Impaired Neural Stress Resistance and Loss of REST in Bipolar Disorder

Abstract

Neurodevelopmental changes and impaired stress resistance have been implicated in the pathogenesis of bipolar disorder (BD), but the underlying regulatory mechanisms are unresolved. Here we describe a cerebral organoid model of BD that exhibits altered early neural development, elevated neural network activity, and a major shift in the transcriptome. These phenotypic changes were reproduced in cerebral organoids generated from iPS cell lines derived in different laboratories. The BD cerebral organoid transcriptome showed highly significant enrichment for gene targets of the transcriptional repressor REST. This was associated with reduced nuclear REST and REST binding to target gene recognition sites. Reducing the oxygen concentration in organoid cultures to a physiological range ameliorated the developmental phenotype and restored REST expression. These effects were mimicked by treatment with lithium. Reduced nuclear REST and derepression of REST targets genes was also observed in the prefrontal cortex of BD patients. Thus, an impaired cellular stress response in BD cerebral organoids leads to altered neural development and transcriptional dysregulation associated with downregulation of REST. These findings provide a new model and conceptual framework for exploring the molecular basis of BD.

INTRODUCTION

Bipolar Disorder (BD) is among the most common major psychiatric disorders with a prevalence of 1–4% and typical onset in young adulthood [1]. Thus far 64 genome-wide significant loci have been associated with elevated risk for BD in extended Genome-wide association studies (GWAS) studies[2]. However, these genetic variants only modestly affect overall risk (Odds Ratio <1.3), suggesting that gene-environment interactions are likely to play a major role[3]. Perinatal risk factors such as cesarean section delivery, maternal influenza infection, maternal smoking, and adverse childhood life events have been implicated in the etiology of BD[4–6]. Furthermore, neuroimaging studies provide evidence for early structural brain abnormalities, including reduction of cortical thickness and volume in multiple neocortical and limbic brain regions, suggesting cell loss and atrophy[7]. Altered response to cellular stress, including oxidative or endoplasmic reticulum stress are potential contributors to cell loss in BD, and may underlie changes in neuronal connectivity[8–11]. However, the mechanistic basis of altered cell resilience and neuronal vulnerability in BD is unresolved.

The transcriptional repressor REST/NRSF plays a key role in early brain development. REST suppresses the expression of genes associated with terminal neuronal differentiation to facilitate precise developmental timing and neural lineage specification[12, 13]. REST is reactivated in neurons of the aging human brain, and regulates the expression of a broad range of gene targets that modulate neural excitation and maintain neuronal viability[12, 14]. REST can also be induced by hypoxia[15] and a variety of stress-related factors, and protects against oxidative and proteostatic stress[12].

Here we describe a cerebral organoid model of BD that exhibits an abnormal induction of synaptic and neural signaling genes resulting in increased neuronal network activity. This was accompanied by a novel neurodevelopmental phenotype characterized by altered structure of developing neural rosettes and impaired neural stress resistance. The neurodevelopmental changes were rescued by culturing cerebral organoids at a lower more physiological oxygen level or treatment with lithium. RNA sequencing showed major changes in the BD transcriptome and transcription factor analysis showed highly significant enrichment for gene targets of the REST transcriptional repressor. This was associated with reduced nuclear REST levels and target gene binding. Furthermore, reduced nuclear REST and derepression of REST targets genes was observed in the prefrontal cortex of adult patients with BD. Thus, impaired stress resistance in the cerebral organoid model of BD is associated with altered neural development and transcriptional dysregulation, which may be mediated in part by downregulation of REST.

MATERIALS AND METHODS

Study Design

The aim of this study was to establish a cerebral organoid model of BD to identify and study disease relevant phenotypes. We performed unbiased RNA sequencing (RNAseq) analysis and analyzed overall structure and morphology of cerebral organoids. Sample size was not calculated prior to experiments. Initial experiments were performed on 5 control and 7 bipolar disorder iPSC lines and later reproduced in additional 5 controls and 5 bipolar disorder iPSCs from different sources. All experiments were repeated 3–5 times by seeding and growing several batches of cerebral organoids from all available lines. Outliers were determined by first, gross morphological observation of organoids, as some lines randomly do not produce viable organoids in each batch and second, by statistical ROUT analysis. RNAseq data is available on Gene Expression Omnibus (GEO) database (Accession number: GSE234795)

Experimental model and subject details

Induced pluripotent stem cells (iPSCs)

iPSCs were generated from human fibroblast cells obtained from different sites. Details of iPSC lines are listed in Table 1.

Table 1.

iPS cell lines

Line	Cohort/Study ID	Age	Sex	Diagnosis
PSC-01–017	Yankner/Tsai	39	M	CONTROL
PSC-01–020	Yankner/Tsai	38	M	CONTROL
PSC-01–0241	Yankner/Tsai	36	F	CONTROL
PSC-01–0262	Yankner/Tsai	24	F	CONTROL
PSC-01–028	Yankner/Tsai	55	F	CONTROL
BD-220–5	Yankner/Tsai	49	F	BD
BD-25–3	Yankner/Tsai	47	M	BD
BD-81–1	Yankner/Tsai	35	F	BD
BD-PSC-01–002	Yankner/Tsai	46	F	BD
BD-PSC-01–003	Yankner/Tsai	51	M	BD
BD-193–3	Yankner/Tsai	32	F	BD
BD-12–33	Yankner/Tsai	35	F	BD
MH0185855	NRGR Study ID 163	22	F	BD
MH0185853	NRGR Study ID 163	28	F	BD
MH0185880	NRGR Study ID 163	35	M	BD
MH0185869	NRGR Study ID 163	26	M	BD
MH01858652	NRGR Study ID 163	24	F	CONTROL
MH01858631	NRGR Study ID 163	36	F	CONTROL
MH0185983	NRGR Study ID 163	29	M	CONTROL
MH0185932	NRGR Study ID 163	19	M	CONTROL
MH0165228	NRGR Study ID 130	39	M	CONTROL
MH0173440	NRGR Study ID 130	62	F	BD

¹These iPS cell lines may be derived from the same donor.

