Dynamic regulation of neural stem cell state transitions by NOTCH1-REST transcriptional axis

Abstract

BACKGROUND

The regulatory role of the neurogenic locus notch homolog protein 1 (NOTCH1)-RE1-silencing transcription factor (REST) transcriptional axis in neural stem cell (NSC) dormancy-activation transitions remains poorly understood.

AIM

To clarify the temporal, epigenetic, and metabolic mechanisms by which the NOTCH1–REST axis regulates NSC/neural progenitor cell (NPC) state transitions and neural repair.

METHODS

Public human embryonic prefrontal cortex single-cell RNA sequencing data, human-induced pluripotent stem cell-derived human NSCs, human neural organoids, multimodal omics, and a mouse spinal cord injury model were integrated. Pseudotime analysis was used to infer NSC/NPC state trajectories, and functional perturbation, organoid intervention, exploratory assay for transposase-accessible chromatin using sequencing/chromatin immunoprecipitation sequencing, metabolomics, live imaging, and in vivo assays were performed for validation.

RESULTS

REST was enriched in quiescent-like NSC/NPC states and declined during activation, whereas NOTCH1 increased during the intermediate activation phase, indicating a sequential rather than simultaneous regulatory relationship. REST inhibition promoted G1/S progression, EdU incorporation, and activation dynamics, while NOTCH1 activation partially counterbalanced these effects. Stage-resolved organoid single-cell RNA sequencing identified SOX2-, NES/Nestin-, and PAX6-expressing NSC/NPC-like populations, supporting the relevance of the organoid model. Exploratory assay for transposase-accessible chromatin using sequencing and chromatin immunoprecipitation sequencing profiles suggested stage-associated changes in chromatin accessibility and transcription factor binding, and metabolomics indicated that REST inhibition was associated with enhanced central carbon metabolism and glycolysis-related remodeling. In the spinal cord injury model, REST inhibition promoted NSC marker re-expression in the lesion area and improved motor recovery.

CONCLUSION

These findings support the NOTCH1-REST axis as a time-dependent regulatory switch governing NSC/NPC state transitions through coordinated transcriptional, epigenetic, and metabolic remodeling, and suggest REST-targeted modulation as a potential strategy for neural repair.

Key Words: Neural stem cells; Neurogenic locus notch homolog protein 1-RE1-silencing transcription factor axis; Epigenetic; Organoid model; Neural repair

Core Tip: This study unveils the neurogenic locus notch homolog protein 1-RE1-silencing transcription factor axis as a pivotal molecular switch controlling neural stem cell dormancy-activation transitions. By integrating transcriptional, epigenetic, and metabolic analyses, it elucidates the dynamic regulatory interplay of neurogenic locus notch homolog protein 1 and RE1-silencing transcription factor, offering crucial insights into neural stem cell state transitions. The findings present promising targets for central nervous system injury repair and regenerative therapies for neurodegenerative diseases.

INTRODUCTION

Neural stem cells (NSCs) possess self-renewal and multipotent differentiation capacities[1,2]. They are widely distributed in embryonic and certain adult regions of the central nervous system (CNS), where they contribute to neural development, injury repair, and functional recovery[3,4]. In adults, NSCs generally remain quiescent under homeostatic conditions but can be activated by pathological stimuli to re-enter the cell cycle and generate neurons or glial cells[5]. However, their activation is inefficient and highly dependent on microenvironmental and signaling cues, limiting their therapeutic potential[6]. Although transcriptional regulation, epigenetic remodeling, and metabolic reprogramming have been implicated in NSC activation[7], the core regulatory networks governing NSC state transitions, particularly in human-relevant models, remain incompletely understood.

Among the known regulatory factors, neurogenic locus notch homolog protein 1 (NOTCH1) and RE1-silencing transcription factor (REST) have emerged as focal points of current research. NOTCH1, a central receptor of the Notch pathway, maintains NSCs in an undifferentiated state and delays neuronal differentiation during neurodevelopment[8,9]. REST, a transcriptional repressor of neuronal lineage genes, is essential for sustaining NSC dormancy during embryonic and adult stages[10,11]. REST can also suppress cell cycle-related genes through epigenetic mechanisms, thereby restraining NSC activation[12,13]. Nevertheless, whether NOTCH1 and REST form a functional transcriptional axis, how they interact during the dormant-to-activated transition, and how their downstream epigenetic and metabolic programs evolve remain unclear.

Recent advances in three-dimensional human neural organoids and multimodal omics technologies have provided powerful tools for dissecting NSC state transitions. Human neural organoids better recapitulate the microenvironment, spatial architecture, and cellular heterogeneity of CNS tissues than conventional two-dimensional cultures, making them valuable models for studying human NSC (hNSC) dynamics[14-16]. Meanwhile, single-cell RNA sequencing (scRNA-seq), assay for transposase-accessible chromatin using sequencing (ATAC-seq), chromatin immunoprecipitation sequencing (ChIP-seq), and metabolomics enable systematic characterization of transcriptional trajectories, chromatin accessibility, transcription factor binding, and metabolic remodeling during NSC fate transitions[17-19]. Together, these approaches provide a multidimensional framework for identifying regulatory mechanisms and candidate targets that promote NSC activation and neural regeneration.

Therefore, this study investigates the mechanistic role of the NOTCH1-REST transcriptional axis in regulating the transition between dormant and activated NSC states. By examining how this axis coordinates epigenetic remodeling and metabolic reprogramming, we aim to define its regulatory function in NSC fate determination and neural regeneration. This work is expected to refine the molecular framework of NSC state regulation, deepen our understanding of endogenous neural repair mechanisms, and provide a basis for the precise activation and control of NSC fate. Moreover, the regulatory factors and pathways identified here may serve as potential therapeutic targets for CNS injury and neurodegenerative diseases, including spinal cord injury (SCI), where effective regenerative strategies remain limited.

MATERIALS AND METHODS

Data acquisition and preprocessing from public databases

scRNA-seq data of the human embryonic prefrontal cortex (PFC) were obtained from the public Gene Expression Omnibus database (GSE104276, https://www.ncbi.nlm.nih.gov/geo/). This dataset, generated using the Smart-seq2 platform and reported by Zhong et al[20], covers neurodevelopmental stages from gestational week 8 to 26. It includes 34 independent embryonic PFC samples, comprising 29 PFC1 and 5 PFC2 samples, and contains 2394 high-quality single cells with an average sequencing depth of approximately 2 × 10⁶ reads per cell. Downloaded files included normalized gene expression matrices in TPM/count format and corresponding cell barcode annotation files. Cell type annotations were based on marker genes provided in the original study, including neural progenitor cells (NPCs), intermediate progenitor cells, excitatory neurons, inhibitory neurons, astrocytes, and oligodendrocyte precursor cells.

Identification of NSC subpopulations and REST activity stratification

The NPC population was extracted using Seurat for downstream analysis. Within this subset, cells were stratified according to REST expression to represent the two ends of the REST activity gradient. Specifically, NPCs with log₂TPM > 2 were defined as REST-high, whereas those with lower REST expression were defined as REST-low. REST-high NPCs were operationally considered “quiescent-like”, while REST-low NPCs were considered “activated-like”. This threshold was used as an analytical strategy to enhance contrast for differential expression analysis and trajectory rooting, rather than to define an absolute biological boundary. Because REST expression showed a continuous distribution within NPCs, these groups should not be interpreted as strictly binary biological subtypes. Differential expression analysis between REST-high and REST-low NPCs was performed using the FindMarkers function in Seurat with the Wilcoxon rank-sum test, applying thresholds of |log₂FC| > 0.25 and adjusted P < 0.05. To evaluate the robustness of downstream findings, sensitivity analyses were performed using alternative stratification strategies, including median split and top/bottom quartile comparisons of REST expression. Key marker-gene changes and neurogenesis-related enrichment patterns were then compared across strategies.

