but not OLIG1, determined the overproduction of GABAergic neurons in DS, mainly of the CR+ subclass. These findings also distinct from the observations in mice, in which mouse Olig1, rather than Olig2, regulates GABAergic neuron differentiation (Furusho et al., 2006; Ono et al., 2008; Petryniak et al., 2007; Silbereis et al., 2014), and overexpression of Olig1 and Olig2 resulted in mainly overproduction of PV+ GABAergic neurons in Ts65Dn mice (Chakrabarti et al., 2010). Taken together, our study has significantly advanced the understanding of the functions of OLIG2 in the pathophysiology of DS. This work complements animal studies but presents a novel and distinct human neuronal context.
Postnatal pharmacotherapy targeting the GABA system has improved learning and memory as well as electrophysiological alternations in affected animals (Contestabile et al., 2017). However, this approach has pro-epileptic side effects (Contestabile et al., 2017; Fernandez et al., 2007). There is a growing interest in the development of prenatal therapy for DS, as part of a ‘fetal personalized medicine’ strategy (Bianchi, 2012; de Wert et al., 2017; Guedj et al., 2014). A prenatal therapy approach may be more effective at rebalancing inhibition and excitation, considering that GABAergic neurons are born during the prenatal period. We demonstrate that inhibiting OLIG2 expression not only rescues the overproduction of GABAergic neurons, but also corrects many of the altered biological processes. Therefore, our results suggest that OLIG2 is potentially an excellent prenatal therapeutic target to reverse the abnormal embryonic brain development as well as improve postnatal cognitive function. Moreover, targeting OLIG2 may also regulate the defective myelination process in DS, which as oppose to the rodent brain, starts prenatally in the human brain (Jakovcevski and Zecevic, 2005b; Tosic et al., 2002). Recently, novel small molecule inhibitors of OLIG2 have been developed to treat glioblastoma (Alton and Kesari, 2016; Tsigelny et al., 2017), which may offer opportunities to alleviate the pathogenesis of DS brain. The hiPSC-based models developed in this study provide powerful tools to develop personalized medicine and precision prenatal treatment for DS. Future studies are warranted to explore prenatal DS therapy aiming to normalize OLIG2 expression with small molecules in the brain organoid model and human neuronal chimeric mouse model, complementing ongoing work with DS transgenic mouse models.
EXPERIMENTAL MODEL AND SUBJECT DETAILS
Generation, culture, and quality control of hPSC lines and their neural derivatives
The control and DS hiPSC lines were generated from patient-derived fibroblasts using the “Yamanaka” reprogramming factors, as reported in our previous study (Chen et al., 2014). DS fibroblasts were obtained from Coriell Institute for Medical Research (supplementary Table 2). The hiPSC lines derived from DS patients include two DS hiPSC lines (DS1, female; and DS2, male) and isogenic Di-DS3 and Tri-DS3 hiPSCs that were generated from a single female patient. Previous studies reported the generation of DS isogenic disomic and trisomic subclones either from the same parental iPSC lines during serial passage (Maclean et al., 2012) or from reprogramming mosaic DS fibroblasts (Weick et al., 2013). As reported in our previous study (Chen et al., 2014), we did not observe mosaicism in the fibroblasts from patient DS3 and the disomic subclones were identified during passaging similar to the study (Maclean et al., 2012). The isogenic Di- and Tri-DS3 iPSCs were then isolated and established by clonal expansion. Moreover, all the hiPSC lines derived from DS patients have been fully characterized by performing karyotyping, teratoma