Exploring roles of neuroglobin in oxygen homeostasis and neurodevelopment based on human cerebral organoid models

Summary

Neuroglobin (NGB) is a neuron-enriched globin known for its neuroprotective role under hypoxic and oxidative stress conditions. However, its function in neurodevelopment remains largely unexplored. Using human cerebral organoids (hCOs), we investigate the impact of NGB silencing on oxygen homeostasis, neurogenesis, and cortical organization. The experimental results demonstrate that NGB regulates intra-organoid oxygen tension, supporting neuronal differentiation, and cortical maturation. Silencing NGB disrupts this process, leading to sustained hypoxia, impaired synaptic connectivity, and activation of stress-response pathways, including DNA damage response (DDR), unfolded protein response (UPR), and mTORC1 signaling. Persistent oxidative stress and metabolic reprogramming further drive neural progenitor cells toward senescence, impairing neurogenic trajectories. These cellular and structural abnormalities in NGB-silenced hCOs are similar to pathological features of neurodegenerative diseases. Our study highlights critical roles of NGB in oxygen-dependent neurodevelopmental processes and suggests that targeting cellular senescence and oxidative stress may offer therapeutic strategies for neurodevelopmental and neurodegenerative disorders.

Graphical abstract

Highlights

• •NGB regulates oxygen homeostasis to sustain cortical neurogenesis in hCOs
• •NGB silencing induces persistent hypoxia and disrupts cortical organization
• •Metabolic stress signaling drives cellular senescence in NGB-silenced hCOs
• •Neural network connectivity and functional activity are impaired in NGB-silenced hCOs

Cellular physiology; Functional aspects of cell biology; Developmental biology

Introduction

Studying neurodevelopment is a key to understanding brain formation, disorders, and early interventions. Direct in vivo investigation of neurodevelopment within a human brain is challenging and impractical due to ethical and technical limitations. Human cerebral organoids (hCOs), derived from pluripotent stem cells, serve as a valuable in vitro model that recapitulates key aspects of early brain development, including cellular organization, gene expression, and neural connectivity, in a self-organizing three-dimensional structure.^1,2 The organoids have been known to lack vascularization that limits their applications for various biological studies. However, the avascular structure of the hCOs makes them an ideal model to investigate neurodevelopment before the vascular system is fully established. In the developing brain, vascularization plays a critical role in supplying oxygen to meet the high metabolic demands required for structural and functional maturation. The absence of a functional blood supply in hCOs provides a well-defined model for studying oxygen transport and regulation in maintaining cellular homeostasis during early neurodevelopment.

Oxygen is crucial for brain development because it supports the high metabolic demands of neural stem cell (NSC) proliferation and differentiation, a key process in neurogenesis and cortical maturation.^3,4,5 During the early developmental stage, the placental gas exchange system is not fully established, creating a naturally low-oxygen environment for maintaining a population of NSCs within specialized stem cell niches. At this stage, cerebral vascularization remains sparse, limiting oxygen availability in most brain regions. As development progresses, the maturation of the vascular system in the cerebral cortex ensures sufficient oxygen supply to meet the metabolic demands of expanding neuronal populations. This transitional phase requires precise regulation of oxygen levels to sustain neurogenesis and neuronal maturation.^4,6,7,8 Insufficient oxygen supply during the developmental stage has been linked to a range of neurodevelopmental disorders such as cerebral palsy, epilepsy, and cognitive impairments.⁹ Given these crucial oxygen-dependent processes, understanding how neurons regulate and adapt to fluctuating oxygen levels is essential for elucidating key mechanisms of neurodevelopment.

Neuroglobin (NGB), a neuron-enriched globin protein that has been extensively studied for its role in counteracting oxidative stress and promoting neuronal survival under hypoxic or ischemic conditions.^{10,11,12,13,14} Initially identified as a protective factor in the central nervous system, NGB is highly expressed in neurons and proposed to function as an oxygen reservoir and transporter for facilitating oxygen diffusion.¹⁰ Beyond its role in oxygen transport, NGB scavenges reactive oxygen species (ROS), mitigates mitochondrial dysfunction, and inhibits apoptotic pathways, further reinforcing its neuroprotective role. In line with these findings, experimental models of neurological disorders, including ischemic stroke, traumatic brain injury, and Alzheimer’s disease, support its broader therapeutic potential in neuroprotection.^15,16 While these studies have established NGB’s role in neuronal survival under stress conditions, its function beyond neuroprotection remains little understood, particularly in the context of early neurodevelopment.

Emerging evidence suggests that NGB not only serves as a stress-response protein but also regulates oxygen homeostasis during the neurodevelopment. Recently, hCOs have been utilized as an in vitro model, demonstrating that NGB plays a pivotal role in regulating intra-organoid oxygen levels, particularly during a specific period of development.¹⁷ During this period, NGB expression coincides with increased intra-organoid oxygenation, enhanced neuronal differentiation, and cortical expansion. It has been further revealed that silencing of NGB disrupts this oxygenation pattern, leading to impaired neurogenesis and abnormal cortical organization. These findings demonstrate the utility of the hCOs for investigating the developmental roles of proteins like NGB within a human-relevant context. Furthermore, beyond its conventional neuroprotective role, it also suggests that NGB is a transportation modulator of oxygen-dependent neurodevelopmental pathways, potentially influencing neuronal fate specification, metabolic regulation, and cortical layer formation.

Research on NGB in neurodevelopment is still in its early stages, with its role in neuronal maturation and network formation largely unexplored. Using a multiomic approach, we employ advanced imaging tools, single-cell RNA sequencing (scRNA-seq), and multielectrode array (MEA) electrophysiology to examine impact of NGB on neurodevelopment (Figure 1). Furthermore, we exploit lentiviral vector to introduce NGB shRNA for generating NGB silencing hCOs (shNGB hCOs). With the shNGB hCOs, we investigate how silencing NGB disrupts long-term neurodevelopment by affecting neuronal differentiation, cortical organization, and synaptic activity. Specifically, we assess whether silencing of NGB leads to sustained hypoxia, impaired synaptic function, and activation of stress-response pathways such as the DNA damage response (DDR), unfolded protein response (UPR), and mTORC1 signaling. Beyond its canonical role in antioxidation and neuronal survival, NGB also regulates oxygen tension to ensure an adequate nutrient supply, which in turn facilitates neuronal maturation and cortical network activity. A deeper understanding of NGB’s role in neurodevelopment may provide novel insights into oxygen-dependent regulatory mechanisms and potential therapeutic strategies for neurodevelopmental disorders associated with hypoxia and oxidative stress.

Figure 1.

Experimental workflow and impact of NGB silencing on cerebral organoids development

Top: Schematic timeline illustrating the stages of cerebral organoid generation from human-induced pluripotent stem cells (hiPSCs), including embryoid body (EB) formation, neural induction, fusion and expansion, and maturation. All measurement points for data analysis were denoted.

Bottom: Experimental workflow comparing control hiPSC-derived cerebral organoids with NGB-silenced organoids. Organoids were analyzed using single-cell RNA sequencing (scRNA-seq) for genotypic profiling, FD-FLIM for functional analysis, confocal microscopy for structural assessment, and microelectrode array (MEA) recordings for functional evaluation.

Results

Comparison of cell population in control and shNGB hCOs across developmental stages using transcriptomic analysis

To silence NGB expression, we transduced the human-induced pluripotent stem cell (hiPSC) with a lentiviral vector expressing a short hairpin RNA (shRNA) targeting the NGB gene (shNGB). A corresponding vector targeting the LacZ gene (shLacZ) was used as a negative control for non-specific viral or shRNA-mediated effects. In evaluating the efficiency of NGB silencing, we exploited quantitative real-time PCR (RT-qPCR) for quantification the expression of NGB within hiPSC cells (Figure S1). After calculating and normalization quantification cycle (Cq), we confirm that NGB expression level of the undifferentiated hiPSC cells is low (Figure S1B). In addition, no significant difference in NGB expression between the untreated control cells and the shLacZ cells is observed (Figure S1B). In contrast, NGB expression in the shNGB cells is reduced, even to undetectable level (Figure S1B). Furthermore, we find that the expression level of the Yamanaka factor KLF4,¹⁸ a key marker for hiPSC pluripotency, is comparable across all three cell lines (Figure S1B), indicating the minimal effect of lentiviral transduction process on their pluripotent state. These results confirm that the expression of NGB in the shNGB hiPSCs is silenced prior to their differentiation into the cerebral organoids.