Pseudotime trajectory reconstruction

Pseudotime analysis was performed using Monocle3 (v1.3.3, Trapnell Lab). The input was an expression matrix converted from a Seurat object using the as.cell_data_set() function. Cells were ordered along the trajectory based on gene expression dynamics using the learn_graph() and order_cells() functions. Transcription factor expression trends were extracted along the trajectory. Visualization was conducted using plot_genes_in_pseudotime(), and gene module heatmaps were generated with ComplexHeatmap (v2.14.0). Co-expression and correlation analysis between REST and NOTCH1 were performed by calculating the Pearson correlation coefficient using the cor.test() function (method = “pearson”) across all NPC samples (n = 578). The REST-high NPC population was selected as the trajectory root based on prior biological evidence that REST functions as a transcriptional repressor associated with low-proliferative progenitor states, together with its higher expression at the early end of the inferred NPC continuum. We emphasize that pseudotime ordering represents an inferred transcriptional-state continuum rather than direct temporal tracking of individual cells.

ATAC-seq analysis

Human-induced pluripotent stem cell (hiPSC)-derived NSCs were cultured under REST inhibition, NOTCH1 activation (NICD overexpression), and control conditions. After dissociation with TrypLE Express (Cat# 12604021, Thermo Fisher Scientific, MA, United States), Tn5 transposition and library preparation were carried out using the TruePrep DNA Library Prep Kit V2 for Illumina (Cat# TD501, Vazyme, China). Library quality was assessed using a Qubit 4.0 Fluorometer (Cat# Q33238, Thermo Fisher Scientific, MA, United States) and an Agilent 2100 Bioanalyzer (Agilent Technologies, CA, United States). Qualified libraries were subjected to paired-end sequencing (PE150) on the Illumina NovaSeq 6000 platform (Illumina, CA, United States), yielding approximately 50 million reads per sample. Raw sequencing data were quality-controlled and filtered using fastp (v0.23.4, OpenGene, China), and aligned to the human reference genome hg38 (UCSC) using Bowtie2 (v2.5.1, Johns Hopkins University). PCR duplicates were removed with Picard (v2.27.5). For downstream signal visualization and comparative analysis, normalized signal tracks (bigWig files) were generated using the bamCoverage function in deepTools (v3.5.3), producing one representative track per condition (n = 1). Signal intensity was normalized as reads per genomic content, and tracks were visualized in Integrative Genomics Viewer (v2.17.1, Broad Institute, Cambridge, MA, United States) with the reference genome set to human (GRCh38/hg38). Comparative visualization was performed at loci corresponding to REST, NOTCH1, DCX, MKI67, and other target genes. Because only one biologically independent ATAC-seq sample was available for each condition, these data were used solely for representative genome-browser visualization and exploratory mechanistic interpretation. No replicate-based differential accessibility testing or statistical significance inference was performed.

ChIP-seq analysis

ChIP-seq was performed using validated antibodies (REST: CST #11987; NOTCH1: Abcam ab52627, United Kingdom). NSCs from the control, REST inhibition, and NOTCH1 activation groups were crosslinked with 1% formaldehyde (Cat# F8775, Sigma-Aldrich, MA, United States) for 10 minutes, and the reaction was quenched with 0.125 M glycine (Cat# G7126, Sigma-Aldrich, MA, United States). After cell lysis, chromatin was fragmented using a Bioruptor Pico sonicator (Diagenode, Belgium) to obtain DNA fragments of approximately 200-500 bp. Immunoprecipitation was carried out at 4 °C by incubating the chromatin with Protein A/G Magnetic Beads (Cat# 88802, Thermo Fisher Scientific, MA, United States) and the corresponding antibodies for 12 hours. Following elution and reverse crosslinking, DNA was extracted and libraries were prepared using the NEBNext Ultra II DNA Library Prep Kit for Illumina (Cat# E7645S, New England Biolabs, MA, United States). Library quality was assessed using a Qubit 4.0 Fluorometer (Cat# Q33238, Thermo Fisher Scientific, MA, United States) and an Agilent 2100 Bioanalyzer (Agilent Technologies, CA, United States). Sequencing was performed on the Illumina NovaSeq 6000 platform (Illumina, CA, United States) with paired-end 150 bp reads (PE150), and one representative signal track file was generated for each condition. Similarly, ChIP-seq data were used as representative binding profiles for selected loci. Differential binding statistics were not calculated because biological replicates were not available for ChIP-seq. Raw sequencing data were quality-checked and filtered using fastp (v0.23.4, OpenGene, China) to remove low-quality reads. Signal visualization was performed in IGV (v2.17.1, Broad Institute, Cambridge, MA, United States), with the reference genome set to human (GRCh38/hg38).

Metabolomic data analysis

Samples were derived from hNSCs in the REST inhibition and control groups, with four independent biological replicates per group (n = 4 per group; total n = 8). Untargeted metabolomic profiling was performed using a Q Exactive™ Orbitrap LC-MS/MS system (Thermo Fisher Scientific, MA, United States). Chromatographic separation was carried out on an ACQUITY UPLC HSS T3 column (2.1 mm × 100 mm, 1.8 μm, Cat# 186003538, Waters, MA, United States) with a flow rate of 0.3 mL/minute, column temperature of 40 °C, and an injection volume of 2 μL. The mobile phase consisted of 0.1% formic acid in water (solvent A; Cat# 06440, Sigma-Aldrich, MA, United States) and acetonitrile containing 0.1% formic acid (solvent B; Cat# 34851, Sigma-Aldrich, MA, United States), with separation achieved through gradient elution. Raw mass spectrometry data were processed using Compound Discoverer v3.2 (Thermo Fisher Scientific, MA, United States) for peak detection, denoising, and alignment. Signal drift was corrected using the QC-RLSC method, and total ion current normalization was applied. Differential metabolites were identified based on variable importance in projection (VIP > 1.0) from the partial least squares discriminant analysis model and a P value < 0.05 (two-tailed Student’s t-test). Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis was performed on the differential metabolites using the BioDeep online platform based on KEGG compound identifiers.

Integrated scRNA-seq analysis of organoids

scRNA-seq was used to characterize transcriptional heterogeneity within NSC/NPC populations in human neural organoids and to identify subpopulations at distinct developmental stages and functional states. The organoid model was generated from hiPSCs through neural induction to form brain organoids. Organoids at different developmental stages (DIV20, DIV30, DIV45, and DIV60; n = 1 per stage) were subjected to scRNA-seq using the Smart-seq2 platform. Because only one organoid scRNA-seq sample was included at each developmental stage, these data were interpreted as stage-resolved descriptive single-cell lineage features rather than replicate-supported differential expression evidence. Functional organoid experiments were validated using independent biological replicates as described below. Sequencing reads were aligned to the human reference genome GRCh38 using STAR (v2.7.11a), and gene expression matrices were quantified using RSEM (v1.3.3). Expression matrices were imported into Seurat (v4.3.0) for downstream analysis. Low-quality cells were removed based on single-cell quality-control metrics, including abnormal numbers of detected genes, abnormal total transcript counts, and abnormal proportions of mitochondrial genes. The data were then normalized, followed by highly variable gene selection, scaling, and principal component analysis. Data from different developmental stages were integrated using Harmony (v1.2.0) to correct batch effects, and the transcriptional distribution of organoid-derived cells was visualized using graph-based clustering and UMAP dimensionality reduction. To evaluate the presence of neural stem/progenitor cell-like populations, canonical NSC/NPC markers SOX2, NES, and PAX6 were selected for visualization. UMAP plots were generated according to developmental stage to show the distribution of DIV20, DIV30, DIV45, and DIV60 cells in low-dimensional space. FeaturePlot was used to display SOX2, NES, and PAX6 expression across the UMAP space. Dot plots summarized normalized average expression levels and proportions of marker-positive cells across developmental stages, while violin plots showed marker expression trends from DIV20 to DIV60. In the scRNA-seq figures, the gene symbol NES was used and annotated as NES/Nestin in the figure legend to correspond to the protein name. Differential expression analysis was performed using the FindMarkers function in Seurat with the Wilcoxon rank-sum test. The screening criteria were set as |log₂FC| > 0.25 and adjusted P < 0.05. Gene set enrichment analysis was conducted using clusterProfiler (v4.10.0), based on the GO_NEUROGENESIS gene set from MSigDB v7.5, to evaluate neurogenesis-related transcriptional programs. Single-cell trajectory trend validation was performed using Monocle3 to reconstruct pseudotime trajectories and assess the dynamic consistency of the REST-NOTCH1 axis across developmental stages.

