Induction of neurodegeneration in the hippocampus of senescence-accelerated mouse-prone 8 (SAMP8) mice by blood-brain barrier-crossing serum amyloid P component

Abstract

Aging is a major risk factor for neurodegeneration. However, the mechanisms by which age-related disruptions in brain homeostasis induce neurodegeneration remain controversial. To clarify this, we analyzed how age-related changes in blood components induce blood-brain barrier (BBB) structure disruption and neurodegeneration using 6-month-old senescence-accelerated mouse-prone 8 (SAMP8) mice. First, we found that 6-month-old SAMP8 mice showed a decline in learning and memory abilities. We used the senescence-associated β-galactosidase marker to visualize which brain regions were responsible for the behavioral change, showing that the hippocampus had a drastic accumulation of senescence-associated β-galactosidase in SAMP8 mice. Next, we performed multi-omics profiling, which demonstrated that SAMP8 mice showed remarkable changes in the expression of BBB maintenance and immune system-related genes at the early stage of aging. Consistent with these results, the structure of occludin-positive tight junctions in the BBB and the number of microglia were altered in the hippocampus of SAMP8 mice. In addition, SAMP8 mice showed an increase in apoptotic hippocampal neurons and a decrease in synaptic density in mossy fibers, leading to impaired learning and memory. Interestingly, proteomic and immunostaining analyses revealed that one of the blood components, the serum amyloid P component (SAP), translocates to the hippocampus while passing through the BBB in SAMP8 mice. In addition, SAP increased BBB permeability by altering the structure of occludin-positive tight junctions, as shown via in vitro analyses. Furthermore, SAP can lead to neuronal cell death and a decline in synaptic density. Overall, our results reveal a previously unrecognized mechanism by which aging induces neuronal death and impairs learning abilities. These results represent an important conceptual advance in that the increase in the serum component SAP disrupts BBB homeostasis, and consequently, SAP leakage into the brain parenchyma leads to neurodegeneration during aging in SAMP8 mice.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12974-026-03704-7.

Introduction

Aging is a primary risk factor for neurodegeneration. During aging, DNA damage, lysosomal dysfunction, epigenetic changes, immune dysfunction, and impaired protein homeostasis accelerate the development of neurodegenerative diseases in the brain [1]. The blood-brain barrier (BBB) separates the brain from the blood with a highly selective semipermeable border composed of endothelial cells. Recently, it has become clear that various factors transfer between the brain parenchyma and the circulating blood. In support of this concept, previous studies have indicated that blood transfer from old to young mice leads to a decline in learning ability [2] and that parabiotic cross-circulation between young and old mice enhances cognitive functions [3]. These studies imply that age-related neurodegeneration is regulated by alterations in circulating blood factors in the vasculature of the neurogenic niche, as well as in the brain-specific environment. However, the mechanisms by which these circulating factors contribute to neurodegeneration remain unclear.

Senescence-accelerated mouse-prone 8 (SAMP8) is a mouse strain developed to exhibit aging-induced phenotypes that has been widely used to study Alzheimer’s disease. SAMP8 mice have half the lifespan of senescence-accelerated mouse resistant 1 (SAMR1) mice, which are used as control strains for SAMP8 mice [4]. In addition, a decline in learning and memory abilities, tau phosphorylation, and neurodegeneration appear in SAMP8 mice before turning 6 months old [4–6]. After being 10 months of age, SAMP8 mice show remarkably abundant cytokine production derived from microglia and amyloid β granule accumulation in neurons [7, 8]. Notably, at 12 months of age, plasma IgG is reported to extravasate to the brain parenchyma, crossing the BBB in the hippocampus, which is not observed at 7 months of age [9]. These studies suggest that increased BBB permeability contributes to accelerated senescence.

Highly selective BBB permeability contributes to brain function homeostasis, whereas disruption of BBB integrity causes neurodegeneration. The vulnerability of the tight junctions between endothelial cells induces the transport of blood-derived toxins and inflammatory molecules into the brain. Even during healthy aging, the BBB integrity is weakened in an age-dependent manner. Aged mice show increased extravasation of IgG into the brain and decreased expression of the tight junction protein occludin-1 and the adhesion protein vascular cell adhesion molecule-1 (VCAM-1) [10]. Additionally, the BBB integrity of patients with Alzheimer’s disease is vulnerable, and blood-derived protein accumulates in their brains [11, 12]. Therefore, the degree of neurodegeneration depends on the amount of blood-derived toxic factors that are regulated by BBB permeability. SAMP8 mice display disease-like phenotype in peripheral organs, particularly in the liver, which synthesizes abundant blood proteins; the liver of SAMP8 mice displays fibrosis and excessive cytokine production [13]. These studies suggest that the blood of SAMP8 mice may contain abundant toxic factors that leak into the brain and accelerate neurodegeneration.

To clarify the mechanisms by which blood-derived toxic factors induce neurodegeneration during aging, this study aimed to compare the proteomic profiles of the serum and hippocampus from SAMP8 mice at an early stage of aging to identify the neurotoxic blood components.

Methods

Mice

Six-month-old male SAMR1 and SAMP8 mice and 8–12-week-old ICR wild-type mice were used. This study was approved by the Institutional Animal Care and Use Committee of Ochanomizu University (protocol No. 24003) and the Experimental Animal Committee of the New Drug Research Center, Inc. (protocol No. 211125 A) and was performed in accordance with the policy for welfare, care, relief of pain, and management of laboratory animals. Mice were housed under controlled conditions at 22 ± 3 °C, 45 ± 5% humidity, and 12-h light/12-h dark cycle.

Passive avoidance test

The learning and memory abilities of 6-month-old SAMR1 and SAMP8 mice were examined using a passive avoidance test in bright- and dark-compartment step-through cages (Muromachi Kikai Co., Tokyo, Japan). The training sessions were conducted on the first day of the examination. The mice were placed in the bright compartment and allowed to explore for 10 s. When the mice entered the dark compartment, the guillotine door was closed and an electrical foot shock (0.2 mA) was delivered for 3 s. The test sessions were performed the following day. Mice were placed in the bright compartment and allowed to explore for 10 s, after which the guillotine door was raised. The latency to enter the dark compartment was recorded up to 300 s.