To examine the chronic impact of NGB silencing on the hCOs, we performed scRNA-seq on the days 24, 35, and 65 control and shNGB hCOs. The transcriptomic profiles of the control and shNGB hCOs at the three time points are visualized as a uniform manifold approximation and projection (UMAP) plot and annotated using canonical markers (Figure 2A). A merged UMAP plot further combined these transcriptomic profiles, and the cell population on the UMAP can be segmented into 7 major cell types: dividing radial glial (Dividing RG; GLI3, MKI67, and CCND2) cells, radial glial (RG; GLI3) cells, basal progenitor cells (BP; EOMES, HES6, NEUROD1, and NEUROD2), excitatory neuron (EXN; TBR1, RELN, SLC17A6, SLC17A7, and GRIN2B), interneuron (IN; DLX5 and SST), unclassified neuron (UN), and oligodendrocyte cells (OLG; PDGFRA) based on the representative markers described in previous literature.^19,20 General neuronal state and maturation are tracked with GAP43, RBFOX3, and DLG4. The overall distribution patterns of cell types are presented as density plots (Figure 2B), and the quantitative compositions are shown as Figure 2C. In addition, the expression patterns of the selected marker genes are visualized as violin plots (Figures 2D, 2E, and S2A). The statistical analysis results are visualized as volcano plots with denoted selected marker genes (Figures S2B–S2E; Table S1).

Figure 2.

Cell type identification in the control and shNGB human cerebral organoids (hCOs) at days 24, 35, and day 65 using scRNA-seq

A total of 11,706, 7,187, and 13,572 single-cell transcriptomes from control hCOs and 4,228, 7,187, and 2,910 transcriptomes from shNGB-treated hCOs at days 24, 35, and 65, respectively, were combined for analysis. The data obtained from the control hCO at day 24 are used as the reference.

(A) Combined data mapped into a UMAP plot and colored using the Leiden graph-clustering method. The expression patterns of particular marker genes (GLI3, MKI67, EOMES, TBR1, SLC17A6, SLC17A7, DLG4, DLX5, SST, and PDGFRA) were also shown as UMAP plots. The distinct groups in the control and shNGB hCOs at days 24, 35, and 65 are clustered into radial glial cells (RG 1 to 4), dividing radial glial cells (dividing RG 1 and 2), basal progenitor cells (BP), excitatory neurons (EXN 1 to 4), interneurons (IN), and undefined neurons (UN).

(B) Density plots showed the cell compositions within the control and shNGB hCOs at days 24, 35, and 65, respectively.

(C) The pie charts showed the proportions of different cell types within control and shNGB hCOs at days 24, 35, and 65.

(D and E) Violin plots illustrated the distribution patterns of selective marker genes within the control and shNGB hCOs at days 24, 35, and 65, respectively. Y axes in all violin plots represented the normalized expression values of the marker genes.

(F) A bar graph showed the proportions of RELN⁺/TBR1⁺ Cajal-Retzius neurons, TBR1⁺ deep layer excitatory neurons, and SATB2⁺ upper layer excitatory neurons as percentages of the total excitatory neuron population within control and shNGB hCOs at days 24, 35, and 65.

In the control hCOs, the expression of NGB is detected across the developmental stages (Figure S2F). The progenitor and proliferation markers GLI3 and MKI67 show a slight decline over time (Figure 2D; Table S1). Similar to this molecular trend, we observe progressive decline in RG and dividing RG populations from day 24 to day 65 (Figure 2C). Within the same period, BP cells are increased (Figure 2C), reflecting the expected neurogenic transition. Moreover, neuronal differentiation markers (HES6, NEUROD1) show a progressive increase across developmental stages (Figures 2D and S2A; Table S1). Notably, CCND2 is strongly upregulated by day 65 (Figure S2A; Table S1), consistent with enrichment of asymmetric RG divisions that sustain neurogenic output rather than simple pool expansion.²¹ The trend illustrates an orderly neurogenic process, with progenitor cells gradually differentiating into mature neuronal cell types as development proceeds. Additionally, the presence of OLGs by day 65 in the control hCOs further suggests the onset of gliogenesis.

In comparison, the shNGB hCOs exhibit an altered developmental trajectory, with absent expression of NGB at all stages. On days 24 and 35, a higher proportion of RG cells and fewer differentiated neurons, particularly excitatory neurons, are found compared to those in the control one, indicating delayed differentiation (Figure 2C). By day 65, the RG and dividing RG cell populations in the shNGB hCOs are decreased (Figure 2C). Although the differentiated neurons are increased slightly at this stage, BP cells are nearly absent (Figure 2C). Furthermore, the expression levels of GLI3 and MKI67 are evidently reduced in the shNGB hCOs on day 65 (Figure 2D; Table S1). The results suggest that the impaired generation of excitatory neurons observed at the earlier stages persists and ultimately hinders the formation of mature neurons at the later stage within the shNGB hCOs.

To further investigate the status of neurogenesis, the proportion of excitatory neuron subtypes is analyzed and the expression patterns of the related marker genes (GAP43, RBFOX3, RELN, TBR1, and SATB2) are visualized as violin plots (Figures 2E and S2A). In the control hCOs, RBFOX3 expression increases steadily across the developmental stages (Figure S2A; Table S1) that agrees with the ongoing neuronal differentiation. In contrast, GAP43 shows only modest at day 35 and day 65 in the control hCOs. Additionally, quantification of RELN⁺/TBR1⁺ subplate neurons (Cajal-retzius cells) shows a decreasing trend across the developmental stages, while TBR1⁺ (deep-layer) and SATB2⁺ (upper-layer) neurons are increased (Figure 2F). These results indicate that the sequential progression of excitatory neuron maturation has been established in the hCOs system. In contrast, RBFOX3 level is significantly reduced in the shNGB hCOs at day 24 and day 35 (compared to that in the day 24 control hCOs) and remains decreased at day 65 (compared to that in the day 65 control hCOs), indicating impaired neuronal differentiation (Figure S2A; Table S1). Although the proportion of RELN⁺/TBR1⁺ subplate neurons in the shNGB hCOs shows the similar decreasing trend as the control hCOs, TBR1⁺ cells increase at a slower rate, and SATB2⁺ cells are nearly absent at day 65 (Figure 2F). These findings suggest that silencing of NGB has a chronic impact on neurogenesis, leading to delayed and incomplete maturation of excitatory neurons.

Chronic effects of NGB silencing on hCO structure and cell composition

Given the established role of NGB in oxygen tension regulation, NGB silencing is found to be associated with potential alterations in cortical structure and neuronal differentiation.¹⁷ While NGB is known to be involved in initial neurogenesis, its chronic effects on later neurodevelopment remain unclear. In the experiments, we performed immunofluorescence analysis on NSC and neuronal markers (NESTIN, SOX2, Ki-67, TBR1, and PSD95) at day 65 to assess neurogenesis status (Figures 3A and 3B). The thicknesses and cell populations of the neuroepithelia regions with ventricular zone/subventricular zone-like (VZ/SVZ) and CP-like structures are analyzed from the fluorescence images (Figures 3C′, 3D′, 3E′, and 3F′) following the established methodologies detailed in the literatures.^22,23,24

Figure 3.

Structure, and cell compositions of the control and shNGB hCOs at Day 65

(A and B) Confocal images of hCO slices stained with SOX2, NESTIN, Ki-67, TBR1, and PSD95 proteins and cell nuclei (DAPI) revealed the distribution of neural stem cells and neurons in the hCOs at day 65 (scale bar is 500 μm). Zoom-in confocal images presented the selected C′, D′, E′, and F′ regions (white squares) within the hCOs (scale bar is 100 μm).

In the control hCOs, the neuroepithelial regions appear well-organized with clear layering, including defined VZ, SVZ, and CP-like regions. SOX2⁺ NSCs and Ki-67⁺ proliferative cells are primarily found in the VZ/SVZ-like region, whereas TBR1⁺ and PSD95⁺ neurons are primarily localized to the CP-like region, indicating proper structural organization and differentiation (Figures 3C′, 3D′, 3E′, and 3F′). In contrast, the shNGB hCOs display disorganized cortical zones with significantly reduced thickness in both the VZ/SVZ and CP-like regions (Figures S4A and S4B). The layers appear less defined, indicating the compromised cortical structure within the day 65 shNGB. To exclude lentiviral and selection effects, we generated shLacZ hCOs using the same transduction workflow as the shNGB hCOs. Fluorescence images of the shLacZ hCOs at Day 65 shows similar pattern of structural organization and differentiation (Figures S3A–S3C′, S4A, and S4B) to the control hCOs. These observations indicate that the structural abnormalities observed in the shNGB hCOs are specific to the NGB silencing rather than the lentiviral manipulation.