State regulation and functional assessment of hNSCs

In vitro experiments involving hNSCs were conducted using well-characterized and authenticated hiPSC lines. Neural differentiation was induced using the dual-Smad inhibition protocol, yielding cell populations that met both morphological and molecular criteria for NSCs and were maintained in an N2B27-based culture system. To model distinct physiological states, NSCs were induced into a dormant state by supplementing the basal medium with bone morphogenetic protein-4 (50 ng/mL) and reducing growth factor concentrations, whereas the activated state was induced by adding epidermal growth factor (20 ng/mL) and fibroblast growth factor 2 (20 ng/mL). The induction duration and culture conditions were standardized based on prior optimization experiments. State induction was initially verified by morphological assessment and expression analysis of key marker genes before subsequent assays.

For genetic manipulation, cells were randomly assigned to four experimental groups: Empty vector control, NOTCH1 activation group with NICD overexpression via plasmid transfection, REST knockdown group using specific siRNA (siREST), and dual perturbation group combining NICD overexpression with siREST treatment. Each group included at least three independent biological replicates. Transfection/transduction methods and reagent dosages were validated in preliminary experiments and kept consistent across groups to minimize technical variability. Functional assays included BrdU incorporation combined with propidium iodide staining and flow cytometric analysis to assess cell cycle distribution, as well as EdU staining to evaluate proliferative activity. Confocal imaging was used to examine the expression and subcellular localization of REST, SOX2, MCM2, and other relevant proteins, with DAPI staining used to visualize chromatin structure. To investigate transcriptional and epigenetic changes, RNA and chromatin were extracted from each treatment group for RNA-seq and ATAC-seq, respectively. All sequencing data underwent standardized quality control and preprocessing before multimodal integrative analysis.

Construction and manipulation of human-derived neural organoids

Experiments involving human-derived neural organoids were established using high-quality hiPSC lines. A three-dimensional culture system was employed to induce organoids exhibiting stratified features characteristic of neural development. The culture period lasted at least 30 days to ensure sufficient tissue maturation. Experimental groups included: Control, DAPT treatment (Notch inhibition), Jagged1 treatment (Notch activation), REST inhibition, REST activation, and a combined REST-Notch co-regulation group. Group allocation followed a randomized design, and the dosage and duration of each intervention were kept consistent across groups to minimize technical bias. The effects of each intervention were preliminarily validated through morphological assessment and expression analysis of key marker genes before downstream analyses.

Western blot

Total protein was extracted from tissues and cells using RIPA lysis buffer containing 1% PMSF (P0013B, Beyotime, Shanghai, China). Membrane proteins were extracted using the ProteoPrep^® Membrane Extraction Kit (PROTMEM-1KT, Merck, Germany). Total protein concentration was determined using a BCA Protein Assay Kit (P0011, Beyotime, Shanghai, China). According to the molecular weights of the target proteins, 8%-12% sodium-dodecyl sulfate gel electrophoresis gels were prepared, and equal amounts of protein were loaded into each lane for electrophoretic separation. Proteins were transferred onto polyvinylidene fluoride membranes (1620177, Bio-Rad, CA, United States), which were then blocked with 5% nonfat milk for 1 hour at room temperature. The membranes were incubated overnight at 4 °C with the following primary antibodies: Anti-p27^Kip1 (Invitrogen, 37-9700, 1:100, CA, United States), anti-CDK2 (Abcam, ab32147, 1:1000, United Kingdom), anti-cyclin D1 (Abcam, ab134175, 1:10000, United Kingdom), anti-cyclin E (Santa Cruz, sc-247, 1:200, TX, United States), anti-β-actin (Abcam, ab8227, 1:2000, United Kingdom), anti-REST (Millipore, 07-579, 1:1000, MA, United States), and anti-NICD (Proteintech, 20687-1-AP, 1:500, Wuhan, Hubei Province, China). The membranes were then washed three times with 1 × TBST for 5 minutes each at room temperature. Subsequently, HRP-conjugated goat anti-rabbit IgG (Abcam, ab6721, 1:2000, United Kingdom) or goat anti-mouse IgG (Abcam, ab6728, 1:2000, United Kingdom) secondary antibodies were added and incubated for 1 hour at room temperature; both secondary antibodies were purchased from Abcam (United Kingdom). The membranes were then washed three times again with 1 × TBST buffer for 5 minutes each at room temperature. Protein bands were visualized using ECL reagent (1705062, Bio-Rad, CA, United States) and imaged with an Image Quant LAS 4000C gel imaging system (GE, United States). β-actin was used as the internal control, and the relative protein expression level was determined by calculating the ratio of the gray value of the target band to that of the internal reference band. Protein expression levels were assessed based on three independent experiments.

Establishment, intervention, and multidimensional functional assessment of the mouse SCI model

The mouse SCI model was established using healthy C57BL/6 mice (8-10 weeks old). A spinal cord hemisection was performed at the T9-T10 level under microsurgical conditions to ensure consistency in lesion location and severity. Before surgery, animals were randomly assigned using a random number table into four groups: Empty vector control, REST inhibition (AAV9-dCas9-KRAB targeting REST), NOTCH1 activation (AAV9-NICD), and dual regulation (AAV9-dCas9-KRAB + AAV9-NICD), with a minimum of 15 mice per group. All AAV9 viral vectors were prepared at high titers. Injection sites and dosages were optimized in pilot experiments and kept constant across groups. Vectors were locally delivered to tissues adjacent to the injury site immediately after surgery to ensure targeted and reproducible intervention.

Functional assessments were conducted on postoperative days 0, 7, 14, and 21. Hindlimb motor recovery was evaluated using the Basso Mouse Scale (BMS), and fine motor coordination was assessed using the balance beam test. For histological analysis, spinal cord sections containing the lesion site and adjacent segments were cryosectioned and subjected to immunofluorescence staining for Nestin, SOX2, GFAP, and related markers. Confocal microscopy was used to evaluate cellular distribution and morphological changes. Micro-computed tomography was further used for three-dimensional reconstruction of spinal cord architecture. The efficiency of in vivo REST and NOTCH1 modulation was validated by western blot analysis, while regenerative responses were evaluated using behavioral and histological assessments. Therefore, the in vivo intervention was interpreted mainly based on protein-level validation, functional recovery, and histological changes, rather than transcriptome-wide specificity analysis.

Statistical analysis

All statistical analyses were conducted within a standardized data processing framework, using R (v4.3) and Python (v3.11) as computational platforms. For comparisons between two groups, the Student’s t-test was used if the data passed the Shapiro-Wilk test for normality and the Levene test for homogeneity of variance. If either assumption was violated, the Mann-Whitney U test was applied. For multiple group comparisons, one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test was used when normality was met. In cases of non-normal distributions, the Kruskal-Wallis test was employed, followed by Dunn’s method for multiple-comparison correction. Statistical significance was reported only for experiments with independent biological replicates. Single-sample omics tracks, including ATAC-seq and ChIP-seq browser views, were not subjected to statistical testing and were interpreted only as exploratory supportive evidence.

