Neuronal Overexpression of Glycogen Synthase Kinase‐3β Accelerates Age‐Related Glial, Vascular, and Inflammatory Remodeling of the Murine Dentate Gyrus

ABSTRACT

Adult hippocampal neurogenesis (AHN)—the lifelong generation of new neurons in the dentate gyrus (DG)—depends on the integrity of the subgranular zone (SGZ) neurogenic niche, which comprises astrocytes, microglia, oligodendroglial cells, and a dense vascular network. Although AHN declines during aging and is impaired in Alzheimer's disease (AD), the cell‐type‐specific alterations within the DG underlying this decline remain poorly characterized. Given the critical role of glycogen synthase kinase‐3β (GSK‐3β) in AD pathogenesis and AHN regulation, we studied how the overexpression of this kinase in dentate granule cells (DGCs) remodels the cellular and vascular composition of the DG throughout life. Using a layer‐resolved approach (SGZ vs. granule cell layer (GCL)) and mice of distinct ages (4–30 months), we quantified glial densities, cellular contacts, vascular architecture, and cytokine profiles in wild‐type (WT) and GSK‐3β‐overexpressing (GSK‐3β‐OE) mice. In the former, aging was associated with a layer‐specific reorganization, characterized by progressive astrocyte loss in the SGZ, midlife increases in GCL microglia, a reduction in oligodendroglial progenitors, and enhanced SGZ vascularization. GSK‐3β overexpression accelerated age‐related changes, triggering premature SGZ astrocyte depletion by 4 months, broader disruption featuring reduced SGZ microglial and mature oligodendrocyte densities accompanied by increased microglial morphological complexity, vascular restructuring, and a shift toward a pro‐inflammatory environment. Therefore, neuronal GSK‐3β overexpression promoted a non‐cell‐autonomous reorganization of the DG, consistent with premature remodeling of the neurogenic niche. These results provide mechanistic insights into AHN failure in AD and point to the modulation of GSK‐3β activity as a strategy to preserve DG integrity.

• Aging induces layer‐specific changes in glial populations and vascular organization across the dentate gyrus.

• Neuronal GSK‐3β overexpression accelerates age‐related remodeling of the dentate gyrus neurogenic niche.

• GSK‐3β‐overexpressing mice show premature subgranular zone (SGZ) astrocyte loss, altered microglia and oligodendroglia, and vascular restructuring.

• Chronic neuronal GSK‐3β overexpression promotes a pro‐inflammatory, non‐cell‐autonomous disruption of dentate gyrus (DG)ed homeostasis.

Abbreviations

AD

Alzheimer's disease

AHN

adult hippocampal neurogenesis

BBB

blood–brain barrier

BSA

bovine serum albumin

CamKII

calcium/calmodulin‐dependent protein kinase II

CBMSO

Centro de Biología Molecular Severo Ochoa

CD31

cluster of differentiation 31

DAPI

4′,6‐diamidino‐2‐phenylindole

DG

dentate gyrus

DGCs

dentate granule cells

GCL

granule cell layer

GFAP

glial fibrillary acidic protein

GFP

green fluorescent protein

GSK‐3

glycogen synthase kinase 3

GSK‐3β‐OE

GSK‐3β‐overexpressing

HC‐AR

heat‐mediated citrate buffer antigen retrieval

Iba1

ionized calcium‐binding adaptor molecule 1

ICAM

intercellular adhesion molecule

IGF

insulin growth factor

IHC

immunohistochemistry

IL

interleukin

JACoP

Just Another Colocalization Plugin

LAS X

Leica Application Suite X

LVCC

large volume computational clearing

NG2

neural/glial antigen 2

NSCs

neural stem cells

Olig2

oligodendrocyte transcription factor 2

OPCs

oligodendrocyte progenitor cells

PB

phosphate buffer

PFA

paraformaldehyde

ROIs

regions of interest

RT

room temperature

SGZ

subgranular zone

UEA‐1

Ulex europaeus agglutinin‐1

VEGF

vascular endothelial growth factors

WT

wild‐type

1. Introduction

The hippocampal dentate gyrus (DG) retains the capacity to generate new neurons throughout life, a process termed adult hippocampal neurogenesis (AHN) (Altman and Das 1965). Anatomically, the DG consists of a densely packed layer of dentate granule cells (DGCs), known as the granule cell layer (GCL), and a thin underlying band of cells called the subgranular zone (SGZ). The SGZ forms the DG neurogenic niche and contains the neural stem cells (NSCs) that give rise to neural progenitors, which eventually differentiate into newborn DGCs (Kempermann et al. 2004; Seri et al. 2001).

The maturation of DGCs is orchestrated by the highly specialized micro‐environment of the SGZ. This niche comprises local interneurons and glial cells (including astrocytes, microglia, and oligodendrocytes) intertwined with a profuse vascular network, all embedded within a specialized extracellular matrix (Vicidomini et al. 2020). Astrocytes are fundamental for maintaining neuronal homeostasis through the secretion of soluble factors and gliotransmitters, as well as through their contribution to maintaining the integrity of the blood–brain barrier (BBB) (Vicidomini et al. 2020). Oligodendrocytes form and maintain myelin sheaths, provide metabolic support, and modulate immune responses (Guo et al. 2025). Microglia, the resident immune cells of the brain, originate from the yolk sac and migrate to the brain during early embryonic stages (Paolicelli et al. 2022). Among their multiple functions, microglia contribute to the regulation and maturation of new DGCs by phagocytosing apoptotic cells (Sierra et al. 2010), and they play a crucial role in neurodegenerative diseases by modulating the neuroinflammatory response (Paolicelli et al. 2022). Finally, the vasculature is integral to the niche; the SGZ is irrigated by a profuse capillary network that provides the trophic factors and metabolic substrates necessary for DGC development (Kim et al. 2021).

The integrity of the neurogenic process is vulnerable to external stimuli. Physiological aging, neurodegenerative diseases, and neuropsychiatric conditions stand out as primary negative regulators of AHN (Alonso‐Moreno et al. 2025; Marquez‐Valadez et al. 2025; Moreno‐Jimenez et al. 2019; Terreros‐Roncal et al. 2021). Exemplifying this vulnerability, Alzheimer's disease (AD), the most prevalent neurodegenerative condition, is characterized by a sharp reduction in the pool of immature neurons (Disouky et al. 2026; Moreno‐Jimenez et al. 2019; Tosoni et al. 2026; Zhou et al. 2022). While aging remains the primary risk factor for AD, patients with this condition experience memory loss, spatial disorientation, and progressive cognitive decline. At the histological level, the AD brain is characterized by two signature hallmarks, namely extracellular deposits of β‐amyloid plaques and intracellular neurofibrillary tangles of hyperphosphorylated tau protein (Braak and Braak 1991). Crucially, the hippocampal formation is one of the earliest regions to undergo degeneration in AD. These alterations closely correlate with the onset of clinical symptoms (Jack Jr. et al. 2010), and it parallels AHN impairments (Moreno‐Jimenez et al. 2019; Tobin et al. 2019). Notably, the persistence of AHN correlates with cognitive resilience in the context of AD pathology (Briley et al. 2016; Disouky et al. 2026), suggesting that maintaining the regenerative capacity of the hippocampus may help to mitigate functional decline.

At the molecular crossroads of AD pathology and neurogenic dysfunction lies glycogen synthase kinase‐3 (GSK‐3). This kinase presents two isoforms in mammals: GSK‐3α and GSK‐3β, the latter being specifically enriched in the brain (Avila et al. 2010). While GSK‐3β contributes to neurofibrillary tangle formation in AD (Avila et al. 2010), it also plays a key role in regulating AHN (Fuster‐Matanzo et al. 2013). Indeed, mice overexpressing GSK‐3β (GSK‐3β‐OE mice) recapitulate key features of AD, including tau hyperphosphorylation (Lucas et al. 2001), impaired AHN (Fuster‐Matanzo et al. 2013), synaptic dysfunction (Hooper et al. 2007; Llorens‐Martin et al. 2013; Moreno‐Jimenez et al. 2023), spatial learning deficits (Hernandez et al. 2002), and altered pattern separation (Engel et al. 2006; Moreno‐Jimenez et al. 2023). Although GSK‐3β overexpression in this model is specifically driven by the calcium/calmodulin kinase II (CamKII) promoter and is restricted to mature excitatory neurons, previous studies point to the occurrence of non‐cell‐autonomous effects, such as increased apoptosis (Sirerol‐Piquer et al. 2011), inhibitory synaptic dysfunction (Moreno‐Jimenez et al. 2023), and aberrant pro‐inflammatory signaling (Fuster‐Matanzo et al. 2013; Llorens‐Martin et al. 2013). These alterations have been linked to the decline in cognitive flexibility and memory performance observed in these mice (Engel et al. 2006; Hernandez et al. 2002; Llorens‐Martin et al. 2014; Rodriguez‐Matellan et al. 2020). However, while these findings suggest that GSK‐3β overexpression influences the surrounding microenvironment, the cell‐type specific impact of chronic GSK‐3β overexpression on the homeostasis of the DG remains largely unexplored.

In this regard, we hypothesized that GSK‐3β overexpression promotes premature and non‐cell‐autonomous dysregulation of the DG microenvironment, which further contributes to AHN failure. To test this notion, we conducted a comprehensive layer‐resolved analysis of the SGZ neurogenic niche and the adjacent GCL, spanning from early adulthood to advanced old age. Our results revealed profound alterations in the cellular and vascular components of the DG, thereby offering a more complete framework for understanding the neurogenic failure and cognitive impairments in this model.

2. Methods

2.1. Experimental Design

To study the effects of aging and GSK‐3β overexpression on the vascular and cellular components of the DG micro‐environment, we performed a comparative analysis across different ages in wild‐type (WT) and GSK‐3β‐OE mice. To this end, male mice were sacrificed at four ages: 4 (WT n = 8, GSK‐3β‐OE n = 8), 12 (WT n = 7, GSK‐3β‐OE n = 12), 18 (WT n = 5, GSK‐3β‐OE n = 6), and 30 (WT n = 4, GSK‐3β‐OE n = 5) months of age. The right hemisphere was used for immunohistochemistry (IHC), and the left hippocampi for cytokine array analyses.