²These iPS cell lines may be derived from the same donor.

Human prefrontal Cortex

Human postmortem paraffin-embedded sections were acquired from the NIH NeuroBioBank at the University of Miami and the Sepulveda Research Corporation. Information on individual cases is listed in Table 2.

Table 2.

Human brain samples

Identifier	Origin	Age	Gender	PMI (hours)	Neurophathology Diagnosis
5702	University of Maryland Brain and Tissue Bank	57	f	33	Bipolar Disorder
5279	University of Maryland Brain and Tissue Bank	43	m	14	Bipolar Disorder
4331	University of Maryland Brain and Tissue Bank	51	f	22	Bipolar disorder, depression
494	University of Maryland Brain and Tissue Bank	37	m	10	Bipolar Disorder, manic depression
A594	University of Maryland Brain and Tissue Bank	37	m	11	Bipolar Disorder
1227	University of Maryland Brain and Tissue Bank	52	m	16	Control
1936	University of Maryland Brain and Tissue Bank	46	m	13	Control
1428	University of Maryland Brain and Tissue Bank	45	m	14	Control
604	University of Maryland Brain and Tissue Bank	43	m	15	Control
1266	University of Maryland Brain and Tissue Bank	42	m	15	Control
544	University of Maryland Brain and Tissue Bank	38	m	11	Control
5621	University of Maryland Brain and Tissue Bank	37	f	9	Control
1104	University of Maryland Brain and Tissue Bank	35	m	12	Control
A66	University of Miami Brain Endowment Bank	66	m	15	Control

Method details

iPSC reprogramming and Culture

iPSCs were generated from human fibroblast cells obtained from different sources using CytoTune Sendai Reprogramming Vectors (Thermo Fisher, Catalog number A1378001). iPSCs were cultured feeder-free on Geltrex LDEV-Free Reduced Growth Factor Basement Membrane Matrix (Gibco) coated plates in mTeSR^™1 or mTeSR^™ Plus Basal Medium (Stem Cell Technologies). The iPSC lines underwent stringent quality control to confirm pluripotency and a normal karyotype. This included alkaline phosphatase assay, karyotype analysis and differentiation into three germ layers.

Generation of Cerebral Organoids

Cerebral organoids were differentiated using an approach adapted from a previously described protocol with modifications [16]. Briefly, 9000 hiPSCs were seeded in each well in a Corning^® 96-well clear round bottom ultra-low attachment microplate (Corning #7007) to form embryoid bodies (EBs) in the presence of 50μM ROCK inhibitor (Sigma #Y-27632). When EBs reached 600 μm in diameter, medium was replaced with neural induction medium for 4 days before embedding EBs in Matrigel (Corning #354234). Embedded EBs were grown in ultra-low attachment plates in differentiation medium without vitamin A for 5 days without shaking. Differentiation medium with vitamin A was used for long term maintenance with shaking on an orbital shaker (90rpm). Medium was changed every 2–3 day.

Treatment of Cerebral Organoids:

Treatment with 200 μM or 500 μM Lithium Carbonate: Two weeks from DIV32-DIV46. Both treatments showed similar results, but organoids supplemented with 500 μM Lithium started to show signs of degeneration, thus we decided to use Lithium concentration of 200 μM, which is comparable to concentrations found in cerebrospinal fluid of BD patients treated with Lithium[17].

Calcium Imaging

Poly(lactide-co-glycolide) copolymer (PLGA) fiber microfilaments were integrated into organoids as a floating scaffold to improve neuroectoderm formation and cortical development as well as maintenance, during long term cultures[18]. After 2 months of culturing, we sectioned PLGA containing organoids to generate acute sections that were cultured at the air-liquid interface. These sections show greatly improved survival and a neuronal tract morphology similar to intracortical and subcortical projections[19]. The live-sections were allowed to mature until day in vitro 160 prior to imaging and were then treated with Fluo-4 calcium indicator dye to performed live cell spinning disk microscopy-based calcium indicator imaging of intact organoid sections. Constrained nonnegative matrix factorization extended (CNMF-E) methods for calcium signal processing were applied to the recorded videos to extract spiking dynamics, which enables unbiased categorization of single-cell calcium dynamics into functional microcircuit clusters[20].

Immunofluorescence Microscopy

Organoid samples were collected on day in vitro 45 and fixed in sterile 4% (vol/vol) paraformaldehyde (PFA) solution. Organoid samples were then equilibrated in sterile 15% (vol/vol) PBS-buffered sucrose solution in 4°C for 72 hours, then transferred into sterile 30% (vol/vol) PBS-buffered sucrose solution at 4°C. Organoids were embedded in optimal cutting temperature (OCT) compound (Sakura Finetek, Tokyo, Japan) and snap-frozen in liquid nitrogen. Sectioning of 20 μM slices was performed by cryostat (Microtome Cryostat, HM505E). Sections were dried at 37°C for 30 mins, then rehydrated with PBS and quenched with 125mM glycine. Sections were blocked and permeabilized with 3% v/v normal donkey serum prepared in 0.3% v/v Triton X-100 in PBS. Sections were then washed with PBS and incubated overnight with primary antibody in blocking solution at 4°C in a humidified chamber (Table 3). After washing, donkey anti-Rabbit IgG (H+L) highly cross-adsorbed secondary antibody, Alexa Fluor Plus 488, 594 and 647 at 1:2000 in blocking solution was added and then incubated with 1X TrueBlack^® lipofuscin autofluorescence quencher (Biotium #23007) and mounted using ProLong^™ Diamond Antifade Mountant (ThermoFisher Scientific #P36961).