RESULTS

Public database analysis reveals stage-specific interaction dynamics between REST and NOTCH1 during hNSC state transitions

At the initial stage of this study, we reanalyzed scRNA-seq data from the PFC in the public dataset GSE104276. This dataset includes developmental samples from gestational week 8 to 26, covering major neural lineage cell types and stem/progenitor cell populations. Using the standardized expression matrix provided by the original authors, we performed UMAP-based dimensionality reduction and clustering, followed by cell-type annotation according to the system described in the source publication[20]. Characteristic marker genes were specifically enriched in their corresponding cell populations (Supplementary Figure 1A), allowing identification of major cell types, including astrocytes, excitatory neurons, inhibitory neurons, oligodendrocyte precursor cells, NPCs, and microglia (Figure 1A). Subpopulation analysis revealed marked heterogeneity in REST expression within NPCs. Based on REST activity, NPCs were stratified into REST-high and REST-low populations, corresponding to quiescent-like and activated-like transcriptional states, respectively (Figure 1B). In this study, “quiescent-like” refers to a REST-high, low-proliferative NPC/NSC-like transcriptional state in developmental and organoid-derived systems, rather than the deep quiescence of adult subventricular zone NSCs in vivo, which involves distinct molecular features such as GFAPδ expression[21,22]. Because REST expression was continuously distributed across NPCs, this grouping was used only as an analytical strategy to enhance state contrast and does not imply a strictly binary classification of NSCs. These findings suggest that REST signaling may participate in balancing self-renewal and activation within heterogeneous NSC/NPC populations.

Figure 1 Inverse dynamic changes of RE1-silencing transcription factor and neurogenic locus notch homolog protein 1 reveal transcriptional state transitions from quiescence to activation in human neural progenitor cells. A: UMAP plot illustrating single-cell clustering results based on the public dataset GSE104276; B: Neural progenitor cells (NPCs) were subdivided into subpopulations according to RE1-silencing transcription factor expression levels; the term “quiescent-like” denotes a low-proliferative NPC-like state inferred from transcriptional features, not adult NSC deep quiescence; C and D: Pseudotime trajectory of NPCs reconstructed using Monocle3, depicting developmental progression; E: Relative expression patterns of key transcription factors - ASCL1, DCX, HES1, NEUROD, neurogenic locus notch homolog protein 1, and RE1-silencing transcription factor - across the pseudotime trajectory; F: Pseudotime-ordered heatmap showing dynamic gene expression changes along the trajectory. OPCs: Oligodendrocyte precursor cells; NPCs: Neural progenitor cells; REST: RE1-silencing transcription factor.

To further elucidate the dynamic progression underlying these state differences, we reconstructed the pseudotime trajectory of NPCs using Monocle3[23-27] (Figure 1C and D). The REST-high population was selected as the trajectory root based on REST’s known transcriptional repressive function and its enrichment in low-proliferative NPC-like cells. Because pseudotime analysis infers cell ordering from transcriptional similarity, the trajectory should be interpreted as a transcriptional state continuum of NPC activation rather than direct real-time lineage progression. Along this inferred trajectory, cells gradually shifted from quiescent-like to activated-like states and toward neuronal differentiation. Key transcription factors, including ASCL1, DCX, HES1, NEUROD, NOTCH1, and REST, showed continuous and stage-specific expression dynamics along pseudotime (Figure 1E). REST was highly expressed at the early quiescent-like stage and declined as cells progressed toward activation, suggesting a potential role in maintaining the low-proliferative state. In contrast, NOTCH1 displayed a biphasic pattern, increasing during the intermediate activation phase, reaching a peak, and then declining in fully activated NSCs. These opposite temporal trends suggest that REST and NOTCH1 may act in a stage-dependent and partially antagonistic manner during NSC state transitions. To validate the reliability of these pseudotime findings, we replotted the continuous expression trends of key genes within the same dataset (Supplementary Figure 1B). ASCL1, DCX, HES1, NEUROD2, NOTCH1, and REST showed trajectories consistent with those in Figure 1E, with smoother dynamic curves. REST declined gradually from early pseudotime points, whereas NOTCH1 peaked at the intermediate phase before decreasing, further supporting their stage-specific inverse dynamics during NSC activation.

The pseudotime heatmap further revealed gradual expression changes across multiple gene modules along the trajectory (Figure 1F). REST-associated genes were enriched at the early quiescent-like stage, whereas NOTCH1-related activation signatures increased during intermediate states, coinciding with the emergence of neuronal differentiation markers. At the single-cell level, REST and NOTCH1 showed a moderate positive correlation across NPCs (Supplementary Figure 1C). However, such co-expression does not demonstrate direct regulation or causality and may reflect shared variation driven by mixed cell states or developmental stages. Combined with their distinct pseudotime dynamics, this correlation suggests a possible time-dependent and stage-specific window of association between REST and NOTCH1, which requires further functional validation[28-31].

Sensitivity analyses using median split and top/bottom quartile definitions of REST expression yielded consistent directional patterns for key neurogenic and cell-cycle-related genes, supporting the robustness of REST-gradient-based stratification rather than dependence on a single arbitrary threshold (Supplementary Table 1). Together, these findings suggest a stage-dependent relationship between REST and NOTCH1 during developmental NPC state transitions, which was further tested in functional hNSC assays.

REST inhibition promotes G1/S progression and hNSC proliferation, whereas NOTCH1 activation partially counterbalances this effect

To validate the inverse dynamic relationship between REST and NOTCH1 during NSC activation revealed by the public database analysis, we performed functional verification in an in vitro NSC model by NICD overexpression and REST inhibition (Figure 2A). Flow cytometric analysis of cell cycle distribution (n = 3 per group) showed that REST knockdown significantly promoted S-phase entry, increasing the proportion of S-phase cells from 18.4% in controls to 32.6% (P = 0.004), while reducing the G1-phase population from 60.3% to 47.9% (P = 0.011). In contrast, NICD overexpression prolonged G1 phase (67.1% vs 53.8%, P = 0.009) and reduced S-phase entry (14.2% vs 18.4%, P = 0.047), consistent with a role for NOTCH1 activation in restraining proliferative progression (Figure 2B and C).

Figure 2 Antagonistic effects of RE1-silencing transcription factor and neurogenic locus notch homolog protein 1 on cell cycle regulation in neural stem cells. A: Western blot analysis was performed to detect the expression of RE1-silencing transcription factor and NICD in each group, with β-actin used as the internal control; B: Representative histograms from flow cytometry analysis showing cell cycle distribution across experimental groups; C: Quantitative analysis of the proportion of cells in G1, S, and G2/M phases; D: Western blot analysis of cyclin D1, cyclin E, CDK2, and p27^Kip1 protein levels; β-actin served as the loading control; E: QPCR analysis of CDC6 and MCM2 mRNA expression levels (normalized to GAPDH); F: Representative EdU staining images across treatment groups. EdU-positive cells indicate proliferating cells, and nuclei were counterstained with DAPI; G: Flow cytometry analysis of cell cycle distribution based on propidium iodide staining; H: Scatter plot of Pearson correlation between RE1-silencing transcription factor/neurogenic locus notch homolog protein 1 protein levels and the proportion of EdU-positive cells. Data are presented as mean ± SD from three independent biological replicates. Statistical analyses were performed using one-way ANOVA followed by Tukey’s post hoc test or two-tailed Student’s t-test where appropriate. ^aP < 0.05, ^bP < 0.01, ^cP < 0.001, ^dP < 0.0001. REST: RE1-silencing transcription factor.