To assess whether GSK‐3β overexpression alter the contacts between glial cells, the vasculature, and newborn DGCs, a Cerulean‐encoding retrovirus was stereotaxically injected into the hippocampus to label newborn DGCs in female mice. Mice were sacrificed 1 (WT n = 3, GSK‐3β‐OE n = 3) or 8 weeks (WT n = 3, GSK‐3β‐OE n = 4) after injection.

2.2. Animals

Transgenic mice overexpressing GSK‐3β under the control of the neuronal CamKII promoter, and their WT littermates were generated as previously reported (Lucas et al. 2001). Female C57BL/6J‐OlaHsd mice (5–6‐week‐old) were purchased from Envigo Laboratories. All mice were maintained in the animal facility at the Centro de Biología Molecular Severo Ochoa (CBMSO) under European Community regulations (Directive 86/609/EEC) and were handled in accordance with both European and local guidelines for animal welfare. The mice were allowed at least 1 week of acclimation before experimental procedures and were housed in groups of four per cage. All protocols were approved by the CBMSO Animal Ethics Committee (AEEC‐CBMSO‐23/172) and the National Committee for Animal Experimentation (PROEX 185.4/20). All experiments complied with the applicable ethical standards for animal research.

2.3. Sacrifice

Mice were deeply anesthetized by intraperitoneal injection of pentobarbital (EutaLender, 60 mg/kg) and sacrificed at 4, 12, 18, and 30 months of age. They were transcardially perfused with 0.9% saline solution. The brains were carefully removed from the skull, and the two hemispheres were separated. The right hemisphere was fixed for 24 h at 4°C in freshly prepared 4% paraformaldehyde (PFA) solution (pH = 7.4), obtained by diluting a commercial 16% PFA solution (Electron Microscopy Sciences, #15710) in 0.2 N phosphate buffer (PB) and bi‐distilled water (1:2:1). The hippocampus from the left hemisphere was dissected and stored at −80°C for cytokine array analysis.

Animals injected with a Cerulean‐encoding retrovirus were transcardially perfused with 0.9% saline, followed by 4% PFA. The brains were then post‐fixed in 4% PFA for 24 h at 4°C before sectioning.

2.4. Brain Tissue Sectioning

Brains were rinsed three times in 0.1 N PB, and 50 μm‐thick coronal sections were cut using a Leica VT1200S vibratome (Leica Biosystems, Wetzlar, Germany). A series of sections was collected systematically, with one slice taken from every eighth section, yielding a final sampling probability of 1/8. The sections were immediately stored at −20°C in 24‐well plastic plates containing a cryopreservative solution composed of 30% polyethylene glycol, 10% 0.2 N PB, 30% glycerol, and 30% bi‐distilled water.

2.5. Immunohistochemistry

Brain sections were rinsed in 0.1 N PB at room temperature (RT) before immunohistochemistry (IHC). For Ulex europaeus agglutinin‐1 (UEA‐1) staining, a heat‐mediated citrate buffer antigen retrieval (HC‐AR) step (Flor‐Garcia et al. 2020) preceded IHC. Briefly, sections from the same animal were placed in 10‐mL glass scintillation vials containing 5 mL of a preheated 1 × citrate buffer (pH 6.0) antigen retrieval solution (prepared by diluting a commercial 10 × stock, Vector, H‐3300, in distilled water). The vials were subjected to five to six short microwave heating cycles (10–20 s each), carefully avoiding boiling of the liquid to prevent tissue damage. Next, the vials were transferred to an 80°C water bath for 20 min and subsequently left at RT for 20 min and rinsed five times in 0.1 N PB before proceeding to triple IHC, as described previously (Llorens‐Martin et al. 2013).

All primary and secondary antibodies (Table S1) were diluted in 0.1 N PB containing 1% Triton X‐100 and 1% bovine serum albumin (BSA). Sections were incubated with primary antibodies under gentle shaking at 4°C for 48–72 h, followed by incubation with Alexa‐conjugated fluorescent secondary antibodies for 24 h at 4°C. After three rinses in 0.1 N PB, nuclei were counterstained with 4′,6‐diamidino‐2‐phenylindole (DAPI; Merck, 1:5000) for 10 min. Finally, sections were mounted on gelatin‐coated glass slides and coverslipped using a non‐commercial anti‐fading mounting medium (33% glycerol and 7.5% mowiol, prepared in Tris–HCl 0.2 M pH = 8.5).

2.6. Retroviral Stock Preparation

To label adult‐born DGCs, we used replication‐incompetent retroviruses that specifically integrate into the genome of cells dividing at the time of stereotaxic injection (Zhao et al. 2006).

We generated a retroviral stock encoding Cerulean (Blue) fluorescent protein on an RSF91 backbone (Gomez‐Nicola et al. 2014; Schambach et al. 2006). The plasmids used to produce these viruses were kindly provided by R. Tsien (Howard Hughes Medical Institute Laboratory at the University of California, San Diego), C. Baum and A. Schambach (Hannover Medical School, Germany), A. Miyawaki (RIKEN Brain Science Institute, Saitama, Japan), and C. Riecken (University Medical Center Hamburg‐Eppendorf, Germany).

The plasmids used to package the retroviral particles were kindly provided by Prof. Gage (Salk Institute, California, USA). Retroviral stocks were concentrated to working titers of 1 × 10⁷–1 × 10⁸ pfu/mL by ultracentrifugation (Zhao et al. 2006).

A primary anti‐green fluorescent protein (GFP) antibody was used to amplify the Cerulean signal, allowing the visualization of the cell body and processes.

2.7. Stereotaxic Injections

Anesthesia was induced in 6‐week‐old female mice in a chamber with isoflurane (3%–4% in oxygen) until the absence of reflexes was confirmed by pinching the hind paw and tail. Mice were then positioned in a stereotaxic frame, and anesthesia was maintained using a nose cone delivering isoflurane (1%–2% in oxygen). The surgical area was shaved, and a 1‐cm incision was made using a sterile blade. Bregma was defined as coordinate 0, and a small craniotomy was performed with a microdrill at −2.0 mm anteroposterior and ±1.4 mm mediolateral relative to bregma, targeting the DG in both hemispheres. A glass micropipette containing 2 μL of the viral suspension was lowered to −2.2 mm dorsoventral from the skull surface. The virus was delivered at a rate of 0.2 μL/min. To minimize backflow, the micropipette was left in place for an additional 5 min before being slowly withdrawn.

To promote cell proliferation, mice were given free access to a running wheel in their cages for 3 days before surgery (Piatti et al. 2011). Following the procedure, they were returned to standard housing conditions and sacrificed 1 or 8 weeks post‐injection.

2.8. Cell Counts

To quantify the density of ionized calcium‐binding adaptor molecule 1 (Iba1)⁺, glial fibrillary acidic protein (GFAP)⁺, Neural/Glial antigen 2 (NG2)⁺, and oligodendrocyte transcription factor 2 (Olig2)⁺ cells, five confocal z‐stacks containing the GCL and SGZ were acquired with a Zeiss LSM900 confocal microscope (Zeiss, Oberkochen, Germany) using a 40X oil‐immersion objective (0.8 zoom; XY dimensions: 199.66 μm; z‐step: 1 μm) and Zen software (version 3.10.103.04000). Cell densities were estimated using unbiased stereological principles, applying the physical dissector method adapted to confocal microscopy (Flor‐Garcia et al. 2020; Llorens‐Martin et al. 2006; Moreno‐Jimenez et al. 2019; Terreros‐Roncal et al. 2021). Briefly, the GCL was delineated on the DAPI channel of each stack in Fiji (ImageJ, NIH), and its area was measured. The reference volumes (Flor‐Garcia et al. 2020; Moreno‐Jimenez et al. 2019) were calculated by multiplying the delineated area by the z‐thickness of the stack. The length of the SGZ was measured in the DAPI channel using the freehand line tool in Fiji. The values were multiplied by the stack thickness to obtain the SGZ reference area (Llorens‐Martin et al. 2006). Cells positive for each marker were counted plane by plane within each stack, and densities were expressed as the number of cells per mm³ (for volumetric estimations in the GCL) or per mm² (for surface estimations in the SGZ), depending on the reference space (Llorens‐Martin et al. 2006).

2.9. Colocalization Analyses

To address whether contacts between astrocytes, microglia, and UEA‐1⁺ blood vessels and capillaries were affected by aging in WT and GSK‐3β‐OE mice, we analyzed four‐channel images (DAPI, Iba1, GFAP, and UEA‐1) acquired as described above from mice aged 4, 12, 18, and 30 months. For each image, the GCL and SGZ were delineated in the DAPI channel and cropped separately. Channels were then split, and colocalization was quantified using the “Just Another Colocalization Plugin” (JACoP) in Fiji, as previously described (Llorens‐Martin et al. 2016). Intensity thresholds were defined for each marker and were consistently applied across all analyses. Within each region (GCL or SGZ), the percentage of colocalization between channels was calculated pairwise as the fraction of the GFAP⁺ area overlapping with the Iba1⁺ area and the fraction of the UEA‐1⁺ area overlapping with the GFAP⁺ or Iba1⁺ area. Five z‐stacks were analyzed per animal.

To examine whether contacts between newborn DGCs, microglia, astrocytes, and UEA‐1⁺ blood vessels and capillaries were altered in GSK‐3β‐OE mice, we analyzed four‐channel images (GFP, Iba1, GFAP, and UEA‐1) obtained from mice injected with the Cerulean encoding retrovirus and sacrificed 1 or 8 weeks post‐infection. Images were acquired with a Zeiss LSM900 confocal microscope (Zeiss, Oberkochen, Germany) using a 40X oil‐immersion objective (0.8 zoom; XY dimensions: 199.66 μm; z‐step: 1 μm). Individual channels were split, and colocalization was quantified using the JACoP plugin in Fiji. Thresholds were defined for each marker and were kept constant across all analyses. The fraction of the GFP⁺ area overlapping with GFAP⁺, Iba1⁺, and UEA‐1⁺ signal was calculated.