Table 3:

Antibodies and concentrations used in this study

Name	Host	Company	Cat.#	Comments
Actin	RB	Sigma	A2066	WB 1:2000
Brachyury	G	R&D	AF2085	ICC 1:250
DCX	RB	CST	4604S	ICC 1:1000
OTX2	G	R&D	AF2086	ICC 1:250
SOX17	G	R&D	AF2087	ICC 1:250
NANOG	RB	CST	4903T	ICC 1:1000
Nestin	M	Millipore	MAB5326	ICC 1:2000
NeuN	M	Millipore	MAB377	ICC 1:150
REST	RB	Millipore	07–579	ICC 1:1000 WB 1:500
REST	RB	Bethyl	IHC #00141	1:100
TRA-1–81	M	CST	TRA-1–81	ICC 1:1000
β-III-Tubulin	M	CST	4466S	ICC 1:1000
ChIPAb REST	RB	Millipore	14–641
TBR1 (D6C6X)	RB	CST	49661	ICC 1:350
Ki-67 (8D5)	M	CST	9449	ICC 1:250
SOX2	G	R&D	AF2018-SP	ICC 1:250
MAP2	G	PhosphoSolutions	1099-MAP2	ICC 1:500
OCT4	RB	CST	2840T	ICC 1:500

Neural rosettes were imaged with a Nikon Ti2 inverted microscope with a W1 Yokogawa Spinning disk scan head with 50 μm pinholes and a Toptica 4 laser launch and a Andor Zyla 4.2 Plus sCMOS monochrome camera. Images were captured using NIS AR 5.02 acquisition software, and 12-bit gain 4 camera setting. All images were thresholded based on the no-primary antibody negative control for all staining. For determining REST REST intensity in SOX2 or NEUN nuclei, the thresholded SOX2/NEUN channel image was converted into 8-bit image to accommodate particle analysis within 5–200 μm² area and 0.1–1.0 circularity parameters. The area for each SOX2/NEUN nucleus was measured. These regions of interest were then used to identify intensity of REST on original, unthresholded 12-bit neural rosette images to calculate the arbitrary intensity of nuclear REST in SOX2+ or NEUN+ cells. The architecture analysis, TBR1, SOX2, Ki67 layer thickness and positive cells was analyzed with ImageJ (1.53c) Software. Convoluted background subtraction was performed with BioVoxxel Toolbox plugin with gausian filter. The extracted image then underwent auto-thresholding for foreground and background identification. The outline and lumen of the rosette was identified with the polygon selection tool. The thickness of SOX2 and TBR1 in the rosette was measured as the average of the 3 arrows drawn from the lumen of SOX2 or TBR1 with the previously drawn rosette outline. Images were converted to binary and underwent a watershed process for SOX2, Ki67 and TBR1 cell quantification using particle analysis tool. Positive cells with size range from 5–100 μ² and 0–1 circularity were included into the counting.

Human brain nuclear REST intensity

Paraffin-embedded human prefrontal cortex sections were deparaffinized and rehydrated in a series of xylene and ethanol solutions. The antigen retrieval step was performed in a boiling 1X DIVA Decloaker solution (Biocare Medical #DV2004MX) for 18.5 mins. Non-specific background was blocked with 2% w/v BSA and 0.1% v/v Triton X-100 in PBS, for one hour at room temperature. Subsequently, the sections were incubated with rabbit anti-REST antibody (Bethyl IHC #00141) in blocking solution overnight at 4°C in a humidified chamber. Then, sections were incubated with secondary antibody, Alexa Fluor Plus 488 at 1:200 in blocking solution at room temperature for 2 h. The sections were washed with PBS (4 × 5 mins) before incubated with 1X TrueBlack^® quencher (Biotium #23007) prepared in 70% v/v ethanol for 2 mins and washed with PBS. The sections were stained with DAPI (1:200 diluted in PBS) for 30 mins at room temperature, mounted in the presence of ProLong^™ Diamond Antifade Mountant (ThermoFisher Scientific #P36961).

Three random micrographs were taken per section by using the Olympus Fluoview 1000 confocal microscope. Images were analyzed in Fiji, briefly five random nuclei were manually outlined to calculate the sum of intensity units in selection after subtracting the average background intensity. Details of brain samples used in this study are summarized in Table 2. BD case 5702 showed higher PMI than the other cases, however, it was not classified as an outlier by ROUT analysis (Q = 5%) and therefore included in the analysis.

Quantitative droplet digital PCR (ddPCR)

Gene expression was measured using a TaqMan based assay in the QX200 Droplet Digital PCR System (Bio-Rad) (Table 4). Probe-sets were used in a multiplexed configuration using three targets in one reaction. The plate was then sealed with heat seal foil (Bio-Rad) and droplets were generated using the AutoDG droplet generator (Bio-Rad). The droplet containing plate was then placed in a thermocycler for one-step RT-PCR (reverse transcription at 42°C for 1h, enzyme activation at 95°C for 10 min, and then subjected to 40 cycles of amplification (30 seconds at 95°C and 60 seconds at 55°C) before enzyme deactivation and a final extension of 98°C for 10 minutes. Data acquisition in the QX200 Droplet Reader (Bio-Rad) and data analysis using QuantaSoft Software (Bio-Rad).