In the dual intervention group (NICD + siREST), the proportion of S-phase cells partially decreased to 25.3% from 32.6% in the siREST group (P = 0.036), while the G1-phase proportion significantly declined compared to the NICD group (59.2% vs 67.1%, P = 0.022). These results indicate that REST inhibition can partially counteract NOTCH1-induced G1 phase prolongation but does not fully reverse its effects. Analysis of cyclin expression (Figure 2D) showed that cyclin E and CDK2 were upregulated in the siREST group, while cyclin D1 was upregulated and p27^Kip1 significantly increased in the NICD group (P < 0.05), further supporting the distinct directional roles of REST and NOTCH1 in G1/S checkpoint regulation. Transcript-level analysis (Figure 2E) revealed significant upregulation of REST target genes CDC6 and MCM2 in the siREST group (both P < 0.01), while the NICD group exhibited a downward trend in their expression (P < 0.05).

EdU incorporation assays further supported these findings. REST inhibition increased the proportion of EdU-positive cells compared with controls (28.7% vs 15.9%, P = 0.002), whereas NICD overexpression reduced EdU incorporation to 11.4% (P = 0.006 vs control). The dual-intervention group showed an intermediate EdU-positive rate of 21.2%, lower than that of the siREST group (P = 0.018) but higher than that of the NICD group (P = 0.041) (Figure 2F and G). Correlation analysis further showed that EdU positivity was negatively associated with REST protein expression and NOTCH1 activation levels, supporting their opposing effects on hNSC proliferative entry (Figure 2H). Together, these results indicate that REST inhibition facilitates G1/S transition and proliferative activation of hNSCs, whereas NOTCH1 activation imposes a counterbalancing delay on this process.

The REST-NOTCH1 transcriptional axis regulates NSC quiescence-activation transitions through dynamic chromatin accessibility

To investigate whether REST and NOTCH1 exert cooperative or antagonistic regulatory effects during NSC activation, we integrated ATAC-seq and ChIP-seq data to examine chromatin accessibility and transcription factor binding at representative target loci. At the REST locus, REST ChIP-seq signals were markedly enriched, suggesting potential autoregulatory activity. NOTCH1 binding was also detected in adjacent regions, while ATAC-seq signals at the REST promoter were visibly increased (Figure 3A). These findings indicate a transcriptionally active chromatin state at the REST locus and suggest that REST autoregulation may be influenced by NOTCH1 signaling.

Figure 3 Chromatin accessibility and binding profiles of RE1-silencing transcription factor and neurogenic locus notch homolog protein 1 at representative target gene loci. A-C: The panels respectively display assay for transposase-accessible chromatin using sequencing (ATAC-seq) and chromatin immunoprecipitation sequencing (ChIP-seq) signal distributions at the genomic loci of RE1-silencing transcription factor (REST), DCX, and MKI67. ATAC-seq data were obtained from human neural stem cells under conditions of high REST expression and neurogenic locus notch homolog protein 1 activation. ChIP-seq data include binding signals for REST and neurogenic locus notch homolog protein 1. Signal intensities are presented as normalized read counts (p-2q range). Genomic annotations and scale bars are based on the hg38 reference genome. The horizontal axis represents the genomic coordinate range, with chromosome position and interval length indicated; the scale is shown in kb, and arrows denote the direction of gene transcription. It should be noted that the ATAC-seq and ChIP-seq data for each condition were derived from a single sample (n = 1). The figure therefore presents representative signals from selected genomic regions for exploratory mechanistic analysis. ATAC-seq and ChIP-seq tracks are representative genome-browser views from one biologically independent sample per condition. These data were used for exploratory visualization only, and no statistical differential accessibility or binding analysis was performed. Signal intensity was normalized as reads per genomic content. REST: RE1-silencing transcription factor; ATAC-seq: Assay for transposase-accessible chromatin using sequencing; ChIP-seq: Chromatin immunoprecipitation sequencing.

At the DCX locus, a marker of neuronal differentiation, REST and NOTCH1 showed distinct ChIP-seq enrichment patterns. REST was mainly enriched at the proximal promoter, whereas NOTCH1 binding was observed near an upstream enhancer (Figure 3B). At the MKI67 locus, a proliferation-associated gene, REST and NOTCH1 displayed complementary binding patterns (Figure 3C). REST signals appeared reduced in the activated-like state, whereas NOTCH1 binding and ATAC-seq accessibility increased, suggesting a chromatin state permissive for cell-cycle-related transcription during NSC activation.

Together, these exploratory chromatin analyses suggest that REST and NOTCH1 are associated with stage-dependent regulatory patterns at representative loci. REST appears to maintain a relatively repressive chromatin environment in the quiescent-like state, whereas reduced REST activity coincides with increased NOTCH1-associated chromatin accessibility during activation. These observations support a model in which REST and NOTCH1 may act in a temporally coordinated, partially antagonistic manner during the NSC quiescence-to-activation transition, particularly around the G1/S checkpoint. However, because the ATAC-seq and ChIP-seq data were derived from single-sample datasets, these findings should be interpreted as supportive and hypothesis-generating rather than as independent causal validation.

Confocal imaging reveals REST-mediated regulation of SOX2 and MCM2 nuclear localization

ATAC-seq and ChIP-seq analyses suggested that REST may regulate cell cycle-related programs by modulating chromatin accessibility. To further validate this mechanism, we examined the subcellular localization of SOX2 and MCM2 in NSCs under different genetic interventions using high-resolution confocal immunofluorescence microscopy (Figure 4A). Quantitative analysis showed that REST inhibition significantly increased nuclear SOX2 fluorescence intensity, with a 1.46-fold elevation compared with the control group (P = 0.001, Figure 4B). Similarly, the proportion of MCM2-positive cells increased by 1.32-fold (P = 0.003, Figure 4C), suggesting enhanced nuclear accumulation of stemness- and replication-associated factors. High-magnification imaging further showed a more punctate or speckled nuclear distribution of SOX2 in the siREST group (Figure 4D), whereas SOX2 signals in the control and NICD overexpression groups appeared more diffuse with less distinct nucleocytoplasmic boundaries.

Figure 4 Nuclear localization changes of SOX2 and MCM2 under RE1-silencing transcription factor and Notch signaling regulation. A: Schematic diagram illustrating the regulatory mechanism; B: Quantification of nuclear fluorescence intensity of SOX2 across groups, presented as bar plots; statistical significance was determined by two-tailed t-test; C: Bar graph showing the percentage of MCM2-positive cells in each group; D: Representative high-magnification images demonstrating enhanced nuclear accumulation of SOX2 in the RE1-silencing transcription factor interference group (white arrows); E: Pearson correlation scatter plot with fitted curve illustrating the relationship between SOX2 nuclear intensity and the proportion of MCM2-positive cells; F: Heatmap of SOX2 and MCM2 nuclear localization levels. Colors represent relative expression levels; rows indicate samples, and columns denote gene/protein signal channels. ^aP < 0.05, ^bP < 0.01, ^cP < 0.001. REST: RE1-silencing transcription factor.

Conversely, in the NICD overexpression group (NOTCH1 activation), nuclear SOX2 signal intensity was significantly reduced (fold change = 0.74, P = 0.012), along with a decrease in the proportion of MCM2-positive cells (fold change = 0.69, P = 0.018), indicating suppression of stemness- and proliferation-associated nuclear programs. In the dual-intervention group (NICD + siREST), SOX2 and MCM2 nuclear localization showed intermediate levels between the single-intervention groups and the control (SOX2: P = 0.021; MCM2: P = 0.037), suggesting that REST inhibition partially counteracts the suppressive effect of NOTCH1 activation, though not completely.