To further investigate changes in NSC and neuronal populations, cell compositions within the VZ/SVZ- and CP-like regions are quantified. At day 65, the control and shLacZ hCOs shows comparable composition patterns across neuroepithelial layers (Figures S4C–S4E), indicating that lentiviral manipulation alone does not alter the cell compositions. In contrast, we find a higher proportion of SOX2⁺ NSCs in the day 65 shNGB hCOs (Figure S4C). Despite the observed NSC population increase, Ki-67⁺ proliferative cells are significantly reduced in the same region, suggesting a decline in NSC proliferative activity (Figure S4D). In the CP-like region, we observe notable reduction in TBR1⁺ excitatory neurons and PSD95⁺ area in the shNGB hCOs (Figure S4E). The reduction correlates with the diminished thickness of the CP-like region indicating impaired neuronal differentiation and maturation.

To obtain further insights of the progenitor cell activities and neuronal maturation, we conduct an integrated analysis by concatenating the scRNA-seq datasets from the day 24 and day 65 control hCOs, and the day 65 shNGB hCOs. The concatenating transcriptomic profile is subclustered into radial glial cells (RG) and excitatory neurons (EXN) to specifically compare these key cellular populations. Gene set enrichment analysis (GSEA) is performed through GO (gene ontology) database to assess changes in gene sets associated with cell proliferation, differentiation, and maturation. GESA results reveal that high levels of cell proliferation-related gene sets are expressed in the RG subcluster of the day 24 and day 65 control hCOs (Figure S5A), indicating robust proliferation activity. By day 65, the control hCO displays increased expression of gene sets related to neuronal differentiation and maturation (Figure S5B), consistent with normal neurogenesis progression.

In contrast, the RG subcluster in the day 65 shNGB hCO shows a significant downregulation of proliferation-related gene sets (Figure S5A), reflecting an impairment of progenitor cell activity. Similarly, gene sets associated with neuronal differentiation and maturation are downregulated in the EXN subcluster of the day 65 shNGB hCO (Figure S5B). These transcriptomic findings align with the structural abnormalities observed in the captured fluorescence images shown in Figure 3, suggesting that developmental trajectory of excitatory neurons is disrupted and persists into later developmental stages of the NGB silencing hCOs.

To examine impacts of the observed structural and cellular abnormalities on synaptic development, the expression levels of genes related to synapse development of excitatory neurons (DLG4, GRIN2B, and SLC17A7) are examined from the transcriptomic profiles of day 24 and 65 control hCOs, as well as day 65 shNGB hCOs (Figures S2A–S2E; Table S1). Violin plots and volcano plots reveal a progressive upregulation of these genes in the day 65 control hCOs compared to the day 24 control hCOs, reflecting normal synaptic maturation. However, in the day 65 shNGB hCOs, GRIN2B and SLC17A7 are significantly downregulated compared to the day 65 control hCOs (Figures S2A–S2E; Table S1). This genetic variation is consistent with the observed structural abnormalities, indicating a failure in synaptic maturation at the later stage.

Pathway activity analysis reveals stress responses and cellular senescence in the shNGB hCOs at later developmental stages

To comprehensively investigate the impact of NGB silencing on cellular homeostasis at later developmental stage, the concatenating transcriptomic profile is subjected to the GSEA and analyzed through MsigDB database for comparison of pathway activity across RG and EXN subclusters. The query results are visualized as a heatmap to display the relative expression levels of the pathway activity (Figure 4A).

Figure 4.

Analysis of pathway activity in the day 24 and 65 control and day 65 shNGB hCOs by scRNA-seq

The combined data were subclustered in to radial glial cell (RG) and excitatory neuron (EXN) groups for detailed comparison analysis.

(A) Datasets from both control and shNGB groups subject to the database. The expression levels of the gene set associated with pathway activity were visualized by a heatmap.

(B) Expression patterns of representative marker genes for hypoxia, reactive oxygen species pathway (ROS), and mammalian target of rapamycin (mTOR) were visualized by a heatmap.

(C) The quantitative expression of each representative gene to its overall expression level for mRNA abundance observation (red: highest expression and blue: lowest expression).

Since the hypoxic environment has been observed to be established in the early-stage shNGB hCOs,¹⁷ we first explore alterations in stress response pathways (Figure 4B). In the control hCOs, hypoxia-related gene set exhibits consistent expression in both RG and EXN subclusters on days 24 and 65, showing a stable normoxic environment. However, these genes are upregulated in the day 65 shNGB hCOs (Figure 4C), confirming the establishment of sustained hypoxic environment. This hypoxic stress is accompanied by upregulation of glycolysis-related pathways, a key metabolic adaptation to oxygen deprivation.²⁵ Representative genes such as ENO1, PGAM1, and PGK1 are strongly expressed in the shNGB hCOs (Figure S6), indicating a metabolic adaptation to chronic hypoxic stress.

To direct evaluate intra-organoids oxygen tension, we applied frequency-domain FLIM (FD-FLIM) with oxygen-sensitive CPOx microbeads to measure oxygen tension within the hCOs from week 3–7 (Figure S7). In the control hCOs, we observe steadily increased oxygen tension from week 4 to week 7 (Figure S7). Similarly, the shLacZ hCOs show a comparable intra-organoids oxygen tension trend and magnitude (Figure S7), indicating that the lentiviral transduction does not affect oxygen dynamics. In the shNGB hCOs, oxygen tension remains low from weeks 4–6 and exhibits a modest increase by week 7 (Figure S7), while oxygen tension remains significantly lower than those observed in the control and the shLacZ hCOs (Figure S7). The results indicate that the NGB knockdown induces a prolonged deficit in internal oxygen with only partial recovery by week 7, which is consistent with the hypoxia signatures detected at later stages.

The sustained hypoxic stress leads to the activation of ROS pathways, further exacerbating cellular stress.²⁶ In the day 65 shNGB hCOs, both RG and EXN subclusters display increased expression of ROS-related gene set compared to the control hCOs (Figures 4A and 4C), confirming elevated oxidative stress levels. To counteract oxidative damage and maintain redox balance,²⁷ we find that the expression levels of antioxidant genes, including PRDX5, TXN, and SOD1, are increased accordingly in the day 65 shNGB hCOs. Together, these results indicate that the elevated expression of antioxidant markers in the day 65 shNGB hCOs reflects an adaptive response to counteract excessive ROS accumulation, a key driver of cellular damage and microenvironmental destabilization.

Along the increase of oxidative stress, pathways associated with DNA repair (DDR) and unfold protein response (UPR) are also activated to counteract the elevated environmental stress.^28,29,30 In the day 65 shNGB hCOs, we observe the increased expression levels of DDR- and UPR-related genes compared to the control hCOs (Figure S6). Both RG and EXN subclusters exhibit upregulation of DNA repair genes such as POLR2A, RNMT, and GTF2F1, indicating a high demand for genomic maintenance (Figure S6). In addition, the day 65 shNGB hCOs display increased expression levels of representative UPR genes, including HSPA5, DDIT4, and SEC31A, in both RG and EXN subclusters (Figure S6). As a consequence of DNA damage response pathway activation,³¹ p53-related genes are also upregulated in the day 65 shNGB hCOs, reflecting an attempt to restore cell cycle progression and genomic stability (Figure S8). These results suggest that the shNGB hCOs undergo a global stress adaptation response to maintain genomic stability and proteostasis at later developmental stages.

Following the prolonged stress adaptation, mTORC1 signaling is activated as a metabolic regulator, integrating nutrient availability, energy status, and cell growth control.³² In the day 65 shNGB hCOs, mTORC1-related genes are upregulated in both RG and EXN subclusters (Figures 4B and 4C). Representative genes associated with mTORC1 signaling, such as HSP90B1, SCD, and SQLE, show increased expression in the shNGB hCOs (Figure 4C). While mTORC1 activation is essential for regulating growth and metabolism under environmental stress, sustained activation of mTORC1 has been linked to premature aging-related processes.^33,34 As stress responses persist, cells tend to enter cellular senescence, a state characterized by permanent cell-cycle arrest and the secretion of inflammatory mediators known as the senescence-associated secretory phenotype (SASP).³⁵ Consistent with this progression, we observe upregulation of SASP-related genes in the shNGB hCOs (Figure S8). To confirm the onset of cellular senescence at the phenotypic level, we conducted SA-β-gal staining using senescence green (Figure 5). The results reveal a higher proportion of senescent cells in the shNGB hCOs, consistent with the SASP enrichment in the transcriptomic data. These results indicate prolonged environmental stress accelerates the transition toward cellular senescence in the shNGB hCOs.