Senescence-associated β-galactosidase staining and image analysis

SA-β-Gal staining was performed on 40-µm free-floating sections of 4% paraformaldehyde-fixed mouse brains using a SA-β-Gal Staining Kit (9860, Cell Signaling Technology, Danvers, MA). Solutions for β-Galactosidase staining were freshly prepared according to the manufacturer’s instructions and applied to the sections, which were incubated at 37 °C for 72 h. Images were captured using a BZ-X710 microscope (Keyence Corporation, Osaka, Japan) with a ×4 objective. The percentage of SA-β-Gal⁺ area and the color intensity in each section were analyzed using NIH ImageJ software [14]. SA-β-Gal⁺ area was normalized with the whole hippocampal area. Mean grey values of SA-β-Gal staining were quantified from six fields (28.6 × 28.6 µm²) in the CA1, CA2, CA3, and dentate gyrus (DG) of each section, and the mean grey value of background was subtracted.

RNA-seq analysis

RNA extraction

Total RNA was extracted from the hippocampal samples (6-month-old SAMR1: n = 4, SAMP8: n = 4) using an RNeasy Mini Kit (74104, Qiagen, Hilden, Germany) according to the manufacturer’s protocol. RNA was eluted in 30 µL RNase-free water, and purity was quantified using a Nanodrop One (Thermo Fisher Scientific, Waltham, MA). Extracted RNA showed an average purity of 260/230 = 2.10 ± 0.005, 260/280 = 1.65 ± 0.13. RNA quality was verified using an Agilent 2100 BioAnalyzer (Agilent Technologies, Santa Clara, CA). All samples used had an RNA integrity value of 8.0 or greater (8.35 ± 0.08). The extracted RNA was stored at -80 °C until use.

Library preparation and sequencing

Library preparation and sequencing were performed by Genome–Lead Co., Ltd. (Kagawa, Japan). mRNA was purified by poly(A) selection using a KAPA mRNA Capture Kit (KK8440, KAPA, Wilmington, MA). The transcriptome libraries were built using the MGIEasy RNA Directional Library Prep Set (1000006386; MGI Tech, Shenzhen, China). Briefly, the RNA was fragmented into 250 bp pieces, reverse transcribed, and the second strand was synthesized. DNA end repair and adapter ligation were conducted using the MGIEasy DNA Adapters-96 Kit (MGI Tech) following the manufacturer’s instructions, and the product was amplified using PCR for 14 cycles. The prepared library was circulated, and DNA nanoballs were prepared by multiple-displacement amplification. Finally, the sequence of the prepared DNB was analyzed on a DNBSEQ-T7RS (MGI Tech) using a DNBSEQ-T7RS high-throughput sequencing set (FCL PE150, 1000028454, MGI Tech), which yielded 150 bp paired-end reads. Libraries from each sample were individually barcoded and loaded onto the flow cell simultaneously. A total of 509 million reads (average 63.7 million reads per sample) were obtained.

Processing and analysis of RNA-seq data

Potential sequencing adapters and low-quality bases in the raw reads were trimmed using fastp v0.23.2, with default parameters [15]. Cleaned high-quality reads were mapped to the UCSC mm39 reference genome obtained from the UCSC genome browser using STAR software v2.7.9a with options “--quantMode TranscriptomeSAM” and “--outSAMtype BAM SortedByCoordinate” [16, 17]. Using the same reference with ncbiRefSeq gene annotations in gtf format, expression levels were then estimated by RSEM v1.3.1 with default parameters [18]. Differential expression analysis between 6-month-old SAMR1 and SAMP8 hippocampi was performed using R package DESeq2 v1.30.1 with default parameters [19]. To control false discovery rate, q-values were calculated using R package v 4.1.0 with default parameters, and genes with q-value < 0.05 were considered differentially expressed [20–24]. Gene Ontology (GO) enrichment analyses of the DEGs were performed using Metascape [25]. Principal component (PC) analysis was performed using the ClustVis tool with the detected transcripts [26].

Proteomic analysis

Sample preparation for proteome analysis

Sample preparation and proteomic analysis were performed as previously described [27, 28]. Briefly, eight hippocampal samples were dissolved in 100 mM Tris-HCl (pH 8.0) containing 4% sodium dodecyl sulfate (SDS) using a BIORUPTOR BR-II (SONIC BIO Co., Kanagawa, Japan). The extracted proteins were quantified using Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, Waltham, MA) at 100 ng/µL. The protein extracts were reduced with 20 mM tris (2-carboxyethyl) phosphine for 10 min at 80 °C, followed by alkylation with 35 mM iodoacetamide for 30 min while being protected from light. Proteins were purified and digested using the SP3 method [28]. The tryptic digestion was performed using 500 ng/µL Trypsin/Lys-C Mix (Promega, Madison, WI) overnight at 37 °C. The digests were purified using GL-Tip SDB (GL Sciences, Tokyo, Japan), according to the manufacturer’s protocol. The peptides were re-dissolved in 2% acetonitrile (ACN) containing 0.1% trifluoroacetic acid (TFA) and quantified using the BCA assay at 200 ng/µL [29]. To remove high-abundance proteins, eight serum samples were treated using the PureProteome Albumin/IgG Depletion Kit (Merck, Darmstadt, Germany), following the manufacturer’s instructions. The filtrates obtained were dissolved in 100 mM Tris-HCl (pH 8.0) containing 4% SDS using a BIORUPTOR BR-II. The reduction and alkylation of proteins and the SP3 method were performed using the aforementioned procedures. Peptides were dissolved in 2% ACN containing 0.1% TFA.

LC-MS/MS

The digested peptides (500 ng) were loaded directly using a 75 μm × 25 cm nanoLC nano-capillary C18 column (Ionopticks., VIC, Australia) at 60 °C and then separated with a 100-min gradient (mobile phase A = 0.1% FA in water, B = 0.1% FA in 80% ACN) consisting of 0 min 7% B, 86 min 37% B, 94 min 65% B, and 100 min 65% B at a flow rate of 200 nL/min using an UltiMate 3000 RSLCnano LC system (Thermo Fisher Scientific). The eluted peptides were detected using a quadrupole Orbitrap Exploris 480 hybrid mass spectrometer (Thermo Fisher Scientific) with a normal DIA window. The MS1 scan range was set as full scan with m/z 495–745 at mass resolution as 15,000 to set an auto gain control (AGC) target of 3 × 10⁶ and maximum injection time of “Auto.” The MS2 were collected at m/z 200-1,800 at 45,000 resolutions to set an AGC target of 3 × 10⁶, maximum injection time of “Auto,” and stepped normalized collision energy of 22%, 26% and 30%. The isolation width for MS2 was set to 4 Th, and for the 500–740 m/z window pattern, an optimized window arrangement was used in Scaffold DIA (Proteome Software, Inc., Portland, OR).