2.10. Area and Diameter Analysis of Ulex europaeus Agglutinin‐1 (UEA‐1)⁺ Blood Vessels and Capillaries

Images of the DG were acquired using a Zeiss LSM900 confocal microscope (Zeiss, Oberkochen, Germany) with a 40X oil‐immersion objective. For each animal, five z‐stacks were acquired (0.8 zoom; XY dimensions: 199.66 μm; z‐step: 1 μm).

Image analysis was performed using Fiji. First, a maximum intensity projection was generated from the first five planes of the z‐stacks. Using the DAPI channel, the GCL and SGZ were manually delineated as regions of interest (ROIs) and their total areas were measured. A threshold was then established and applied to all images to segment and measure the area occupied by the UEA‐1⁺ signal within each ROI. Finally, the area occupied by the vessels was calculated as the percentage of the total ROI area covered by the UEA‐1⁺ signal.

Additionally, the capillary diameter was measured by drawing lines perpendicular to the longitudinal axis of the vessel using Fiji's straight‐line tool in 50 randomly selected vessels from each ROI (GCL or SGZ) per mouse.

2.11. Analysis of Cluster of Differentiation 31 (CD31) Fluorescence Intensity

Images of the DG were acquired using a Zeiss LSM900 confocal microscope (Zeiss, Oberkochen, Germany) equipped with a 40X oil‐immersion objective. For each animal, five 16‐bit z‐stacks were acquired (0.8 zoom; XY dimensions: 199.66 μm; z‐step: 1 μm). To ensure reproducibility, all acquisition parameters, including the laser intensity, offset, and digital gain, were kept constant across all samples.

To quantify CD31 fluorescence intensity, the GCL and SGZ were manually delineated on the DAPI channel as ROIs using Fiji. A fixed, consistent fluorescence threshold was established and applied to all images to segment the CD31⁺ signal. Finally, the mean fluorescence intensity was quantified within each ROI.

2.12. Morphometric Analysis of Microglia and Astrocytes

The 3D morphology of Iba1⁺ microglia and GFAP⁺ astrocytes was analyzed using Neurolucida 360 (v. 2024.2.1; MBF Bioscience) for tracing and Neurolucida Explorer (v. 2024.1.1; MBF Bioscience) for quantification.

For the 12‐month‐old cohort, a total of 20 astrocytes and 20 microglia, equally distributed among the mice that composed each group and located within the SGZ, were fully reconstructed per genotype. Cells were selected for reconstruction based on clear staining, location within the SGZ, and having processes that were fully contained within the z‐stack and a lack of overlap with neighboring cells.

The cells were manually traced using the ‘user‐guided mode’ in Neurolucida. The following morphological parameters were then extracted for each cell using Neurolucida Explorer: total cell process length (μm), cell surface (μm²), cell volume (μm³), number of nodes, number of endings, number of primary trees, complexity of the cell (calculated as the sum of the total number of branch endings and their respective terminal orders, multiplied by the ratio of the total process length to the number of primary dendrites), convex‐hull 3D surface (μm²) and convex‐hull 3D volume (μm³) (representing the surface area and volume of the convex polyhedron generated by connecting the endings of the distal branches of the cell), Sholl's analysis (using 2‐μm concentric sphere intervals), and branch ordering analysis. To calculate the latter, numerical values were assigned to each segment based on its hierarchical position within the arbor. A centrifugal ordering method was applied, where primary branches emerging from the soma were designated as order 1, and the order increased by one with each successive bifurcation moving away from the soma.

2.13. Morphometric Analysis of Cluster of Differentiation 31 (CD31)⁺ Blood Vessels

For 3D vascular network analysis, the entire DG was imaged. For each mouse, three sections (rostral, medial, and caudal) were acquired as large‐area tile scans (mosaic merge) using a Leica THUNDER Imager Tissue microscope equipped with a 20X air objective (z‐step: 10 μm). Acquired images were then post‐processed using the ‘large volume computational clearing (LVCC)’ method in the Leica Application Suite X (LAS X v. 3.7.5.24914) software.

The CD31⁺ vessel network was reconstructed from the computationally cleared z‐stacks using Vesselucida (Neurolucida 360 Ultra v. 2024.2.1; MBF Bioscience). To optimize detection, images were first processed with a closing filter (4‐pixel structuring radius). Reconstruction was then performed semi‐automatically: an initial automatic trace was manually refined using the user‐guided tracing mode to ensure all vessels were accurately captured. The tracing parameters were set as follows: 70% seed sensitivity, 90% trace sensitivity, and 4 μm for the typical process width.

Once all tracings had been saved, further analyses were performed using Vesselucida Explorer (v. 2022.1.1; MBF Bioscience). In this regard, the network totals dataset provided total values for the number of branching nodes, endings, vessel network length (μm), surface area (μm²), volume (μm³), and loop count, which were normalized by the total volume of the DG. The segment details dataset included the determination of segment length (μm), average diameter (μm), and tortuosity for each detected segment. A segment was defined as the continuous vessel portion between two branch points, or between a branch point and an ending. The network nodes dataset provided measurements of the diameter (μm), surface area (μm²), and volume (μm³) of bifurcation nodes, defined as the intersections of two segments. The vessel loops dataset reported the length (μm), surface area (μm²), volume (μm³), and average diameter (μm) of each individual loop, which was defined as a closed path starting and ending at the same node.

2.14. Analysis of Cytokine Expression Levels

A longitudinally sectioned half of the left hippocampus was homogenized, and total protein concentration was estimated using the BCA Protein Assay Kit (Pierce, Rockford, IL). For each experimental group, equal amounts of protein from each animal were combined to generate a pooled extract, ensuring that all mice contributed equally to the final sample. Cytokine profiling was performed using a mouse cytokine antibody array (RayBio C‐series Mouse Cytokine Antibody Array C1000; Ray Biotech Inc.), which contains 96 cytokine antibodies immobilized on a PVDF membrane. Membranes were incubated following the manufacturer's instructions and previously published procedures (Llorens‐Martin et al. 2016). Signal intensity for each cytokine spot was quantified as net gray‐level density using Fiji, and relative cytokine expression was calculated for each experimental group.

2.15. Heatmap Visualizations

To visualize comparative profiles of morphological data, heatmaps were constructed using features from CD31⁺ blood vessels, Iba1⁺ microglia, and GFAP⁺ astrocytes. Data were standardized by calculating the z‐score across the different genotypes and ages for each feature of interest. Z‐scores were calculated using the ‘z‐score’ function from NumPy (v. 1.23.5) in Python 3 (v. 3.9.13), and data frame operations were managed with the Pandas library (v. 1.5.3). The resulting z‐score matrix was then plotted using the ‘heatmap’ function from the Seaborn library (v. 0.12.2).

To visualize the relative expression levels of cytokines, all group‐level data were normalized by dividing them by the corresponding mean of the 4‐month‐old WT (control) group, generating relative fold‐change values. A heatmap of these fold‐change values was generated using the ‘heatmap’ function from the Seaborn library. The colormap was set as a divergent map (“PiYG”) centered at 1.0, where values above 1.0 represent an increase and values below 1.0 represent a decrease relative to the control group.

2.16. Statistics and Reproducibility

All statistical analyses were conducted using GraphPad Prism 10 (v. 10.4.0). Data are presented in graphs as mean values ± SEM. A 95% confidence interval was used for statistical comparisons.

Normality of the data distribution was assessed using the Kolmogorov–Smirnov test, or the Shapiro–Wilk test in those cases in which n ≤ 4. Outliers were identified and removed, when necessary, based on the ROUT method.

For comparisons between two independent groups (e.g., WT vs. GSK‐3β‐OE mice within a specific age and layer), an unpaired Student's t‐test was used for normally distributed data. In cases where the assumption of normality was not met, the non‐parametric Mann–Whitney U test was applied.

For single‐factor comparisons (e.g., comparing ages within WT groups), a one‐way ANOVA was performed for normally distributed data with equal standard deviations (SDs). When equal SDs could not be assumed, the Brown‐Forsythe and Welch ANOVA test was used. For non‐parametric data, the Kruskal‐Wallis test was applied. Significant results were followed by Tukey's (for parametric) or Dunn's (for non‐parametric) multiple comparison post hoc tests.

For two‐factor comparisons (e.g., GSK‐3β overexpression and age), a two‐way ANOVA was performed. Statistically significant results were followed by Tukey's multiple comparison post hoc tests.

For Sholl's and branch ordering analyses, a two‐way repeated‐measures ANOVA, treating distance or branch order as the repeated measure, was performed, followed by Šídák's multiple comparison post hoc test to identify differences at specific intervals.

3. Results

3.1. Aging Was Associated With Changes in the Density of Glial Cell Populations and the Vasculature of the Dentate Gyrus (DG) in Wild‐Type (WT) Mice

To unveil the effects of aging on the cellular and vascular architecture of the DG micro‐environment in WT mice, we quantified cell densities and parameters related to the DG vascularization and to the physical contacts established between glial cells, as well as between the latter and the vasculature at 4, 12, 18, and 30 months of age (Figure 1). The SGZ and GCL were analyzed independently to assess layer‐specific changes.

FIGURE 1.

