Entorhinal Cholecystokinin in Alzheimer’s Disease: Its Earliest Vulnerability and Rescue Effects Across Different Disease Stages in a Mouse Model

Abstract

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder characterized by cognitive decline. The entorhinal cortex (Ent) is among the earliest affected regions, and its neuropeptide cholecystokinin (CCK) supports neocortical-associated memory. Although our previous work demonstrated that CCK treatment rescues cognition and neuroplasticity in aged AD mice, the correlations among CCK expression, synaptic function, and cognitive decline with aging remain poorly understood. Using 3xTg-AD mice (2–18 months), we performed stereological, histological, and molecular analyses to evaluate brain atrophy, neuronal loss, glial responses, and gene expression. Cognitive function was assessed using the novel object recognition test (NOR), motor learning was evaluated using the rotarod test, and synaptic integrity was measured via electrophysiological recordings. We also investigated the therapeutic potential of CCK-B receptor (CCK-BR) agonists on learning and memory and neuroplasticity across disease stages of AD. In 3xTg-AD mice, the Ent exhibits significant atrophy and excitatory neuronal loss as early as 7 months of age, confirming its role as one of the earliest and most severely affected regions in AD. CCK was downregulated earlier than other synaptic genes in the Ent and other brain regions. Cholecystokinin tetrapeptide (CCK-4) treatment rescued deficits in synaptic plasticity, cognition, and motor learning in 3xTg-AD mice across multiple disease stages. Long-term administration of HT-267, a CCK-BR agonist with a prolonged half-life, in young 3xTg-AD mice delayed cognitive decline, enhanced synaptic scaffolding, and restored long-term potentiation in both the cortex and hippocampus (HPC). Our findings identify CCK downregulation as an early biomarker in AD and demonstrate the therapeutic effects of CCK-BR agonists in mitigating cognitive and synaptic deficits from mild to severe disease stages. The ability of long-term CCK-4 analogue treatment to decelerate AD progression indicates its promise as an early intervention strategy.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1007/s10571-026-01689-8.

Background

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder with a complex etiology, posing significant challenges for early diagnosis and treatment. Distinct from normal cognitive aging, AD is defined by hallmark neuropathological features, including the aggregation of amyloid-beta (Aβ) plaques and neurofibrillary tangles (NFTs) (DeTure and Dickson 2019). These pathological changes initially appear in the entorhinal cortex (Ent) and hippocampus (HPC), key areas for memory and learning, before spreading to broader cortical and subcortical regions (Trejo-Lopez et al. 2022). Understanding the spatiotemporal progression of AD pathology is critical for identifying early biomarkers and developing targeted interventions.

The medial temporal lobe (MTL), which includes the HPC, Ent, and adjacent cortical regions such as the perirhinal and postrhinal cortices, plays a central role in memory formation. Bilateral resection of the MTL in humans results in severe anterograde amnesia and profound memory deficits in patients (Squire 2009). Although brain atrophy occurs gradually during normal aging, the process accelerates markedly in AD, with the MTL being specifically vulnerable (Chauveau et al. 2021). These brain regions show early AD-related pathology, including Aβ deposition, tau aggregation, synaptic loss, and neuronal death, ultimately resulting in their progressive volume reduction (Jack et al. 1998). As the disease advances, atrophy extends beyond the MTL to other regions, such as the prefrontal cortex (PFC), insula, and anterior cingulate gyrus (Salat et al. 2001; Tan et al. 2013). Furthermore, patients with mild cognitive impairment (MCI) frequently display early deficits in spatial navigation, suggesting functional disruption within the Ent-HPC circuit, one of the first neural networks affected in AD (Hort et al. 2007). Therefore, elucidating the mechanisms underlying this selective vulnerability is important for advancing our understanding of AD pathogenesis and developing new therapeutic strategies.

Cholecystokinin (CCK) is a neuropeptide abundantly expressed in the brain, where it acts primarily through its CCK B receptor (CCK-BR) in the central nervous system (CNS) to regulate synaptic transmission, memory encoding, and emotion regulation (Lee and Soltesz 2011; Rehfeld 2021). As the gateway between the HPC and neocortex, the Ent establishes strong reciprocal connections via CCK-positive projections to both hippocampal and cortical areas (Li et al. 2014; Chen et al. 2013). In mouse models, disruption of CCK signaling, whether via genetic ablation (CCK knockout, CCK^−/−) or pharmacological blockade, compromises neocortical long-term potentiation (LTP) and associative learning, especially in tasks that require integration of environmental cues. Notably, these deficits can be reversed by exogenous CCK administration (Lee et al. 2011), indicating that CCK signaling in the Ent supports memory encoding in the neocortex by promoting local synaptic plasticity. In our previous study, CCK^−/− mice exhibited severe learning and memory deficits comparable to those in aged 3xTg-AD mice, a widely used triple transgenic AD mouse model. Importantly, these impairments were rescued by exogenous administration of CCK-BR agonists, suggesting a potential role of CCK in AD-related cognitive decline (Zhang et al. 2024).

In the present study, we investigated the age- and region-specific progression of AD neuropathology. We demonstrated the pronounced vulnerability of the Ent in 3xTg-AD mice, accompanied by a marked reduction in local CCK expression. This led us to hypothesize that CCK downregulation exacerbates excitatory synaptic dysfunction and thereby contributes to cognitive decline in AD. To test the therapeutic potential of enhancing CCK signaling, we employed Cholecystokinin tetrapeptide (CCK-4), a selective CCK-BR agonist with minimal affinity for CCK A receptor (CCK-AR) and proven ability to cross the blood-brain barrier (Berna et al. 2007; Feng et al. 2021). However, its utility is limited by a short plasma half-life (< 5 min in rodents) (Koulischer et al. 1982), necessitating frequent administration. To overcome this, we developed HT-267, a synthetic analogue of CCK-4 with structural modifications that confer metabolic stability, a prolonged half-life (> 90 min), and retained high potency and selectivity for CCK-BR (EC50 = 11.55 nM) (Zhang et al. 2024). We evaluated the rescue effects of both CCK-4 and HT-267 on cognitive, motor, and synaptic impairments in 3xTg-AD mice across different ages. We further examined the therapeutic potential of long-term administration of HT-267 for preventing AD progression. Our findings highlight the promise of CCK-BR-targeted therapy in mitigating cognitive decline in AD and provide a pharmacological rationale for developing stable CCK-BR agonists as a novel treatment strategy.

Methods

Animals

Male and female 3xTg-AD mice (strain: B6;129-Tg(APPSwe, tauP301L)1Lfa Psen1tm1Mpm/Mmjax; stock #034830-JAX) and age-matched wild-type (WT) mice of the 129S1/SvImJ strain (Stock #002448), which contributes to the genetic background of the 3xTg-AD model, were obtained from The Jackson Laboratory. Mice ranging from 2 to 18 months of age were used in this study. They were group-housed under standard specific pathogen-free (SPF) conditions at a constant temperature of 23 ± 1 °C, with a 12-hour light/dark cycle (lights on from 08:00 to 20:00), and provided with food and water ad libitum. All animal experiments were conducted in accordance with protocols approved by the Animal Research Ethics Sub-Committee of the City University of Hong Kong (Approval Nos. A-0661 and AN-STA-00000653).

Behavioral Analyses

Novel Object Recognition (NOR) Test

The novel object recognition (NOR) test assesses recognition memory by exploiting the innate preference of rodents for novelty (Lueptow 2017). The test consisted of three phases: habituation, training, and testing. On the first day (habituation), each mouse was allowed to freely explore the empty arena for 10 min. Training was performed 24 h later, in which mice were placed in the arena containing two identical objects for 10 min. Following a 1-hour retention interval, one familiar object was replaced with a novel object of a different shape and texture, and mice were allowed to explore both objects for 5 min in the testing session.

All objects were pre-validated to ensure no inherent preference and were thoroughly cleaned with 70% ethanol between trials to eliminate olfactory cues. The positions (left/right) of the novel and familiar objects during testing were counterbalanced across animals. Locomotor activity and animal trajectories were automatically tracked using SMART v3.0 video tracking software, while object exploration time was manually scored by an experimenter blinded to the experimental groups.

Exploratory behavior was strictly defined as the mouse directing its nose toward and within 2 cm of the object, with the head oriented toward it. Passive contact (e.g., sitting on or leaning against the object) or brief encounters lasting less than 1 s were excluded from analysis.

The recognition index (RI) is used to assess the recognition memory of mice and is calculated as follows:

To control for potential spatial bias, the location preference (LP) during the training session was assessed using the formula:

An LP value approximating 50% indicates the absence of inherent location preference.

Rotarod Test

The motor coordination and motor learning ability of mice were analyzed using an accelerating rotarod apparatus (Shiotsuki et al. 2010). During the training period, three trials were conducted daily with a 15-minute interval between each trial for five consecutive days. In each trial, mice were placed on a rotating rod accelerating from 5 to 50 rpm for a maximum of 300 s. Up to five mice were tested simultaneously, with each mouse separated by a black plastic barrier. The trial ended when the last mouse fell or when the maximum observation time (300 s) was reached. Fall latency was recorded per trial, with daily means of 3 trials representing motor performance. Body weights were recorded daily after the test.