Data processing

Raw data were searched against in silico predicted spectral libraries using DIA-NN (version1.8.1, https://github.com/vdemichev/DiaNN) [29]. First, an in silico predicted spectral library was generated from the mouse protein sequence database (UniProt id UP000000589, reviewed, canonical, 21986 entries, April 1, 2021, downloaded) using DIA-NN. To generate the spectral library, the parameters were set as follows: digestion enzyme, trypsin; missed cleavage, 1; peptide length range, 7–45; precursor charge range, 2–4; precursor m/z range, 495–745; fragment ion m/z range, 200–1800. Additionally, “FASTA digest for library-free search/library generation,” “Deep learning-based spectra, RTs, and IM prediction,” “n-term M excision,” and “C carbamidomethylation” were enabled. The DIA-NN search parameters were as follows: protease, trypsin; missed cleavages, 1; peptide length range, 7–45; precursor charge range, 2–4; precursor mass range, 490–750; fragment ion m/z range, 200–1800; mass accuracy, 10 ppm; static modification, cysteine carbamidomethylation; enabled “Heuristic protein interferences,” “Use isotopologues,” “MBR,” and “No shared spectra.” Additional commands were set as follows: “mass acc cal 10,” “peak translation,” and “matrix spec q.” The protein identification threshold was set at < 1% for both peptide and protein false discovery rates. Statistical calculations and Pearson’s correlation coefficient heatmap analysis with hierarchical clustering were performed using Perseus v1.6.15.0 [30]. GO enrichment analyses of the differentially expressed peptides (DEPs) was performed using Metascape [25]. PC analysis was performed using the ClustVis tool with the detected peptides [26].

Immunohistochemical staining and image analysis

For immunoenzymatic staining, the brain sections were incubated with the primary antibodies overnight, then with secondary biotin-labeled antibodies for 1 h, and incubated with the avidin-biotin complex (Vectastain Elite ABC Standard Kit, Vector Laboratories, Burlingame, CA) for 1 h, followed by 3,3-diaminobenzidine solution to detect signals. The primary antibodies used were anti-PSD95 (1:1000 dilution; MA1-045, RRID: AB_325399, Thermo Fisher Scientific), anti-Iba1 (1:1000 dilution; 019–19741, RRID: AB_839504, FUJIFILM Wako Pure Chemical Corporation, Tokyo, Japan), anti-P2RY12 (1:500 dilution; 011-28873, FUJIFILM Wako Pure Chemical Corporation), anti-S100β (1:3000 dilution; 15146-1-AP, RRID: AB_2254244, Proteintech), and anti-GFAP (1:1000 dilution; G9269, RRID: AB_477035, Sigma-Aldrich). Secondary antibodies included anti-mouse IgG (1:300 dilution; BA-9200, RRID: AB_2336171, Vector Laboratories), anti-rabbit IgG (1:300 dilution; BA-1000, RRID: AB_2313606, Vector Laboratories) and anti-guinea pig IgG (1:300 dilution; 706-065-148, RRID: AB_2340451, Jackson ImmunoResearch, West Grove, PA). To analyze PSD95 staining intensity, images of mossy fibers in the hippocampus were acquired using a BZ-X710 microscope (Keyence Corporation) with a ×40 objective. The mean gray values of PSD95 staining in mossy fibers were acquired using NIH ImageJ and normalized to the background intensity. The mean gray values were quantified using five randomly selected square field (31.8 × 31.8 µm²) from each mouse. To count the microglia and astrocytes, images were captured using a BZ-X710 microscope. The numbers of Iba1⁺ and P2RY12⁺ microglia were quantified in six fields (364 × 272 µm²) of the CA1, CA2, CA3, and DG in each mouse brain. GFAP⁺ and S100β⁺ astrocytes were quantified in four fields (500 × 500 µm²) of the CA1, CA2, CA3, and DG in each mouse brain. Sholl analysis of Iba1⁺ and P2RY12⁺ microglia was performed by Neuroanatomy Plugin Macro of NIH ImageJ using 10 microglia for each hippocampus [31].

For immunofluorescence staining, sections were incubated with primary antibodies overnight, followed by incubation with secondary antibodies for 1 h. The primary antibodies included anti-NeuN (1:400 dilution; MAB377, RRID: AB_2298772, Merck, Darmstadt, Germany), anti-cleaved-caspase 3 (1:500 dilution; 9664, RRID: AB_2070042, Cell Signaling Technology), anti-serum amyloid P component (SAP) (1:200 dilution; 50113-T24, Sino Biological, Beijing, China), anti-CD31 (1:200 dilution; AF3628, RRID: AB_2846942, R&D Systems, Minneapolis, MN), anti-GFAP (1:1000 dilution; ab4674, RRID: AB_304558, Abcam, Cambridge, England), anti-Iba1 (1:200 dilution; NB100-1028, RRID: AB_3148646, Novus Biologicals, Centennial, CO), anti-Tuj1 (1:1000 dilution; MMS-435P, RRID: AB_2313773, Covance), and anti-occludin (1:200 dilution; 33-1500, RRID: AB_87033, Thermo Fisher Scientific). The secondary antibodies were donkey anti-rabbit IgG Alexa Fluor 568 (1:300 dilution; A10042, RRID: AB_2534017, Thermo Fisher Scientific), donkey anti-mouse IgG Alexa Fluor 488 (1:300 dilution; A21202, RRID: AB_141607, Thermo Fisher Scientific), donkey anti-rabbit IgG Alexa Fluor 488 (1:300 dilution; A21206, RRID: AB_2768318, Thermo Fisher Scientific), donkey anti-goat IgG Alexa Fluor 568 (1:300 dilution; A11057, RRID: AB_2534104, Thermo Fisher Scientific), goat anti-chicken IgY DyLight 649 (1:300 dilution; NBP1-72815, RRID: AB_11042652, Novus Biologicals), and goat anti-mouse IgG Alexa Fluor 680 (1:300 dilution; A32729, RRID: AB_2633278, Thermo Fisher Scientific). DAPI was used for nuclear staining. Fluorescent images of the hippocampus were captured using an LSM710 or LSM980 confocal microscope (Carl Zeiss). The number of cleaved-caspase 3-stained cells were quantified in two fields (200 × 200 µm²) of the CA1, CA2, CA3, and DG in each mouse brain. The integrated intensity values of SAP in CA3 astrocytes were calculated using the formula CTCF = integrated density (area of selected cell × mean fluorescence of background) using at least nine images from each mouse, which included 1–3 astrocytes per image. The lengths of the tight junctions were quantified using the Measure Skeleton Length plugin in NIH ImageJ [32]. For each mouse, 10 regions containing at least one vessel were randomly chosen for quantification. The colocalized area of SAP with CD31⁺, GFAP⁺, or Tuj1⁺ area in hippocampi were quantified using the coloc2 Plugin Macro in NIH ImageJ [33]. For each mouse, at least five regions were randomly chosen for quantification.