assay, DNA fingerprinting STR (short tandem repeat) analysis, gene expression profiling, and Pluritest (www.PluriTest.org), a robust open-access bioinformatic assay of pluripotency in human cells based on their gene expression profiles (Muller et al., 2011), as described in our previous study (Chen et al., 2014). In the current study, fluorescence in situ hybridization (FISH) analysis with a human chromosome 21 (HSA21)-specific probe (Vysis LSI 21 probe; Abbott Molecular) was routinely performed to examine the copy number of HSA21 in the DS and control hiPSC-derived pNPCs, as previously described (Chen et al., 2014). In addition, OLIG1 and OLIG2 relative DNA copy number were examined routinely by using TaqMan Copy Number Assay (Applied Biosystems). Briefly, genomic DNA from DS and control pNPCs was collected by using DNA Extract All Reagents Kit (Applied Biosystems) and gene copy number of OLIG1, OLIG2, and RNase P were examined with primers Hs05549528, Hs02041572 or Copy Number Reference Assay human RNase P, respectively, using TaqMan Gene Expression Master Mix kit (Applied Biosystems). OLIG1 and OLIG2 relative DNA copy number were calculated by normalizing to RNase P DNA copy number, according to the manufacturer’s instructions. The OLIG2-GFP knockin hESC and hiPSC reporter lines were generated using a gene-targeting protocol as reported in our previous studies (Liu et al., 2011; Xue et al., 2009). The hPSCs were maintained under feeder-free condition and cultured on dishes coated with hESC-qualified Matrigel (Corning) in mTeSR1 media (STEMCELL Technologies). The hPSCs were passaged approximately once per week with ReLeSR media (STEMCELL Technologies). All the hPSC studies were approved by the committees on stem cell research at Rutgers University.
Animals
Rag1-/- mice (B6.129S7-Rag1tm1Mom/J, The Jackson Laboratory) were used for cell transplantation. All animal work was performed without gender bias under the Institutional Animal Care and Use Committee (IACUC) protocol approved by Rutgers University IACUC Committee.
METHOD DETAILS
Ventral forebrain organoid generation and culture
To generate ventral forebrain organoids, we used pNPCs as the starting population, because utilization of these neural fate-restricted progenitors has the advantage of efficient differentiation into desired brain regions (Monzel et al., 2017). Similar to previous studies (Mariani et al., 2012; Pasca et al., 2015; Xiang et al., 2017), we did not include Matrigel as supportive matrix in forming organoids, as we did not find significant improvement in the formation of ventral forebrain organoids by Matrigel embedding in our system. Briefly, the pNPCs were derived from hPSCs using small molecule-based protocols (Chen et al., 2016; Li et al., 2011). Neural differentiation in the embryoid bodies (EB) was induced by dual inhibition of SMAD signaling (Chambers et al., 2009), using inhibitors SB431542 (5 μM, Stemgent) and noggin (50 ng/ml, Peprotech) in neural induction medium composed of DMEM/F12 (HyClone) and 1× N2 (Thermo Fisher Scientific) for 1 week. EBs were then grown on dishes coated with growth factor-reduced Matrigel (BD Biosciences) in the medium consisting of DMEM/F12, 1× N2, and laminin (1 μg/ml; Sigma-Aldrich). pNPCs in the form of rosettes developed for another 7 days (week 2). Next, rosettes were manually isolated from surrounding cells and expanded for 7 days (week 3) in pNPC medium, composed of a 1:1 mixture of Neurobasal (Thermo Fisher Scientific) and DMEM/F12, supplemented with 1× N2, 1× B27-RA (Thermo Fisher Scientific), FGF2 (20 ng/ml, Peprotech), human leukemia inhibitory factor (hLIF, 10 ng/ml, Millipore), CHIR99021 (3 μM, Stemgent), SB431542 (2 μM), and ROCK inhibitor Y-27632 (10 μM, Tocris).