Figure 5.

Analysis of cellular senescence within control and shNGB hCOs at day 65 Confocal images of hCO slices stained with SOX2 and cell nuclei (DAPI) revealed the distribution of neural stem cells and overall structure within the control, shLacZ, and shNGB hCOs at day 65

The senescence green staining represented the level of senescence within the hCOs (scale bar is 400 μm).

Reduced spontaneous electrical activity of day 65 shNGB hCOs

To evaluate the functional consequences of NGB silencing on the hCOs, we performed spontaneous electrical activity recordings using a multi-electrode array (MEA) system (MED64 system, Alpha MED Scientific Inc., Ibaraki, Japan) on Day 65. Brightfield images (Figure 6A) confirm proper sample placement on the chips equipped with planar microelectrodes recording neural activity across the control and shNGB hCOs. Within the 5-min recording interval, analysis of local field potential (LFP) traces (Figure 6B, top) reveals significantly reduced frequency of spontaneous neural activity from the shNGB hCO compared to the control one, indicating diminished cortical network functionality. Line chart of spike frequency (Figure 6B, middle) shows consistently lower activity levels in the shNGB hCOs throughout the 5-min recording interval. Additionally, heatmaps visualizing the spatial distribution of mean spike rates across all recording electrodes (Figure 6B, bottom) demonstrate broad reductions in neural activity in the shNGB hCO compared to the control one. This observation is further supported by quantitative analysis, which shows a significantly lower spike rate in the shNGB hCOs compared to the control one (Figure 6C). The MEA results indicate the impairment of neuronal differentiation and maturation in the NGB silencing hCOs causes disruption of synaptic connectivity and leads to global reduction in neural network functionality.

Figure 6.

Analysis of spontaneous electrical activity in control and shNGB hCOs at day 65 using multi-electrode array (MEA) recordings

(A) Brightfield (BF) images demonstrating the MEA recording setup with 64 planar microelectrodes used to record spontaneous electrical activity in day 65 control and shNGB hCOs (scale bar is 150 μm).

(B) Spontaneous electrical activity traces and analyses in day 65 control and shNGB hCOs. Recordings were taken in 5-min intervals for each experimental hCO. Top: representative local field potential (LFP) traces taken from a short time interval in the day 65 control and shNGB hCOs, with baseline threshold set between ±10 μV. Middle: line charts showing spike frequency over the 5-min recording period for control and shNGB hCOs. Bottom: heatmap plots illustrating mean spike rates recorded by each microelectrode, demonstrating spatial distribution of electrical activity across the MEA for control and shNGB hCOs at day 65.

(C) Quantification of electrical activity in control and shNGB hCOs at day 65. The spike rate was measured as the number of spikes per second. Independent t test was performed for statistical analysis. Data were presented as boxplots and mean (SD) with all data points (n = 4, ^∗∗p ≤ 0.01).

Discussion

Human iPSCs-derived cerebral organoids (hCOs) have been wildly used in studying progression of neural development and maturation, as they closely mimic the physiological structures and biochemical responses of tissues.^1,2,36 During early developmental stage, we have demonstrated that elevated oxygen tension during a specific time period promotes neurogenesis to expand necessary neuronal population for constructing an intact organoid structure.¹⁷ A well-oxygenated microenvironment further supports synaptic maturation and facilitates neural connectivity during the development of the organoids. In contrast, silencing of NGB disrupts the process, attenuating the timed elevation of oxygen tension and establishing a sustained hypoxic environment within the hCOs.

Sustained hypoxia is a significant environmental risk factor that impairs brain development by disrupting neuronal proliferation, differentiation, and maturation.³⁷ In the developing cerebral cortex, hypoxia has been associated with structure deficits, including reduced neuronal density, delayed synaptogenesis, and disrupted myelination.⁹ These structural abnormalities are often accompanied by increased oxidative stress, apoptosis, and proteostasis imbalances, ultimately leading to long-term functional deficits such as impaired motor coordination and cognitive abilities.³⁸ Our study demonstrates that NGB silencing disrupts neuronal differentiation and synaptic connectivity in the day 65 shNGB hCOs.

NGB has been found as an oxygen-binding globin producing protective effect against hypoxic damage in neurons, especially for mature neural tissues. Previous studies highlight NGB as a hypoxia sensor that mounts a reactive response to environmental stressors, such as ischemic stroke, traumatic brain injury, or oxygen deprivation.^11,12,16,39 Under these conditions, NGB scavenges ROS, mitigates mitochondrial dysfunction, and inhibits apoptosis thereby safeguarding neurons and prioritizing survival during acute stress. This protective role is particularly evident in mature neural tissues, where NGB’s upregulation under hypoxia serves as an adaptive mechanism to limit cellular damage and maintain homeostasis.^12,40

In contrast, the role of NGB in neurodevelopment remains largely unexplored. It has been identified that NGB plays an important role in regulating oxygen homeostasis to support the metabolic demands of neurogenesis during a specific period.¹⁷ This period coincides with rapid cortical expansion and the transition from progenitor proliferation to neuronal maturation, emphasizing the role of NGB in maintaining optimal oxygen tension in shaping brain structure. Unlike its protective function in mature neurons in response to environmental challenges, the function of NGB in neurodevelopment appears to maintain optimal oxygen conditions for supporting neuronal differentiation and cortical maturation. This dual role suggests that NGB is not just a stress-response protein but also a crucial regulator of oxygen dynamics, ensuring both the progression of neurodevelopment and the long-term integrity of neural networks.

Electrophysiological trajectories are key indicators of neurodevelopmental maturation. In the experiments, the trajectories within organoids captured using multi-electrode array (MEA) systems offer a powerful tool for assessing synaptic maturation and network connectivity.^41,42 Under normal developmental conditions, increasing neuronal firing rates indicate the progressive establishment of functional synaptic networks. The experimental MEA data within the 5-min recording interval from the day 65 control hCOs reveals robust neuronal activity that agrees to previous studies.^41,42 The results confirm the development of functional neuronal networks in our organoid model. In contrast, the day 65 shNGB hCOs exhibit significantly reduced neuronal firing rates suggesting the impaired synaptic connectivity. Transcriptomic profiles further supported this observation showing a marked downregulation of synapse-related genes including DLG4, SLC17A7, and GRIN2B. The agreement between electrophysiological and transcriptomic findings highlights the utility of multiomic approaches in identifying developmental disruptions based on the hCO models. The results are capable of providing valuable insights into the molecular mechanisms underlying impaired neurogenesis and network functionality.

Furthermore, several interconnected pathways are activated to alleviate cellular damage triggered by exceed oxidative stress.^28,29,30 An imbalance in ROS activity deteriorate protein misfolding within the endoplasmic reticulum (ER), leading to unfold protein response (UPR) activation.²⁹ In our study, transcriptomic profile illustrates prominently upregulation of UPR related gene set in the day 65 shNGB hCOs. The observation indicates establishment of persistent environmental stress within the hCOs. Concurrently, the upregulation of DNA damage response (DDR)-related genes is observed in the day 65 shNGB hCOs, aligning with previous findings that associate stress responses with hypoxia-induced damage.^28,30 Persistent activation of DNA damage response enforces cell-cycle arrest to prevent the propagation of damaged cells.³⁰ Our findings of reduced proliferation and differentiation capacity in progenitor cells of the day 65 shNGB hCOs demonstrates that the sustained DDR signaling enforces cell-cycle arrest to prevent the propagation of damaged cells.