Dentate gyrus (DG) neurogenic niche in wild‐type (WT) mice across lifespan. (A–D) Representative images showing staining with ionized calcium‐binding adapter molecule 1 (Iba1) (A), glial fibrillary acidic protein (GFAP) (B), oligodendrocyte transcription factor 2 (Olig2) and nerve/glial‐antigen 2 (NG2) (C) and Ulex europaeus agglutinin‐1 (UEA‐1)⁺ blood vessels and capillaries (D) in the DG of WT mice at 4, 12, 18 and 30 months of age. (E) Densities of Iba1⁺ microglia, GFAP⁺ astrocytes, and Olig2⁺ oligodendrocytes in the granule cell layer (GCL). (F) Densities of NG2⁺ cells, NG2⁻ Olig2⁺ mature oligodendrocytes, and NG2⁺ Olig2⁺ oligodendrocyte progenitor cells (OPCs) in the GCL. (G) Densities of Iba1⁺ microglia, GFAP⁺ astrocytes, and Olig2⁺ oligodendrocytes in the subgranular zone (SGZ). (H) Densities of NG2⁺ cells, NG2⁻ Olig2⁺ mature oligodendrocytes, and NG2⁺ Olig2⁺ OPCs in the SGZ. (I–L) Representative images showing cell contacts between GFAP⁺ astrocytes, Iba1⁺ microglia, and UEA‐1⁺ blood vessels and capillaries in WT mice at 4 (I), 12 (J), 18 (K), and 30 (L) months of age. (M) Area of the GCL and SGZ occupied by UEA‐1⁺ blood vessels and capillaries. (N) Fraction of UEA‐1⁺ capillaries contacting Iba1⁺ microglia in the GCL and SGZ. (O) Fraction of GFAP⁺ astrocytes contacting Iba1⁺ microglia in the GCL and SGZ. (P) Fraction of UEA‐1⁺ capillaries contacting GFAP⁺ astrocytes in the GCL and SGZ. H, hilus; GCL, granule cell layer; ML, molecular layer; SGZ, subgranular zone. Yellow scale bar: 20 μm. White scale bar: 10 μm. Graphs represent mean values ± SEM. *0.05 > p ≥ 0.01; **0.01 > p ≥ 0.001; ***0.001 > p ≥ 0.0001; ****p < 0.0001. Detailed results of statistical comparisons are shown in Extended Data S1.

In the GCL, the density of Iba1⁺ microglia varied with age (Figure 1A,E; F _3,20 = 14.07; p < 0.0001). Compared to 4‐month‐old mice, microglial density peaked at 12 months of age (p < 0.0001) and returned to basal levels at 30 months (p = 0.0085) of age. The density of GFAP⁺ astrocytes also changed with age (Figure 1B,E; F _3,20 = 4.367; p = 0.0161), showing an increase at 18 months (p = 0.0157). In contrast, the density of NG2⁺ Olig2⁺ OPCs decreased with age (Figure 1C,F; F _3,20 = 4.749; p = 0.0161), specifically at 12 months of age (p = 0.0132). The densities of total Olig2⁺ oligodendrocytes, total NG2⁺ cells, NG2⁻ Olig2⁺ mature oligodendrocytes (Figure 1C,E,F), and NG2⁺ Olig2⁻ cells (Figure S1B) remained unchanged.

In the SGZ, the density of Iba1⁺ microglia was not affected by aging (Figure 1A,G). However, that of GFAP⁺ astrocytes progressively decreased with age (Figure 1B,G; F _3,20 = 11.35; p = 0.0001), reaching statistical significance in 18‐ (p = 0.0008) and 30‐month‐old mice (p = 0.0010) compared to their 4‐month‐old counterparts. Regarding the oligodendrocyte lineage, age affected the total Olig2⁺ oligodendrocyte density (Figure 1C,G; F _3,20 = 3.286; p = 0.042) and increased that of NG2⁻ Olig2⁺ mature oligodendrocytes (Figure 1C,H; F _3,20 = 4.374; p = 0.016), specifically at 12 months of age (p = 0.0197). In contrast, the densities of total NG2⁺ cells, NG2⁺ Olig2⁺ OPCs (Figure 1C,H), and NG2⁺ Olig2⁻ cells (Figure S1C) remained stable.

Given the age‐related regional differences in the density of distinct cell populations, we next calculated the SGZ/GCL cell density ratio to assess putative changes in cellular distribution. This ratio changed with age for both Iba1⁺ microglia (Figure S1A; F _3,20 = 5.025; p = 0.0093) and GFAP⁺ astrocytes (Figure S1A; F _3,20 = 21.67; p < 0.0001). These results indicate not only an alteration in cell density but also in the spatial distribution of distinct cell populations across ages.

Aging was also associated with increased SGZ vascularization, as indicated by the area occupied by UEA‐1⁺ blood vessels and capillaries (Figure 1D,M; F _3,20 = 5.645; p = 0.0057), which was higher in 12‐ (p = 0.0163) and 30‐month‐old mice (p = 0.0133) compared to 4‐month‐old counterparts. Conversely, the vascularization of the GCL did not change with age (Figure 1D,M).

Finally, we analyzed the overlap between cell markers of distinct cell populations as a proxy for physical contacts established between these cells. In the GCL, aging was paralleled by changes in the contacts between UEA‐1⁺ capillaries and Iba1⁺ microglia (Figure 1I–L,N; KW = 10.73; p = 0.0133), GFAP⁺ astrocytes and Iba1⁺ microglia (Figure 1I–L,O; KW = 14.3; p = 0.0025), and UEA‐1⁺ capillaries and GFAP⁺ astrocytes (Figure 1I–L,P; F _3,20 = 3.161; p = 0.0472). Post hoc analysis revealed a transient increase in the contact at 12 months of age for both UEA‐1⁺ capillaries with Iba1⁺ microglia (p = 0.0355) and GFAP⁺ astrocytes with Iba1⁺ microglia (p = 0.04).

In the SGZ, the contacts between UEA‐1⁺ capillaries and Iba1⁺ microglia (Figure 1I–L,N; F _3,20 = 14.96; p < 0.0001), and those between GFAP⁺ astrocytes and Iba1⁺ microglia (Figure 1I–L,O; F _3,20 = 9.523; p = 0.0004) transiently increased at 12 months of age (p < 0.0001; p = 0.0043, respectively), whereas those between UEA‐1⁺ capillaries and GFAP⁺ astrocytes remained unchanged.

Together, these data reveal a significant cellular and vascular reorganization of the DG during physiological aging, characterized by localized alterations within the SGZ neurogenic niche and the adjacent GCL.

3.2. The Overexpression of Glycogen Synthase Kinase 3‐β (GSK‐3β) Caused Premature and Non‐Linear Age‐Associated Changes in the Densities and Morphologies of Glial Cell Populations of the Dentate Gyrus (DG)

To assess the effects of GSK‐3β overexpression on the glial populations that compose the DG micro‐environment, we studied the SGZ and the GCL separately. Specifically, we quantified cellular densities and morphometric characteristics of these cells in the GSK‐3β‐OE mouse model and compared them to those of WT animals at 4, 12, 18, and 30 months of age (Figure 2).

FIGURE 2.

Hippocampal neurogenic niche in wild‐type (WT) and glycogen synthase kinase 3‐β‐overexpressing (GSK‐3β‐OE) mice across lifespan. (A) Representative images showing staining with ionized calcium‐binding adapter molecule 1 (Iba1) in WT and GSK‐3β‐OE mice at 4, 12, 18, and 30 months of age. (B, C) Density of Iba1⁺ microglia in the granule cell layer (GCL) (B) and subgranular zone (SGZ) (C). (D–F) Representative images (D) and Neurolucida 360 tridimensional reconstructions of microglia from the SGZ of WT (E) and GSK‐3β‐OE mice (F). (G, H) Sholl's (G) and branch ordering (H) analyses of microglia. (I) Heatmap showing the morphological features of microglia. (J) Representative images showing staining with glial fibrillary acidic protein (GFAP) in WT and GSK‐3β‐OE mice at 4, 12, 18, and 30 months of age. (K, L) Density of GFAP⁺ astrocytes in the GCL (K) and SGZ (L). (M) Representative images showing staining with oligodendrocyte transcription factor 2 (Olig2) and nerve/glial‐antigen 2 (NG2) in WT and GSK‐3β‐OE mice at 4, 12, 18, and 30 months of age. (N, O) Density of Olig2⁺ oligodendrocytes in the GCL (N) and the SGZ (O). (P, Q) Density of NG2⁺ cells in the GCL (P) and SGZ (Q). (R, S) Density of NG2⁻ Olig2⁺ cells in the GCL (R) and SGZ (S). (T, U) Density of NG2⁺ Olig2⁺ cells in the GCL (T) and SGZ (U). H, hilus; GCL, granule cell layer; ML, molecular layer; SGZ, subgranular zone. Yellow scale bar: 20 μm. Black scale bar: 10 μm. Graphs represent mean values ± SEM. *0.05 > p ≥ 0.01; **0.01 > p ≥ 0.001; ***0.001 > p ≥ 0.0001; ****p < 0.0001. Detailed results of statistical comparisons are shown in Extended Data S1.

In the GCL, the density of Iba1⁺ microglia (Figure 2A,B; age: F _3,47 = 2.839; p = 0.0480; GSK‐3β: F _1,47 = 4.418; p = 0.0409; Interaction: F _3,47 = 11.05; p < 0.0001), as well as the SGZ/GCL density ratio of this cell population (Figure S2X; Age: F _3,47 = 3.362; p = 0.0264; GSK‐3β: F _1,47 = 45.93; p < 0.0001; interaction: F _3,47 = 3.978; p = 0.0132), were affected by the interaction between GSK‐3β overexpression and age. In contrast, microglial density in the SGZ was predominantly altered by GSK‐3β overexpression, regardless of age (Figure 2A,C; age: F _3,47 = 1.935; p = 0.1369; GSK‐3β: F _1,47 = 21.85; p < 0.0001; interaction: F _3,47 = 2.801; p = 0.0501).

To further characterize microglial alterations, we performed 3D reconstructions and examined 11 morphometric features (Figure 2D–I and Figure S2A–I) of these cells, which have been proposed to serve as a proxy for their functional state (Paolicelli et al. 2022).