Then the expanded pNPCs were dissociated into single cells using TrypLE Express (Thermo Fisher Scientific). 10,000 cells were cultured in suspension in the presence of ROCK inhibitor Y-27632 (10 μM) in ultra-low-attachment 96-well plates to form uniform organoids and then those organoids were transferred to ultra-low-attachment 6-well plates. To pattern these organoids to the fate of ventral forebrain, we treated them with sonic hedgehog (SHH) (50 ng/mL, Peprotech) and purmorphamine (1 μM, Cayman Chem), an agonist of sonic hedgehog signal pathway from week 3 to 5 (Fig. 1A). The media were replenished every day. Starting from week 5, the cell culture plates were kept on orbital shaker with speed of 80 rpm/min. Maximum 8 organoids were cultured in each well of the ultra-low-attachment 6-well plate in neuronal differentiation medium containing a 1:1 mixture of Neurobasal and DMEM/F12, supplemented with 1× N2 (Thermo Fisher Scientific), 1× B27 (Thermo Fisher Scientific), BDNF (20 ng/ml, Peprotech), GDNF (20 ng/ml, Peprotech), dibutryrl-cyclic AMP (1mM, Sigma), and ascorbic acid (200 nM, Sigma) and media was replenished every other day. For 2D neuronal differentiation, 5,000 dissociated cells were plated onto poly-L-ornithine (0.002%, Sigma) and laminin (10 μg ml-1) pre-coated coverslips in the same neuronal differentiation medium. The neurons cultured in the neuronal differentiation medium for 4–6 weeks were used for experiments.
Cell transplantation
The 5-week-old hiPSC-derived ventral forebrain organoids were dissociated and suspended as single cells at a final concentration of 100,000 cells per μl in PBS. The cells were then injected into the brains of P0 rag1 -/- immunodeficient mice (B6.129S7-Rag1tm1Mom/J on a C57BL/6 background, Jackson Laboratories). The precise transplantation sites were bilateral from the midline = ± 1.0 mm, posterior bregma = -1.0 mm, and dorsoventral depth = -1.0 or -1.8 mm (Fig. 4A and supplementary Fig. 5A). The mouse pups were anesthetized by placing them on ice for 5 minutes. Once cryoanesthetized, the pups were placed on a digital stereotaxic device (David KOPF Instruments), equipped with a neonatal mouse adaptor (Stoelting). The pups were then injected with 1 μl of cells into each site by directly inserting Hamilton needles through the skull into the target sites. The pups were weaned at 3 weeks and were kept up to 6 months before they were tested for the engraftment of human cells.
RNA isolation and qPCR
Total RNA was prepared from organoids with RNAeasy kit (Qiagen) (Chen et al., 2014). Complementary DNA was prepared with a Superscript III First-Strand kit (Invitrogen). The qPCR was performed with Taqman primers (Life Technologies) on an ABI 7500 Real-Time PCR System. All primers used are listed in supplementary Table 4. All experimental samples were analyzed and normalized with the expression level of housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Relative quantification was performed by applying the 2-∆∆Ct method(Chen et al., 2014).
ChIP
Chromatin immunoprecipitation (ChIP) was performed using Magna-CHIPTM A Chromatin Immunoprecipitation Kit (Millipore, USA) according to the manufacturer’s instructions. Briefly, after being washed with cold PBS, the 5-week-old organoids were incubated in PBS with 1% formaldehyde at room temperature for ten minutes to cross-link chromatin proteins to genomic DNA. Glycine was used to quench unreacted formaldehyde. Following cold PBS washing, organoids were resuspended in lysis buffer with protease inhibitor Cocktail II and subjected to bioruptor sonication to shear DNA length to 200 -1000 bp. 5 μg of anti-OLIG2 (Phosphosolutions 1538), anti-IgG (Kit included) or anti-Acetyl-Histone H3 antibody (positive control, included in the kit) and Protein A magnetic beads were added to the chromatin solution and incubated overnight at 4°C with rotation. After washing and elution, crosslinks of the protein/DNA complexes were reversed to free DNA by the addition of proteinase K and incubation at 62°C for 2 h followed by further incubation at 95°C for 10 min. Then, the DNA fragments were purified and used for PCR. The primers targeting human GAPDH was provided by the kit, working as the positive control. All primer sequences were listed in supplementary Table 5. The PCR products were analyzed by agarose gel electrophoresis.