Correlation analysis further revealed a strong positive association between nuclear SOX2 intensity and the proportion of MCM2-positive cells across all experimental groups (r = 0.82, P < 0.001), which was even stronger under siREST conditions (r = 0.89, P < 0.001; Figure 4E). Heatmap visualization showed the highest SOX2 and MCM2 nuclear signals in the siREST group, the lowest signals in the NICD group, and intermediate levels in the dual-intervention group (Figure 4F). Together with the cell-cycle and EdU results, these findings indicate that REST inhibition promotes nuclear accumulation of stemness- and replication-associated factors, whereas NOTCH1 activation restrains this response.

Metabolomics analysis reveals REST suppression facilitates the metabolic shift from quiescence to activation in NSCs

To investigate the metabolic effects of REST signaling, we performed metabolomic profiling of REST-suppressed and control NSCs, followed by KEGG pathway enrichment analysis of differentially abundant metabolites. REST suppression led to significant enrichment of pathways related to central carbon metabolism, the pentose phosphate pathway, amino acid metabolism, and glycerophospholipid metabolism (Figure 5). These changes indicate enhanced energy production and biosynthetic activity, providing metabolic support for the increased G1/S progression and EdU incorporation observed after REST inhibition. In particular, activation of central carbon metabolism and the pentose phosphate pathway may facilitate ATP production, NADPH generation, and nucleotide biosynthesis required for proliferative entry. Together with the chromatin accessibility and transcriptional findings, these data suggest that REST inhibition coordinates metabolic remodeling with transcriptional derepression to support hNSC activation.

Figure 5 Enrichment and distribution analysis of metabolic pathways affected by RE1-silencing transcription factor inhibition in human neural stem cells. A: Bar plot of enriched differential metabolic pathways between RE1-silencing transcription factor inhibition and control groups. Differential metabolites were identified via liquid chromatography tandem mass spectrometry and analyzed for Kyoto Encyclopedia of Genes and Genomes pathway enrichment using the BioDeep platform. Enrichment significance is expressed as -log₁₀(P); B: Bubble plot displaying enriched differential pathways. Significance [-log₁₀(P)] and impact values were calculated via topological analysis; bubble size corresponds to the number of differential metabolites involved in each pathway. Data were obtained from four independent biological replicates per group (n = 4 per group; total n = 8).

Live imaging confirms that REST inhibition accelerates hNSC activation dynamics

To determine whether the cell-cycle and EdU-based proliferation changes reflected altered activation dynamics in a three-dimensional context, we performed multiphoton live-cell imaging of Nestin/Ki67 double-positive cells over 48 hours. REST inhibition accelerated the transition of dormant-like hNSCs into an activated state, whereas NICD overexpression delayed this process. The dual-intervention group showed an intermediate activation pattern, consistent with the partial counterbalancing effect observed in the cell-cycle and EdU assays (Figure 6). These time-resolved imaging data support the conclusion that REST inhibition promotes early hNSC activation, while NOTCH1 activation delays the kinetics of this transition.

Figure 6 Multiphoton imaging reveals that RE1-silencing transcription factor inhibition accelerates the transition of neural stem cells from dormancy to activation. A: Representative multiphoton images of Nestin+/Ki67+ cells in the NICD overexpression and control groups at 0 hour and 48 hours (Nestin: Green; Ki67: Red; DAPI: Blue), along with bar graphs showing the proportion of activated cells; B: Representative imaging and quantification of activation proportions in the RE1-silencing transcription factor (REST) inhibition and control groups; C: Comparison of activation rates across four treatment groups (vector control, REST inhibition, NICD activation, dual intervention), with statistical significance indicated; D: Time-course activation curves of neural stem cells based on 48 hours continuous imaging data, showing a marked increase within the first 24 hours in the REST inhibition group and a delayed activation in the NICD activation group; E: Kaplan-Meier-style analysis of neural stem cell activation latency, comparing the time distribution for transition from dormancy to activation across treatment groups. ^aP < 0.05, ^bP < 0.01, ^dP < 0.0001. REST: RE1-silencing transcription factor.

Single-cell analysis of organoids reveals dynamic shifts in REST activity accompanying NSC activation and neurogenesis

Following the identification of dynamic REST-NOTCH1 interactions in public developmental datasets, we further validated these findings at the single-cell level using human neural organoids. Integrated scRNA-seq analysis across DIV20, DIV30, DIV45, and DIV60 revealed partially overlapping but stage-associated distributions of organoid-derived cells in UMAP space, indicating transcriptional heterogeneity during organoid maturation (Supplementary Figure 2A). Canonical neural stem/progenitor cell markers, including SOX2, NES/Nestin, and PAX6, were detectable across developmental stages and enriched in specific cell populations (Supplementary Figure 2B-D). Dot plot and violin plot analyses further showed that these markers were most abundant at DIV20 and gradually decreased during maturation, while remaining detectable at later stages (Supplementary Figure 2E-H). These results confirm the presence of NSC/NPC-like populations in human neural organoids from DIV20 to DIV60, supporting the suitability of this model for studying neural stem/progenitor state transitions.

Differential expression analysis (Figure 7A) revealed substantial transcriptomic divergence between REST-high and REST-low expression groups. Hierarchical clustering (Figure 7B) showed that REST-high cells predominantly localized to early developmental stages (DIV20-30), with expression profiles enriched for stemness-related and repressive signaling genes. In contrast, REST-low cells became increasingly prevalent in mid-to-late stages (DIV45-60), characterized by upregulation of genes associated with neuronal differentiation and energy metabolism.

Figure 7 Single-cell RNA sequencing of organoids reveal RE1-silencing transcription factor activity variations and their association with neurogenesis. A: Volcano plot illustrating the differentially expressed genes between RE1-silencing transcription factor (REST) high-expression and low-expression groups; B: Heatmap of differentially expressed genes between REST high- and low-expression cells. Data were integrated and normalized across developmental stages (DIV20, DIV30, DIV45, DIV60), and visualized using hierarchical clustering; C: Expression profiles of key developmental marker genes (ASCL1, DCX, HES1, NEUROD2, neurogenic locus notch homolog protein 1, REST) across different developmental stages; D: Gene set enrichment analysis enrichment curve for the neurogenesis-related gene set (GO_NEUROGENESIS), calculated based on ranked differential gene expression, showing significant enrichment of neurogenesis pathways in REST low-expression cells. REST: RE1-silencing transcription factor.

Further analysis of stage-specific marker genes (Figure 7C) demonstrated sequential, phase-dependent expression patterns of ASCL1, DCX, HES1, NEUROD2, and NOTCH1 during development. REST exhibited high expression during early stages and progressively declined as neurogenesis advanced. Conversely, NOTCH1 peaked during the intermediate phase (DIV30-45) before tapering off, revealing an opposing temporal trend relative to REST. Gene set enrichment analysis (Figure 7D) further confirmed this pattern: REST-low cells were significantly enriched in neurogenesis-related gene sets (GO_NEUROGENESIS, NES = 1.23, Padj = 0.0148), indicating a strong association between REST downregulation and the activation of neuronal differentiation programs.

In summary, the organoid single-cell analysis mirrored the trends observed in developmental datasets. REST-high states were associated with early NSC/NPC-like populations, whereas REST downregulation accompanied NOTCH1 dynamics and neurogenic progression. These findings support a temporally coordinated and partially antagonistic relationship between REST and NOTCH1 signaling during human neural progenitor activation and neurogenesis.