Both branch ordering and Sholl's analyses revealed a regional effect of GSK‐3β overexpression on microglial morphological complexity (Figure 2D–H). Specifically, the microglia of GSK‐3β‐OE mice exhibited a higher number of branches at intermediate branching orders (Figure 2D–G; branch order: F _3.871,147.1 = 5.671; p < 0.0001; GSK‐3β: F _1,38 = 16.25; p = 0.0003; interaction: F _3.871,147.1 = 5.671; p = 0.0003; Extended Data S1) and an increased number of intersections at intermediate distances from the soma (Figure 2D–F,H; distance from the soma: F _4.507,171.3 = 206.1; p < 0.0001; GSK‐3β: F _1,38 = 16.59; p = 0.0002; interaction: F _40,1520 = 4.762; p < 0.0001; Extended Data S1). These cells also showed increases in convex hull volume (t ₃₈ = 3.863; p = 0.0004) and surface (t ₃₈ = 3.36; p = 0.0018), number of nodes (t ₃₈ = 3.952; p = 0.0003), number of terminals (t ₃₈ = 4.075; p = 0.0002), total process length (t ₃₈ = 4.204; p = 0.0002), and total cell surface (t ₃₈ = 3.983; p = 0.0003), volume (t ₃₈ = 3.367; p = 0.0018), and complexity (U = 43; p < 0.0001). Conversely, the number of primary trees was reduced (U ₃₈ = 127; p = 0.0448). These results indicate that the overexpression of GSK‐3β not only reduced microglial cell density in the SGZ of 12‐month‐old mice but also triggered morphological changes in these cells that are putatively related to altered function.

The density of GFAP⁺ astrocytes was affected by GSK‐3β overexpression in both the GCL (Figure 2J,K; age: F _3,47 = 3.052; p = 0.0375; GSK‐3β: F _1,47 = 41.45; p < 0.0001; interaction: F _3,47 = 0.3127; p = 0.8161) and SGZ (Figure 2J,L; age: F _3,47 = 21.36; p < 0.0001; GSK‐3β: F _1,47 = 17.62; p = 0.0001; interaction: F _3,47 = 2.129; p = 0.1092). Moreover, the SGZ/GCL astrocyte density ratio showed a significant interaction between GSK‐3β overexpression and age (Figure S2Y; age: F _3,47 = 37.66; p < 0.0001; GSK‐3β: F _1,47 = 88.47; p < 0.0001; interaction: F _3,47 = 7.328; p = 0.0004).

In sharp contrast to microglia, no changes in the morphology of SGZ GFAP⁺ astrocytes in 12‐month‐old GSK‐3β‐OE mice were observed (Figure S2J–W; see Extended Data S1 for detailed statistics).

Regarding the oligodendroglial cell lineage, the density of Olig2⁺ oligodendrocytes was altered by GSK‐3β overexpression in both the GCL (Figure 2M,N; age: F _3,47 = 0.4323; p = 0.7308; GSK‐3β: F _1,47 = 10.03; p = 0.0027; interaction: F _3,47 = 0.0416; p = 0.9886) and SGZ (Figure 2M,O; age: F _3,47 = 8.285; p = 0.0002; GSK‐3β: F _1,47 = 32.48; p < 0.0001; interaction: F _3,47 = 1.372; p = 0.2630), although no changes in the SGZ/GCL Olig2⁺ oligodendrocyte density ratio were found (Figure S2Z; age: F _3,47 = 2.042; p = 0.1208; GSK‐3β: F _1,47 = 0.9741; p = 0.9741; interaction: F _3,47 = 0.2709; p = 0.8460).

Both the density of total NG2⁺ cells in the GCL (Figure 2M,P; age: F _3,44 = 2.1150; p = 0.1120; GSK‐3β: F _1,44 = 0.9995; p = 0.3229; interaction: F _3,44 = 1.994; p = 0.1288) and the SGZ/GCL NG2⁺ cell density ratio (Figure S2AA; age: F _3,43 = 3.541; p = 0.0223; GSK‐3β: F _1,43 = 1.291; p = 0.2622; interaction: F _3,43 = 1.349; p = 0.271) remained unchanged. However, the density of NG2⁺ cells in the SGZ was affected by GSK‐3β overexpression (Figure 2M,Q; age: F _3,44 = 1.451; p = 0.2409; GSK‐3β: F _1,44 = 7.468; p = 0.009; interaction: F _3,44 = 2.741; p = 0.0545).

Moreover, in the GCL, the density of NG2⁻ Olig2⁺ mature oligodendrocytes was not modified by GSK‐3β overexpression (Figure 2M,R; age: F _3,44 = 2.838; p = 0.0488; GSK‐3β: F _1,44 = 3.191; p = 0.0809; interaction: F _3,44 = 0.2053; p = 0.8922). However, this parameter was decreased in the SGZ (Figure 2M,S; age: F _3,44 = 10.62; p < 0.0001; GSK‐3β: F _1,44 = 36.59; p < 0.0001; interaction: F _3,44 = 2.031; p = 0.1233). The SGZ/GCL mature oligodendrocyte density ratio was also affected (Figure S2AB; age: F _3,43 = 4.281; p = 0.001; GSK‐3β: F _1,43 = 5.508; p = 0.0236; interaction: F _3,43 = 1.167; p = 0.3332).

The density of NG2⁺ Olig2⁺ OPCs in the GCL (Figure 2M,T; age: F _3,44 = 3.317; p = 0.0284; GSK‐3β: F _1,44 = 3.803; p = 0.0576; interaction: F _3,44 = 2.434; p = 0.0775) and SGZ (Figure 2M,U; age: F _3,44 = 1.777; p = 0.1654; GSK‐3β: F _1,44 = 0.0028; p = 0.8675; interaction: F _3,44 = 1.084; p = 0.3656), as well as the SGZ/GCL density ratio (Figure S2AC; age: F _3,42 = 1.635; p = 0.1958; GSK‐3β: F _1,42 = 0.915; p = 0.3443; interaction: F _3,42 = 0.9412; p = 0.4294) remained unaffected by GSK‐3β overexpression.

Conversely, the density of NG2⁺ Olig2⁻ cells in GSK‐3β‐OE mice was altered in both the GCL (Figure S2AE; Age: F _3,44 = 3.093; p = 0.0365; GSK‐3β: F _1,44 = 4.283; p = 0.0444; interaction: F _3,44 = 0.5726; p = 0.6361) and SGZ (Figure S2AF; age: F _3,44 = 2.188; p = 0.1029; GSK‐3β: F _1,44 = 8.541; p = 0.0055; interaction: F _3,44 = 1.316; p = 0.2813). Despite these changes, the SGZ/GCL density ratio was not affected (Figure S2AD; age: F _3,44 = 2.032; p = 0.1232; GSK‐3β: F _1,44 = 0.9458; p = 0.3361; interaction: F _3,44 = 1.288; p = 0.2903).

Taken together, these data suggest that GSK‐3β overexpression triggers a layer‐specific remodeling of glial populations within the DG. Specifically, the SGZ neurogenic niche is characterized by a reduction in the density of astrocytes, mature oligodendrocytes, and microglia, alongside an increase in microglial morphological complexity. In contrast, the GCL exhibits an early increase in microglial and astrocytic densities.

3.3. The Overexpression of Glycogen Synthase Kinase 3‐β (GSK‐3β) Altered the Vascular Network of the Dentate Gyrus (DG)

To assess the effects of GSK‐3β overexpression on the vasculature of the SGZ and the GCL, we studied the area occupied by UEA‐1⁺ blood vessels and capillaries and the diameter of these structures. We then performed a detailed 3D morphometric analysis of CD31⁺ vessel network in WT and GSK‐3β‐OE mice at 4, 12, 18, and 30 months of age (Figure 3).

FIGURE 3.

Vascular network of the dentate gyrus (DG) in wild‐type (WT) and glycogen synthase kinase 3‐β‐overexpressing (GSK‐3β‐OE) mice across lifespan. (A) Representative images showing staining with Ulex europaeus agglutinin‐1 (UEA‐1) in WT and GSK‐3β‐OE mice at 4, 12, 18, and 30 months of age. (B, C) Area of the granule cell layer (GCL) (B) and subgranular zone (SGZ) (C) occupied by UEA‐1⁺ blood vessels and capillaries. (D) SGZ/GCL ratio of the area occupied by UEA‐1⁺ blood vessels and capillaries. (E, F) Diameter of UEA‐1⁺ blood vessels and capillaries in the GCL (E) and SGZ (F). (G) Representative images showing staining with cluster of differentiation 31 (CD31) paired with their corresponding 3D reconstructions in WT and GSK‐3β‐OE mice at 4, 12, 18, and 30 months of age. (H, K) Heatmaps showing CD31⁺ blood vessel morphological features at 4 (H), 12 (I), 18 (J), and 30 (K) months of age. 1: Number of isolated elements/μm³. 2: Number of branching nodes/μm³. 3: Number of endings/μm³. 4: Total length (μm)/μm³. 5: Surface area (μm²)/μm³. 6: Total volume (μm³)/μm³. 7: Loop count/μm³. 8: Segment length (μm). 9: Segment average diameter (μm). 10: Segment tortuosity. 11: Node diameter (μm). 12: Node volume (μm³). 13: Node surface area (μm²). 14: Loop length (μm). 15: Loop surface area (μm²). 16: Loop average diameter (μm). 17: Loop volume (μm³). 18: Loop nodes. H, hilus. GCL: Granule cell layer. ML: Molecular layer. SGZ: Subgranular zone. Yellow scale bar: 20 μm. White scale bar: 10 μm. Green scale bar: 100 μm. Graphs represent mean values ± SEM. *0.05 > p ≥ 0.01; **0.01 > p ≥ 0.001; ***0.001 > p ≥ 0.0001; ****p < 0.0001. Detailed results of statistical comparisons are shown in Extended Data S1.