RNA sequencing
Organoids (5-week old) generated from hiPSC lines derived from two DS patients (patient DS1 and DS3 and corresponding cell lines include DS1, control1, isogenic Tri-DS3 and Di-DS3, Tri-DS3+ContshRNA, and Tri-DS3+OLIG2shRNA hiPSCs) were used for RNA-sequencing. In order to reduce technical variations among different batches of cultures, a large number of organoids (total 40-60 organoids for each cell line) collected from three batches of organoid cultures were pooled together for RNA extraction and sample preparation.
Total RNA was prepared with RNAeasy kit (Qiagen) (Chen et al., 2014) and libraries were constructed using 600 ng of total RNA from each sample and the TruSeqV2 kit from Illumina (Illumina, San Diego, CA) following manufacturers suggested protocol. The libraries were subjected to 75 bp paired read sequencing using a NextSeq500 Illumina sequencer to generate approximately 30 to 36 million paired-reads per sample. Fastq files were generated using the Bc12Fastq software, version 1.8.4. The genome sequence was then indexed using the rsem-prepare-reference command. Each fastq file was trimmed using the fqtrim program, and then aligned to the human genome using the rsem-calculate-expression (version 1.2.31) command to calculate FPKM (fragments per kilobase of transcript per million mapped reads) values.
In order to analyze the transcripts, FPKM > 1 was set as cutoff to filter transcripts. |Fold change | > 1.5 was set as criteria to filter differential expressed genes (DEGs). Upregulated and downregulated DEGs were mapped to each chromosome and the number of DEGs in each chromosome was counted. Functional enrichment analysis was performed using DAVID online tool (Huang da et al., 2009a, b) and visualized using GraphPad. Heatmaps were generated using pheatmap package of R. Correlations with Brainspan data were performed in R using only genes common to both experiments. The average of the Spearman correlation coefficient is presented for all the combinations of organoid sample replicates or Brainspan samples matching the labeled condition.
Immunostaining and cell counting
Mouse brains and organoids fixed with 4% paraformaldehyde were processed and cryo-sectioned for immunofluorescence staining (Liu et al., 2011; Pasca et al., 2015). The primary antibodies were listed in supplemental Table 7. Slides were mounted with the anti-fade Fluoromount-G medium containing 1, 40,6-diamidino-2-phenylindole dihydrochloride (DAPI) (Southern Biotechnology). Diaminobenzidine (DAB) immunostaining was performed with antibody against GFP or a human-specific marker STEM121 as described in our previous study (Chen et al., 2016; Jiang et al., 2013b). Images were captured with Zeiss 710 confocal microscope. The analysis of fluorescence intensity was performed using ImageJ software (NIH Image). The relative fluorescence intensity was presented as normalized value to the control group. The cells were counted with ImageJ software. For organoids, at least three fields of each organoid were chosen randomly and at least 16 organoids from each group were counted. For brain sections, at least five consecutive sections of each brain region were chosen. The number of positive cells from each section was counted after a Z projection and at least 7 mice in each group were counted. Engraftment efficiency and degree of chimerization were assessed by quantifying the percentage of hN+ cells among total DAPI+ cells in sagittal brain sections, as reported in the previous studies (Chen et al., 2015; Han et al., 2013). The cell counting was performed on every fifteenth sagittal brain section with a distance of 300 μm, covering brain regions from 0.3 to 2.4 mm lateral to the midline (seven to eight sections from each mouse brain were used).
Electron microscopy
The selected vibratome sections that contained DAB-labeled cells were post-fixed and processed for electron microscopy (EM) as described in our previous studies (Chen et al., 2016; Jiang et al., 2016; Jiang et al., 2013a). EM images were captured using a high-resolution charge-coupled device (CCD) camera (FEI).