Cellular senescence and quiescence both involve cell-cycle arrest, but they lead to distinct outcomes. Quiescence is a reversible pause in proliferation that preserves stem cell reservoirs and allows re-entry into the cell cycle under favorable conditions.⁴³ In contrast, senescence is an irreversible arrest triggered by stress, such as oxidative stress or DNA damage.⁴⁴ Cells in this state remain metabolically active but lose their proliferative ability, ultimately contributing to aging and dysfunction.^{30,45,46,47,48} The transition between senescence and quiescence is regulated by key signaling pathways, notably the p53 and mTORC1 pathways, which response to elevated environmental stress.^49,50 Strong activation of p53 generally induces reversible quiescence, whereas partial p53 activation coupled with mTORC1 activity promotes irreversible senescence.^51,52,53

In the developing cerebral cortex, mTORC1 is a key signal for neural stem cells (NSCs) to exit self-renewal and commit to neuronal differentiation, regulating the balance of stem cell expansion and neuronal differentiation to maintain the structural integrity.⁵⁴ It also drives dendritic growth and synaptic development, both critical for functional cortical circuits.⁵⁵ Our findings in the day 65 shNGB hCOs, however, present a paradoxical outcome. Despite the upregulation of mTORC1-related genes, we observed a higher population of NSCs with diminished proliferation activity and a corresponding decrease in the neuronal population. This outcome is inconsistent with the canonical pro-neurogenic role for mTORC1 and suggests a stalled differentiation process. Consequently, we propose that the prolonged cellular stress from NGB silencing, including chronic hypoxia and amplified oxidative stress, leads to a pathological and persistent activation of the mTORC1 pathway. This is directly supported by identification of the persistent mTORC1 signaling as a hallmark of cellular senescence.³³ The sustained activation of stress response pathways, particularly p53 and mTORC1, instead of promoting neurogenesis, contributes to NSC divergence into an irreversible, senescent-like state under prolonged environmental stress. This progression impairs neurogenesis and results in structural deficits in cortical development.

Neurogenesis persists into adulthood, supporting neuronal population dynamics and plasticity.²³ However, this process declines with age and is further accelerated by neurodegenerative diseases, including Alzheimer’s and Parkinson’s diseases.^56,57 In neurodegenerative conditions, sustained cellular senescence and the accumulation of senescence-associated secretory phenotype (SASP) are commonly observed in the cerebral cortex.^48,58 The impaired neurogenic process, reduced synaptic connectivity and stress-induced senescent acceleration observed in the day 65 shNGB hCOs closely resemble key pathological features of neurodegenerative diseases. These findings suggest that targeting cellular senescence and reducing oxidative stress can help restore neurogenic trajectories in NGB-silenced hCOs, offering a potential therapeutic strategy for neurodegenerative disorders.

In conclusion, our findings demonstrate that NGB plays a pivotal role in maintaining oxygen homeostasis during neurodevelopment, supporting neuronal differentiation, cortical organization, and synaptic maturation. Silencing NGB in the hCOs disrupts oxygen dynamics, leading to sustained hypoxia and amplified environmental stress, which in turn activates multiple stress response pathways, including DDR, UPR, p53, and mTORC1 signaling. This chronic stress response drives progenitor cells toward senescence, resulting in reduced proliferative capacity, impaired differentiation, and structural disorganization. The upregulation of antioxidant gene expression observed in the shNGB hCOs is consistent with the homeostatic response to the chronic oxidative stress and hypoxia induced by the NGB silencing. However, our findings suggest that this elevated expression represents a compensatory failure rather than a successful adaptation. While the system attempts to mitigate the stress by increasing the expression of genes, such as PRDX5, TXB, SOD1, and MAP1LC3B, this response is ultimately insufficient to prevent the long-term cellular damage. The progressive decline in neurogenic activity and the induction of irreversible cellular senescence demonstrates that the protective mechanisms were overwhelmed. Therefore, the increased expression of antioxidant genes serves as molecular signature of a system failing to cope with chronic stress, leading to compromise cortical structure.

The cellular and structural deficits observed in the shNGB hCOs closely resemble those seen in neurodegenerative diseases, underscoring the critical role of oxygen regulation in sustaining long-term neuronal function. By elucidating the mechanistic link between sustained hypoxia, stress-induced senescence, and impaired neurogenesis, the study highlights potential therapeutic targets for mitigating hypoxia-related neurodevelopmental impairments. Future research should explore strategies to modulate these pathways, restore progenitor cell function, and support proper cortical development, with implications for both neurodevelopmental and neurodegenerative disorders.

Limitations of the study

Our study employs hCOs to model oxygen homeostasis during neurodevelopment. Although the absence of a functional vascular system provides a controlled environment to investigate oxygen diffusion within organoids,¹ it does not fully recapitulate the complex neurovascular interactions present in vivo.⁵⁹ In addition, while our organoids model reflects early cortical development, it lacks certain non-neural cell types, such as microglia, which may affect cellular stress responses.⁶⁰

Methodologically, the oxygen measurements were performed using FD-FLIM on a widefield inverted microscope with LED-based excitation. This configuration substantially reduces phototoxicity compared to the laser-scanning confocal microscopy, which is critical for maintaining organoid viability⁶¹; however, it lacks the optical sectioning capability. Consequently, the spatial resolution of oxygen measurements using CPOx beads is limited, particularly in larger or more matured organoids where increased tissue thickness and light scattering degrade the signal fidelity.

Furthermore, the transcriptomic profiles were conducted using single-cell RNA sequencing (scRNA-seq). The technique enables high-resolution cellular profiling at single cell level, but the tissue dissociation results in the loss of spatial information.⁶² Recently, emerging spatial transcriptomics technologies are developed to address this limitation; however, they involve trade-offs between resolution and sensitivity compared to the dissociated scRNA-seq.⁶³ As a result, precise reconstruction of transcriptomic landscapes within the microanatomical structures of cerebral organoids remains challenging.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Dr. Yi-Chung Tung (tungy@gate.sinica.edu.tw).

Materials availability

All materials generated in this study are available from the lead contact upon reasonable request with a completed materials transfer agreement.

Data and code availability

• •Data: scRNA-seq data have been deposited at GEO as GSE253940 and GSE293664, and the data are publicly available as of the date of publication.
• •Code: This article does not report original code.
• •Additional information: Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We would like to thank the RNA technology Platform and Gene Manipulation core Facility (RNAi core) of the National core Facility for Biopharmaceuticals at Academia Sinica in Taiwan for providing shRNA reagents and related services. This work was supported by the National Science and Technology Council (NSTC) Taiwan (Grant 110-2221-E-001-005-MY3 to Y.-C.T.), Academia Sinica Career Development Award (grant AS-CDA-106-M07 to Y.-C.T.), and Academia Sinica Neuroscience Core Facility (grant AS-CFII-113-A6 to Y.-J.C.). The graphical abstract and Figure 1 were created in BioRender (https://BioRender.com/q96b800, and https://BioRender.com/4wedwbb).