ELISA

We quantified the concentration of the SAP in the sera and hippocampal lysates of SAMP8 mice using an ELISA kit (ab235639, Abcam). Samples were diluted 1:2,000, and 0.5 µL was applied to ELISA. The measured values were then used to calculate the original SAP concentrations. Total protein was quantified with the BCA Protein Assay Kit (A65453, Thermo Fisher Scientific).

In vitro 3D BBB model

In vitro 3D BBB model was prepared by co-culturing primary mouse astrocytes, pericytes, and brain endothelial cells.

Primary astrocyte culture

Primary astrocyte cultures were prepared from the cerebral cortices of postnatal day (P) 3–6 ICR mice using previously established protocols [34]. Briefly, cerebral cortices were dissected and cultured in Dulbecco’s modified Eagle medium (DMEM) with 20% fetal bovine serum (FBS) and penicillin-streptomycin on T75 flasks, which were pre-coated with 1 mg/mL poly-L-lysin (PLL) (P2636, Sigma) in borate buffer solution, at 37 °C, 5% CO₂. On day in vitro (DIV) 3, the primary astrocytes were replaced with fresh media. On DIV14, astrocytes isolated with trypsin were seeded in 24-well plates, which were pre-coated with PLL in borate buffer solution at 500,000 cells/well, and cultured in DMEM with 20% FBS and penicillin-streptomycin.

Primary pericyte culture

Primary pericytes were prepared from the cerebral cortex of 8–12-week-old ICR mice using previously established protocols [35, 36]. Briefly, five cortices were dissected in ice-cold phosphate buffered saline (PBS), and the meninges and choroid plexuses were completely removed. Then, the cortices were minced in a 15 mL collagenase II solution (1 mg/mL collagenase II, 15 µg/mL DNase I in DMEM high glucose), shaken for 1.5 h at 37 °C, supplemented with 10 mL DMEM high glucose, and centrifuged for 10 min at 1,000 g To remove myelin, 10 mL of 20% bovine serum albumin in DMEM high glucose was added to the pellet, gently pipetted 25 times, and then centrifuged for 25 min at 1,000 g, followed by removal of the milky upper layer. The pellet was suspended with a collagenase/dispase solution (1 mg/mL collagenase/dispase, 6.7 µg/mL DNase I in 13 mL DMEM high glucose) and then incubated for 30 min at 180 rpm, 37 °C. After incubation, 10 mL of DMEM high glucose was added, and the mixture was centrifuged for 10 min at 1,000 g. The pellet was gently suspended with 2 mL of a brain vascular endothelial cell culture medium (20% FBS, 100 µg/mL Heparin, 1.5 ng/mL bFGF, 500 nM hydrocortisone, insulin-transferrin-selenium, and penicillin-streptomycin in DMEM/F12), filtered with 40-µm cell strainer, and seeded on a 30-mm dish and cultured at 37 °C, 5% CO₂. On DIV4, trypsinized pericytes were transferred to a 60-mm dish and cultured in brain vascular endothelial cell culture medium. Cultured pericytes were replaced with fresh medium every 2–3 days. On DIV11, trypsinized 5,000 cells pericytes were seeded on inverted membrane of a cell insert with 0.4-µm pore size and 33.6 mm² filtration area and then incubated in the brain vascular endothelial cell culture medium for 2 h, at 37 °C, 5% CO₂. After the pericytes were attached to the inverted membrane, the cell inserts were placed in a 24-well plate in which astrocytes were cultured in 500 µL of brain vascular endothelial cell culture medium, and 500 µL of brain vascular endothelial cell culture medium was added to the inside of the cell inserts.

Primary endothelial cell culture

Primary brain endothelial cells were prepared from the cerebral cortices of 8–12-week-old ICR mice. The same procedures were performed until the myelin removal for primary pericytes, and the pellets in the collagenase/dispase solution were incubated for 1 h at 180 rpm and 37 °C. After incubation, 10 mL of DMEM high glucose was added, and the mixture was centrifuged for 10 min at 1,000 g. The pellet was gently suspended in 2 mL of brain vascular endothelial cell culture medium containing 10 µg/mL puromycin to remove pericytes, then filtered through a 40-µm cell strainer and seeded onto a 30-mm dish, which was pre-coated with collagen IV and fibronectin. On DIV1, 500 µL of brain vascular endothelial cell culture medium containing puromycin was added to the dish. On DIV2, the endothelial cells were transferred to brain vascular endothelial cell culture medium without puromycin. On DIV7, 1.3 × 10⁵ trypsinized endothelial cells were seeded in the luminal compartment of the cell insert, which was pre-coated with collagen IV and fibronectin and seeded with pericytes. After the insert was positioned in a 24-well plate, in which primary astrocytes were cultured, 500 µL of brain vascular endothelial cell culture medium was added to the luminal compartment of the cell insert.

In vitro 3D BBB model

The prepared co-cultures were cultured for 2 days after endothelial cell seeding and then replaced with fresh medium. After 2 days, co-cultures were changed to an enhanced medium (2 mM L-glutamine, 550 nM hydrocortisone, 312.4 µM cAMP, 17.5 µM phosphodiesterase inhibitor, 1 µM retinoic acid, 10% FBS, insulin-transferrin-selenium, and penicillin-streptomycin in DMEM/F12) and incubated for 24 h. Before the experiments, the co-cultures were cultured in a serum-free medium for 1 h.

For trans-endothelial electrical resistance (TEER) analysis, in vitro 3D BBB models were treated with 30, 60, or 120 nM recombinant mouse SAP (50113-M08H; Sino Biological). At 1, 4, 12, and 24 h after SAP treatment, the resistance values inside and outside the cell inserts were quantified using a NEPA21 type II (NEPA GENE, Chiba, Japan) with a parallel needle electrode (CUY560-5-0.5, NEPA GENE). TEER was calculated using the following formula: TEER = (resistance value in in vitro 3D BBB model – resistance value in cell inserts without cells) × cell insert area. For each insert, the resistance values were measured at least three times and the averages were used.