Open Field Test (OFT)

The open field test (OFT) was conducted in a uniform white plastic arena (50 cm x 50 cm x 50 cm), under low-light conditions. The central zone (25 cm x 25 cm) occupies 25% of the total area. Locomotor activity was recorded by a video camera positioned above the arena and analyzed by SMART v3.0 video tracking software. At the beginning of each trial, the mouse was placed in the lower right corner of the arena and allowed to move freely for 5 min. The total distance traveled (cm), ambulatory time (s), and speed (cm/s) of each mouse’s movement were measured to evaluate its locomotor ability. Rearing behavior was defined as the number of times the mouse stood up on its hind limbs. The central activity was assessed using two parameters: the central/total distance ratio (%) and the number of entries into the central area (Seibenhener et al. 2015; Sturman et al. 2018).

Tissue Preparation and Histological Analyses

Brain Section Preparation and Immunohistochemistry (IHC)

Mice were deeply anesthetized by intraperitoneal injection of sodium pentobarbital (50 mg/kg) and transcardially perfused with 0.01 M phosphate-buffered saline (PBS), followed by 4% paraformaldehyde (PFA) in PBS. The perfusion with PFA lasted for at least 20 min to ensure adequate fixation of brain tissue. Brains were subsequently removed and post-fixed overnight in 4% PFA at 4 °C. Coronal Sect. (50 μm thick) were prepared using a cryostat (HM525 NX; Thermo Fisher Scientific) and stored at −20 °C in a cryoprotectant solution containing 50% PBS, 20% ethylene glycol, and 30% glycerol. For IHC, free-floating sections encompassing the regions of interest were thoroughly rinsed four times in PBS and blocked for 2 h at room temperature with 10% normal goat serum in PBS containing 0.2% Triton X-100. Sections were then incubated with primary antibodies diluted in the blocking buffer at 4 °C overnight. After incubation, sections were washed five times with PBS and then incubated with secondary antibodies prepared in a 5% blocking buffer for 2 h at room temperature. Following five final washes in PBS, sections were mounted onto glass slides. Image acquisition was performed using the Nikon A1HD25 confocal microscope. All images were processed and analyzed using ImageJ software.

RNAscope in Situ Hybridization Assay

RNAscope in situ hybridization was performed using the RNAscope^® Multiplex Fluorescent Reagent Kit v2 with TSA^® Vivid™ Dyes and the following probes: mouse Gad1 (C3 channel) and Slc17a7 (O2 channel). For RNAscope analysis, an independent group of mice was perfused, and their brains were rapidly frozen. Coronal Sect. (20 μm thick) were cut on a cryostat and mounted directly onto slides. Brain Sect. (20 μm thick) were dehydrated by incubation at −20 °C for 60 min, then baked at 60 °C for 20 min in a HybE™ oven. Sections were fixed in 4% PFA for 10 min, rinsed twice with distilled water (ddH₂O), dehydrated through graded ethanol (50%, 70%, 100%), and then air-dried for 10 min at room temperature. RNAscope Hydrogen Peroxide (3–5 drops) was applied to cover the entire surface of the slides, followed by incubation at room temperature for 10 min and 3–5 washes with ddH₂O. Slides were placed in preheated RNAscope 1x Target Retrieval solution and boiled at > 95 °C for 5 min, followed by rinsing with ddH₂O. After drying, a hydrophobic barrier was drawn around each brain slide, and Protease III was applied for 30 min at 40 °C. Slides were incubated with a pre-prepared 1x mixed target probe solution at 40 °C for 2 h. Subsequently, slides underwent sequential incubation with amplifier probes (Amp1, Amp2, and Amp3), followed by RNAscope Multiplex FL v2 HRP-C2 and HRP-C3, which developed signals corresponding to the C2 and C3 probes. HRP blockers were added after each HRP signal development. Finally, slides were mounted and stored in the dark at 4 °C. For quantification, two sections per mouse were hybridized. Each data point represents one section analyzed for Slc17a7 + and GAD1 + neuron densities.

Quantification and Morphological Analysis of NeuN-Positive Neurons

Neuronal quantification was performed using a systematic random sampling approach, analyzing every sixth serial section at 300 μm intervals to ensure unbiased coverage of the target regions. Sections immunostained with anti-NeuN antibody were imaged using a Nikon confocal microscope with a 40x objective. Z-stacks were acquired at 1 μm intervals to analyze neuronal somas in three dimensions.

For neuronal density analysis, three non-overlapping 100 × 100 μm sampling fields were randomly selected within each cortical layer (I, II, III, V/VI) of the Ent and hippocampal subregion (CA1, CA3). The dentate gyrus (DG) was excluded from analysis due to high neuronal density, which precluded accurate soma segmentation at the current imaging resolution. Neuronal density (expressed as 10⁵ cells/mm³) was calculated for each sampling field. Data from all fields and sections were averaged to make a single mean density value per layer or subregion for each animal, which was considered one biological replicate in subsequent statistical analyses. The overall neuronal density for the Ent or HPC was derived by averaging values across all respective layers or subregions per mouse.

For morphological analysis, the cross-sectional area of NeuN-positive somas was measured within the same sampling fields used for density quantification. Soma size was automatically calculated using ImageJ software after manual delineation of the sampling regions. Measurements from all sampling areas and sections were averaged to generate a single mean soma size per region for each animal.

Throughout all analyses, consistent imaging parameters, field selection criteria, and unbiased stereological principles were applied to ensure the reproducibility and accuracy of both density and morphological measurements.

Quantification and Morphological Analysis of GFAP-Positive Astrocytes

Analysis of astrocytes was systematically performed on every sixth consecutive coronal section (spaced 300 μm apart), to ensure unbiased sampling of the entire target area. Sections were immunolabeled with anti-GFAP antibodies to visualize the astrocytic cytoskeleton. High-resolution z-stack images (1 μm interval) were acquired using a Nikon confocal microscope equipped with a 60x oil immersion objective to capture the full three-dimensional morphology.

Density quantification: For the HPC, one systematically sampled 300 × 300 μm field per section was analyzed from each subregion (CA1, CA3, and DG). The mean astrocyte density across these subregions was calculated to represent the overall HPC astrocytic density for each mouse. For the Ent, astrocyte density was quantified in one 300 × 300 μm field per section (spanning all cortical layers I-VI), avoiding layer-specific analysis due to low cell density. Data from all analyzed fields across multiple sections were averaged for each brain to derive a single mean astrocytic density value.

Morphological analysis: The volume of the GFAP-positive cytoskeleton was quantified as a primary indicator of astrocytic reactivity. Z-stack images were imported into ImageJ software for 3D reconstruction and analysis using the “3D Objects Counter” plugin. This involved applying an intensity threshold to the original grayscale image to objectively segment the GFAP-positive signal from the background, followed by conversion into a binary image (black and white) for quantification. The volume (in µm³) of each individual astrocyte was then automatically calculated. In the HPC, one 200 × 200 μm field per section from each subregion (CA1, CA3, and DG) was analyzed. For layered morphological analysis in the Ent, two 100 × 100 μm fields were sampled per layer (I, II, III, V/VI) per section. The mean astrocytic volume value was calculated from all analyzed layers/fields to represent the overall astrocytic volume for each mouse. The consistent application of imaging parameters, field selection criteria, and analysis methods ensured the reproducibility of results across experiments.

Quantification and Morphological Analysis of Iba1-Positive Microglia

Analysis of microglia was systematically performed on every sixth consecutive section (spaced 300 μm apart). Microglia were immunostained with anti-Iba1 antibody. High-resolution z-stack images (1 μm interval) were acquired using a Nikon confocal microscope with a 60x oil immersion objective.

Density quantification: To quantify the density of Iba1-positive microglia, two non-overlapping 250 × 250 μm fields were systematically sampled per section for analysis in the global Ent or HPC. Data from all sampled fields from multiple sections of each brain were averaged to obtain a single mean microglial density value.

Morphological analysis: Microglial activation was assessed using a well-established classification system based on soma size and process morphology, which strongly correlates with functional state (Franco-Bocanegra et al. 2019; Kettenmann et al. 2011; Soltys et al. 2001; Kreutzberg 1996). Microglia were classified into three states by a trained, blinded experimenter: Resting: small somas (< 30 µm²) with thin, highly ramified processes; Intermediate: moderate soma sizes (30–60 µm²) with partially retracted, thicker processes; Reactive: large soma sizes (> 60 µm²) with severely retracted, thickened processes or amoeboid morphology.

The cross-sectional area of the Iba1-positive soma was quantitatively measured using ImageJ software after manual delineation. For each animal, at least 100 cells were classified across multiple fields and sections. The proportion of cells in each state was then calculated.

Quantification of Glial Association with Pathological Aggregates

To quantitatively assess the association of astrocytes and microglia with AD pathology, we performed a dual immunofluorescence labeling for phosphorylated tau (pTau) and amyloid-β (Aβ), followed by a proximity analysis based on established methods (Serrano-Pozo et al. 2013; Su et al. 2025; Olabarria et al. 2011).

Brain sections were co-immunostained with a mixture of two mouse-derived primary antibodies: anti-Phospho-Tau (AT8) and anti-β-Amyloid (6E10) to simultaneously label pTau tangles and Aβ plaques, respectively. This allowed us to treat both types of pathologies as a unified category of “pathological aggregates” for subsequent quantification.