Organoid-based experiments reveal the time-dependent regulation of REST and NOTCH1 in a three-dimensional context

Single-cell analyses of organoids suggested a stage-dependent interaction between REST and NOTCH1 during neural progenitor development. To functionally validate this relationship in a three-dimensional context, we performed perturbation experiments in human-derived neural organoids cultured for ≥ 30 days (n = 3 per group, independent biological replicates) after standardized quality control and morphological maturity assessment. Organoids were treated with the Notch inhibitor DAPT, the Notch-activating ligand Jagged1, a REST-specific inhibitor, or REST overexpression, followed by assessment of transcriptional changes and proliferative status at defined time points. Functionally, REST inhibition led to a 1.52-fold increase in the proportion of Ki67+ cells (P = 0.004), suggesting a substantial release of proliferative potential. In contrast, REST overexpression reduced Ki67+ cell proportions to 0.81-fold (P = 0.021), indicating that REST activation tended to maintain a low-proliferation state (Figure 8A). At the molecular level, REST mRNA expression decreased by 22% in the DAPT treatment group compared to the control group (fold change = 0.78, P = 0.017), while Jagged1 treatment significantly increased REST expression by 35% (fold change = 1.35, P = 0.009), indicating a bidirectional regulation of REST by Notch signaling (Figure 8B). Temporal profiling further revealed that Jagged1 treatment upregulated REST within the 7-14 day window, with a non-significant negative trend observed between REST expression and the proportion of Ki67+ cells (r = -0.39, P = 0.258, Figure 8C). Collectively, these findings demonstrate that within the three-dimensional organoid microenvironment, the Notch-REST signaling axis exerts a time-dependent regulatory effect on NSC states, establishing a dynamic balance between proliferative and quiescent phases.

Figure 8 The organoid model reveals time-dependent regulation of cell states by the Notch-RE1-silencing transcription factor axis in a three-dimensional microenvironment. A: Immunofluorescence images and quantitative analysis of Ki67+ cell proportions following RE1-silencing transcription factor (REST) inhibition and activation treatments. Ki67 is shown in green, DAPI in blue. Scale bar = 50 μm; B: Effects of REST inhibition and REST activation on REST mRNA expression levels (n = 3); C: Correlation analysis between REST expression and Ki67+ cell proportions at different time points under Jagged1 treatment, showing a non-significant negative trend (r = -0.39, P = 0.258). ^cP < 0.001. REST: RE1-silencing transcription factor.

REST inhibition promotes re-expression of NSC markers in the lesion area

In a C57BL/6J mouse model of complete SCI at the T9 level, AAV9 vectors were used for in vivo genetic interventions, including control, REST inhibition, NICD overexpression, and combined intervention groups. The efficiency of these manipulations was verified by western blot (Figure 9A). Immunofluorescence imaging (Figure 9B) revealed a significant increase in the proportion of Nestin+ cells within the lesion area in the REST inhibition group, reaching 38.4% ± 4.5%, nearly 1.8-fold higher than that in the control group (21.7% ± 3.8%, P < 0.001). These Nestin+ cells exhibited a band-like aggregation along the periphery of the lesion cavity. The proportion of SOX2+ cells also increased markedly (27.6% ± 3.1% vs 18.4% ± 2.9%, P = 0.002; Figure 9C).

Figure 9 RE1-silencing transcription factor inhibition promotes the re-expression of neural stem cell markers in the lesion area. A: Western blot analysis of RE1-silencing transcription factor (REST) and NICD protein expression in each group, with β-actin used as the internal control; B: Immunofluorescence images showing the distribution of Nestin (green) and DAPI (blue) signals in the control and REST inhibition groups. Scale bar = 25 μm. The bar graph on the right shows the quantitative comparison of the proportion of Nestin+ cells (n = 6, three fields per group); C: Immunofluorescence images displaying SOX2 (red) and DAPI (blue) signals in the control and REST inhibition groups. Scale bar = 25 μm. The bar graph on the right shows the quantitative comparison of the proportion of SOX2+ cells (n = 6, three fields per group); D: Co-localization images of Nestin (green) and SOX2 (red) double-positive cells. Scale bar = 25 μm. The bar graph below shows the quantitative comparison of the proportion of Nestin (green) and SOX2 (red) double-positive cells (n = 6, three fields per group). Statistical analysis was performed using one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test, with significance set at P < 0.05. ^aP < 0.05, ^cP < 0.001, ^dP < 0.0001. REST: RE1-silencing transcription factor.

Co-localization analysis (Figure 9D) further demonstrated a significant rise in the proportion of Nestin+/SOX2+ double-positive cells in the REST inhibition group (15.2% ± 2.3% vs 8.1% ± 1.9%, P < 0.01). These cells displayed elongated processes and diffuse chromatin patterns, consistent with the morphology of activated NSCs. The increase in double-positive cells was positively correlated with Ki67+ proliferative signals (r = 0.68, P = 0.004), suggesting that REST inhibition promotes the activation and potential regenerative capacity of NSCs within the lesion site.

REST inhibition significantly improves motor function recovery in a mouse SCI model

To evaluate whether REST inhibition promotes regenerative outcomes in vivo, we established a complete T9 SCI model in C57BL/6J mice (n = 15 per group), followed by genetic interventions including control, REST inhibition, NICD overexpression, and dual intervention groups. Motor function was assessed longitudinally using the BMS (Figure 10A). On day 7 post-injury, all groups showed similarly low BMS scores (approximately 0-1), indicating comparable acute motor deficits. By day 21, the REST inhibition group reached a BMS score of 6.2 ± 0.7, significantly higher than the control group (4.3 ± 0.8, P < 0.001), indicating partial hindlimb weight-bearing and improved gait coordination. In contrast, the NICD overexpression group showed only modest improvement (3.9 ± 0.9, P = 0.031 vs control). The dual intervention group scored 5.4 ± 0.6, suggesting partial antagonism between the two signaling pathways. Gait analysis (Figure 10B) further illustrated differences in hindlimb movement trajectories among groups. Quantitative analysis of stride length (Figure 10C) showed that REST inhibition markedly improved hindlimb stride length, approaching levels observed in uninjured controls, whereas NICD overexpression led to only minor improvement, and the dual intervention group showed intermediate effects. Histological examination revealed a noticeable reduction in cavity formation at the injury site in the REST inhibition group, as shown by hematoxylin and eosin staining; quantitative analysis confirmed that the cavity area was significantly smaller than in controls (Figure 10D). GFAP immunofluorescence staining demonstrated that REST inhibition reduced astroglial scar density surrounding the lesion, with quantification consistent with the observed functional recovery trend (Figure 10E). In summary, REST inhibition promotes neural regeneration and functional recovery after SCI, whereas NICD overexpression partially attenuates REST inhibition’s beneficial effects.

Figure 10  RE1-silencing transcription factor inhibition significantly enhances motor function recovery following spinal cord injury in mice. A: Basso Mouse Scale score time-course line graph showing significantly higher scores in the RE1-silencing transcription factor inhibition group compared to controls, with the dual-intervention group exhibiting intermediate levels; B: Schematic illustration of footprint-based gait analysis; C: Quantification of hindlimb stride length, revealing a marked increase in the RE1-silencing transcription factor inhibition group; D: Representative hematoxylin and eosin staining images and quantitative analysis of cavity area; E: Representative GFAP immunofluorescence images and quantification of astroglial scar density. ^aP < 0.05, ^cP < 0.001, ^dP < 0.0001. BMS: Basso Mouse Scale; REST: RE1-silencing transcription factor.

DISCUSSION

Current research has advanced our understanding of NSC dormancy and activation, yet limitations remain in model fidelity and mechanistic integration[32,33]. Previous studies have mainly focused on individual pathways, such as Notch and Wnt signaling[34,35], but often lack a multidimensional view of how transcriptional, epigenetic, and metabolic programs are coordinated during NSC state transitions[36]. More critically, conventional two-dimensional culture systems and animal models do not fully capture the complex microenvironment of hNSCs, limiting the translational relevance of mechanistic findings[37]. In this study, we integrated public developmental single-cell datasets, human-derived neural organoids, multimodal omics, and in vivo SCI models to investigate the regulatory role of the NOTCH1-REST axis in NSC state transitions.