For immunostaining, coverslips were incubated with primary antibodies for anti-CD31 (1:200 dilution; AF3628, RRID: AB_2846942, R&D Systems), anti-PDGFRβ (1:100 dilution; 3169s, RRID: AB_2162497, Cell Signaling), and anti-GFAP (1:1000 dilution; G9269, RRID: AB_477035, Sigma). Secondary antibodies for donkey included anti-goat IgG Alexa Fluor 488 (1:300 dilution; A11055, RRID: AB_2534102, Thermo Fisher Scientific), donkey anti-rabbit IgG Alexa Fluor 568 (1:300 dilution; A10042, RRID: AB_2534017, Thermo Fisher Scientific), and donkey anti-rabbit IgG Alexa Fluor 488 (1:300 dilution; A21206, RRID: AB 2535792, Thermo Fisher Scientific). Images were captured using a Zeiss LSM 700 confocal microscope.

For tight junction analysis in primary endothelial cells, trypsinized 5.0 × 10⁷ primary endothelial cells were seeded on coverslips in 24-well plates, which were pre-coated with collagen IV and fibronectin and cultured for 4 days. On DIV7, 120 nM of recombinant mouse SAP was added for 4 h. PBS was used as the vehicle. Images were captured using a Zeiss LSM 980 confocal microscope with a ×63 objective. For each coverslip, 10 regions were randomly chosen for quantification. The area of occludin-positive tight junctions was quantified by particle analysis using the NIH ImageJ software. Occludin-positive area was normalized to the entire image area.

Primary cortical neuron culture

Primary cortical neurons were isolated from the cerebral cortex of embryonic day 15.5 ICR mouse embryos. Briefly, the cortical plates were dissected from embryos in ice-cold Hanks’ balanced salt solution, then transferred to 2 mL of 0.25% trypsin solution, and incubated at 37 °C for 15 min. The dissociated cells were placed in DMEM high glucose with 10% FBS, penicillin-streptomycin, and 2 mM L-glutamine, as previously reported [37]. Approximately 80,000 cells were plated on 12-mm glass coverslips pre-coated with 1 mg/mL PLL (P2636, Sigma) in a borate buffer solution in 24-well plates. On DIV1, the culture media for cortical neurons were changed to a neurobasal medium with 2% B-27 supplement (Invitrogen) and 200 µM L-glutamine twice to completely remove serum from the culture media. On DIV2, 2 µM Ara-C was added to culture media to inhibit glial cell growth. Cortical neurons were fixed on DIV14 for all experiments. Sera from control SAMR1 and SAMP8 mice were added to culture media at final total serum protein concentrations of 100, 250, 500, or 1,000 µg /mL for 48 h. Recombinant mouse SAP (50113-M08H, Sino Biological, Kawasaki, Japan) was added at a final concentration of 60 nM for 6 h. PBS served as the vehicle for both recombinant SAP and serum. The BCA Protein Assay Kit (A65453, Thermo Fisher Scientific) was used to quantify the total serum protein concentration in the sera. The average total serum protein concentration was 102.665 ± 13.880 mg/mL in SAMR1 sera, and 89.540 ± 13.151 mg/mL in SAMP8 sera. Sera were diluted in culture media to achieve the indicated treatment concentrations. The SAP-neutralizing antibody (50113-T24, Sino Biological) and normal rabbit IgG (CR1, Sino Biological) were added at a concentration of 1 µL/well.

For apoptotic analysis, cortical neuron cultures were immunostained with primary antibodies for anti-cleaved-caspase 3 (1:500 dilution; 9664, RRID: AB_2070042, Cell Signaling Technology) and anti-MAP2 (1:3000 dilution; ab5392, RRID: AB_2138153, Abcam) and secondary antibodies for donkey anti-rabbit IgG Alexa Fluor 568 (1:300 dilution; A10042, RRID: AB_2534017, Thermo Fisher Scientific) and goat anti-chicken IgY Alexa Fluor 488 (1:300 dilution; A11039, RRID: AB_142924, Thermo Fisher Scientific). Images were captured using a Zeiss LSM 700 confocal microscope with a ×20 objective and ×0.5 digital zoom. For each coverslip, 10 regions containing at least 10 MAP2⁺ neurons were randomly selected for quantification. The total number of MAP2⁺ neurons and cleaved caspase 3 and MAP2 double-positive (cleaved caspase 3⁺; MAP2⁺) neurons was counted to determine the ratio of neuronal death.

For synaptic density analysis, cortical neuron cultures were immunostained with primary antibodies for synaptophysin (1:500 dilution; 101 006, RRID: AB_2622239, Synaptic Systems, Goettingen, Germany), PSD95 (1:1000 dilution; MA1-045, RRID: AB_325399, Thermo Fisher Scientific), MAP2 (1:500 dilution; MAB3418, RRID: AB_94856, Merck Millipore), and MAP2 (1:3000 dilution; ab5392, RRID: AB_2138153, Abcam) and secondary antibodies for donkey anti-mouse IgG Alexa Fluor 568 (1:300 dilution; A11037, RRID: AB_2534095, Thermo Fisher Scientific) and goat anti-chicken IgY Alexa Fluor 488 (1:300 dilution; A11039, RRID: AB_142924, Thermo Fisher Scientific). Images were captured using a Zeiss LSM 700 or a Zeiss LSM980 confocal microscope at ×40 objective. For each coverslip, 10 regions were randomly chosen for quantification. The colocalized area of synaptophysin⁺ presynapses or PSD95⁺ postsynapses with MAP2⁺ neuronal area were quantified using the coloc2 Plugin Macro in NIH ImageJ [33] .

Statistical analysis

Detailed information on the number (n) and statistical tests used for each experiment is provided in the manuscript. All statistical analyses were performed using the Prism Software 10.0 (GraphPad Software).