For each animal (n = 4 3xTg-AD mice per age group), three non-overlapping fields of view (300 × 300 μm) containing at least one pathological aggregate (either Aβ + or pTau+) were selected from the Ent and HPC for analysis. A glial cell was considered “associated” with a pathological aggregate if its cell soma or primary processes were located within a 50 μm radius from the edge of the aggregate. For each selected field, we manually counted the total number of pathological aggregates (either Aβ plaques or pTau tangles) and the number of those aggregates that had at least one associated glial cell (GFAP + for astrocytes or Iba1 + for microglia) as defined above. The data are presented as the percentage of aggregates associated with glia for each animal, calculated as:

For each mouse, the percentage values from the three fields were averaged to generate a single biological replicate.

Immunofluorescence Quantification of Postsynaptic Density Protein-95 (PSD-95)

Brain sections were collected from 3xTg-AD mice after the long-term treatments with the same method as described in IHC and were double-stained with anti-PSD-95 and anti-NeuN antibodies. Quantification of hippocampal PSD-95 fluorescence intensity was performed on 5 coronal sections per mouse. For each section, three 150 μm x 150 μm fields were systematically sampled in the HPC. The mean fluorescence intensity from the three sampled fields was calculated for each brain section. Fluorescence intensity was normalized to adjacent white matter regions (PSD-95-negative) using the formula:

For cortical PSD-95 positive neuron density quantification, five coronal sections per mouse were immunostained and imaged at 1 μm z-step intervals (total 15 μm depth). Within each section, three non-overlapping 200 μm x 200 μm fields were systematically sampled in the cortex. Section-level densities were obtained by averaging triplicate measurements.

Image Acquisition and Analysis

All confocal images were acquired using a Nikon AX R confocal microscope system controlled by NIS-Elements Advanced Research software (version 6.02.03). Image acquisition was performed at room temperature with a pixel dwell time of 2.0 µs and a pinhole set to 1.0 Airy Unit (AU). The following objectives were used for different quantitative analyses: a 20x objective (NA 0.75) for brain region volumetry (Fig. 1), a 40x objective (NA 0.95) for neuronal and synaptic density quantification (Figs. 2, 3, 4, 5, 6 and 7), and a 60x oil immersion objective (NA 1.40) for high-resolution three-dimensional morphological analysis of glial cells (Figs. 2 and 3). Z-stacks were collected at 1.0 μm intervals for morphological reconstructions. All subsequent image processing and quantitative analyses were performed using ImageJ software.

Fig. 1.

Age-dependent morphological changes in the cerebrum and medial temporal lobe subregions of 3xTg-AD mice. All volumetric data and representative images in this figure were obtained from male mice. A Representative dorsal views of whole brains from 18-month-old WT (left) and 3xTg-AD male mouse (right), including the olfactory bulb, cerebrum, and cerebellum. Dashed lines outline the cerebrum for comparison. Scale bar, 2 mm. B Schematic of cerebrum length measurement: Coronal Sect. (50 μm thickness) were collected from + 3.08 mm (rostral) to −4.96 mm (caudal) relative to Bregma. Cerebrum length = N x 50 μm (“N” is the total number of sections). C Cerebrum length across age groups. D Dorsal view of a representative coronal section from a 2-month-old WT male mouse illustrating the dimensions measured for cerebrum width and thickness (dashed lines). Quantified values in (E) and (F) represent the mean width and thickness derived from multiple systematically sampled sections per brain (see Methods for details). Scale bar, 1 mm. E, F Quantification of cerebrum width (E) and thickness (F) at indicated ages. G Volume of the cerebral cortex across ages. H–L Age-related volumetric changes in the Ent (H), PrC (I), POR (J), HPC (K), and PFC (L) across ages. M Relative volume changes in the HPC, Ent, and cerebral cortex of WT (left) and 3xTg-AD (right) mice, normalized to 2-month-old WT controls. Data are derived from n = 4 biologically independent mice per group, except for panel (K) (HPC volume), where n = 5. Technical replication details are provided in the Methods and Supplementary Table S1. Data are presented as mean ± SD. Statistical significance was assessed by two-way repeated measures (RM) ANOVA with post hoc Bonferroni’s test. At the right side of the curves, the significance levels for the genotype × age interaction are displayed. *P < 0.05, **P < 0.01, ***P < 0.001 for between-genotype comparisons at the same age; #P < 0.05, ##P < 0.01, ###P < 0.001 for within-genotype comparisons relative to 2-month-old counterparts. NS, not significant (P > 0.05)

Fig. 2.

Neuronal and glial alterations occur in the Ent of 3xTg-AD mice. A Density (x10⁵ cells/mm³) of NeuN+ neurons in the global Ent. B Mean neuronal soma area in the Ent. C Representative coronal sections showing Slc17a7+ (green) excitatory and GAD1+ (red) inhibitory neurons in the Ent. The dashed lines demarcate the boundaries of the entorhinal cortical layers, which were identified based on established cytoarchitectonic features. Scale bar, 250 μm. D Density (cells/mm²) of Slc17a7 + and GAD1 + neurons in the Ent. n = 3 mice per group per age, with two brain sections hybridized per mouse (six sections analyzed per group). E Representative images of GFAP+ (green) astrocytes in the Ent. Scale bar, 20 μm. F Volume (µm³) of GFAP+ astrocytes in the Ent. G Density (cells/mm³) of GFAP+ astrocytes in the Ent. H Representative images of Iba1+ (red) microglia in the Ent categorized as ramified (white triangle), reactive (green triangle), or amoeboid (blue triangle). Scale bar, 50 μm. I Proportion (%) of activated (reactive and amoeboid) microglia in the Ent. J Density (cells/mm³) of microglia in the Ent. K Representative images of GFAP+ (green) astrocytes and Iba1+ (red) microglia in relation to Aβ (magenta, left) and pTau (magenta, right) deposits in the Ent of 7-, 12-, and 18-month-old 3xTg-AD mice. Squares indicate areas within 50 μm of pathological aggregates. L Quantification of the percentage of pathological aggregates associated with glia in the Ent. Data are derived from n = 4 biologically independent mice per group, except for (C) and (D) (RNAscope in situ hybridization), where n = 3. Technical replication details are provided in the Methods and Table S1. Data are presented as mean ± SD. Statistical analysis was performed by two-way RM ANOVA with Bonferroni’s post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001; NS, not significant

Fig. 3.

Impaired neurogenesis and glial responses in the hippocampus of 3xTg-AD mice. A Representative images of DCX+ (green) neurons in the DG. Scale bars: 300 μm (overview), 50 μm (magnified views). B Density (cells/mm²) of DCX+ cells in the DG. C Representative images of NeuN+ (red) neurons in CA1 and CA3 of 18-month-old WT and 3xTg-AD mice. Scale bars: 100 μm (overview), 50 μm (magnified views). D Density (x10⁵ cells/mm³) of NeuN+ neurons in the global HPC across ages. E Representative images of GFAP+ (green) astrocytes in CA1, CA3, and DG at 2 and 18 months of age. Scale bar, 50 μm. F Volume (µm³) of GFAP+ astrocytes in the global HPC across ages. G Density (cells/mm³) of GFAP+ astrocytes in the global HPC across ages. H Representative images of Iba1+ (red) microglia in ramified (white triangle), reactive (green triangle), and amoeboid (blue triangle) states. Scale bar, 40 μm. I Proportion (%) of activated (reactive and amoeboid) microglia in the HPC across ages. J Density (cells/mm3) of microglia in the HPC across ages. K Representative images of GFAP+ (green) astrocytes and Iba1+ (red) microglia adjacent to Aβ (magenta, left) and pTau (magenta, right) deposits in the HPC of 7-, 12-, and 18-month-old 3xTg-AD mice. Squares indicate the area within 50 μm of aggregates. L Quantification of the percentage of pathological aggregates associated with glia in the HPC. Data are derived from n = 4 biologically independent mice per group per age. Detailed information on technical replicates is provided in the Methods and Table S1. Data are presented as mean ± SD. Statistical significance was assessed by two-way RM ANOVA with Bonferroni’s post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001 for between-genotype comparisons at the same age; #P < 0.05, ##P < 0.01 for within-genotype comparisons relative to the 12-month-old group. NS, not significant

Fig. 4.

CCK was downregulated age-dependently and earlier than other synaptic function-associated genes in multiple brain regions. A–F Relative mRNA expression levels of CCK, CCK-BR, and synaptic function-associated genes (BDNF, GAD1, SLC1A2, SYN1) in the Ent (A), AC (B), VC (C), HPC (D), PFC (E), and MC (F) of 7-, 12-, and 18-month-old 3xTg-AD mice. Gene expression was normalized to GAPDH and is presented as fold change relative to age- and region-matched WT controls (set to 1.0). Data are derived from n = 6 biologically independent mice per group; each sample was analyzed in technical triplicate, and the mean value was used for analysis. Data are presented as mean ± SD. Statistical significance was assessed by two-way RM ANOVA with Bonferroni’s post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001; NS, not significant

Fig. 5.