Western blotting
Western blotting was performed as described previously (Xu et al., 2014). Lysates from organoids or human brain tissue were prepared using RIPA buffer and the protein content was determined by a Bio-Rad Protein Assay system. Proteins were separated on 4-12% SDS-PAGE gradient gel and transferred onto nitrocellulose membrane. Then the membrane was incubated with antibodies (supplemental Table 7). Appropriate secondary antibodies conjugated to HRP were used and the ECL reagents (Pierce ECL Plus Western Blotting Substrate, Thermoscientific) were used for immunodetection. For quantification of band intensity, blots from at least three independent experiments for each molecule of interest were used. Signals were measured using ImageJ software and represented as relative intensity versus control. β-tubulin was used as an internal control to normalize band intensity. The human brain tissues used for western blotting analyses are de-identified by encoding with digital numbers and obtained from University of Maryland Brain and Tissue Bank which is a Brain and Tissue Repository of the NIH NeuroBioBank. All the human brain tissues were derived from the cerebral cortex of patients at the ages from 186 to 339 days (supplementary Table 3). Additional information on the human brain tissue, including sex, race, and post mortem interval (PMI), was also provided.
Coomassie blue staining
Similar to previous studies involving human tissue for western blotting (Blair et al., 2016), we performed denatured SDS-PAGE and Coomassie blue staining using the human samples, prior to examining the expression of proteins of interest. The sample preparations that exhibited the absence of smear (little or no degradation) and sharp resolution of protein bands were used for experiments. SDS-PAGE gradient gel was stained by staining solution (0.1% Coomassie Brilliant Blue R-250, 50% methanol and 10% glacial acetic acid) with shaking overnight and was de-stained using de-staining solution (40% methanol and 10% glacial acetic acid) for 1 hour.
shRNA knockdown experiments
OLIG2 shRNA (sc-38147-V), non-targeting control shRNA (sc-108080), and copGFP (sc-108084) lentiviral particles were purchased from Santa Cruz Biotechnology. OLIG2 shRNA lentiviral particles carry three different shRNAs that specifically target OLIG2 gene expression. Undifferentiated hiPSCs were infected with lentivirus and selected with puromycin treatment. The optimal concentration of puromycin for the selection of transduced hiPSC colonies was established by puromycin titration on the specific hiPSC line used for the experiments. Puromycin was added at a concentration of 0.5 ?g/ml for two weeks to select transduces hiPSCs. Stable hiPSCs were then used for neural differentiation with puromycin kept in all the media at 0.3 ?g/ml as previously described (Mariani et al., 2015). Knockdown of OLIG2 mRNA expression was confirmed by qPCR after SHH and Pur treatment.
FACS
Sorting of OLIG2+/GFP+ cells and data acquisition were performed in a BD FACS Calibur (BD Bioscience). Acquired data were subsequently analyzed by calculating proportions/percentages and average intensities using the FlowJo software (Treestar Inc)(Jiang et al., 2013b).
Cell fractionation
The isolation of cell nuclear and cytosolic fractionation was performed by using Nuclear/Cytosolic Fractionation Kit (Cell Biolabs Inc, AKR171). The lysates of fractions were subjected to western blotting analyses. β-tubulin was used as the cytosolic marker and phospho-Histone H3 was used as the nuclear marker.
Open field test
Mice were placed into a white Plexiglas chamber (60 long×60 wide×30 cm high) under dim ambient light conditions and activity was monitored for 5 min in a single trial. Four squares were defined as the center and the 12 squares along the walls as the periphery. The movements were monitored by an overhead video camera and analyzed by an automated tracking system (San Diego Instruments, CA). Results were averaged for total distance traveled and number of entries into the center of the field(Belichenko et al., 2009; Ichiki et al., 1995).
Elevated plus maze
The elevated plus maze was made of a black Plexiglas cross (arms 40 cm long × 6 cm wide) elevated 55 cm above the floor. Two opposite arms were enclosed by the transparent walls (40 cm long × 15 cm high) and the two other arms were open. A mouse was placed in the center of the apparatus facing an enclosed arm and allowed to explore the apparatus for 5 min. The total traveling distance and the time spent on the open arms was scored (Walf and Frye, 2007).