We observed that the area occupied by UEA‐1⁺ blood vessels and capillaries was affected by GSK‐3β overexpression in both the GCL (Figure 3A,B; age: F _3,47 = 3.887; p = 0.0146; GSK‐3β: F _1,47 = 14.85; p = 0.0004; interaction: F _3,47 = 1.845; p = 0.1520) and SGZ (Figure 3A,C; age: F _3,47 = 13.11; p < 0.0001; GSK‐3β: F _1,47 = 20.23; p < 0.0001; interaction: F _3,47 = 1.174; p = 0.3295). Similarly, the SGZ/GCL vascularization ratio was also altered in GSK‐3β‐OE mice (Figure 3A,D; age: F _3,47 = 6.939; p = 0.0006; GSK‐3β: F _1,47 = 33.63; p < 0.0001; interaction: F _3,47 = 0.2032; p = 0.8936).

Regarding capillary diameter, the impact of GSK‐3β overexpression in the GCL was distinctly modulated by age (Figure 3A,E; age: F _3,47 = 7.257; p = 0.0004; GSK‐3β: F _1,47 = 21.44; p < 0.0001; interaction: F _3,47 = 5.688; p = 0.0021) but not in the SGZ (Figure 3A,F; age: F _3,47 = 3.165; p = 0.033; GSK‐3β: F _1,47 = 2.3420; p = 0.1326; interaction: F _3,47 = 1.99; p = 0.1283).

Moreover, 3D reconstructions of CD31⁺ blood vessels (Figure 3G) revealed widespread alterations in GSK‐3β‐OE mice (Figure 3H–K and Figure S3A–R). General network parameters (Figure S3A–G), including the number of isolated elements (Figure S3A; age: F _3,47 = 5.81; p = 0.0018; GSK‐3β: F _1,47 = 19.72; p < 0.0001; interaction: F _3,47 = 2.999; p = 0.0399), vessel endings (Figure S3C; age: F _3,46 = 12.78; p < 0.0001; GSK‐3β: F _1,46 = 14.8; p = 0.0004; interaction: F _3,46 = 3.911; p = 0.0143), total vascular length (Figure S3D; age: F _3,47 = 8.451; p = 0.0001; GSK‐3β: F _1,47 = 4.867; p = 0.0323; interaction: F _3,47 = 4.098; p = 0.0115), and surface area (Figure S3E; age: F _3,47 = 8.09; p = 0.0002; GSK‐3β: F _1,47 = 3.469; p = 0.0688; interaction: F _3,47 = 5.042; p = 0.0041) were modified by age and GSK‐3β overexpression, showing interactions between both factors. Conversely, the loop length (Figure S3N; age: F _3,47 = 0.604; p = 0.6157; GSK‐3β: F _1,47 = 4.334; p = 0.0428; interaction: F _3,47 = 0.5694; p = 0.6379) and number of nodes in the loops (Figure S3R; age: F _3,47 = 0.1868; p = 0.9049; GSK‐3β: F _1,47 = 5.376; p = 0.0248; Interaction: F _3,47 = 0.944; p = 0.427) were affected only by GSK‐3β overexpression. Other parameters, including the number of branching nodes, the total volume, the loop count, the segment length, the average segment diameter, and the vessel node diameter, were altered only by age (see Extended Data S1 for detailed statistics).

Finally, quantification of CD31 fluorescence intensity revealed an interaction between age and GSK‐3β overexpression in both the GCL (Figure S3S; age: F _3,47 = 7.959; p = 0.0002; GSK‐3β: F _1,47 = 3.842; p = 0.0559; interaction: F _3,47 = 8.531; p = 0.0001) and SGZ (Figure S3T; age: F _3,47 = 4.119; p = 0.0113; GSK‐3β: F _1,47 = 8.9; p = 0.0045; interaction: F _3,47 = 4.574; p = 0.0068). The SGZ/GCL CD31⁺ vessel intensity ratio was also impacted by GSK‐3β overexpression (Figure S3U; age: F _3,47 = 0.3977; p = 0.7552; GSK‐3β: F _1,47 = 4.856; p = 0.0325; interaction: F _3,47 = 0.1522; p = 0.9278).

Collectively, these data reveal a premature and non‐linear remodeling of the vascular architecture in GSK‐3β‐OE mice, characterized by increased capillary density and diameter, along with alterations in the vascular network structure.

3.4. The Overexpression of Glycogen Synthase Kinase 3‐β (GSK‐3β) Altered the Physical Contacts Between Glial Cells, Blood Vessels, and Newborn Dentate Granule Cells (DGCs)

GSK‐3β overexpression did not affect the contacts between UEA‐1⁺ capillaries and Iba1⁺ microglia in the GCL (Figure 4A,B; age: F _3,46 = 4.894; p = 0.0049; GSK‐3β: F _1,46 = 0.1422; p = 0.7078; interaction: F _3,46 = 0.2740; p = 0.8438). However, an interaction between age and GSK‐3β overexpression was observed in the SGZ (Figure 4A,C; age: F _3,46 = 5.502; p = 0.0026; GSK‐3β: F _1,46 = 0.1410; p = 0.7091; interaction: F _3,46 = 9.693; p < 0.0001).

FIGURE 4.

Hippocampal cell contacts and cytokine expression in wild‐type (WT) and glycogen synthase kinase 3‐β‐overexpressing (GSK‐3β‐OE) mice. (A) Representative images showing staining with ionized calcium‐binding adapter molecule 1 (Iba1), glial fibrillary acidic protein (GFAP), and Ulex europaeus agglutinin‐1 (UEA‐1) in WT and GSK‐3β‐OE mice at 12 months of age. (B, C) Fraction of UEA‐1⁺ capillaries overlapping with Iba1⁺ microglia in the granule cell layer (GCL) (B), and subgranular zone (SGZ) (C) at 4, 12, 18, and 30 months of age. (D, E) Fraction of GFAP⁺ astrocytes contacting Iba1⁺ microglia in the GCL (D) and SGZ (E) at 4, 12, 18, and 30 months of age. (F, G) Fraction of UEA‐1⁺ capillaries contacting GFAP⁺ astrocytes in the GCL (F) and the SGZ (G) at 4, 12, 18, and 30 months of age. (H) Representative images showing staining with GFAP and green fluorescent protein (GFP) in 1‐ and 8‐week‐old newborn dentate granule cells (DGCs) in WT and GSK‐3β‐OE mice. (I) Fraction of GFP⁺ newborn DGCs contacting GFAP⁺ astrocytes. (J) Fraction of GFP⁺ newborn DGCs overlapping with Iba1⁺ microglia. (K) Fraction of GFP⁺ newborn DGCs overlapping with UEA‐1⁺ capillaries. (L) Heatmap showing the fold‐change in the levels of 96 anti‐ and pro‐inflammatory cytokines and other soluble factors in WT mice at 12, 18, and 30 months of age and in GSK‐3β‐OE mice at 4, 12, 18, and 30 months of age relative to the 4‐month‐old WT group. H, hilus; GCL, granule cell layer; ML, molecular layer; SGZ, subgranular zone. Yellow scale bar: 20 μm. White scale bar: 10 μm. Graphs represent mean values ± SEM. *0.05 > p ≥ 0.01; **0.01 > p ≥ 0.001; ****p < 0.0001. Detailed results of statistical comparisons are shown in Extended Data S1.

Similarly, the contacts between GFAP⁺ astrocytes and Iba1⁺ microglia were modulated by a significant interaction between genotype and age in both the GCL (Figure 4A,D; age: F _3,47 = 9.111; p < 0.0001; GSK‐3β: F _1,47 = 1.926; p = 0.1717; interaction: F _3,47 = 6.677; p = 0.0008) and the SGZ (Figure 4A,E; age: F _3,47 = 12.39; p < 0.0001; GSK‐3β: F _1,47 = 4.15; p = 0.0473; interaction: F _3,47 = 5.881; p = 0.0017).

Conversely, contacts between UEA‐1⁺ capillaries and GFAP⁺ astrocytes were not altered by GSK‐3β overexpression in either the GCL (Figure 4A,F; age: F _3,46 = 5.783; p = 0.0019; GSK‐3β: F _1,46 = 0.4798; p = 0.492; interaction: F _3,46 = 0.8184; p = 0.4394) or SGZ (Figure 4A,G; age: F _3,47 = 3.632; p = 0.0194; GSK‐3β: F _1,47 = 0.0186; p = 0.8921; interaction: F _3,47 = 1.336; p = 0.2739).

To address how GSK‐3β overexpression impacted the physical contacts between newborn DGCs and distinct components of the DG, we labeled 1‐ and 8‐week‐old newborn DGCs using a retrovirus encoding Cerulean. The contacts between GFP⁺ newborn DGCs and GFAP⁺ astrocytes showed an interaction between cell age and GSK‐3β overexpression (Figure 4H,I; age: F _1,131 = 90.55; p < 0.0001; GSK‐3β: F _1,131 = 4.21; p = 0.0422; interaction: F _1,131 = 4.646; p = 0.033). In contrast, those between GFP⁺ newborn DGCs Iba1⁺ microglia were not affected by GSK‐3β overexpression (Figure 4H,J; age: F _1,131 = 4.315; p = 0.0397; GSK‐3β: F _1,131 = 3.203; p = 0.0758; interaction: F _1,131 = 0.0048; p = 0.9447). Similarly, no effect of GSK‐3β overexpression on the contacts between GFP⁺ newborn DGCs and UEA‐1⁺ capillaries was observed (Figure 4H,K; age: F _1,119 = 4.775; p = 0.0308; GSK‐3β: F _1,119 = 1.186; p = 0.2784; interaction: F _1,119 = 2.082; p = 0.1517).

These results demonstrate that GSK‐3β overexpression alters the physical contacts between microglia and blood vessels within the SGZ, as well as the contacts between the former and astrocytes in both layers. Notably, while GSK‐3β overexpression specifically affected the early contact between astrocytes and newborn DGCs, the contacts between the latter and microglia or capillaries were modulated solely by cell age.