Results

SAMP8 mice showed premature senescence in the hippocampus

SAMP8 mice have approximately half the lifespan of common mouse strains [4]. We first analyzed the learning and memory abilities of 6-month-old SAMP8 mice, which did not show abundant microglial cytokine production or neuronal amyloid β granule accumulation, using the passive avoidance test [7, 8]. Both SAMR1 and SAMP8 mice showed short latency in the bright compartment in the training session (Fig. 1A, B). In the test session, control SAMR1 mice showed the longest latency in the bright compartment to avoid electrical foot shock, whereas SAMP8 mice showed a shorter latency (Fig. 1A, B), indicating that 6-month-old SAMP8 mice showed a decline in learning and memory despite being at an early stage of aging. To identify the brain regions that caused a decline in learning and memory in 6-month-old SAMP8 mice, SA-β-Gal staining was performed, which showed that the neuronal layers in the hippocampal region of SAMP8 mice were drastically stained (Fig. 1C). The SA β-Gal⁺ area in whole hippocampal region of SAMP8 mice increased compared to that in control SAMR1 mice (Fig. 1D). Furthermore, the staining intensity of SA β-Gal in the neuronal layers of the CA1, CA2, CA3, and DG regions was significantly increased (Fig. 1E). These results suggest that the decline in learning and memory at an early stage of aging in SAMP8 mice might be caused by accelerated senescence of hippocampal neurons.

Fig. 1.

Early senescence in the hippocampus induces learning and memory decline in 6-month-old SAMP8 mice. (A) Schematic diagram of passive avoidance test for learning and memory of 6-month-old control SAMR1 and SAMP8 mice. In the training session, a painful memory is developed when the mice enter the dark compartment. In the test session, the latency to enter the dark compartment is recorded for up to 300 s. (B) Learning and memory abilities are analyzed by latency time in the bright compartment in training (left) and test (right) sessions using control SAMR1 and SAMP8 mice. Data represent mean ± SEM, and dots indicate the results of each mouse. Statistic uses unpaired and two-tailed t test. (C) Stains for senescence cells using SA-β-Gal in the hippocampi of 6-month-old control SAMR1 and SAMP8 mice, which revealed accelerated senescence in hippocampal neuron layers in SAMP8 mice. (D) Quantification of SA-β-Gal⁺ area in the whole hippocampi of SAMR1 and SAMP8 mice. Data represent mean ± SEM, and dots indicate the results of each mouse. Statistic uses unpaired and two-tailed t test. (E) Quantification of SA-β-Gal staining mean gray value in the neuronal layers of CA1, CA2, CA3, and DG in SAMR1 and SAMP8 mice. Data represent mean ± SEM, and dots indicate the results of each mouse. Statistics uses two-way ANOVA

The hippocampus of SAMP8 mice showed expression changes for BBB-, immune response-, and neuronal survival-related factors in multi-omics profiling

Bulk RNA-seq profiling was performed using the hippocampi of 6-month-old SAMR1 and SAMP8 mice to understand the mechanisms of accelerated senescence in the hippocampus of SAMP8 mice (Fig. 2A). First, in the bulk RNA-seq profile, we performed PC analysis using the transcripts per million of the detected RNAs to visualize the transcriptomic differences between the hippocampi of SAMR1 and SAMP8 mice, showing that the hippocampal replicates distantly clustered (Fig. 2B). In addition, we extracted DEGs from the hippocampus of SAMP8 mice using the DESeq2 function with hippocampal samples of SAMR1 mice as baselines, which showed that the SAMP8 hippocampal group represented 383 DEGs, including 227 and 156 upregulated and downregulated genes, respectively (Fig. 2C). Functional annotations of DEGs in SAMP8 mice showed much more expanded GO terms, including cellular response to interferon-beta, maintenance of the BBB, and chemokine receptor-binding chemokines (Fig. 2C, D). Next, in the proteomic profile, PC analysis was performed using the detected peptides, which indicated that the hippocampal replicates of SAMR1 and SAMP8 mice were clustered remotely (Fig. 2E). In addition, we obtained DEPs from the hippocampus of SAMP8 mice, which indicated that the hippocampal cluster of SAMP8 mice represented 87 DEPs (Fig. 2F). GO enrichment analysis showed that the DEPs in the hippocampus of SAMP8 mice included cognition-related peptides (Fig. 2F, G). Of these DEPs, 17 shared peptides showed the same alteration as the transcriptomic changes in the hippocampus of SAMP8 mice, including upregulated peptide Dpp7, which regulates apoptosis; Abcb1a, which regulates BBB transport; and downregulated peptide Sypl2, which codes for a synaptic vesicle membrane peptide (Fig. 2H). These profiles suggest that the hippocampus of SAMP8 mice exhibited BBB maintenance, immune responses, and neuronal loss-related dysfunction at the RNA and protein levels.

Fig. 2.

Multi-omics showing BBB maintenance dysfunction and neuronal loss in the hippocampus of SAMP8 mice. (A) Schematic diagram indicating the histological region used for RNA-seq and proteomics in 6-month-old control SAMR1 and SAMP8 mice. (B) PC analysis of the hippocampal transcripts from SAMR1 and SAMP8 mice. (C) Volcano plot showing significantly upregulated and downregulated genes in the hippocampus of SAMP8 mice, with SAMR1 mice as the baseline. DEGs included (magenta) BBB maintenance-, (orange) chemokine-, and (green) interferon-related genes. (D) Heatmap of GO enrichment analysis using DEGs in the hippocampus of SAMP8 mice. (E) PC analysis of peptides detected in the hippocampi of SAMR1 and SAMP8 mice. (F) Volcano plot showing upregulated and downregulated peptides in the hippocampus of SAMP8 mice. DEPs included (cyan) cognition-related peptides. (G) Heatmap of GO enrichment analysis using DEPs in the hippocampus of SAMP8 mice. (H) Venn diagrams indicating the upregulated (left) and downregulated (right) factors shared between DEGs and DEPs from the hippocampus of SAMP8 mice

Neuronal apoptosis was increased in the hippocampus of SAMP8 mice

We analyzed neuronal apoptosis to understand the pathological features in the hippocampus of SAMP8 mice. First, we quantified the ratio of cleaved-caspase 3⁺ neurons in each region of the hippocampus of SAMP8 mice, which indicated that SAMP8 mice showed an increase in cleaved caspase 3⁺ apoptotic neurons in the CA3 and DG regions (Fig. 3A, B). In addition, the staining intensity of the postsynaptic marker PSD95 in the CA3 mossy fiber region was decreased in SAMP8 mice (Fig. 3C, D). As excessive glial cell activation induces neurodegeneration [38], the numbers of microglia and astrocytes in the hippocampi of 6-month-old SAMP8 mice were quantified, revealing that Iba1⁺ and P2RY12⁺ microglial density decreased, despite our expectations (Fig. 3E-G). It is reported that P2RY12 is a more specific marker for brain-resident microglia [39], demonstrating a clear decrease in microglial number in SAMP8 hippocampi. Microglial process morphology in SAMR1 and SAMP8 hippocampi was quantified by Sholl analysis, showing that there were no changes of Iba1⁺ and P2RY12⁺ microglial morphology (Fig. 3H, I). Furthermore, the GFAP⁺ astrocyte density in the hippocampus of SAMP8 mice significantly increased only in the CA3 region (Fig. 3J, K). Meanwhile, the S100β⁺ astrocyte density was no change in all hippocampal regions (Fig. 3J, L). Though both GFAP and S100β are expressed abundantly in hippocampal astrocytes [40], astrocyte density data in the CA3 region were inconsistent. These data suggest no significant increase in astrocytes in SAMP8 hippocampi.