Author contributions

Y.-H.L. and Y.-C.T. conceptualized the study, wrote the manuscript, and prepared the figures. Y.-H.L., M.-T.C., H.-C.L., H.-L.L., H.L., Y.-J.C., and H.-M.W. performed experiments and analyzed the majority of the results. T.-W.C. and Y.-C.T. acquired the funding and supervised the study.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Human/Mouse/Rat SOX2 Antibody	R&D Systems	Cat# AF2018, RRID: AB_355110
Anti-PSD95 Antibody	Abcam	Cat# ab18258, RRID: AB_444362
Purified Mouse Anti-Ki-67 Clone B56	BD Biosciences	Cat# 550609, RRID: AB_393778
Anti-NESTIN Antibody	Merck Millipore	Cat# ABD69, RRID: AB_2744681
Anti-TBR1 Antibody	Abcam	Cat# ab31940, RRID: AB_2200219
Goat anti-Mouse IgG (H + L) Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 488	Thermo Fisher Scientific	Cat# A-11001, RRID: AB_2534069
Goat Anti-Rabbit IgG (H + L) Highly Cross-adsorbed Antibody, Alexa Fluor™ 546	Thermo Fisher Scientific	Cat# A-11035, RRID: AB_143051
Donkey Anti-Goat IgG (H + L) Antibody, Alexa Fluor™ 647	Thermo Fisher Scientific	Cat# A-21082, RRID: AB_141493
Chemicals, peptides, and recombinant proteins
Stemflex™	Thermo Fisher Scientific	Cat# A3349401
AggreWell™ EB formation medium	STEMCELL Technologies	Cat# 05893
DMEM/F12	Thermo Fisher Scientific	Cat# 12660012
Neurobasal	Thermo Fisher Scientific	Cat# 21103049
Matrigel	Corning	Cat# 354277
Polybrene	Sigma-Aldrich	Cat# H9268
Puromycin	Sigma-Aldrich	Cat# P8833
Accumax	Sigma-Aldrich	Cat# A7089
ROCK inhibitor Y-27632	Sigma-Aldrich	Cat# SCM075
Glutamax	Thermo Fisher Scientific	Cat# 35050061
MEM-NEAA	Thermo Fisher Scientific	Cat#11140050
Heparin	Sigma-Aldrich	Cat# H0878
Cyclopamine A	Sigma-Aldrich	Cat# H0878
IWP-2	Sigma-Aldrich	Cat# I0536
SAG	Sigma-Aldrich	Cat# 566660
N2 supplement	Thermo Fisher Scientific	Cat# 17502048
B27 supplement without vitamin A	Thermo Fisher Scientific	Cat#12587010
B27 supplement	Thermo Fisher Scientific	Cat#17504044
2-mercaptoethanol	Thermo Fisher Scientific	Cat#21985023
Insulin	Sigma-Aldrich	Cat# I9278
Calcium- and magnesium-free DPBS	Thermo Fisher Scientific	Cat# 14190144
Cell Recovery Solution	Corning	Cat# 354253
Neural Tissue Dissociation Kit	Miltenyi Biotec	Cat# 130-092-628
Calcein AM	Thermo Fisher Scientific	Cat# C1430
Polyethyleneimine	Sigma-Aldrich	Cat# P3143
Laminin	Sigma-Aldrich	Cat# 23017015
Paraformaldehyde	Sigma-Aldrich	Cat# 158127
Sucrose	Sigma-Aldrich	Cat# S0389
Gelatin	Sigma-Aldrich	Cat# G2500
Optimal cutting temperature, OCT compound	Leica Biosystems	Cat# 3801480
Glycine	Sigma-Aldrich	Cat# G8898
Triton X-100	Sigma-Aldrich	Cat# X100
BSA	Sigma-Aldrich	Cat# A7906
Sodium azide	Sigma-Aldrich	Cat# S8032
Normal goat serum	Thermo Fisher Scientific	Cat# 31872
Normal donkey serum	Sigma-Aldrich	Cat# D9663
ProLong Gold antifade reagent	Thermo Fisher Scientific	Cat# P36930
Ruthenium-based phosphorescent microbeads	Colibri Photonics	CPOx-50-RuP
Tris(4,7-diphenyl-1,10-phenanthroline) ruthenium(II) bis(perchlorate) (Ru(dpp)3(ClO4)2	Toronto Research Chemicals	Cat# 75213-31-9
CellEvent™ Senescence Green Detection Kit	Thermo Fisher Scientific	Cat# C10850
RNase-free H2O	Thermo Fisher Scientific	Cat# J70783.AP
Critical commercial assays
Chromium Single Cell 3′ Library & Gel Bead Kit v3.1	10x Genomics	Cat# PN-1000121
64-channel multi-electrode array (MEA) system	Alpha MED Scientific	MED64 system
PureLink™ RNA Mini Kit	Thermo Fisher Scientific	Cat# 12183020
ToolsQuant II Fast RT Kit	BIOTOOLS	Cat# KRT-BA06-2
TOOLS 2xSYBR qPCR Mix Kit	BIOTOOLS	Cat# FPT-BB05
Deposited data
Raw and analyzed scRNA-seq data	This paper	GSE253940 and GSE293664
Experimental models: Cell lines
Human induced pluripotent stem cells (hiPSCs)	RIKEN BioResource Research Center, Ibaraki, Japan	Cat# 409B2
Oligonucleotides
NGB-targeting shRNA sequence: GTGATGCTCGTGATTGATGCT	National RNAi Core Facility at Academia Sinica, Taipei, Taiwan	TRCN0000059495
LacZ-targeting shRNA sequence for scramble control: CGCGATCGTAATCACCCGAGT	National RNAi Core Facility at Academia Sinica, Taipei, Taiwan	TRCN0000072224
GAPDH forward primer sequence: CTCTGACTTCAACAGCGACA	Biotools	designed through IDT PrimerQuest and OligoAnalyzer tool
GAPDH reverse primer sequence: GTAGCCAAATTCGTTGTCATACC	Biotools	designed through IDT PrimerQuest and OligoAnalyzer tool
KLF4 forward primer sequence: ACCTACACAAAGAGTTCCCATC	Biotools	designed through IDT PrimerQuest and OligoAnalyzer tool
KLF4 reverse primer sequence: ATCTGAGCGGGCGAATTT	Biotools	designed through IDT PrimerQuest and OligoAnalyzer tool
NGB forward primer sequence: CCCTCTTCCAGTACAACTGC	Biotools	designed through IDT PrimerQuest and OligoAnalyzer tool
NGB reverse primer sequence: AATCACGAGCATCACCTTCC	Biotools	designed through IDT PrimerQuest and OligoAnalyzer tool
Software and algorithms
Cell Ranger v6.1.2	10x Genomics	https://www.10xgenomics.com/support/software/cell-ranger/latest
Scanpy v1.8	Wolf et al.64	https://scanpy.readthedocs.io/en/1.8.x/
GSEApy	Fang et al.65	https://gseapy.readthedocs.io/en/latest/introduction.html
decoupleR	Badia et al.65	https://saezlab.github.io/decoupleR/
Mobius Offline software v1.4.5	Alpha MED Scientific	https://www.med64.com/products/med64-mobius-software/mobius-offline-toolkit/
MATLAB R2017a	MathWorks	https://www.mathworks.com/products/compiler/matlab-runtime.html
ImageJ	U.S. National Institutes of Health	https://imagej.net/nih-image/about.html
SPSS v16.0	IBM	https://www.ibm.com/products/spss-statistics
Sigma plot v14.0	Grafiti	https://grafiti.com/sigmaplot-detail/
BioRender	BioRender	https://help.biorender.com/hc/en-gb
PrimerQuest	Integrated DNA Technologies	https://www.idtdna.com/pages/tools/primerquest
OligoAnalyzer	Integrated DNA Technologies	https://www.idtdna.com/pages/tools/oligoanalyzer
CFX MaestroTM Software v2.3	BIO-RAD	https://www.bio-rad.com/en-tw/product/cfx-maestro-software-for-cfx-real-time-pcr-instruments?ID=OKZP7E15
Other
Sequencing system	Illumina	NovaSeq 6000
Cryostat microtome	Leica Microsystems	CM3050S
Confocal microscope	Zeiss	LSM880
Inverted fluorescence microscope	Leica Microsystems	DMI6000B
High-power LED	Thorlabs	M470LP-C2
Dual tap CMOS FLIM camera	Excelitas Technologies	PCO.FLIM
Narrow band filter cube	Leica Microsystems	AOR
5X, NA = 0.12 objective	Leica Microsystems	N PLAN
Real-time PCR Detection System	BIO-RAD	CFX Connect
6-well plates	Corning	Cat# 3516
Ultra-low attachment 96-well round-bottom plates	Corning	Cat# 7007
Ultra-low attachment 24-well plates	Corning	Cat# 3473
40 μm cell strainer	Corning	Cat# 352340
Chromium Single Cell 3′ Chip G	10x Genomics	Cat# PN-2000177
MEA probes	Alpha MED Scientific	Cat# MED-P515A
Silane-coated coverslips	Muto Pure Chemicals	Cat# 5116

Experimental model and study participant details

Culture of human iPS cells

Feeder-free human induced pluripotent stem cells (hiPSCs) were obtained from and authenticated by the RIKEN Bioresource Center (Cell No. HPS0076, clone 409B2) and maintained on human embryonic stem cell (hESC)-qualified Matrigel (Corning, NY) coated 35 mm dishes in Stemflex (Thermo Fisher Scientific, Waltham, MA) medium and incubated at 37 °C in a humidified atmosphere containing 5% CO₂. Routine passaging and culture conditions were followed according to the manufacturer’s guidelines. The cells were regularly tested for mycoplasma contamination.