CCK-4 rescues age-related cognitive and motor learning deficits in 3xTg-AD mice. A Schematic of the NOR test procedure. B Location preference (left) during NOR training phase and recognition index (right) during NOR testing phase of 3xTg-AD and WT mice at different ages. C NOR test protocol with intraperitoneal (i.p.) administration of CCK-4 or VEH in 3xTg-AD mice 2 min before training. D Location preference (left) and recognition index (right) in the NOR test of CCK-4- or VEH-treated 3xTg-AD mice. E Schematic of the rotarod test procedure. F Learning curve (latency to fall) across training days for 3xTg-AD and WT mice at different ages. G Latency to fall on the final training day (Day 5) of 3xTg-AD and WT mice. H Rotarod test protocol with daily i.p. administration of CCK-4 or VEH in 3xTg-AD mice 2 min before training. I Learning curve for CCK-4- or VEH-treated 12- and 18-month-old 3xTg-AD mice. Sample sizes: n = 10 biologically independent mice per group for all NOR experiments (A-D); n = 14 per group for all rotarod experiments (F-I). Data are presented as mean ± SD. Statistical significance was assessed by two-way RM ANOVA with Bonferroni’s post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001; NS not significant

Fig. 6.

CCK-4 application restores age-dependent neuroplasticity impairments in 3xTg-AD mice. A, C Representative cortical (A) and hippocampal (C) slices showing recording sites (black dots) and stimulation site, and the corresponding TBS paradigm. B, D Scatter plots show the normalized fEPSP amplitudes before and after TBS in cortical (B) and hippocampal (D) slices with or without CCK-4 infusion of 3xTg-AD mice at 7-, 12-, and 18-month-old. E-F Summary bar charts of normalized fEPSP amplitudes averaged over the last 15 min of recordings in cortical (E) and hippocampal (F) slices. For ex vivo electrophysiological recordings (A–F), N = 4 mice per group per age were used, with 3 slices recorded per mouse (total n = 12 slices per group). For (B) and (D), data are presented as mean ± SEM in the scatter plots. For (E) and (F), data are presented as mean ± SD in the bar charts. For (B) and (D), at the right side of the curves, the significance levels of the “treatment × time” interaction effect on fEPSP amplitude are displayed, reflecting whether the time course of potentiation differs statistically across treatment conditions. Statistical significance was assessed by two-way RM ANOVA with Bonferroni’s post hoc test. ***P < 0.001, **P < 0.01, *P < 0.05; NS not significant

Fig. 7.

Long-term HT-267 treatment improves cognitive function, synaptic integrity, and neuroplasticity in 3xTg-AD mice. A Schematic timeline of the 90-day HT-267 treatment and experiments. B Representative movement traces of 3xTg-AD mice in the OFT before and after long-term treatment. C Changes in total distance traveled (left), speed (middle), and percentage of distance in the center zone (right) across 7 OFT sessions. D Location preference (left) and recognition index (right) in the NOR test before and after treatment. E–F Left: Scatter plots of normalized fEPSP amplitudes before and after TBS in cortical (E) and hippocampal (F) slices, recorded from 9-month-old 3xTg-AD mice following the long-term HT-267 or VEH treatment. Right: Summary bar charts of normalized fEPSP amplitudes over the last 15 min of recordings. G Representative images of PSD-95 (green) and NeuN (red) immunoreactivity in the HPC of HT-267- or VEH-treated 3xTg-AD mice. Scale bar, 50 μm. H Relative PSD-95 fluorescence intensity in hippocampal apical dendrites of HT-267-treated mice compared to VEH-treated controls. I Representative images of PSD-95 and NeuN co-labeling in cortical neurons. Scale bar, 50 μm. J Density (cells/mm³) of PSD-95 + neurons in the cortex of HT-267- or VEH-treated 3xTg-AD mice. Sample sizes: n = 10 mice per group for all behavioral tests (C, D); N = 6 mice per group for electrophysiology (E, F), with 3–4 slices per mouse (n = 20 slices per group per region); N = 4 mice per group for IHC (H, J), with n = 20 sections analyzed per group. Data in bar charts are presented as mean ± SD, except for the scatter plots in (E, F), which are presented as mean ± SEM. For (C), (E), and (F), the significance markers at the right side of the curves represent the respective interaction effects; for (C), the “treatment × session” effect on open field behavior, and for (E, F), the “treatment × time” effect on fEPSP amplitude, reflecting whether the longitudinal profiles differ statistically across treatment conditions. Behavioral data (C, D) were analyzed by two-way RM ANOVA with Bonferroni’s test. Electrophysiological summary data (E, F bar charts) and immunohistochemical data (H, J) were analyzed by two-tailed unpaired Student’s t-test. *P < 0.05, **P < 0.01, #P < 0.05, ##P < 0.01, NS not significant

A comprehensive table detailing the laser wavelengths, power settings, emission filters, and detector gains for each specific experiment and fluorescent channel is provided in Supplementary Table S3.

Stereological and Volumetric Measurements

Measurement of Brain Dimensions

Brain size was quantified in three orthogonal dimensions: anteroposterior (AP) length, mediolateral (ML) width, and dorsoventral (DV) thickness, using systematic sampling of serial sections.

AP Length: The cerebrum was serially sectioned in the coronal plane at a thickness of 50 μm, spanning from + 3.08 mm anterior to Bregma to −4.96 mm posterior to Bregma. The total AP length was calculated as the product of the total number of collected sections (N) and the section thickness (50 μm).

ML Width: The ML width was assessed by measuring the distance along the mediolateral axis in coronal sections sampled at 300 μm intervals (every 6th section), ensuring comprehensive coverage of the entire cerebrum (25–30 sections per brain). The ML width for each brain was determined as the mean value across all sampled sections.

DV Thickness: Similarly, the DV thickness was measured along the dorsoventral axis in the same set of sections sampled at 300 μm intervals. The mean DV thickness across all sampled sections was taken as the representative value for each brain.

All measurements were performed using ImageJ software, ensuring consistent anatomical landmarks were used across all samples.

Volumetric Measurement of Brain Regions

The volumes of specific brain regions were estimated using the Cavalieri principle, an unbiased stereological method, on anti-NeuN-immunostained sections (Gundersen et al. 1987; West et al. 1990; Karaca et al. 2020; Sonmez et al. 2010; Altunkaynak et al. 2009). Serial sections were collected systematically in a rostral-to-caudal direction, and every section containing the region of interest within the sampling frame was included in the analysis. The total number of sections analyzed for each region was determined by its rostrocaudal extent, which was identified using the Paxinos and Franklin mouse brain atlas. Images of the selected sections were acquired with a Nikon confocal microscope. The boundaries of each brain region were manually delineated in ImageJ software according to the Paxinos and Franklin mouse brain atlas, and the cross-sectional area was measured after proper spatial calibration (Paxinos and Franklin 2019). The total volume (V) of the target brain region was calculated as:

“t” is the section thickness (50 μm in this study), and “ΣS” is the sum of the measured cross-sectional areas across all sampled sections.

Molecular Analyses

Quantitative Real-Time PCR (qRT-PCR)

Quantitative real-time PCR (qRT-PCR) was performed to quantify the expression levels of target genes in specific brain regions. Mice were deeply anesthetized with isoflurane and then decapitated for brain collection. The regions of interest were freshly dissected and immediately homogenized in TRIzol™ reagent for total RNA extraction. Subsequently, RNA was reverse-transcribed to complementary DNA (cDNA) by PrimeScript™ RT Reagent Kit. qRT-PCR amplification was carried out on the QuantStudio™ 3 Real-Time PCR System. Sequences of the primers are listed in Supplementary Table S2 of the Supplementary Material. Each cDNA sample was run in triplicate as technical replicates, and the mean cycle threshold (Ct) value was used for subsequent calculations. Relative gene expression levels were calculated using the comparative ΔΔCt method, with GAPDH as the internal reference gene. For each target gene, the expression level in 3xTg-AD mice was normalized to the mean value of the corresponding samples from region- and age-matched WT mice, which was set as 1.0.

Ex Vivo Electrophysiological Recording

Mice were deeply anesthetized with isoflurane and decapitated. Brains were quickly extracted and immersed in ice-cold oxygenated (95% O₂ − 5% CO₂) artificial cerebrospinal fluid (ACSF) (124 mM NaCl, 26 mM NaHCO₃, 10 mM glucose, 3 mM KCl, 2 mM CaCl₂, 1.25 mM KH₂PO₄, and 1.25 mM MgSO₄, pH 7.4). After cooling for 30 s, the brain was fixed on a vibrating slicer and immersed in ice-cold oxygenated ACSF. Coronal brain Sect. (250 μm thick) containing cortical and hippocampal regions were prepared and incubated in oxygenated ACSF at 35 °C for 2 h of recovery before electrophysiological recording. Extracellular field potential recordings were performed using a MED64 multielectrode array system. Brain slices were fixed on a MED probe with a fine-mesh anchor, and electrodes were positioned over the target region. During field excitatory postsynaptic potential (fEPSP) recordings, slices were continuously perfused with oxygenated ACSF at 35 °C. One microelectrode channel was selected for stimulation, with intensity adjusted to elicit 30% − 50% of the maximal response during baseline recording. After establishing a stable fEPSP baseline for at least 15 min, theta-burst stimulation (TBS) was applied to induce long-term potentiation (LTP) using an intensity that can evoke 75% − 90% of the maximal response. Following TBS, responses were recorded at baseline intensity for 1 h. Electrophysiological data were analyzed with MED Mobius software. fEPSP amplitudes were normalized and expressed as a percentage change from the baseline level. LTP magnitude was determined as the mean of the normalized fEPSP over the last 15 min of recording.