Novel object recognition
As previously described (Belichenko et al., 2015; Bevins and Besheer, 2006), two sample objects in one environment were used to examine learning and memory with 24 hours delays. Before testing, mice were habituated in a black Plexiglas chamber (40 long×30 wide×30 cm high) for 10 minutes on 2 consecutive days under ambient light conditions. The activities were monitored by an overhead video camera and analyzed by an automated tracking system (San Diego Instruments, CA). First, two identical objects, termed as ‘familiar’, were placed in the chamber, and a mouse was placed at the mid-point of the wall opposite the sample objects. After 10 minutes to explore the objects, the mouse was returned to the home cage. After 24 hours, one of the ‘familiar’ objects used for the memory acquisition was replaced with a ‘new’ object similar to the ‘familiar’ one. The mouse was again placed in the chamber for 3 minutes to explore the objects; The total traveling distance and the time spent investigating the objects was assessed. The preference to the novel object was calculated as Time exploring new object / (Time exploring new object + Time exploring familiar object).
Vector construction, transfection, and luciferase reporter assay
Promoter regions of human DLX1 (from TSS upstream -10 to -1250 bp) or human LHX8 (from TSS upstream 0 to -2000 bp) were amplified by PCR from human genomic DNA extracted from control hiPSC-derived pNPCs with the primers shown in supplementary Table 6. Then promoter regions were inserted into the pGL3-Basic luciferase-reporter vector (Promega, USA). Control pNPCs were used for luciferase reporter assay. The pNPCs were seeded at 3X105 per well in 24 well plates one day before transfection. The cells were co-transfected with 0.3 μg firefly luciferase-reporter construct (DLX1-pGL3, LHX8-pGL3 or pGL3-Basic), 0.02 μg pRL-TK Renilla luciferase-reporter plasmid (Promega, USA), and 0.6 μg pLenti-Olig2-IRES-GFP vector or control vector using Lipofectamine 3000 (Invitrogen, USA). The luciferase activity was examined using a dual-luciferase reporter assay system (Promega, USA) according to the manufacturer’s instructions, and the signal was normalized to the internal Renilla control for transfection efficiency.
Data analysis
All data represent mean ± s.e.m. The organoid experiments were replicated at least three times and each experiment was performed in triplicate using organoids derived from the three pairs of DS and control hiPSCs. When only two independent groups were compared, significance was determined by two-tailed unpaired t-test with Welch’s correction. When three or more groups were compared, one-way ANOVA with Bonferroni post hoc test was used. A P value less than 0.05 was considered significant.
The analyses were done in GraphPad Prism v.5.
Competing Financial Interests
The authors declare no competing financial interests.
Acknowledgements
This work was in part supported by grants from the NIH (R21HD091512 and R01NS102382 to P.J.), an Institutional Development Award (IDeA) from the National Institute of General Medical Sciences of the NIH under grant number P30GM110768 (to P.J.), and Edna Ittner Pediatric Research Support Fund (to P.J.). Y.L. was supported by the start-up fund from The Vivian L. Smith Department of Neurosurgery, McGovern Medical School, the University of Texas Health Science Center at Houston, Memorial Hermann Foundation-Staman Ogilvie Fund (to Y.L), the Bentsen Stroke Center (to Y.L), the TIRR Foundation through Mission Connect (014-115 to Y.L.), and the Craig H. Neilsen Foundation (338617 to Y.L.). R.P.H. was supported by R01ES026057, R01AT008458, and R21DA039686. We thank DNA Sequencing Core at UNMC for performing RNA-seq experiment, which receives partial support from the National Institute for General Medical Science INBRE - P20GM103427-14 and COBRE - 1P30GM110768-01 grants as well as The Fred & Pamela Buffett Cancer Center Support Grant - P30CA036727. We thank the Bioinformatics and Systems Biology Core at UNMC for providing RNA-seq data analysis services, which receives support from Nebraska Research Initiative and NIH (2P20GM103427 and 5P30CA036727). We thank Dr. Minhan Ka from UNMC for technical assistance with ChIP-PCR assay and thank Dr. Alice Liu from Rutgers University and Dr. Guang Yang from Beckman Research Institute of City of Hope for suggestions and technical assistance with luciferase reporter assay.