Table 4:

BioRad probes used in this study

Target	Dye Quencher	Assay ID
DCX	5’ 6-FAM, 3’ Iowa Black FQ	dHsaCPE5045338
SYP	5’ 6-FAM, 3’ Iowa Black FQ	dHsaCPE5033786
TBP	5’ HEX, 3’ Iowa Black FQ	dHsaCPE5058363
DLG4	5’ 6-FAM, 3’ Iowa Black FQ	dHsaCPE5036504
REST	5’ HEX, 3’ Iowa Black FQ	dHsaCPE5045815
ASCL1	5’ 6-FAM, 3’ Iowa Black FQ	dHsaCPE5027636

Chromatin immunoprecipitation

A total of 3–5 Cerebral organoids were collected for each chromatin preparation and extraction was performed using the SimpleChIP Kit (CST) according to the manufacturer’s protocol with some modifications. First, chromatin was cross-linked with 1.5% PFA at RT for 20 min (1.5% PFA in PBS, supplemented with protease inhibitors (CST)) and crosslinking was stopped by adding Glycine. Fixed organoids were chopped using a razor blade and lysed in the appropriate lysis buffer (CST). DNA was digested using 0.8 uL Micrococcal Nuclease (CST) per IP. Before each IP, chromatin concentration was measured with a NanoDrop 1000 and quality was determined by assessing DNA fragment size by electrophoresis on a 1% agarose gel. For REST ChIP, the REST antibody was incubated with 5–10 ug chromatin over night at 4°C on a rotation device. The ChIP DNA was quantified using (ChIP qPCR primers F-SYN-RE1: GGT GCT GAA GCT GGC AGT and R-SYN-RE1: TGG GTT TTA GGA CCA GGA TG) the Luna Universal qPCR Master Mix according to the manufacturer’s guidelines, and run in a one-step RT-PCR cycle in an iQ5 (Bio-Rad) [denatured at 95° for 60 sec, and then subjected to 45 cycles of amplification (15 seconds at 95°, 30 seconds at 60°)]. The purity of the PCR products was determined by single peak melting curves.

Fluorescence-activated Cell Sorting

Organoids were dissociated into single cells by pipetting through a 1ml tip, centrifuged at 1000xg for 1 min and washed with PBS. Dissociated cells were immediately fixed with 4% w/v PFA for 30 mins on ice, washed and stored in cell storage buffer (1% w/v BSA, 0.5mM EDTA and 0.1% sodium azide in PBS) for immunolabeling.

Western Blotting

Western blot analysis was performed as described previously[21].

REST target gene analysis

REST target genes. The REST RE1 motif position-specific weight matrix MA0138.2 was obtained from JASPAR. FIMO was used with the Homo sapiens genome sequence GRCh38 to predict REST binding sites. A gene was defined as a REST target if it had a RE1 motif with motif pvalue < 1e-7 that was +/−10kb from the transcription start site of any transcript of the gene in the Ensembl GRCh38.86 gene models. This procedure identified 2632 REST target genes before filtering for expressed genes in each data set during gene set enrichment analysis.

Statistical analysis

The student’s two-tailed t-test or an ordinary one-way ANOVA with Tukey or Dunnett’s multiple comparison test was used to analyze parametric data. The Mann Whitney U test was used to analyze non-parametric data. To determine if datasets show a normal distribution, the Shapiro-Wilk test was used. Outlier analysis was performed using the ROUT method.

RESULTS

Altered neuronal development and activity in bipolar disorder cerebral organoids

To gain insight into pathogenic mechanisms in BD, induced pluripotent stem cell (iPSC) lines were established from BD patients and age-matched controls. In addition, iPSC cell lines derived by other laboratories were obtained from the NIMH repository and genome resource (NRGR) https://www.nimhgenetics.org[22–24] (Table 1). The new iPSC lines underwent rigorous quality control including alkaline phosphatase activity, karyotyping and differentiation into all three germline layers (Figure S1A–C). Additionally, expression of the pluripotent stem cell markers Nanog, Oct4 (octamer-binding transcription factor 4) and Tra1–60 were confirmed (Figure S1D). Cerebral organoids generated from the iPSC lines were validated by immunolabeling for the neural progenitor cell marker SRY-Box Transcription Factor 2 (SOX2), the early neuronal markers doublecortin (DCX) and T-Box Brain Transcription Factor 1 (TBR1), and the later neuronal markers NeuN (RNA Binding Fox-1 Homolog 3) and β-III-tubulin (Figure S1E–H).

Cerebral organoids derived from 5 controls and 7 BD iPSC lines were assessed for 90 days and showed no significant differences in gross morphology and size (Figure S2A–C). However, RNA sequencing showed significant transcriptome differences between CTR and BD organoids (Fig. 1A). Gene ontology enrichment analysis of biological pathways (GoBP) of upregulated genes in the BD cerebral organoids showed significant enrichment in categories related to neuronal activity and function. The most significant changes related to synaptic signaling (Fig. 1B). Furthermore, gene ontology enrichment analysis of cellular compartment (GoCC) showed that most upregulated genes are associated with plasma membrane and synapses (Tables S1, S2). Interestingly, this is the same gene category that showed the greatest enrichment among BD risk alleles identified by GWAS[2].

Figure 1. Increased neuronal gene expression and network activity in BD cerebral organoids.