Drug Treatment

For in vivo behavioral experiments, CCK-4 and HT-267 were dissolved in saline (0.9% NaCl) containing less than 0.2% DMSO to facilitate dissolution. The vehicle (VEH) control consisted of saline with an equivalent DMSO concentration. Intraperitoneal (i.p.) injection volumes were standardized based on the estimated total blood volume (TBV) of mice, calculated as 56.5 ml/kg body weight (Cole et al. 2000). For mice weighing 30–40 g, the TBV was estimated to be 1.68–2.24 ml. Accordingly, the i.p. volume of CCK-4 (20 µM) was 8.4–11.2 µl, and the i.p. volume of HT-267 (320 µM) was about 20.0–26.7 µl. In the rotarod and NOR tests, a single i.p. injection of CCK-4 or VEH was administered 2 min prior to the training session. For the long-term intervention study, mice received a daily i.p. injection of HT-267 or VEH over a 90-day period. For ex vivo electrophysiological recordings, CCK-4 or HT-267 was dissolved in artificial cerebrospinal fluid (ACSF) with < 0.2% DMSO at a final concentration of 100 nM. The VEH control contained ACSF with an equivalent DMSO concentration to the drug-treated group. Brain slices were incubated in either drug or VEH solution for 5 min before TBS.

Statistical Analysis

Data are presented as mean ± SD, except for the scatter plots of electrophysiological recordings (fEPSP amplitudes before and after TBS), which are presented as mean ± SEM. All statistical analyses were performed by GraphPad Prism 8.0.1 software.

The choice of parametric tests was justified as follows. Prior to analysis, the normality of data distribution for each group was confirmed using the Shapiro-Wilk test, and homogeneity of variances was verified using Levene’s test. For all datasets subjected to parametric analysis, including those from qPCR, behavioral, electrophysiological, and pharmacological studies, the assumptions of normality and homogeneity of variance were met (all P > 0.05), validating the use of tests such as t-tests and ANOVA. For brain morphological and cellular quantification studies, some datasets derived from small sample sizes showed deviations from normality. Although non-parametric tests (Mann-Whitney U test) were subsequently performed and drew conclusions fully consistent with the parametric ANOVA results, we primarily report the two-way RM ANOVA findings for the following critical reasons. First, two-way RM ANOVA is suited to address our central hypothesis regarding the interaction between “Genotype” and “Time” across multiple age points. This statistical model allows for the simultaneous evaluation of both main effects and their interaction, important for studying progressive neurodegeneration that cannot be directly assessed using standard non-parametric alternatives. Second, the robustness of the parametric approach is supported by several factors: (1) the large anticipated effect sizes (Cohen’s d ≈ 2.0) observed in our pilot data and supported by the literature, and (2) the high precision of our stereological methods, which involved systematic sampling of numerous sections per animal, thereby increasing the reliability of each data point. The consistency between parametric and non-parametric results further reinforces the credibility of our findings.

The following statistical tests were applied: Two-way RM ANOVA with post hoc Bonferroni’s test (for comparisons involving two independent variables); ordinary one-way ANOVA with post hoc Dunnett’s test (for single-variable comparisons against a control group across three or more groups); two-tailed unpaired Student’s t-test (for direct comparisons between two groups).

The number of biological replicates (independent samples or animals) and technical replicates for each figure panel is summarized in Supplementary Table S1.

Results

Distinct Spatiotemporal Trajectories of Brain Volume in 3xTg-AD Mice: Rapid Entorhinal Atrophy and Early Hippocampal Growth Arrest

Identifying region-specific brain atrophy patterns is critical for early detection and intervention of AD. As an important region regulating learning and memory, MTL is known to be particularly vulnerable to early neurodegeneration (Chauveau et al. 2021). To systematically characterize these changes, we investigated global and region-specific brain volume changes in 3xTg-AD and WT mice across multiple ages.

We first quantified the whole-brain dimensions, including anteroposterior (AP) length, mediolateral (ML) width, and dorsoventral (DV) thickness, in WT and 3xTg-AD mice at 2, 5, 7, 12, and 18 months of age. Given the large workload of this comprehensive stereoscopic analysis across multiple ages, we initially characterized volume changes using male mice. In 3xTg-AD mice, all three dimensions followed a similar trajectory: a slight increase up to 7 months of age, reflecting normal postnatal brain development, followed by progressive shrinkage. By 18 months, all measures were significantly reduced in 3xTg-AD mice compared to WT controls (Fig. 1C, E, F), indicating a transition from early developmental growth to later neurodegeneration in overall brain size.

Next, we examined volume changes in the cerebral cortex and MTL subregions. Significant atrophy across the cerebral cortex emerged in 3xTg-AD mice at 12 and 18 months, indicating widespread degeneration in late AD stages (Fig. 1G and S1A-B). Notably, MTL subregions showed distinct temporal patterns of volume loss. In 3xTg-AD mice, the Ent showed the earliest and most severe atrophy, with significant reduction detectable by 7 months of age and progressive decline thereafter (Figs. 1H and 6.17 mm³ in 3xTg-AD vs. 7.25 mm³ in WT; 14.9% decrease, P = 0.0036). In contrast, the perirhinal cortex (PrC) and postrhinal cortex (POR) exhibited delayed atrophy, becoming significant only at 12 and 18 months (Fig. 1I, J). These data suggest that neurodegeneration spreads sequentially from the Ent to adjacent cortical areas as AD progresses.

Hippocampal volume also displayed distinct developmental trajectories between genotypes. In WT mice, hippocampal volume increased steadily during early postnatal stages and stabilized thereafter, indicating normal postnatal development (Hammelrath et al. 2016; Zhang et al. 2005; Calabrese et al. 2013). In 3xTg-AD mice, however, hippocampal growth was impaired early in development, leading to significantly smaller volume by 5 months of age relative to WT controls (Figs. 1K and 16.54 mm³ in 3xTg-AD vs. 18.99 mm³ in WT; 12.9% decrease, P = 0.0119). These findings imply that the smaller hippocampal volume in 3xTg-AD mice stems from an early neurodevelopmental growth arrest, rather than from the progressive degenerative atrophy seen in the Ent.

Beyond the MTL, significant volume loss was also observed in the PFC, a region important for cognitive regulation (Salat et al. 2001). However, different from the Ent, PFC atrophy in 3xTg-AD mice emerged later, with significant reductions detected only at 18 months of age (Fig. 1L and S1C).

To conclude, morphological analysis of 3xTg-AD brains revealed distinct region-specific volume changes during AD progression. The Ent showed rapid progressive atrophy, while the HPC exhibited early growth arrest (Fig. 1M). These findings suggest that the AD pathogenesis in this model may involve not only progressive neurodegeneration, but also early neurodevelopmental defects. Furthermore, the late-stage atrophy observed in multiple brain regions (PrC, POR, PFC) indicates progressive disease spread throughout the brain.

Early Excitatory Neuron Loss and Distinct Glial Alteration Underlie Entorhinal Vulnerability in 3xTg-AD Mice

To further investigate the cellular basis of Ent atrophy, we analyzed neuronal density, cell-type-specific apoptosis, and glial activation within this region. Histological analysis revealed progressive neuronal degeneration in the Ent of 3xTg-AD mice. Overall neuronal density declined with age and was significantly lower in 3xTg-AD mice than in age-matched WT mice by 7 months of age (Fig. 2AA, 1.34 × 10⁵ vs. 1.57 × 10⁵ cells/mm³, P = 0.0278). Layer-specific analysis further demonstrated that this neuronal loss was most pronounced in layer II of the Ent, which exhibited a 17.4% reduction in density compared to WT controls (Figure S2 A-B, 2.06 × 10⁵ vs. 2.49 × 10⁵ cells/mm³, P = 0.0237). By later stages of AD, neuronal degeneration extended to all Ent layers. In parallel, the average neuronal soma size progressively decreased (Fig. 2B), and frequency distribution analysis revealed a shift toward smaller soma profiles (Figure S2C), suggesting continuous neuronal shrinkage. Cell-type-specific analysis via RNA in situ hybridization further revealed preferential reduction in excitatory neurons in the Ent of 3xTg-AD mice at 7 months, with a 17.2% decrease relative to WT controls (Figs. 2D, top, and 432.3 vs. 521.8 cells/mm², P = 0.0472). Inhibitory neurons remained unchanged at this stage. By later time points, however, both excitatory and inhibitory neuronal populations were substantially diminished, reflecting widespread neuronal network disruption (Fig. 2D).