Fig. 3.

The hippocampus of SAMP8 mice shows the increase in neuronal apoptosis and the reduction in synaptic density. (A) Confocal images showing apoptotic neurons stained with cleaved-caspase 3 (red) and NeuN (green) in the CA3 regions of 6-month-old SAMR1 and SAMP8 mice. (B) Ratio of apoptotic neurons in neuronal layers of CA1, CA2, CA3, and DG in SAMR1 and SAMP8 mice. Data represent mean ± SEM, and dots indicate the results of each mouse. Statistics use two-way ANOVA. (C) Immunostained images for PSD95 on mossy fibers in the hippocampi of 6-month-old SAMR1 and SAMP8 mice. (D) Mean gray values of stained PSD95 in mossy fibers of SAMR1 and SAMP8 mice. Data represent mean ± SEM, and dots indicate the results of each mouse. Statistics uses unpaired and two-tailed t test. (E) Immunostained images for Iba1 and P2RY12 in the hippocampi of 6-month-old SAMR1 and SAMP8 mice. (F, G) Numbers of Iba1⁺ (F) and P2RY12⁺ (G) microglia in CA1, CA2, CA3, and DG of SAMR1 and SAMP8 mice. Data represent mean ± SEM, and dots indicate the results of each mouse. Statistics use two-way ANOVA. (H, I) Sholl analysis of Iba1⁺ (H) and P2RY12⁺ (I) microglia, indicating the number of intersections between concentric circles and microglial processes. A graph shows the area under the curve. Data represent mean ± SEM, and dots indicate the results of each mouse. Statistics uses unpaired and two-tailed t test. (J) Immunostained images for GFAP and S100β in the hippocampi of 6-month-old SAMR1 and SAMP8 mice. (K, L) Numbers of GFAP⁺ (K) and S100β⁺ (L) astrocytes in CA1, CA2, CA3, and DG of SAMP8 mice. Data represent mean ± SEM, and dots indicate the results of each mouse. Statistics use two-way ANOVA

Although SAMP8 mice are commonly used as an Alzheimer’s disease mouse model [41], the hippocampi of 6-month-old SAMP8 mice did not show excessive glial activation, unlike other Alzheimer’s disease mouse models. Therefore, we profiled the similarity of gene expression using the DEGs extracted from the transcriptomic data of hippocampus from 6-month-old SAMP8 mice, an Alzheimer’s disease model 8-month-old 5×FAD mice, and a frontotemporal dementia model 7-month-old Grn^−/− mice [42, 43]. The SAMP8 profile showed 52 and 26 DEGs shared by 5×FAD and Grn^−/− mice, respectively (Fig. S1A). The shared DEGs included Myo1f but not signature genes, such as Apoe and C1qa (Fig. S1B). Although multiple GO enrichment analyses showed that DEGs from the hippocampi of SAMP8 and 5×FAD mice shared GO terms, including regulation of defense response, Epstein-Barr virus infection, response to virus, response to interferon-beta, inorganic ion transmembrane transport, and cellular cation homeostasis, shared GO terms were a small part of them in the 5×FAD group (Fig. S1C). These results indicated that aging in the SAMP8 hippocampus progressed through a specific mechanism distinct from that of Alzheimer’s disease and frontotemporal dementia. Furthermore, SAMP8 mice showed no changes in the expression of inflammatory response-related genes, which is a characteristic phenotype in Alzheimer’s disease and frontotemporal dementia. This is consistent with the quiescent glial cells in the hippocampus of SAMP8 mice and supports the idea that SAMP8 mice develop neurodegeneration without drastic glial activation.

Blood-derived SAP was accumulated in the hippocampus of SAMP8 mice

As aged SAMP8 mice show a drastic increase in BBB permeability, which leads to the extravasation of blood components into the brain parenchyma [9], the neuronal cell death might be induced by blood component factors transferred into the parenchyma due to the disruption of selective BBB permeability. Here, we identified the blood component factors transferred into the brains of 6-month-old SAMP8 mice. First, we performed proteomic analysis using the sera from 6-month-old SAMP8 mice. PC analysis of the protein intensity for the detected peptides showed that serum replicates clustered distantly between SAMR1 and SAMP8 mice (Fig. S2A). GO enrichment analysis using the DEPs, which included 53 and 143 downregulated and upregulated peptides, respectively, indicated that axon guidance- and adaptive immune system-related peptides were upregulated, and positive regulation of immune response-related peptides were downregulated in the SAMP8 mouse serum (Fig. S2B), implying that SAMP8 mice showed immune system alterations in the peripheral organs. Next, to identify the blood component factors that were transferred into the brain, shared DEPs between the sera and the hippocampus of SAMP8 mice were extracted, revealing that six peptides were common (Fig. 4A, B). Interestingly, Apcs (common name: SAP), Ces1b (common name: carboxylesterase 1), and Orm1 (common name: orosomucoid 1) were not expressed in the hippocampi of either SAMR1 or SAMP8 mice at the RNA level (Fig. S2C), suggesting that SAP, carboxylesterase 1, and orosomucoid 1 may be detected in the hippocampus via transport from the blood to the parenchyma.

Fig. 4.