(A) Unsupervised hierarchical clustering of genes differentially expressed between CTR and BD cerebral organoids after 46 days of maturation. Differentially expressed genes (rows) and organoids (columns) were clustered, and gene expression was transformed to a z-score and represented in the heat map. CTR, n=7; BD, n=9, sample IDs provided in Table 1. (B) GO biological process (GoBP) groups enriched in differentially expressed genes in BD compared to CTR organoids. (C) Representative image of two live-sections from cerebral organoids at DIV60 (left) and representative image showing live calcium imaging in live sections. (D) Mean firing rate of neurons that do not participate in synchronous firing events (Un-synchronized) and neurons that participate in synchronous, burst firing (Synchronized). Data is shown from 4–6 field of views, from two independent organoid preparations from CTR, n=3 and BD, n=3. (E-F) Width and Interval in seconds of each firing peak of synchronized neurons derived from data collected in (D). (G) Amplitude of individual firing peaks of synchronized neurons derived from data collected in (D). *P <0.05, **P <0.01, ***P<0.001

Neuronal activity was examined in the organoid model using live-cell calcium imaging. Viable sections of BD and CTR organoids were generated and calcium imaging was performed (Fig 1C). Constrained nonnegative matrix factorization extended (CNMF-E) methods for calcium signal analysis were applied to the recorded videos to extract spiking dynamics, which enables unbiased categorization of single-cell calcium dynamics into functional microcircuit clusters[20]. We allowed sections to mature until day in vitro (DIV) 160 when neuronal network activity was most pronounced, as indicated by repetitive waves of nearly complete synchronization of calcium transients. Comparison of the mean firing rate of calcium transients in synchronized and un-synchronized neurons in CTR and BD cerebral organoids showed more frequent, synchronized neural network activity in BD organoids (Fig. 1D). Further characterization of synchronization clusters showed significantly increased peak duration, reduced interval, and elevated amplitude indicative of elevated neural activity (Fig. 1E–G). Thus, BD organoids exhibit hyperactive neural network activity (Fig. 1E–G), raising the possibility that this system might model the manic phase of BD.

Altered development and degeneration in bipolar disorder cerebral organoids

We next explored the cytoarchitecture of developing BD and control cerebral organoids. Immunolabeling for the neural progenitor cell (NPC) marker SOX2, the cell proliferation marker Ki67, and the early neuronal marker TBR1 demonstrated the distinctive structural features of neural rosettes (NRs), a prominent structure of cerebral organoids analogous to the developing neural tube which gives rise to mature neuronal cell types [16]. BD organoids showed a markedly increased number of NRs with abnormal structure relative to controls (Fig. 2 A–C). These abnormal structures were characterized by loss of the stereotypical cell layers with diffuse admixture of cell types that are typically spatially segregated (Fig. 2 B). The abnormal NRs were interspersed with a smaller number of NRs with intact structure in BD organoids, as determined by quantitative analysis of the TBR1 and SOX2 cellular layers (Figures S2D–F). These findings were reproduced using organoids generated from iPSC lines derived in two independent studies[22, 24] that were obtained from the NIMH repository and genome resource (NRGR) (Table 1). Both newly derived and NRGR iPSC lines were used for subsequent organoid analysis. Despite the abnormal NR structure in BD, overall numbers of TBR1, Ki67 and SOX2 positive cells did not differ significantly from control organoids (Figure S2 G–I). These results suggest a potential neurodevelopmental vulnerability associated with the development of BD.

Figure 2. Abnormal neural development in BD cerebral organoids.

(A) Representative SOX2 (green) localization and organization in BD cerebral organoids showing normal and abnormal neural rosettes (NR). The yellow box marks the area for higher magnification images. Scale bar, 20 μm (B) Representative localization and organization of SOX2 (red), TBR1 (green) and Ki67 (white) in CTR and BD cerebral organoids. Yellow circle marks the NR lumen (C) Representative quantification of abnormal rosettes in CTR and BD organoids. Data represent the mean ± SEM percentage of abnormal NR from 3 organoids from one line. Lines used in this representative experiment CTR, n=4; BD, n=5. *P <0.05, **P <0.01, ***P<0.001

Oxygen-related stress and mitochondrial function in BD organoids

Several lines of evidence suggest that mitochondrial function may be altered in BD[25–28]. In addition, increased mitochondrial activity and number has been reported in an iPSC-derived neuronal model of BD[29]. To begin to assess mitochondrial function, we examined mitochondrial genes in the organoid RNA-seq analysis. We found many significantly upregulated genes that encode proteins essential for mitochondrial respiration (Figure 3A), consistent with increased mitochondrial mass as indicated by significantly increased protein levels of the structural protein mitochondrial translocase of outer membrane (TOM20) and the mitochondrial inner membrane component ATP synthase F1 subunit alpha (ATP5A - CV) (Figure 3B). These results suggest that mitochondrial function may be altered in BD organoids.

Figure 3. Oxygen-related stress and mitochondrial function in BD organoids.

(A) Gene expression (logFC) of nuclear and mitochondrially encoded genes in BD vs CTR cerebral organoids from RNA-Seq data shown in Fig. 1A and Table S2. (B) Western blot and densitometric analysis using TOM20 and ATP5A antibodies in CTR and BD organoids in 20% O₂ conditions. (C) Gene expression (logFC) of nuclear and mitochondrially encoded genes in BD-5% O₂ vs CTR cerebral organoids at 20% O₂ from RNA-Seq data. (D) Flow cytometric analysis of JC-1 Red/Green Ratio in CTR and BD under standard and 5% O₂ conditions. (E) Flow cytometric analysis of DCFDA relative mean fluorescent intensity (MFI) in CTR and BD under standard and 5% O₂ conditions. (F) Measurement of abnormal neural rosettes in CTR and BD under standard and 5% O₂ conditions. *P <0.05, **P <0.01, ***P<0.001