Glial cells in the Ent also exhibited distinct, age-dependent responses. Astrocytes displayed early and persistent atrophy, with significantly reduced somatic volume in 3xTg-AD as early as 2 months of age (Figs. 2F and 143.0 µm³ in 3xTg-AD vs. 230.6 µm³ in WT; 38.0% decrease, P < 0.001), a pattern consistent across all Ent layers (Figure S3B). Microglia, in contrast, underwent a gradual activation process (Fig. 2H-I). A growing proportion adopted a reactive phenotype by 5 months of age, characterized by enlarged somata and shortened, thickened processes (Figs. 2I and 14.25% in 3xTg-AD vs. 35.95% in WT; P = 0.017). By later stages (12 and 18 months of age), most microglia adopted a fully activated amoeboid morphology. Although astrocyte and microglial density in the Ent remained unchanged across ages (Fig. 2G and J), the absolute cell number likely decreased in proportion to overall tissue atrophy. In 3xTg-AD mice with established pathology, we also assessed glial associations with Aβ plaques and phosphorylated tau (pTau) tangles (Fig. 2K). Quantification of these interactions revealed a clear divergence in glial responses: as early as 7 months of age, microglia were found near the vast majority (86.7%) of aggregates, whereas astrocytes were associated with significantly fewer (44.0%) aggregates (P < 0.001; Fig. 2L). This profound disparity persisted at 12 and 18 months of age. Notably, entorhinal astrocytes remained distant from these aggregates even at advanced disease stages, indicating a limited reactive response, while microglia were frequently localized near Aβ and pTau deposits, consistent with a role in active immune surveillance.

Taken together, these results identify the Ent as an early and critically vulnerable site in the 3xTg-AD mouse model. In this model, the pathogenesis involves not only progressive neurodegeneration, as exemplified by the rapid atrophy of the Ent, but also early neurodevelopmental defects, as indicated by the hippocampal growth arrest. The Ent is characterized by early selective excitatory neuronal loss and a distinct glial response, wherein sustained astrocytic atrophy likely contributes to limited anti-inflammatory effects, thereby exacerbating local circuit dysfunction.

Early Neurogenesis Deficits and Distinct Glial Activation Characterize Hippocampal Pathology in 3xTg-AD Mice

The HPC, a key region for learning and memory, exhibits a different pathological trajectory from that of the Ent in 3xTg-AD mice (Fig. 1). We further investigated the specific vulnerability of the HPC in this AD model.

Neurogenesis in the dentate gyrus (DG) is essential for memory processes (Mu and Gage 2011; Hainmueller and Bartos 2020). Using doublecortin (DCX) as a marker of immature neurons in the DG, we found that neurogenesis was significantly impaired in 3xTg-AD mice as early as 2 months of age. DCX-positive cell density in the DG was 46.1% lower in 2-month-old 3xTg-AD mice than in WT controls (Figs. 3A-B and 223.0 vs. 120.3 cells/mm², P < 0.001). By 7 months of age, DCX-positive cells were nearly absent in 3xTg-AD mice, indicating the cessation of neurogenesis in the DG.

Analysis of hippocampal subregions revealed an age-dependent loss of NeuN-positive neurons in the CA1 and CA3 of 3xTg-AD mice, which became statistically significant by 18 months of age (Figure S4 A-B, 1.85 × 10⁵ vs. 1.55 × 10⁵ cells/mm³, P = 0.0134). At this late stage, representative images clearly illustrated the marked neuronal loss in these subfields compared to WT controls (Fig. 3C). Consequently, the overall neuronal density across the entire HPC was significantly reduced in 18-month-old 3xTg-AD mice (Fig. 3D).

Astrocytes in the HPC also underwent progressive morphological changes with age (Fig. 3E and S4C). By 18 months of age, astrocytes in both WT and 3xTg-AD mice showed significant morphological remodeling, characterized by enlarged cell bodies and thickened processes, compared with their 12-month-old counterparts (Fig. 3F and S4E, WT: 383.5 µm³ vs. 463.3 µm³, P = 0.0016; 3xTg-AD: 335.6 µm³ vs. 418.0 µm³, P = 0.0011). The density of astrocytes in the HPC remained stable across ages and genotypes (Fig. 3G and S4D). However, microglia in the HPC exhibited earlier activation than astrocytes. By 7 months, 3xTg-AD mice showed a significant increase in microglial reactivity, with a higher proportion of cells adopting a reactive morphology (Figs. 3I and 30.1% in 3xTg-AD vs. 56.5% in WT; P = 0.0038). This activation became widespread by 18 months (Fig. 3I). In 3xTg-AD mice with established pathology, both astrocytes and microglia were frequently found in close proximity to Aβ plaques and pTau tangles (Fig. 3K). Quantitative analysis confirmed this observation but revealed a difference from the Ent: from the initial emergence of detectable aggregates at 7 months, both glial cell types were closely associated with the pathology, and no significant difference was observed in the percentage of aggregates associated with either cell type at any age point examined (Fig. 3L). This suggests that, in contrast to the Ent, the hippocampal environment fosters the active engagement of both astrocytes and microglia with pathological aggregates from the early stages.

Taken together, 3xTg-AD mice exhibit severe deficits in hippocampal neurogenesis and early hippocampal microglial activation. These alterations provide insights into the unique vulnerability of the HPC in AD pathology and identify potential cellular targets for reliable interventions.

CCK Presents the Earliest and Region-Specific Downregulation in 3xTg-AD Mice

CCK, a neuropeptide involved in synaptic plasticity and memory formation (Li et al. 2014; Chen et al. 2013, 2019; Zhang et al. 2005; Feng et al. 2021) shows reduced levels in brain regions vulnerable to AD, suggesting its potential as an early biomarker of cognitive decline (Zhang et al. 2024; Plagman et al. 2019). To explore the impact of AD pathology on CCK expression in the brain, we quantified mRNA levels of CCK and its receptor CCK-BR across different ages and brain regions in 3xTg-AD and WT mice. For comparison, we simultaneously assessed the expression of other key genes associated with synaptic and neuronal function, including brain-derived neurotrophic factor (BDNF), glutamate decarboxylase 1 (GAD1), solute carrier family 1 member 2 (SLC1A2), and synapsin I (SYN1), which are known to decline with age and in AD (Mirza and Zahid 2018; Gao et al. 2022; Hill and Gammie 2022).

At 7 months of age, expression levels of BDNF, GAD1, SLC1A2, and SYN1 remained comparable between 3xTg-AD and WT mice across all regions examined (Fig. 4A-F). Notably, CCK was significantly downregulated specifically in the Ent of 3xTg-AD mice (Fig. 4A, 0.803 in 3xTg-AD vs. 1.000 in WT; P = 0.0421), whereas no such change was observed in the HPC or other cortical regions (Fig. 4B-F), suggesting the specific earliest vulnerability of CCK in the Ent. In the earlier stage of AD (2–5 months of age), expression levels of CCK and CCK-BR in the Ent of 3xTg-AD mice were comparable to those in age-matched WT mice (Figure S5).

By 12 months, as the disease progressed, significant downregulation of CCK and CCK-BR extended to multiple brain regions in 3xTg-AD mice, accompanied by marked reductions in BDNF, GAD1, SLC1A2, and SYN1 expression (Fig. 4A-F). By 18 months, these degenerative trends became more pronounced, with CCK exhibiting the most pronounced reductions across all brain regions analyzed (Fig. 4A-F).

Together, these results suggest that CCK expression is selectively and earliest impaired in the Ent during AD progression, supporting its potential role as a regional-specific biomarker for early AD detection.

CCK-4 Rescues Cognitive Function, Motor Learning, and Synaptic Function Across Disease Stages in 3xTg-AD Mice

To evaluate the therapeutic potential of CCK in AD, we first longitudinally characterized the progression of cognitive, motor, and synaptic deficits in 3xTg-AD mice from 2 to 18 months of age. The results of this characterization (presented in Fig. 5B, F, G and Figure S7) defined the specific ages for subsequent therapeutic intervention. Based on the established onset of significant deficits, separate groups of 3xTg-AD mice at these defined ages received acute administration of the CCK-BR agonist CCK-4 or vehicle (VEH) prior to behavioral training or electrophysiological recording.

We used the novel object recognition (NOR) test to assess cognitive function and the rotarod test to evaluate motor learning in mice (Fig. 5). During NOR training, all groups showed no inherent preference for object location (Fig. 5B, D), confirming unbiased exploration. In the test phase, however, 7- to 18-month-old 3xTg-AD mice exhibited significantly lower recognition indices (RI) than age-matched WT controls, indicating progressive impairment in recognition memory (Fig. 5B). Acute CCK-4 administration significantly improved NOR performance in 3xTg-AD mice across all ages tested, as shown by higher RI values than those in VEH-treated controls (Figs. 5D and 7 months: CCK-4 68.1% vs. VEH 60.9%, P = 0.0052; 12 months: 60.4% vs. 53.4%, P = 0.0065; 18 months: 58.1% vs. 50.9%, P = 0.0047). Loss of motivation or apathy often accompanies memory impairment in AD (Förstl and Kurz 1999). Consistent with this, 3xTg-AD mice spent less time exploring objects than WT mice from 7 months of age (Figure S6B, C), a motivational deficit not rescued by CCK-4 (Figure S6E, F). Sex-based analysis revealed no significant differences between males and females in NOR performance (Figure S6A-F).