Cerebral organoid formation

Cerebral organoids were generated using an established protocol with slight modifications.^1,2 The experimental workflow is summarized in Figure 1. First, hiPSC colonies were dissociated into single-cell suspensions using Accumax (Sigma-Aldrich Inc., St. Louis, MO). The dissociated cells were then seeded into ultra-low attachment 96-well round-bottom plates (Corning) at a density of 12,000 cells per well for the formation of embryoid bodies (EBs). These EBs were cultured in AggreWell EB formation medium (STEMCELL Technologies Inc., Cambridge, MA) supplemented with 10 μM ROCK inhibitor Y-27632 (Sigma-Aldrich Inc.) to support initial cell aggregation. After six days, DMEM/F12 (Thermo Fisher Scientific)-based neural induction (NI) medium containing N2 supplement (Thermo Fisher Scientific) with a ratio of 1:100, Glutamax (Thermo Fisher Scientific), 1:100 MEM-NEAA (Thermo Fisher Scientific), and 1 μg/mL Heparin (Sigma-Aldrich Inc.) was used to initiate the neural differentiation for patterning the EBs. To induce dorsal forebrain identity, a 5 μM Sonic Hedgehog (SHH) inhibitor Cyclopamine A (Sigma-Aldrich Inc.) was added to the NI medium. Concurrently, ventral forebrain patterning was promoted using a combination of 2.5 μM Wnt signaling inhibitor IWP-2 (Sigma-Aldrich Inc.), and 100 nM SHH agonist SAG (Sigma-Aldrich Inc.). Following six days of neural induction, a dorsal-patterned EB and a ventral-patterned EB were embedded together in a single droplet of the Matrigel. These EB-containing droplets were then transferred to ultra-low attachment 24-well plates (Corning) with differentiation medium. The differentiation medium consisted of a 1:1 mixture of DMEM/F12 and Neurobasal Medium (Thermo Fisher Scientific) supplemented with 1:200 N2 supplement, 1:100 B27 supplement without vitamin A (Thermo Fisher Scientific), 550 μM 2-mercaptoethanol (Thermo Fisher Scientific; 1:3000 dilution), 1:4000 insulin (Sigma-Aldrich Inc.), 1:100 Glutamax, and 1:100 MEM-NEAA (Thermo Fisher Scientific). After four days of stationary growth, the medium was modified by replacing the B27 supplement without vitamin A with 1:200 B27 supplement containing vitamin A (Thermo Fisher Scientific) to support further neuronal differentiation. The developing cerebral organoids were then maintained on an orbital shaker until the collection date.

Method details

Lentiviral transduction

To silence the NGB gene, hiPSCs were transduced with lentiviral vectors carrying short hairpin RNA (shRNA). The NGB-targeting shRNA (shNGB, TRCN0000059495) and the scrambled shRNA control (shLacZ, TRCN0000072224) were applied in this study. For stable transduction, hiPSCs were seeded in 6-well plates (Corning) at approximately 40% confluence in antibiotic-free medium. The appropriate lentiviral vectors were introduced into the fresh culture medium containing 6 mM polybrene (Sigma-Aldrich Inc.) to enhance infection efficiency. After 48 h, 2 μg/mL puromycin (Sigma- Aldrich Inc.) was added for selection, and cells were maintained in the selection medium for 10 days with media changed every other day.

Single-cell preparation, RNA sequencing, and data analyzing

Organoids were collected at Days 24, 35, and 65 for transcriptomic profiling. After a brief rinse in calcium- and magnesium-free DPBS (Thermo Fisher Scientific), organoids were incubated in Cell Recovery Solution (Corning) to dissolve the polymerized Matrigel. The organoids were then mechanically dissected into smaller fragments (approximately 2–4 pieces) using a scalpel and subjected to multiple washes in DPBS to eliminate residual debris. Fragmented organoids were dissociated into a single-cell suspension through the Neural Tissue Dissociation Kit (Miltenyi Biotec) according to the manufacturer’s instruction. After enzymatic digestion, the single-cell suspension was passed through a 40 μm cell strainer (Corning) to remove clumps and debris. To assess cell viability, the dissociated cells were stained with Calcein AM (Thermo Fisher Scientific), and only samples with over 80% viable cells were processed further. The cell suspension with adjusted density (1000 cells/μL) was loaded onto a Chromium Single Cell 3′ Chip G (10x Genomics, Pleasanton, CA) and processed using the Chromium Controller to generate single-cell Gel Beads in Emulsion (GEMs). Library construction for single-cell RNA sequencing (scRNA-seq) was performed using the Chromium Single Cell 3′ Library & Gel Bead Kit v3.1 (10x Genomics). Sequencing was conducted by a sequencing system (NovaSeq 6000, Illumina, San Diego CA) following the prescribed protocol (R1: 28 cycles, R2: 91 cycles, and i7 index: 8 cycles).

The raw sequencing reads were aligned to the human reference genome (GRCh38) using the software Cell Ranger v6.1.2 (10x Genomics) under default settings to create a cell-by-gene count matrix. Further analysis was performed using a custom Python code pipeline incorporating the Scanpy v1.8 toolkit.⁶⁴ Quality control was applied by filtering out cells with fewer than 200 or more than 6000 detected genes to remove potential empty droplets and doublets, respectively. Additionally, cells containing over 5% mitochondrial RNA reads, indicative of cell death, were excluded. Normalization of expression data was carried out using Unique Molecular Identifier (UMI) reads for each cell and log-transformed. To identify highly variable genes, principal component analysis (PCA) was applied for dimensionality reduction. Uniform Manifold Approximation and Projection (UMAP) coordinates were computed using the built-in function Leiden graph-clustering method⁶⁶ in Scanpy for visualizing distinct cellular populations. Cluster annotation was performed based on established marker genes^19,20 validated by the Wilcoxon rank-sum test.⁶⁷ The gene sets used for classification included the following: neural stem/progenitor cells (FABP7, GLI3, SOX2, PAX6, VIM, and HES1), basal progenitor cells (NEUROG2, EMX2, EOMES, NEUROD1, and NEUROD2), dividing cells (CCNB1, HMMR, MKI67, ASPM, and CCND2), interneuron progenitors (DLX2 and ASCL1), interneurons (GAD1, DLX5, and SST), differentiated neurons (DCX, SOX4, CD24, STMN2, and MAP2), and excitatory neurons (TBR1, SLC17A6, GRIN2B, SATB2, STMN2, and MAP2).

Excitatory neuron subtypes were identified from the single-cell transcriptomic dataset based on the established marker gene expression profiles.^68,69,70 Cells expressing TBR1 were defined as excitatory neurons. Within this population, the Cajal–Retzius (CR) cells were classified as those co-expressing TBR1 and RELN. The upper-layer excitatory neurons were defined as cells co-expressing TBR1 and SATB2. The deep-layer excitatory neurons were classified as TBR1 expressing cells that did not co-express RELN or SATB2. For each developmental stage and experimental condition, the number of cells in each subtype was quantified and expressed as a percentage of the total TBR1 expressing excitatory neuron population.

For batch effect correction in integrated datasets, the built-in function “ingest” in Scanpy was utilized to map new data onto reference datasets, while the batch balanced k nearest neighbor (BBKNN) method⁷¹ was applied for further batch correction. Gene set enrichment analysis (GSEA) was conducted using the EnrichR tool⁷² from the GSEApy package,⁶⁵ employing the “GO_Biological_Process_2021” library to infer biological functions associated with hCOs. Additionally, pathway analysis was carried out using the decoupler package,^73,74 with the MSigDB database⁷⁵ conducted to query the gene sets for matching the pathway activity.

Measurement of intra-organoid oxygen tension by frequency-domain fluorescence lifetime imaging microscopy (FD-FLIM)

Frequency-domain fluorescence lifetime imaging microscopy (FD-FLIM) was performed on an inverted fluorescence microscope (DMI 6000B, Leica Microsystems, Wetzlar, Germany). The microscope was equipped with a high-power LED excitation source (470 nm nominal wavelength; Thorlabs, M470LP-C2) and a dual-tap CMOS FLIM camera (PCO.FLIM, Excelitas Technologies Corp., Waltham, MA). The excitation source and detector were synchronized and modulated at 50 kHz via a digital signal generated by the camera. A narrow-band filter cube (AOR, Leica Microsystems) was used to isolate emission from the oxygen-sensitive beads, and fluorescence was collected through a 5× objective (N PLAN, NA 0.12; Leica Microsystems) before detection.