Most cells are exposed to oxygen partial pressures equivalent to 1–6% O₂ in vivo, while standard cell culture conditions expose cells to hyperoxic conditions of 18–21% O₂. Such supraphysiological O₂ levels in cell culture impact many O₂ consuming reactions and cell viability[30], and culturing cells at physiological O₂ levels was shown to reduce oxidative, metabolic, and ER stress, and reduce DNA damage[30–33]. To examine the role of oxygen-related stress on mitochondrial gene expression and organoid development, we reduced the O₂ level to 5% to achieve a more physiological O₂ concentration. Assessment of mitochondrial gene expression in physiological O₂ conditions showed significant downregulation of metabolic and mitochondrial ribosomal genes in BD when compared to CTR and BD at 20% O₂ (Figure 3 C and Figure S2J–K). Notably, we could not detect differentially expressed mitochondrial genes in BD vs CTR organoids at 5% O_2, suggesting an altered gene expression in response to supraphysiological O₂ levels in BD. We then assessed mitochondrial function in cerebral organoids by measuring the mitochondrial membrane potential (MMP) using the JC-1 assay (Figure 3D). Flow cytometric analysis showed no difference between BD and CTR organoids but a significant reduction of MMP for both in low oxygen conditions (Figure 3D). In addition, we could not detect any significant differences in reactive oxygen species as measured by the 2’,7’-dichlorodihydrofluorescein diacetate (H₂DCFDA) assay (Figure 3E). However, examination of organoid rosette structure showed that low O₂ completely rescued the abnormal neural rosette phenotype in BD cerebral organoids (Figure 3F). Thus, supraphysiological O₂ levels lead to an altered metabolic response and neurodevelopmental changes in BD but not CTR organoids.

Loss of REST in BD organoids is rescued by low oxygen

To identify regulatory factors that mediate gene expression changes in BD cerebral organoids, we assessed enrichment of transcription factor binding sites in affected genes using ENCODE chromatin immunoprecipitation assays followed by sequencing (ChIP-seq) libraries. The most significantly predicted transcription factors for genes upregulated in BD were REST and the Polycomb Repressive Complex 2 (PRC2) components Enhancer of zeste homolog 2 (EZH2) and Polycomb Repressive Complex 2 Subunit (SUZ12) (Figure 4A, S3). Notably, REST interacts with EZH2 and SUZ12 and recruits the PRC2 complex to target genes[34, 35]. The most significantly predicted transcription factors targeting downregulated genes in BD are Myc and the repressor E2F Transcription Factor 4 (E2F4).

Figure 4. Reduced REST expression and activity in BD cerebral organoids.

(A) Transcription factor prediction based on enrichment in BD differentially expressed genes using the ENCODE ChIP-seq database. Extended list provided in Table S3. Predictions are stratified for genes that were upregulated or downregulated. (B) Immunocytochemistry of SOX2 (red), NeuN (magenta) and REST (green) in DIV45 cerebral organoids. The circles indicate the outlines for SOX2+ and NeuN+ staining used to quantify REST signal intensity. Scalebar 10 μm. (C, D) Quantitative measurements of REST expression in SOX2 (C) and NeuN (D) positive cells. (E) REST ChIP-PCR (REST antibody Millipore # 17–641) was performed for a SYN1 RE1 site in BD organoids. Data represent the mean ± SEM from CTR, n=3; BD, n=3. ***P<0.001 by Student’s t test. (F, G) Quantification of nuclear REST expression in NEUN (F) and SOX2 (G) positive cells after immunocytochemistry and confocal imaging of SOX2, NeuN, and REST in DIV45 cerebral organoids in 20% and 5% O₂ culturing conditions. *P <0.05, **P <0.01, ***P<0.001 by one-way ANOVA.

REST has been shown to protect against neurodegeneration and plays a central role in the regulation of neuronal differentiation[12, 13]. Enrichment of REST targets in genes upregulated in BD organoids predicts loss-of-function of the REST transcriptional repressor. To confirm this prediction, nuclear REST was assessed in neural progenitor cells (NPCs) and neurons in CTR and BD organoids, the predominant cell types in cerebral organoids at this stage. Nuclear REST levels were significantly reduced in both cell populations in BD organoids relative to controls (Fig. 4B–D). REST mRNA expression was not changed, suggesting posttranscriptional regulation of REST (Figure S3A). REST function was further assessed by ChIP-PCR with primers targeting the RE-1 site of the REST target gene synapsin 1 (SYN1). There was significantly reduced REST-RE1 site binding in BD organoids relative to controls (Figure 4E). Thus, BD cerebral organoids exhibit reduced REST nuclear translocation and RE1 site binding. Importantly, culturing BD organoids in the low oxygen condition restored REST expression (Figure 4F, G). This finding highlights the sensitivity of neurodevelopmental trajectories to O₂ levels, raising the possibility that oxygen fluctuation during preterm birth might predispose to neuropsychiatric disorders later in life[36].

Lithium treatment rescues abnormal NR structures in BD cerebral organoids

The mood-stabilizer lithium is the first-line treatment for managing BD and its neuroprotective and neurotropic effects are well documented in a variety of cellular models[37]. Thus, we explored whether lithium ameliorates the structural changes in BD organoids by treatment of control and BD organoids with 200 μM Li₂CO₃ and assessment of organoid cytoarchitecture. Lithium did not alter the phenotype of control organoids, but reduced the number of abnormal NR structures resulting in non-significant differences between BD and CTR organoids (Figure 5A).

Figure 5. Lithium modulates the development and neural activity of BD cerebral organoids.