In the rotarod test, 3xTg-AD mice performed similarly to WT controls at 2–7 months of age, but showed significantly shorter latencies to fall at 12 and 18 months, indicating impaired motor coordination and learning in later disease stages (Fig. 5F-G). Notably, acute CCK-4 treatment significantly improved motor learning performance in older 3xTg-AD mice, evident in steeper learning curves and longer fall latencies on Day 4 and Day 5 (Figs. 5I and 12 months: Day 4 CCK-4 256 s vs. VEH 208 s, P = 0.0157; Day 5 CCK-4 277 s vs. VEH 219 s, P = 0.0014. 18 months: Day 4 CCK-4 148 s vs. VEH 113 s, P = 0.0116; Day 5 CCK-4 159 s vs. VEH 123 s, P = 0.0172). Sex-grouped analysis showed no difference between males and females in the rotarod test (Figure S6G-H).

To assess synaptic function across different disease stages, we investigated LTP, a cellular correlate of learning and memory, in cortical and hippocampal slices using ex vivo electrophysiology. We focused on these regions as they are primary targets of entorhinal CCK projections, allowing us to evaluate the functional impact of entorhinal pathology on downstream memory circuits. After obtaining a stable baseline of field excitatory postsynaptic potential (fEPSP), LTP was induced by theta-burst stimulation (TBS). At 2 months of age, 3xTg-AD mice showed similar TBS-induced LTP in both the cortex and HPC compared to age-matched WT mice, indicating intact synaptic plasticity at this pre-symptomatic stage (Figure S7A, C). However, significant impairments in LTP magnitude emerged in 3xTg-AD mice from 5 months of age, became pronounced at 7 months, and progressively worsened at 12 and 18 months (Figure S7B, D), reflecting the progression of cognitive deficits. We next asked whether pharmacological activation of the CCK-BR could rescue these synaptic impairments. Remarkably, infusion application of CCK-4 during recording effectively restored synaptic plasticity in 3xTg-AD mice from 7 to 18 months of age (Fig. 6A-D). The rescue effect was evident shortly after TBS and was sustained throughout the one-hour recording period. Quantitative analysis of the normalized fEPSP amplitude averaged over the final 15 min confirmed the robust effect of CCK-4 (Fig. 6E, F). Relative to the pre-TBS baseline, cortical LTP in CCK-4-treated slices reached 134.9%, 130.6%, and 118.0% at 7, 12, and 18 months, respectively, values that were significantly higher than those in VEH-treated slices (Fig. 6E). Similarly, in the HPC, CCK-4 treatment potently enhanced LTP to 163.2%, 154.9%, and 136.1% of baseline at the same ages (Fig. 6F). These results demonstrate that acute CCK-4 treatment can significantly ameliorate deficits in synaptic plasticity across mild to severe stages of the disease in both important brain regions.

In summary, acute CCK-4 treatment significantly improved cognitive performance, motor learning, and synaptic plasticity in 3xTg-AD mice across mild to severe disease stages, demonstrating the remarkable therapeutic effects of CCK-BR agonist on AD.

HT-267, an Optimized CCK-4 Analogue, Is a Promising Drug Candidate for AD

Given the selective downregulation of CCK in the Ent during early AD, we next explored whether long-term CCK treatment could mitigate disease progression. We conducted a 90-day preventive treatment study in a group of 6-month-old 3xTg-AD mice (n = 20), a stage preceding overt cognitive and synaptic dysfunction (Fig. 5B, G and Figure S7E-F), randomly assigned to receive daily intraperitoneal injections of either HT-267 (n = 10) or VEH (n = 10) (Fig. 7A). HT-267 is an optimized CCK-BR agonist with a prolonged half-life, which we characterized in the present study. Throughout the treatment period, biweekly OFT revealed no significant differences in locomotor activity or motivational behavior between the HT-267 and VEH groups, indicating the absence of adverse behavioral effects from long-term HT-267 administration (Fig. 7B, C).

The NOR test was conducted for all mice before and after the 90-day treatment to assess cognitive changes. Although both groups showed comparable baseline performance (pre-treatment RI: HT-267 vs. VEH, P > 0.05), HT-267-treated mice exhibited a significantly higher RI than VEH-treated controls after treatment (61.04% vs. 53.32%, P = 0.0109) (Fig. 7D), demonstrating that HT-267 treatment attenuated age-related cognitive decline. Following the final behavioral assessment, animals from each group were allocated to terminal procedures: six mice per group for ex vivo electrophysiology and the remaining four for immunohistochemistry (IHC).

To evaluate synaptic plasticity, we conducted ex vivo electrophysiological recordings on hippocampal and cortical slices. Long-term HT-267 treatment resulted in significantly enhanced LTP compared to VEH controls (Fig. 7E-F; Cortex: HT-267 118.3% vs. VEH 111.3%, P = 0.0021; HPC: HT-267 132.3% vs. VEH 126.0%, P = 0.0171). We further assessed synaptic structural integrity by quantifying the expression of postsynaptic density protein 95 (PSD-95), a key scaffolding protein essential for maintaining excitatory synapses (Keith and El-Husseini 2008; Chen et al. 2011; Sturgill et al. 2009). In the HPC, HT-267 treatment increased dendritic PSD-95 immunolabeling by 25.7% compared to VEH controls (Fig. 7G-H; P = 0.0018). Consistently, the density of PSD-95-positive neurons was significantly higher in the cortex of HT-267-treated mice (Fig. 7I-J; HT-267 1042 cells/mm³ vs. VEH 895 cells/mm³, P = 0.0376), suggesting improved synaptic connectivity.

Taken together, long-term treatment of HT-267 improved memory function, enhanced synaptic plasticity, and preserved synaptic structure in 3xTg-AD mice. These results support the potential of CCK-BR agonists as disease-modifying therapies for AD, capable of delaying cognitive decline and strengthening neural circuitry integrity.

Discussion

This study employed the 3xTg-AD mouse model to investigate age- and region-specific neuropathological changes in AD. Unbiased stereological analyses revealed significant age-dependent reductions in cranial size and regional brain volumes in 3xTg-AD mice. Among these regions, the Ent exhibited early and pronounced progressive atrophy, indicating its role as a primary vulnerable site in this AD model. Early degeneration of the Ent is known to trigger a cascade of AD-related pathological events, disrupting cortical-hippocampal circuits and accelerating cognitive decline (Braak and Braak 1998; de Toledo-Morrell et al. 2000). At the cellular level, we observed that excitatory neurons in the Ent of 3xTg-AD mice were reduced earlier than inhibitory neurons. This imbalance possibly results from the gene expression profile of excitatory neurons, which renders them more susceptible to tau pathology, and probably contributes to widespread network dysfunction (Fu et al. 2017; Fu et al. 2019).

Glial responses in the Ent were distinct. Astrocytes exhibited early and persistent atrophy with weak spatial associations with Aβ plaques or tau tangles. This impaired reactivity probably exacerbates localized aggregation of neurotoxic proteins, reduces synaptic coverage in the Ent, and impairs neuronal communication (Chung et al. 2015). Microglia in the Ent were reactive and closely associated with pathological aggregates, suggesting that microglia are involved in synaptic remodeling through protective clearance mechanisms and potentially synaptic pruning (Ohyagi et al. 2005; Franco-Bocanegra et al. 2019).

In contrast to the progressive entorhinal atrophy, the HPC in 3xTg-AD mice exhibited a distinct phenotype of early developmental growth arrest and impaired neurogenesis, particularly within the vulnerable DG. This observation carries several important implications. First, it suggests that the hippocampal deficit in this model may originate from an early-life impairment in circuit assembly or maturation, rather than solely from a later degenerative process. Second, the severe impairment of adult neurogenesis, a key source of hippocampal plasticity, likely compromises the circuit’s capacity for adaptive rewiring from the outset, predisposing it to functional failure even before significant neuron loss (Aniol et al. 2022; Kawles et al. 2023). Although progressive hippocampal atrophy is a hallmark of advanced clinical AD, our findings imply that earlier, subtler developmental vulnerabilities may establish a reduced cognitive reserve and accelerate decline upon subsequent pathological insults. This suggests the potential value of strategies aimed at enhancing neural circuit robustness during pre-symptomatic stages, rather than merely targeting later neurodegeneration. Microglia in the HPC of 3xTg-AD mice displayed a reactive phenotype from early stages, frequently co-localizing with Aβ and tau aggregates, indicating their persistent engagement in neuroinflammatory processes. Astrocytic activation emerged later, with marked hypertrophy observed in both WT and 3xTg-AD mice by 18 months. In WT mice, this may be a compensatory response to the age-related increasing oxidative stress and metabolic burden (Clarke et al. 2018). In AD mice, this activation likely serves a protective function, enhancing clearance of neurotoxic proteins and maintaining homeostasis (Fuller et al. 2009; Cabral-Miranda et al. 2025).

Collectively, our results revealed two distinct pathological trajectories in the 3xTg-AD mouse brain: a severe, early-onset progressive degeneration in the Ent characterized by the vulnerability of both neurons and astrocytes, and a pathological phenotype in the HPC marked by early neurogenesis deficits, arrested growth, and distinct glial responses. These patterns underscore the need for region-targeted strategies in AD.