3.5. The Overexpression of Glycogen Synthase Kinase 3‐β (GSK‐3β) Modified the Expression of Cytokines and Soluble Factors

To shed further light on the molecular mechanisms underlying the aforementioned changes in the DG caused by GSK‐3β overexpression, we screened 96 cytokines and soluble factors in WT and GSK‐3β‐OE mice and normalized expression values to those of the 4‐month‐old WT baseline (Figure 4L). Anti‐inflammatory cytokines, such as interleukin (IL)‐10, were reduced in GSK‐3β‐OE mice from 12 months of age onwards, and in WT mice at 18 months and onwards. In contrast, several pro‐inflammatory cytokines (e.g., IL‐7, IL‐15, IL‐17B R, IL‐1β) were upregulated in GSK‐3β‐OE mice at 12 and 18 months of age, anticipating the age‐related increase observed in WT mice at 30 months. Additionally, we detected an increased expression of adhesion molecules (intercellular adhesion molecule (ICAM)‐1 and E‐selectin) in GSK‐3β‐OE mice at 12 and 18 months of age, whereas these markers were increased in WT animals only at 30 months. An increase in vascular endothelial growth factor (VEGF)‐D and its receptors (VEGF‐R1, −R2, and ‐R3) was also found in 12‐ and 30‐month‐old control mice, being more pronounced at 18 months in the GSK‐3β‐OE group. Interestingly, the pro‐neurogenic factors insulin growth factor I and II (IGF‐I and IGF‐II) and basic fibroblast growth factor (bFGF) were also elevated in the GSK‐3β‐OE model at 12 and 18 months of age, contrasting with the intermittent increases observed at 12 and 30 months in their WT counterparts (see Raw data).

In summary, GSK‐3β overexpression modulates the expression of anti‐ and pro‐inflammatory cytokines and other soluble factors in a non‐linear manner shaped by the age of the animals, likely underlying the alterations in the cellular and vascular components of the DG neurogenic micro‐environment.

4. Discussion

AHN is a fundamental process for maintaining brain plasticity and cognitive functioning, yet it remains highly vulnerable to the effects of aging and neurodegenerative diseases (Boldrini et al. 2018; Moreno‐Jimenez et al. 2019; Terreros‐Roncal et al. 2021). In the present study, we examined whether aging and neuronal GSK‐3β overexpression induce coordinated alterations in the glial and vascular components of the DG micro‐environment. By analyzing the SGZ and GCL separately, we identified widespread layer‐specific modifications throughout the lifespan of WT and GSK‐3β‐OE mice.

Within the neurogenic niche, microglia play a pivotal role in regulating the integration and maturation of newborn DGCs (Sierra et al. 2010; Vicidomini et al. 2020). During aging, these cells transition through states transcriptionally associated with inflammatory activation and cognitive decline (Shea and Villeda 2025). Concomitantly, they adopt a reactive phenotype (Kohman et al. 2013) and enter an altered metabolic state (Minhas et al. 2021). While previous studies on microglial density within the aging DG have yielded conflicting results, ranging from stability (Long et al. 1998) to decline (Adachi et al. 2010), our observations in the GCL align with reports of a mid‐life increase (Gebara et al. 2013) and extend these observations, previously reported in females (Mouton et al. 2002), to male mice.

Interestingly, while GSK‐3β overexpression induced an early GCL‐restricted increase in microglial density at 4 months of age, both WT and GSK‐3β‐OE mice converged on a microglial cell density reduction within the SGZ by 12 months. Notably, the remaining microglia in the SGZ of transgenic mice exhibited a larger and more ramified morphology, contrasting with the classical amoeboid phenotype of acute activation (Edler et al. 2021). It has been proposed that such structural remodeling is linked to altered subcellular mRNA distribution (Henze et al. 2026). In our model, this morphological restructuring correlated with increased apoptosis (Rodriguez‐Matellan et al. 2020) and persistent AHN alterations (Fuster‐Matanzo et al. 2013) previously described in this model at this age, suggesting microglial functional exhaustion. These data are further supported by a secretory profile characterized by elevated IL‐15 and IL‐1β, which impair neural progenitor cell proliferation and differentiation (Borsini et al. 2015), alongside reduced IL‐10, a pro‐neurogenic cytokine (Navarro Negredo et al. 2026), indicating a shift toward a non‐permissive micro‐environment that actively compromises the neurogenic potential of the niche.

In addition to microglia, astrocytes play an indispensable role in maintaining niche homeostasis and promoting the integration and dendritic refinement of newborn DGCs (Vicidomini et al. 2020). With advancing age, astrocytes progressively adopt a reactive phenotype (Labarta‐Bajo and Allen 2025; Wu et al. 2005), characterized by functional decay and morphological shifts (Popov et al. 2021), despite the preservation of communication with neurons (Gomez‐Gonzalo et al. 2017). Although both increases (Mouton et al. 2002) and decreases (Cole et al. 2022) in astrocytic density have been reported in the aging DG, our layer‐resolved analysis reconciles these findings by showing that a localized decrease in the SGZ coincides with a concomitant increase in the GCL.

Notably, GSK‐3β overexpression significantly accelerated and sustained this process, driving persistent GCL astrocytosis from 4 to 30 months of age and SGZ astrocytic reduction from 4 to 12 months. Notably, this SGZ astrocytic depletion at 12 months occurred without accompanying morphological alterations, in marked contrast to the structural remodeling observed in microglia at the same age. This pattern suggests that astrocytes and microglia follow distinct response trajectories in the SGZ, with astrocytic alterations manifesting mainly as cell loss and microglial alterations involving overt structural remodeling. This layer‐specific imbalance may also explain the transient alteration in GFAP⁺ astrocyte–Iba1⁺ microglia contacts observed at 12 months, as concomitant changes in astrocytic and microglial densities may modify the local spatial relationships between both cell populations.

In this pathological environment, the increased contact between astrocytes and newborn DGCs that we describe may represent a compensatory clustering aimed at providing survival cues to neuroblasts through factors such as IGF‐I and bFGF, both of which are released by astrocytes (Kirby et al. 2013; Pitt et al. 2017) under stress conditions and were found to be increased in GSK‐3β‐OE mice. However, such abnormal clustering may paradoxically contribute to the maturation failure observed in this model (Fuster‐Matanzo et al. 2013; Llorens‐Martin et al. 2013).

This disruption of glial homeostasis also extends to the oligodendroglial lineage. In the aging brain, this population exhibits a progressive impairment in progenitor differentiation and maintenance of myelin integrity (Sams 2021). While the overall density of mature oligodendrocytes remains largely stable across the lifespan in the DG (Adachi et al. 2010; Cole et al. 2022; Long et al. 1998), OPCs undergo structural dystrophy and a steady decline in both their number and proliferative capacity (Chacon‐De‐La‐Rocha et al. 2020; DeFlitch et al. 2022; Vanzulli et al. 2020; Yamada and Jinno 2014). Our results reveal a spatially compartmentalized dynamic within the neurogenic niche at mid‐life: the density of OPCs decreased selectively in the GCL, while that of mature oligodendrocytes increased within SGZ. This pattern suggests that local OPCs actively differentiate, progressively depleting their pool to sustain oligodendrogenesis, possibly as a compensatory response to age‐related axonal fragility within the niche (Young et al. 2013).

Against this backdrop of age‐related adaptation, neuronal GSK‐3β overexpression disrupted this balance, driving a reduction in mature oligodendrocytes within the SGZ from 12 months of age onwards. This loss coincided with the accumulation of NG2⁺ Olig2⁻ cells. As Olig2 is required for oligodendrocyte specification, its absence in OPCs redirects their glial commitment toward the astrocytic lineage (Zhu et al. 2012). This switch in lineage may account for the concurrent SGZ mature oligodendrocyte depletion and GCL astrocytosis described, implying that the inflammatory milieu induced by neuronal GSK‐3β overexpression alters the lineage commitment of NG2 glia through indirect, non‐cell‐autonomous mechanisms. This hypothesis is reinforced by previous findings in a viral‐mediated GSK‐3β S9A model, in which sustained kinase activity shifted progenitor identity toward astrogliogenesis at the expense of the neuronal lineage (Flor‐Garcia et al. 2021).

Additionally, changes within the niche also impacted the vascular scaffold that provides the essential metabolic and structural support required by the neurogenic environment. Aging is characterized by a decline in vascularization (Bell et al. 2010; Cole et al. 2022), thereby reducing oxygen supply and impairing nutrient exchange, with consequent alterations in AHN and cognitive performance (Perosa et al. 2020). Furthermore, aging blood vessels acquire a pro‐inflammatory secretory profile that drives detrimental crosstalk with surrounding neural progenitors and glial populations (Yousef et al. 2019). In this context, our observation of an increased SGZ vascular area in WT mice at 12 and 30 months of age, further supported by the upregulation of VEGF‐D and its receptors, likely indicates a compensatory angiogenic response to the reduced cerebral blood flow characteristic of the aging brain (Bell et al. 2010; Condrau et al. 2026; Rojas‐Vazquez et al. 2021). Interestingly, at 12 months, we also identified an increased overlap between microglia and blood vessels. This transient increase may be favored by the local availability of both cellular and vascular elements at this stage, coinciding with increased microglial density in the GCL and increased UEA‐1⁺ vascular area in the SGZ. This phenomenon, which has been observed in other studies (Cole et al. 2022), may reflect a targeted recruitment of microglia to the vasculature, possibly to mitigate or repair the BBB disruption associated with the onset of brain aging (Rojas‐Vazquez et al. 2021). The disappearance of this overlap at 30 months further suggests that glial–vascular interactions may be dynamically remodeled during middle age before declining at more advanced stages of senescence.