NOTCH1 and REST have traditionally been studied as separate regulators of NSC fate[38]. NOTCH1 is widely recognized as a key regulator of NSC undifferentiation[39], whereas REST suppresses premature neuronal gene expression and helps preserve a low-activation state[10]. Our findings suggest that these two factors are dynamically coordinated during NSC activation. REST expression was enriched in quiescent-like or low-proliferative states and declined during activation, while NOTCH1 showed a stage-specific peak during the intermediate phase. This pattern supports a model in which REST downregulation acts as an initial permissive step, relieving transcriptional repression and enabling proliferative entry, whereas NOTCH1 provides a time-dependent modulatory signal that shapes the activation trajectory. Thus, rather than acting independently, REST and NOTCH1 appear to form a temporally coordinated regulatory axis that controls the balance between NSC maintenance and activation.

At the epigenetic level, previous studies have primarily focused on REST-mediated gene silencing through the HDAC complex[11], whereas its relationship with chromatin accessibility during NSC state transitions remains less well defined. In this study, integrated ATAC-seq and ChIP-seq analyses suggested stage-associated changes in REST and NOTCH1 binding profiles, accompanied by alterations in chromatin openness at representative target loci. Promoter accessibility of several key genes appeared to increase after relief of REST-associated repression or in association with NOTCH1 activity, indicating that these factors may contribute to transcriptional remodeling through chromatin-level regulation. These findings support a model in which dynamic chromatin plasticity participates in NSC fate regulation, while also highlighting the need for further replicated epigenomic validation.

In addition, metabolomic profiling showed that REST inhibition was associated with enrichment of glycolysis-related pathways, central carbon metabolism, and biosynthetic metabolic programs. These changes are consistent with the increased energy and anabolic demands required for NSC transition from a low-proliferative state to activation, and may help explain the enhanced EdU incorporation, increased MCM2-associated replication activity, and accelerated activation dynamics observed in vitro. Thus, REST may influence NSC fate not only through transcriptional regulation but also by supporting metabolic remodeling during activation. However, metabolomics was performed only under REST knockdown conditions and not in NOTCH1 activation or inhibition groups. Moreover, metabolomic profiling was not conducted on injured spinal cord tissues in the SCI model. Therefore, these findings should be interpreted as mechanistic support for REST-driven hNSC activation rather than direct evidence that the NOTCH1-REST axis regulates metabolism in vivo. Future metabolomic analyses under NOTCH1 modulation and in SCI tissues will be required to determine whether NOTCH1 synergizes with or antagonizes REST-dependent metabolic regulation.

From a technical perspective, this study integrates human-derived neural organoids with multimodal omics to provide a multidimensional framework for investigating NSC state transitions. The organoid model offers a human-relevant three-dimensional context that partially preserves neural tissue architecture and microenvironmental features, while scRNA-seq, ATAC-seq, ChIP-seq, and metabolomics enable the coordinated assessment of transcriptional activity, chromatin accessibility, transcription factor binding, and metabolic remodeling. However, several technical limitations should be noted. The ATAC-seq and ChIP-seq analyses included only one sample per condition. Therefore, the observed signal changes should be regarded as exploratory and hypothesis-generating. Although these patterns were consistent with transcriptomic and functional results, future studies with biological replicates are required to confirm the reproducibility of binding events and chromatin accessibility changes. In addition, although the organoid model provided stage-resolved single-cell transcriptomic evidence for NSC/NPC-like populations, independent immunofluorescence validation of SOX2, Nestin, and PAX6 across organoid developmental stages was not performed. Future studies incorporating independent organoid batches, spatial immunostaining, and quantitative marker distribution analysis will further strengthen model validation.

NSC populations exhibit substantial functional and molecular heterogeneity. After neural injury, distinct NSC subpopulations may occupy quiescent, primed, or activated states[40,41]. Therefore, single-cell transcriptomics was used in this study to resolve NSC/NPC population heterogeneity rather than to treat these cells as a homogeneous population. At single-cell resolution, we identified REST-high and REST-low subpopulations and further showed, through pseudotime analysis, that they represent different positions along a continuous state-transition trajectory. Thus, REST expression should be interpreted as a continuous regulatory gradient rather than a discrete classification marker. In this study, REST-based stratification was used as an analytical tool to capture differences in regulatory-axis activity. This interpretation is supported by convergent evidence from transcriptional changes, chromatin accessibility patterns, and functional experiments. Moreover, this dynamic regulatory pattern was further supported in human neural organoids and the SCI model, indicating that the NOTCH1-REST axis regulates NSC state transitions at the population level. These findings provide a molecular framework for understanding heterogeneous NSC regulation during neural regeneration.

Importantly, this study further demonstrated the functional relevance of REST inhibition in a mouse SCI model. REST inhibition increased the re-expression of Nestin+/SOX2+ NSC/progenitor-like markers in the lesion area and was associated with improved motor function recovery. These findings extend the in vitro and organoid-based mechanistic observations to an in vivo injury context, suggesting that REST may influence regenerative responses after CNS injury. Although further validation is required to confirm cell-type specificity and therapeutic safety, these results support REST as a potential target for NSC-based neuroregeneration strategies.

Despite these findings, several limitations should be considered. First, pseudotime analysis using Monocle3 provides an inferred model of transcriptional state transitions rather than direct temporal tracking. The trajectory direction depends on root-node selection, and the REST-high population was designated as the root based on prior biological knowledge of REST as a transcriptional repressor, rather than unbiased lineage tracing. Second, the “quiescent-like” state defined in the developmental and organoid models represents a slow-cycling progenitor-like state, which differs from the deep quiescence of adult subventricular zone NSCs characterized by markers such as GFAPδ and profound metabolic suppression[21,22]. Therefore, our findings primarily reflect the dynamic transition within the NPC continuum during expansion or regeneration, rather than reactivation of long-term dormant adult NSCs. Another limitation concerns the specificity of AAV9-dCas9-KRAB-mediated REST inhibition in the SCI model. Although western blot confirmed reduced REST protein expression after vector delivery, transcriptome-wide RNA-seq or in vivo ChIP-seq was not performed in injured spinal cord tissues. Therefore, potential sgRNA off-target effects, AAV9 tropism, and broader dCas9-KRAB-mediated epigenetic repression cannot be fully excluded. Accordingly, the in vivo findings should be interpreted primarily as functional and histological evidence linking REST inhibition with improved regenerative responses. Future studies using multiple independent sgRNAs, non-targeting controls, rescue experiments, cell-type-specific delivery, and single-nucleus transcriptomic validation will be needed to confirm the specificity and safety of REST-targeted intervention in vivo. Finally, although this study identifies the REST-NOTCH1 axis as a key regulator of NSC state transitions, the broader upstream signals, downstream effectors, and cell-type-specific regulatory networks remain to be further delineated.

CONCLUSION

In summary, this study identifies the NOTCH1-REST transcriptional axis as an important regulator of NSC quiescence-like-to-activation transitions and provides a multidimensional framework linking transcriptional regulation, chromatin remodeling, and metabolic adaptation (Figure 11). These findings advance our understanding of NSC state control and suggest that REST inhibition may create a permissive state for NSC activation and regenerative responses. From a translational perspective, REST-targeted modulation may represent a potential strategy for promoting neural repair after SCI and other CNS injuries. Future studies integrating spatial transcriptomics, patient-derived models, and more specific in vivo validation will further clarify the therapeutic potential of this regulatory axis.

Figure 11 Graphical abstract. Schematic of the multi-omics mechanism by which the neurogenic locus notch homolog protein 1-RE1-silencing transcription factor transcriptional axis regulates the dormancy-to-activation transition in human neural stem cells through epigenetic and metabolic reprogramming. Notch1: Neurogenic locus notch homolog protein 1; NSC: Neural stem cell; ATAC-seq: Assay for transposase-accessible chromatin using sequencing.