These region-specific cascades reflect both conserved and model-divergent features of AD pathophysiology. The early, severe entorhinal atrophy and loss of layer II excitatory neurons directly parallel the established vulnerability of this region in human AD, as shown by neuroimaging and post-mortem studies (Jack et al. 1998; Gómez-Isla et al. 1996). In contrast, the hippocampal phenotype diverges: it shows a developmental growth impairment rather than the progressive atrophy typical of late-stage human AD. This aligns with observations in other amyloidogenic models (e.g., APP/PS1 mice, Delatour et al. 2006), suggesting a shared model-specific developmental component. Furthermore, we observed divergent glial responses within this AD model: decreased astrocytic reactivity in the atrophying Ent alongside robust microglial activation in the HPC. Although widespread gliosis is a hallmark of both human AD and many mouse models, the regional difference in glial engagement within the same brain we observed is less commonly reported. This regional disparity indicates that neuroinflammatory responses are highly context-dependent, shaped by local pathology and disease stage. Consequently, glia-targeting therapeutic strategies may need to be tailored not only to cell type but also to the specific pathological environment of each affected brain region.

The early and specific downregulation of entorhinal CCK is a key finding of our study, aligning with growing evidence for impaired CCK signaling in AD pathogenesis. Human post-mortem studies show reduced cortical CCK immunoreactivity in AD (Mazurek and Beal 1991; Perry et al. 1981), while higher cerebrospinal fluid (CSF) CCK levels correlate with reduced AD risk, better cognition, and greater grey matter volume (Plagman et al. 2019). Transcriptomic analyses further identify downregulated CCK receptors in AD brains (Hokama et al. 2014; Lin et al. 2014). Complementing this, AD animal models like APP/PS1 mice exhibit reduced hippocampal CCK mRNA (Liu et al. 2021).

We found that CCK downregulation in the Ent of 3xTg-AD mice at 7 months precedes widespread cortical atrophy and correlates with the onset of hippocampal synaptic dysfunction. This early loss, occurring prior to overt neuronal death and potentially stemming from activity-dependent mechanisms (Li et al. 2014; Chen et al. 2013), was followed by progressive CCK/CCK-BR reduction across brain regions by 12 months. Based on the Ent’s early vulnerability and the specific downregulation of CCK, a key neuropeptide for synaptic modulation and memory, we hypothesized that this deficit would disrupt entorhinal control over its downstream hippocampal and cortical targets, thereby contributing to cognitive decline (Lee and Soltesz 2011; Rehfeld 2021; Li et al. 2014; Chen et al. 2013). This rationale guided our focus on assessing and rescuing synaptic function in the HPC and neocortex, the primary recipients of entorhinal projections. We therefore hypothesized that deficient entorhinal CCK signaling disrupts its control over downstream hippocampal and cortical targets. Critically, administering CCK-BR agonists restored synaptic plasticity and cognition in these downstream regions, demonstrating that compensating for entorhinal CCK deficiency can effectively reverse network-level dysfunction and providing a mechanistic link between CCK loss and early circuit impairment in AD.

To evaluate the therapeutic potential of CCK for AD, we treated 3xTg-AD mice at different ages with CCK-4, a natural CCK-BR agonist. We found that CCK-4 significantly improved recognition memory, motor learning, and cortical and hippocampal LTP across mild to severe AD stages. These therapeutic effects verify that CCK-4 enhances synaptic plasticity and neural network connectivity, addressing key pathological features of AD. However, the poor pharmacokinetic properties of CCK-4, including its short plasma half-life (less than 1 min in rodents) and low bioavailability, hinder its clinical translation and application (Koulischer et al. 1982). To overcome these limitations, we developed HT-267, a novel CCK-4 analogue with optimized pharmacological properties. HT-267 exhibits enhanced CCK-BR binding affinity, increased potency, a substantially extended half-life, and favourable rescue effects on cognition and neuroplasticity of aged 3xTg-AD mice (Zhang et al. 2024). The 90-day treatment with HT-267 in 3xTg-AD mice significantly improved their cognitive function and synaptic plasticity. Meanwhile, HT-267 enhanced synaptic scaffolding, as indicated by increased PSD-95 expression in hippocampal dendrites and cortical neurons (Chen et al. 2011; Sturgill et al. 2009). Together, these findings support that pharmacological activation of CCK-BR can mitigate synaptic deficits and cognitive decline in AD, and HT-267 represents a promising drug candidate for maintaining neural circuit function and delaying AD progression. Future studies should focus on elucidating the exact molecular mechanisms underlying CCK-BR-mediated neuroprotection, evaluating the long-term safety and efficacy of CCK-BR agonist applications, and exploring combination strategies to maximize the therapeutic potential of CCK-based therapies in AD.

Some limitations in this study warrant consideration. First, regarding anatomical quantification, the regional volume changes were quantified using ex vivo stereology on histological sections. Although this approach was important for achieving direct spatial correlations between atrophy, cellular pathology, and molecular alterations within the same tissue, it lacks the precision and longitudinal capacity of in vivo imaging techniques such as magnetic resonance imaging (MRI). Future MRI studies could refine these anatomical findings. Second, regarding therapeutic mechanisms, the systemic (intraperitoneal) delivery of CCK-4 and HT-267, though effective, acts globally. This precludes attributing the rescue effects definitively to the Ent versus other brain regions. Regarding causal neural circuits, although our data strongly link entorhinal CCK downregulation to downstream synaptic deficits, the definitive role of entorhinal CCK neurons remains to be established. Cell-type-specific neuromodulatory approaches (e.g., chemogenetics) to selectively manipulate entorhinal CCK neuronal activity could determine if this population is sufficient to elicit or rescue the observed circuit-level dysfunctions.

Conclusions

This study strengthens the evidence that the Ent is a critical early site of degeneration in AD, and that downregulation of CCK constitutes an early molecular signature of AD pathogenesis. We established the rescue effects of CCKBR agonists on cognitive, synaptic, and motor impairments across mild to severe AD progression stages. Additionally, we confirmed that long-term treatment with the optimized analogue HT-267, initiated at a young age in 3xTg-AD mice, decelerates the progression of cognitive and synaptic deficits compared to VEH-treated controls. Taken together, these findings position CCK as a potential biomarker for early AD detection, and the CCK-BR as a promising therapeutic target for modifying disease course.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

AD

Alzheimer’s disease

3xTg-AD

Triple-transgenic Alzheimer’s disease mouse model

Aβ

Amyloid-beta

AC

Auditory cortex

ACSF

Artificial cerebrospinal fluid

AP

Anteroposterior

AC

Auditory cortex

BBB

Blood-brain barrier

BDNF

Brain-derived neurotrophic factor

CAS

Chinese Academy of Sciences

Cb

Cerebellum

CCK

Cholecystokinin

CCK-BR

Cholecystokinin B receptor

CCK-4

Cholecystokinin tetrapeptide

CNS

Central nervous system

DCX

Doublecortin

DV

Dorsoventral

DG

Dentate gyrus

EC50

Half maximal effective concentration

Ent

Entorhinal cortex

fEPSP

Field excitatory postsynaptic potential

GAD1

Glutamic acid decarboxylase 1

GAPDH

Glyceraldehyde 3-phosphate dehydrogenase

GFAP

Glial fibrillary acidic protein

HFS

High-frequency stimulation

HPC

Hippocampus

IHC

Immunohistochemistry

Iba1

Ionized calcium-binding adapter molecule 1

i.p.

Intraperitoneal injection

LTP

Long-term potentiation

LP

Location preference

MC

Motor cortex

MCI

Mild cognitive impairment

ML

Mediolateral

MTL

Medial temporal lobe

NFTs

Neurofibrillary tangles

NMDA

N-Methyl-D-aspartic acid

NOR

Novel object recognition

OFT

Open field test

PBS

Phosphate-buffer saline

PFA

Paraformaldehyde solution

PFC

Prefrontal cortex

Pir

Piriform cortex

POR

Postrhinal cortex

PrC

Perirhinal cortex

PSD-95

Postsynaptic density protein-95

pTau

Phosphorylated tau

qRT-PCR

Quantitative real-time PCR

RI

Recognition index

SLC1A2

Solute Carrier Family 1 Member 2

SYN1

Synapsin I

TBS

Theta-burst stimulation

TBV

Total blood volume

VEH

Vehicle control

VC

Visual cortex

WT

Wild-type mouse

Author Contributions

N.Z., J.F.H., and Y.X.S. designed the experiments. Y.X.S. collected and analyzed the data. Y.X.S. and N.Z. interpreted the data and wrote the manuscript. All authors reviewed the manuscript. J.F.H. conducted the funding acquisition for this study.

Funding

This work was funded by the following: Hong Kong Research Grants Council, General Research Fund: CityUHK 11103922, CityUHK 11104923, CityUHK 11104524. Hong Kong Research Grants Council, Collaborative Research Fund: C1043-21G, C1002-24 W. Hong Kong Research Grants Council, Senior Research Fellow Scheme: SRFS2324-1S02. Hong Kong Health Bureau, Health and Medical Research Fund: 09203656. Innovation Technology Commission of the Hong Kong SAR, China: Health@InnoHK program. We also thank the following charitable foundations for their generous support to Prof. Jufang HE: Wong Chun Hong Endowed Chair Professorship, Charlie Lee Charitable Foundation, and Fong Shu Fook Tong Foundation.

Data Availability

All data that support the findings of this study are available within the paper and its supplementary materials. Additional source data is available from the corresponding author upon reasonable request.

Declarations

Competing Interests

The authors declare no competing interests.

Ethical Approval

All experimental procedures were conducted in accordance with the Animal Subjects Ethics Sub-Committees of the City University of Hong Kong (A-0661 and AN-STA-00000653).