(A) Quantification of abnormal rosettes in cerebral CTR and BD organoids treated with 200 μM Li₂CO₃ for 14 days. Data represent the mean ± SEM from rosettes from 6 organoids from CTR/CTR- Li₂CO₃, n=5; BD/BD- Li₂CO₃ n=7 lines. *P <0.05, **P <0.01, ***P<0.001 by One-Way ANOVA with Tukey multiple comparison correction. (B) GO biological process (GoBP) groups enriched in differentially expressed genes in BD compared to CTR organoids before and after lithium treatment. CTR/CTR- Li₂CO₃ n=4; BD/BD- Li₂CO₃, n=4 lines. (C) Average firing rate (Hz) from fluo-4 calcium imaging of synchronously firing clusters in cerebral organoid live-sections in untreated CTR and BD, and BD after treatment with 200 μM Li₂CO₃ for 1 week. *P <0.05, **P <0.01, ***P<0.001 by One-Way ANOVA

We then performed RNA sequencing to investigate the gene ontology pathways most affected by lithium. BD organoids exhibit upregulation of genes involved in synapse function and neural development. These gene expression changes were selectively reversed by lithium (Figure 5B). In contrast, pathways that are downregulated in BD organoids, including protein targeting, did not change upon lithium treatment (Figure 5B). These results suggest that lithium reduces the expression of genes involved in neuronal development and synaptic function in BD organoids. Consequently, we investigated whether lithium treatment could normalize the abnormal neuronal network activity observed in BD organoid live-sections. Calcium imaging and analysis of the average firing rate of control and BD organoids with or without treatment with Li₂CO₃ revealed a complete restoration of synchronized neuronal activity upon lithium treatment (Figure 5C).

REST Expression in the BD Brain

These observations led us to analyze REST levels in postmortem prefrontal cortex of BD patients and age-matched controls by quantitative immunofluorescence microscopy (Table 2). A decrease in nuclear REST levels was detected in neurons of the prefrontal cortex in every BD case examined relative to controls (Figure 6A, B and S3B). REST mRNA expression did not significantly differ, suggesting posttranscriptional dysregulation of REST in the BD prefrontal cortex (Figure 6C and S3C). Expression of the canonical REST target genes ASCL1 (Achaete-Scute Family BHLH Transcription Factor 1) and PSD95 (Postsynaptic density protein 95) was elevated in BD, consistent with reduced activity of the REST repressor (Figure 6D, E). Thus, REST function may be reduced in the brain of patients with BD.

Figure 6. REST expression is reduced in the prefrontal cortex in BD.

(A) Representative immunostaining from postmortem human prefrontal cortex (BA11) from control and BD patients stained with antibodies against REST. Scale bar, 20 μm. (B) Quantification of nuclear REST intensity from (A) using ImageJ (1.53c) software. Data represent the mean ± SEM from 5 fields per section. CTR, n=9; BD, n=5 (Table 2). Analysis from individual samples is shown in Figure S3A. (C) Expression of REST mRNA in postmortem human PFC from CTR and BD individuals. Expression was determined by ddPCR and normalized to TBP. Data represent the mean ± SEM from CTR, n=5; BD, n=4. Analysis from individual samples shown in Figure S3B. (D-E) Expression of ASCL1 and PSD95 mRNA in postmortem human PFC from CTR and BD individuals as measured by ddPCR and normalized to TBP. (F-G) Enrichment of REST target genes among candidate BD risk loci identified by GWAS. Shown are cell lines in which REST target genes were identified by ChIP-seq in ENCODE. P-values represent the enrichment for genome-wide significant BD risk loci among these REST targets in each cell line. Data is shown from two different GWAS analyses - Mullins et al., 2021 (2) (F), and Stahl et al., 2019 (20) (G). The dotted lines indicate P-value = 0.05. For B-E, *P <0.05 by Student’s t test.

Genome-wide association studies have identified an increasing number of significant associations between BD and genetic variants. The most recent comprehensive meta-analysis identified 64 autosomal loci that achieved genome-wide significance and 161 protein-expressing genes associated with these risk loci[2]. Interestingly, a number of genome-wide significant BD risk variants are in well-established REST target genes involved in neural development and the regulation of synaptic function, including the Sodium channel protein type 2 (SCN2A) and the neuron-specific microRNA miR-124. The Calcium Voltage-Gated Channel Subunit Alpha1 C (CACNA1C), a predicted REST target based on ENCODE ChIP-seq, has been strongly linked to both BD and schizophrenia[38]. Unbiased analysis of the ENCODE ChIP-seq database showed highly significant enrichment for REST target genes among genes associated with genome-wide significant risk loci for BD from meta-analysis of 41,917 BD and 371,549 control cases[2] (Figure 6F). Enrichment for REST target genes was confirmed in another study that identified 30 genome-wide significant variants[39] (Figure 6G). Thus, REST may regulate the expression of many genes associated with candidate risk loci for BD.

Discussion

Perinatal risk factors and the observation of structural changes in the brains of patients with BD suggest that early alterations in neural development may contribute to the later onset of disease. To explore neurodevelopmental changes associated with BD, we established a cerebral organoid model using patient-derived iPS cell lines. Transcriptome analysis of BD organoids showed upregulation of synaptic signaling and neurodevelopmental genes. Moreover, BD organoids showed neuronal hyperactivity characterized by increased burst firing frequency and spike amplitude, consistent with altered neural network activity. These findings are consistent with neural hyperexcitation observed in neurons differentiated from BD iPS cells[29] raising the possibility that neural network instability and hyperexcitation associated with mood instability in BD can be modeled in iPSC-derived neurons and cerebral organoids.

Cerebral organoids generated from iPSC lines derived by different laboratories from multiple BD patients reproducibly exhibited a novel neurodevelopmental phenotype characterized by abnormal neural rosette structure. Brain imaging studies have shown structural cortical abnormalities associated with BD, including reduced thickness and volume of prefrontal and temporal cortical regions and the cingulate gyrus[40–43]. Moreover, a recent longitudinal study provided evidence for progressive temporal cortical thinning during the course of BD[43]. In addition, this study demonstrated a fixed non-progressive frontal cortical change, as well as progressive thinning of the inferior frontal cortex in patients with multiple manic episodes.

Functional imaging studies of brain activity in BD show altered neural activity in multiple brain areas. Increased activity in the left dorsal anterior cingulate and the left head of caudate were shown during mania, while subcortical/limbic hyperreactivity was found during depression and in the remitted phase[44–50]. Thus, both pre-existing and progressive structural and functional changes can appear in multiple brain regions in BD, and may relate to the altered neural development and function observed under conditions of oxygen-related stress in the cerebral organoid model.