In GSK‐3β‐OE mice, vascular remodeling followed a distinct, non‐linear trajectory. These animals showed a more pronounced increase in SGZ vascular area at 12 and 18 months, accompanied by increased capillary diameter, altered CD31 fluorescence intensity, and elevated levels of VEGF‐D. However, this profile was not maintained at very advanced ages. By 30 months, GSK‐3β‐OE mice no longer exhibited increased vascular expansion relative to age‐matched WT animals, coinciding with reduced VEGF‐D and VEGFR expression. This dissociation indicates that the vascular changes observed in GSK‐3β‐OE mice should not be interpreted as a simple amplification of the compensatory angiogenic response proposed for WT aging. Rather, neuronal GSK‐3β overexpression appears to alter the temporal coordination between vascular remodeling and angiogenic signaling in the DG niche. In this scenario, the increased vascular area observed at earlier stages may reflect premature vascular restructuring, whereas the decline in VEGF‐D‐related signaling at 30 months suggests that this angiogenic response is not sustained during advanced aging.

Crucially, this structural remodeling coincides at 12 months with the accumulation of the NG2⁺ Olig2⁻ population described in the preceding paragraphs. Given that pericytes express NG2 and naturally lack Olig2 (Ozerdem et al. 2001), the expansion of this population may reflect the reactive proliferation of mural cells aimed at stabilizing a compromised vascular compartment (Kisler et al. 2017). Concomitantly, the upregulation of the endothelial adhesion molecules ICAM‐1 and E‐selectin points to a permissive environment for peripheral immune cell infiltration (Munoz‐Castro et al. 2023). These changes align with the elevated IL‐7 levels detected in our model. This cytokine is essential for the survival and proliferation of T‐lymphocytes, and its downstream effectors, including IFN‐γ, have been shown to exert deleterious effects upon infiltrating aged neurogenic niches (Dulken et al. 2019).

5. Conclusions

Collectively, our findings demonstrate that the DG micro‐environment is highly dynamic and continuously shaped by both aging and disease‐associated signals. The remodeling across the lifespan involves a layer‐specific reorganization of the cellular and vascular components, suggesting that pathological changes in the GCL are intrinsically linked to the progressive dysregulation of the SGZ niche. Notably, GSK‐3β overexpression both accelerates age‐related alterations and elicits additional pathological changes, underscoring the importance of age as a modifying variable in the pathogenesis of neurodegenerative diseases.

The temporal dissection of this pathological condition reveals a premature and non‐linear disruption of the DG pro‐neurogenic microenvironment. While alterations within the astrocytic population emerge as an early response at 4 months of age, a more comprehensive disruption involving microglial homeostasis, vascular architecture, and the oligodendroglial lineage becomes fully manifest by 12 months. Critically, the observation that GSK‐3β overexpression is restricted to mature DGCs in our model (Llorens‐Martin et al. 2013) demonstrates that cell‐autonomous neuronal dysregulation is associated with a cascade of non‐cell‐autonomous alterations that compromise the neurogenic niche. This finding expands upon previous descriptions (Fuster‐Matanzo et al. 2013; Llorens‐Martin et al. 2013; Lucas et al. 2001) by establishing the extrinsic dimension of GSK‐3β‐driven neurodegeneration at the niche level throughout life.

These results reinforce the central role of GSK‐3β in maintaining the structural and functional integrity of the DG. Given that the dysregulation of this kinase simultaneously destabilizes progenitor cell fate and disrupts the supporting glial and vascular environment, therapeutic strategies targeting GSK‐3β activity hold the potential to protect the neurogenic niche through a dual mechanism: directly preserving progenitor identity and indirectly restoring the homeostatic balance of the DG.

Author Contributions

A.V.P.‐S. and M.L.‐M. conceived and designed the study. A.V.P.‐S. and C.B.R.‐M. performed the experiments. A.V.P.‐S. analyzed the data. M.L.‐M. obtained funding. A.V.P.‐S. and M.L.‐M. wrote the manuscript. All the authors discussed the data and revised the final version of the manuscript.

Funding

This work was supported by European Research Council (ERC‐CoG‐2020‐101001916), Spanish Ministry of Economy and Competitiveness (PID2020‐113007RB‐I00, PID2023‐146572OB‐I00), BrightFocus Foundation (A2024021S), Center for Networked Biomedical Research on Neurodegenerative Diseases, and Fundación Tatiana Pérez de Guzmán el Bueno.

Ethics Statement

All protocols were approved by the CBMSO Animal Ethics Committee (AEEC‐CBMSO‐23/172) and the National Committee for Animal Experimentation (PROEX 185.4/20). All experiments complied with the applicable ethical standards for animal research.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: glia70248‐sup‐0001‐Supinfo1.xlsm.

Data S2: glia70248‐sup‐0002‐Supinfo2.xlsx.

Figure S1: Glial cells in the dentate gyrus (DG) of wild‐type (WT) mice across lifespan. (A) Subgranular zone (SGZ)/granule cell layer (GCL) cell density ratio of ionized calcium‐binding adapter molecule 1 (Iba1)⁺ microglia, glial fibrillary acidic protein (GFAP)⁺ astrocytes, oligodendrocyte transcription factor 2 (Olig2)⁺ oligodendrocytes, nerve/glial‐antigen 2 (NG2)⁺ cells, NG2⁻ Olig2⁺ mature oligodendrocytes, NG2⁺ Olig2⁺ oligodendrocyte progenitor cells (OPCs), and NG2⁺ Olig2⁻ cells in WT mice at 4, 12, 18, and 30 months of age. (B, C) Cell density of NG2⁺ Olig2⁻ cells in the GCL (B) and SGZ (C) of WT mice at 4, 12, 18, and 30 months of age. Graphs represent mean values ± SEM. *0.05 > p ≥ 0.01; **0.01 > p ≥ 0.001; ***0.001 > p ≥ 0.0001; ****p < 0.0001. Detailed results of statistical comparisons are shown in Extended Data S1.

Figure S2: Cell populations in the dentate gyrus of wild‐type (WT) and glycogen synthase kinase 3‐β‐overexpressing (GSK‐3β‐OE) mice. (A–I) Morphometric characteristics of ionized calcium‐binding adapter molecule 1 (Iba1)⁺ microglia from the subgranular zone (SGZ) of WT and GSK‐3β‐OE mice at 12 months of age. Convex (Cx) hull 3D volume (μm³) (A), Cx hull 3D surface (μm²) (B), number of primary trees (C), number of nodes (D), number of terminals (E), total process length (μm) (F), total cell surface (μm²) (G), total cell volume (μm3) (H), and cell complexity (I). (J–R) Morphometric characteristics of glial fibrillary acidic protein (GFAP)⁺ astrocytes from the subgranular zone (SGZ) of WT and GSK‐3β‐OE mice at 12 months of age. Cx hull 3D volume (μm³) (J), Cx hull 3D surface (μm²) (K), number of primary trees (L), number of nodes (M), number of terminals (N), total process length (μm) (O), total cell surface (μm²) (P), total cell volume (μm3) (Q), and cell complexity (R). (S, T) Sholl's (S) and branch ordering (T) analysis of astrocytes. (U, V): Neurolucida 360 tridimensional reconstructions of astrocytes from the SGZ of WT (U) and GSK‐3β‐OE mice (V) at 12 months of age. (W) Heatmap showing the morphological features of astrocytes. (X–AD) SGZ/granule cell layer (GCL) cell density ratio of Iba1⁺ microglia (X), GFAP⁺ astrocytes (Y), oligodendrocyte transcription factor 2 (Olig2)⁺ oligodendrocytes (Z), nerve/glial‐antigen 2 (NG2)⁺ cells (AA), NG2⁻ Olig2⁺ mature oligodendrocytes (AB), NG2⁺ Olig2⁺ oligodendrocyte progenitor cells (OPCs) (AC), and NG2⁺ Olig2⁻ cells (AD) in WT and GSK‐3β‐OE mice at 4, 12, 18, and 30 months of age. (AE–AF) Density of NG2⁺ Olig2⁻ cells in the GCL (AE) and the SGZ (AF) in WT and GSK‐3β‐OE mice at 4, 12, 18, and 30 months of age. Black scale bar: 10 μm. Graphs represent mean values ± SEM. *0.05 > p ≥ 0.01; **0.01 > p ≥ 0.001; ***0.001 > p ≥ 0.0001; ****p < 0.0001. Detailed results of statistical comparisons are shown in Extended Data S1.

Figure S3: Cluster of differentiation 31 (CD31)⁺ blood vessels in wild‐type (WT) and glycogen synthase kinase 3‐β‐overexpressing (GSK‐3β‐OE) mice. (A–R) Morphometric characteristics of CD31⁺ vessels in WT and GSK‐3β‐OE mice at 4, 12, 18, and 30 months of age. Number of isolated elements/μm³ (A), number of branching nodes/μm³ (B), number of endings/μm³ (C), total length (μm)/μm³ (D), surface area (μm2)/μm³ (E), total volume (μm3)/μm³ (F), loop count/μm³ (G), segment length (μm) (H), average segment diameter (μm) (I), segment tortuosity (J), node diameter (μm) (K), node volume (μm³) (L), node surface area (μm²) (M), loop length (μm) (N), loop surface area (μm²) (O), loop average diameter (μm) (P), loop volume (μm³) (Q), and loop nodes (R). (S, T) CD31⁺ vessel mean fluorescence intensity (arbitrary units [A.U.]) in the granule cell layer (GCL) (S) and subgranular zone (SGZ) (T) in WT and GSK‐3β‐OE mice at 4, 12, 18, and 30 months of age. U: SGZ/GCL CD31⁺ vessel mean intensity ratio in WT and GSK‐3β‐OE mice at 4, 12, 18, and 30 months of age. Graphs represent mean values ± SEM. *0.05 > p ≥ 0.01; **0.01 > p ≥ 0.001; ***0.001 > p ≥ 0.0001; ****p < 0.0001. Detailed results of statistical comparisons are shown in Extended Data S1.

Table S1: Antibodies used for immunohistochemistry. The following primary and secondary antibodies were used at the indicated dilutions.

Data Availability Statement

All data generated or analyzed in this study are included in this article and its Supporting Information files and are available from the corresponding author on reasonable request.