Oxygen-sensitive microbeads (CPOx, 50 μm diameter; Colibri Photonics GmbH) were prepared at 10 mg/mL in DPBS (Thermo Fisher Scientific, 14190144) containing 1% bovine serum albumin (BSA; Sigma-Aldrich, 05470). On day 12 of differentiation, the bead suspension was mixed 1:20 (v/v) with the liquid hydrogel and incorporated during the fusion of dorsal and ventral embryoid bodies (EBs). Fluorescence lifetime imaging was performed using an excitation range of 420–490 nm and an emission range of 570–670 nm. Due to dynamic quenching, the presence of molecular oxygen reduces fluorescence intensity and shortens lifetime. To calibrate lifetime measurements, a droplet of oxygen-sensitive fluorescent dye, 25 μM tris(4,7-diphenyl-1,10-phenanthroline) ruthenium(II) bis(perchlorate) (Ru(dpp)₃(ClO₄)₂; 75213-31-9, Toronto Research Chemicals, Ontario, Canada), was utilized as reference with a known lifetime (τ = 1.185 μs).⁷⁶

Quantitative real-time PCR

Total RNA was extracted from the iPSC cells (at least 1 × 10⁶ cells) using the PureLink RNA Mini Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. 1 μg of each RNA sample was used in reverse transcription for cDNA synthesis following the manual of ToolsQuant II Fast RT Kit (BIOTOOLS Co., Taiwan). The synthesized cDNA was then diluted to the concentration of 10 ng/μL 10 μL of the PCR mix contained 1× TOOLS 2xSYBR qPCR Mix Kit (BIOTOOLS Co.), 2 μM forward primer, 2 μM reverse primer (primer sequences were listed in STAR★Methods), 1 μL RNase-free H₂O, and 2 μL of cDNA template (10 ng/μL). Quantitative real-time PCR (qRT-PCR) was performed using a CFX Connect Real-Time PCR Detection System (BIO-RAD Laboratories Inc., Hercules, CA). The two-step amplification program was set as 95 °C for 15 min, followed by 40 cycles of 95 °C for 10 s and 62 °C for 30 s, and a melt-curve analysis from 65 °C to 95 °C with 0.5 °C increments every 5 s. Experiments were conducted in technical triplicate across three independent biological replicates in compliance with the criteria of the MIQE guidelines.⁷⁷

Electrophysiological analysis using multi-electrode array (MEA)

To evaluate spontaneous electrophysiological activity of the hCOs, we utilized a 64-channel multi-electrode array (MEA) system (MED64 system, Alpha MED Scientific Inc., Ibaraki, Japan). Prior to the experiments, MEA probes (Alpha MED Scientific) were coated with 0.005% polyethyleneimine (Sigma-Aldrich Inc.) and subsequently treated with 2 mg/mL laminin (Thermo Fisher Scientific) to enhance cell adherence and optimize signal detection.

For each recording experiment, a single organoid was carefully placed at the center of the MEA probe and positioned within the 8 × 8 electrode grid on the probe. The organoid was submerged in 150 μL of the organoid culture medium to preserve physiological conditions. Electrical signal recording was conducted at a sampling frequency of 20 kHz per channel, with the system maintained at 37°C using a temperature-controlled setup. Data acquisition was carried out over a 5-min period to capture spontaneous extracellular field potentials.

Histology, cryosectioning, and immunofluorescence staining

Cerebral organoids were fixed in 4% paraformaldehyde (PFA) (Sigma-Aldrich Inc.) at 4°C overnight. After fixation, the samples were rinsed twice with DPBS and then incubated in a 30% sucrose solution (Sigma-Aldrich Inc.) overnight for cryoprotection. The organoids were subsequently immersed in an embedding medium consisting of 7.5% gelatin (Sigma-Aldrich Inc.) and 15% sucrose for 1 h before being embedded in an optimal cutting temperature (OCT) compound (Leica Biosystems, Deer Park, IL) filled mold. The embedded tissues were stored at −80°C until further processing. Frozen sections with thicknesses of 20 μm were obtained using a cryostat microtome (Leica Microsystems) and mounted onto silane-coated coverslips (Muto Pure Chemicals, Japan).

To prepare the sections for immunofluorescence staining, coverslips were first washed with Tris-buffered saline (TBS) containing 1% glycine (Sigma-Aldrich Inc.), 0.4% Triton X-100 (Sigma-Aldrich Inc.), 3% BSA (Sigma-Aldrich Inc.), and 0.1% sodium azide (Sigma-Aldrich Inc.). To prevent non-specific antibody binding, tissue sections were blocked for 1 h at room temperature in a buffer containing either 10% normal goat serum (Thermo Fisher Scientific) or 10% normal donkey serum (Sigma-Aldrich Inc.) diluted in TBS. Primary antibody incubation was performed at 4°C for 24 h using species-appropriate antibody combinations. The primary antibodies and their respective dilutions included goat anti-SOX2 (1:250) (R&D Systems Inc., Minneapolis, MN), rabbit-PSD95 (1:300) (Abcam Limited, Cambridge, UK), mouse anti-Ki-67 (1:250) (BD Pharmingen, Franklin Lakes, NJ), rabbit anti-NESTIN (1:250) (Merck Millipore, Temecula, CA), and rabbit anti-TBR1 (1:300) (Abcam Limited). After primary antibodies incubation, sections were thoroughly washed with TBS before being incubated with the appropriate secondary antibodies at 4°C for 24 h. The secondary antibodies used for the staining included Alexa 488, 546, 633-conjugated goat anti-mouse (Thermo Fisher Scientific), anti-rabbit immunoglobulin G (IgG) (Thermo Fisher Scientific), and donkey anti-goat IgG (Thermo Fisher Scientific). Following another washing step with TBS, the coverslips were mounted with ProLong Gold antifade reagent (Thermo Fisher Scientific) to preserve fluorescence signals. All images were captured using an LSM880 confocal microscope (Zeiss GmbH, Oberkochen, Germany).

Definition of VZ/SVZ- and CP-like regions in hCOs

Neuroepithelial organization within human cerebral organoids (hCOs) was assessed based on established histological criteria for cortical development.^1,78 The ventricular zone/subventricular zone (VZ/SVZ)-like region was identified as a dense progenitor layer arranged in a rosette-like structure, characterized by high nuclear density and enriched expression of radial glia marker SOX2 and proliferative marker Ki-67. This rosette morphology has been previously described in early cortical neuroepithelium.⁷⁹ The cortical plate (CP)-like region was defined as the outer neuronal layer containing postmitotic neurons derived from the VZ/SVZ, marked by the neuronal marker TBR1.

Quantification and statistical analysis

Oxygen concentrations were calculated from lifetime measurements using the Stern–Volmer equation:

$$\frac{{\tau }_0}{\tau }=1+K_q\times \left[{\mathrm{O}}_2\right]$$

where τ₀ and τ are the fluorescence lifetimes in the absence and presence of oxygen, respectively, and K_q was a quenching constant. The quenching constant was determined using an established calibration process.⁷⁶ The spatial positions of beads within the organoids were identified through intensity-based thresholding using MATLAB R2017a (MathWorks, Natick, MA), and mean lifetime and oxygen concentration per bead were calculated. Characterization and data analysis were performed according to the reported protocol.⁷⁶

For Quantitative real-time PCR, data were analyzed with CFX Maestro Software (BIO-RAD). Relative gene expression was determined by the 2^−ΔΔCq method⁸⁰ using GAPDH as a reference gene.

For electrophysiological analysis, spontaneous extracellular field potentials were recorded over 5-min intervals. Following the signal collection, data were processed using MED64 Mobius Offline software (version 1.4.5) to extract electrophysiological parameters.

For morphometric analysis, the VZ/SVZ-like thickness was measured radially from the apical/luminal surface of the rosette to the outer boundary of the SOX2⁺/Ki-67⁺ domain. The CP-like thickness was measured from this boundary to the pial surface, defined by the extent of TBR1⁺ neuronal labeling. Within the defined neuroepithelial regions, total cell numbers were calculated based on the DAPI nuclear staining. The proportion of neural progenitor/stem cells (NSCs) was calculated as the number of SOX2⁺ cells relative to total cells. The proportion of proliferating NSCs was calculated as Ki-67⁺ cells relative to total cells, and the proportion of neurons was calculated as TBR1⁺ cells relative to total cells. For each experimental condition (control, shLacZ, shNGB), four hCOs (n = 4) were analyzed. Within each organoid, three rosettes meeting the specified quality criteria (clear lumen, well-arranged inner SOX2⁺/Ki-67⁺ and outer TBR1⁺ regions) were selected. The independent radial measurement was taken per rosette-containing neuroepithelial region (see white squares in Figures 3 and S3) and averaged to yield a rosette-level value. These values were then averaged to generate organoids-level means, which were used for statistical analysis. All measurements were performed on high-resolution confocal images using image ImageJ software (NIH).

Statistical analysis was performed using the software SPSS (IBM). Comparisons between two groups were assessed using two-tailed Student’s t-tests, while multiple group comparisons were evaluated using one-way ANOVA followed by Tukey’s post-hoc test. The data were expressed as mean ± standard deviation (SD). Sample size was described in each figure caption, and the levels of statistical significance were denoted as ^∗p < 0.05, and ^∗∗p < 0.01.

Published: February 14, 2026