Culturing BD organoids at standard room oxygen concentration resulted in increased mitochondrial gene and protein expression. However, neurons in vivo are exposed to much lower levels of oxygen. Oxygen partial pressures in the brain are in the range of 1–6% O₂, whereas oxygen partial pressures in cell culture are considerably higher in the range of 18–21% O₂[51]. Such supraphysiological O₂ levels in cell culture settings impact many O₂ consuming reactions and cell viability, and culturing cells at physiological O₂ levels has been shown to reduce oxidative, metabolic and endoplasmic reticulum-related stress, as well as DNA damage[30–32]. Interestingly, culturing organoids at 5% O₂ led to reduced mitochondrial gene expression and mitochondrial membrane potential. Moreover, these conditions prevented the appearance of abnormal neural rosettes. Mitochondrial dysfunction may play a key role in the pathophysiology of BD as components of the mitochondrial electron transport chain are down regulated in the postmortem frontal cortex. In addition, several magnetic resonance spectroscopy (MRS) studies showed increased lactate and decreased phosphocreatine indicating reduced mitochondrial function[27, 52–54]. Our observations suggest that BD cerebral organoids are more sensitive to oxygen-related cellular stress with reduced mitochondrial gene expression, similar to observation in the brains of adults with BD. These findings raise the possibility that altered neural development may be precipitated by early stress-related events during perinatal development, as well as later stages of brain maturation, in individuals destined to develop BD.

Transcription factor enrichment analysis showed significant enrichment of REST targets among genes that were upregulated in BD, consistent with loss-of-function of the REST repressor. This was confirmed in BD organoids and human BD prefrontal cortex, which showed reduced nuclear REST levels, reduced REST-RE1 site binding in chromatin, and elevated REST target gene expression. REST plays a key role in brain development by regulating neural stem cell function and the timing of terminal neuronal differentiation[13, 55]. Loss-of-function of REST in BD organoids would therefore be predicted to accelerate neural differentiation, consistent with upregulation of neural and synaptic genes and increased neuronal electrical activity in BD organoids. However, this acceleration might disrupt the normal temporal sequence of events and ordered layering of neurons, consistent with the disrupted structure of neural rosettes observed in BD organoids. Altered expression of neurodevelopmental genes has also been observed in another study of iPSC-derived neural progenitor cells from BD patients[56].

REST has been shown to be neuroprotective in the adult brain by repressing genes involved in apoptosis and by activating genes that promote stress resistance and neuronal survival[12, 57]. Lithium is a REST activator that has been used as a primary treatment for BD[12]. Another study has identified REST as a regulator of transcriptional networks in response to lithium and another mood stabilizer, sodium valproate, in a tissue culture model[58]. Notably, lithium completely prevented abnormal differentiation in BD cerebral organoids. These findings raise the possibility that drugs which target the REST repressor might have therapeutic efficacy in BD.

The role of neurodegenerative processes in the pathogenesis of BD is unresolved. BD has been associated with a 3-fold elevated incidence of Alzheimer’s disease[59]. Moreover, lithium reverses the elevated risk of dementia in BD patients, raising the possibility of a neuroprotective mechanism[60]. Similar to BD, REST expression is reduced in Alzheimer’s disease with loss-of-function in the prefrontal cortex and hippocampus[12, 21]. Thus, dysregulation of REST may predispose to BD or Alzheimer’s disease, depending on developmental stage and the affected neuronal populations.

The identification of multiple genetic variants associated with increased risk of BD provides an opportunity to define underlying molecular pathways of pathogenesis. Many of the recently identified genome-wide significant genetic variants are known or predicted REST target genes. These risk loci are associated with genes that are not only regulated by REST but also, in turn, regulate REST expression, such as the neuron-specific microRNA miR-124. Predicted REST-regulated BD risk variants are involved in neural development and the regulation of neurotransmission and synaptic function. Thus, REST may be a regulatory component of a gene network that contributes to BD through control of neural activity. In this regard, loss-of-function of REST can globally elevate cortical neural activity[14], potentially predisposing to mood instability. The reduced nuclear REST in both cerebral organoids and postmortem prefrontal cortex from BD patients suggests that it may be involved in both early and late stages of the disease.

Abbreviations:

BD

Bipolar Disorder

CTR

control

iPSC

induced pluripotent stem cells

NPC

neural progenitor cell

NR

Neural Rosettes

GoBP

Gene Ontology Biological Pathways (GoBP)

GoCC

Gene Ontology Cellular Compartment

GWAS

Genome-Wide Association Studies

DIV

day in vitro

MRS

Magnetic Resonance Spectroscopy (MRS)

ER

Endoplasmic Reticulum

REST

RE1 Silencing Transcription Factor

OCT4

Octamer-Binding Transcription Factor 4

SOX2

SRY-Box Transcription Factor 2

DCX

doublecortin

TBR1

T-Box Brain Transcription Factor 1

NeuN

Neuronal Nuclear Protein

TOM20

Mitochondrial Translocase of Outer Membrane

ATP5A-CV

ATP synthase F1 subunit alpha

PRC2

Polycomb Repressive Complex 2

EZH2

Enhancer of Zeste Homolog 2

E2F4

E2F Transcription Factor 4

SYN1

Synapsin 1

ASCL1

Achaete-Scute Family BHLH Transcription Factor 1

PSD95

Postsynaptic Density Protein 95

SCN2A

Sodium Channel Protein Type 2

CACNA1C

Calcium Voltage-Gated Channel Subunit Alpha1 C

Data and materials availability:

Supplementary information is available at MP’s website