Blood-derived SAP is transported through the BBB into the hippocampus of SAMP8 mice. (A) Volcano plot showing upregulated and downregulated peptides in the sera of SAMP8 mice. Cyan-colored dots indicate shared DEPs between the sera and hippocampus of SAMP8 mice. (B) Venn diagram indicating the shared DEPs between the sera and hippocampus of 6-month-old SAMP8 mice. The peptides that are not detected in the transcriptomic data of hippocampus of SAMR1 and SAMP8 mice are highlighted with red. (C) Violin graphs showing ELISA for SAP volume using the lysates of the hippocampi of SAMR1 and SAMP8 mice. The SAP volume is normalized with respect to the total volume of protein in lysates. The dashed line indicates the median, dot lines indicate quartiles, and dots indicate the results of each mouse. Statistic uses unpaired and two-tailed t test. (D) Confocal and IMARIS-3D images of blood vessels stained with SAP (green), CD31 (red), and GFAP (blue) in the hippocampi of 6-month-old SAMR1 and SAMP8 mice, showing the extravasation of SAP from the blood to the astrocytes (arrows) and brain parenchyma. (E) Quantification of SAP staining intensity in the astrocytes of the hippocampus of SAMP8 mice. The dashed line indicates the median, dot lines indicate quartiles, and dots indicate the results of each mouse. Statistic uses unpaired and two-tailed t test. (F) Confocal images of blood vessels stained with occludin (green) and CD31 (magenta) in the hippocampi of 6-month-old SAMR1 and SAMP8 mice, showing the gap of vulnerable tight junction (arrowheads) in SAMP8 mice. (G) Quantification of occludin-positive tight junction length in CD31⁺ endothelial cell area. The dashed line indicates the median, dot lines indicate quartiles, and dots indicate the results of each mouse. Statistic uses unpaired and two-tailed t test. (H) Schematic diagram of the study design to quantify the changes in BBB resistance after SAP treatment into Boyden chambers using an in vitro 3D BBB model, which is composed by primary endothelial cells, pericytes, and astrocytes. (I) Line graph showing the changes of electrical resistance values in the in vitro BBB model treated with 30, 60, and 120 nM SAP at 0, 1, 4, 12, and 24 h after treatment (left). Data represent mean ± SEM. Violin graph showing the area under the curve (AUC) of the electrical resistance value changes for each SAP concentration (right). The dashed line indicates the median, dot lines indicate quartiles, and dots indicate the results from each chamber prepared with four independent cultures. Statistics use two-way ANOVA. (J) Violin graphs showing ELISA for SAP volume using the media in the outer compartment of in vitro BBB model treated with 60 nM SAP into the inner chamber. The SAP volume is normalized with respect to the total volume of protein in media. The dashed line indicates the median, dot lines indicate quartiles, and dots indicate the results of each culture. Statistic uses unpaired and two-tailed t test. (K) Confocal images of blood vessels stained with occludin (green) and CD31 (magenta) in primary brain endothelial cells treated with 120 nM SAP for 4 h showing the disorder of occludin-positive tight junction. (L) Quantification of occludin-positive tight junction area in CD31⁺ primary brain endothelial cell area. The dashed line indicates the median, dot lines indicate quartiles, and dots indicate the results from each independent culture. Statistic uses paired and two-tailed t test

A previous study has reported that SAP induces neuronal apoptosis and amyloid β accumulation [44]; therefore, we studied whether SAP passes through selective permeability-disrupted BBB in the hippocampus of SAMP8 mice. First, we confirmed whether the SAP concentration was increased in the hippocampus of SAMP8 mice using ELISA, showing that the hippocampus of SAMP8 mice had more SAP than the hippocampi of control SAMR1 mice (Fig. 4C). In addition, we observed SAP localization in the hippocampus, showing that SAMR1 mice showed the localization of a small amount of SAP in blood vessels composed of CD31⁺ endothelial cells using 3D rendered images. In contrast, in SAMP8 mice, while some SAP was retained in the blood vessels, the majority was localized outside the blood vessels, particularly colocalizing with GFAP⁺ astrocytes (Fig. 4D, Fig. S3A-D). The staining intensity of SAP in GFAP⁺ astrocytes was significantly increased in SAMP8 mice (Fig. 4E). SAP also co-localized with Tuj1⁺ neurons, with no localization to Iba1⁺ microglia in the hippocampi of SAMP8 (Fig. S3E-G). These results suggest that SAP is transported from the blood to the brain while crossing the BBB in the hippocampus of SAMP8 mice.

Although it has been reported that SAP selectively passes through the endothelial cell layer [45], it is possible that SAP may, which is upregulated in the sera of SAMP8 mice, weakens BBB integrity. Here, to reveal whether BBB integrity is disrupted in the hippocampus of SAMP8 mice, we stained for occludin, a tight junction factor, indicating that occludin-positive tight junctions represent a vulnerable morphology and gaps in SAMP8 mice (Fig. 4F). Although proteomic analysis did not detect differences in expression levels, the length of occludin-positive tight junctions decreased in SAMP8 mice (Fig. 4G). Consistent with this data, the previous studies indicated that SAMP8 mice show higher Evans Blue-related BBB permeability from 5-month-old [46, 47]. These results imply that the disruption of BBB integrity leads to remarkable SAP transport into the hippocampus in 6-month-old SAMP8 mice. Next, to evaluate whether SAP regulates BBB function, we prepared an in vitro 3D BBB model that mimicked the BBB structure using a co-culture of endothelial cells, pericytes, and astrocytes in Boyden chambers (Fig. S4A). First, to confirm whether the three cell types were isolated from mouse brains, the cells were stained with CD31, platelet-derived growth factor receptor β (PDGFRβ), or GFAP as markers for endothelial cells, pericytes, and astrocytes, respectively (Fig. S4B), indicating that all cell types were pure. Next, we applied various concentrations of recombinant SAP to the inner chambers and quantified BBB permeability for 24 h using resistance (Fig. 4H). We found that in the vehicle condition, the TEER was commonly enhanced at 1 and 4 h after changes to the serum-free condition, but SAP treatment inhibited TEER enhancement and attenuated the TEER of in vitro BBB membrane at all concentrations of SAP (Fig. 4I). It has been suggested that SAP enhances BBB permeability. In addition, after 4 h after 60 nM SAP treatment into the inner chambers, SAP was detected in the outer compartment (Fig. 4J), suggesting that SAP could cross the BBB in the direction from blood to the brain. Therefore, to determine whether SAP affects the structure of tight junctions between endothelial cells, primary brain endothelial cells were treated with 120 nM SAP and stained with occludin to visualize tight junctions. SAP treatment induced tight junction disorder, which might be caused by the enlargement of endothelial cells and the vulnerability of occludin-positive tight junctions (Fig. 4K). Supporting this observation, SAP treatment decreased the area of occludin-positive tight junction (Fig. 4L). These results suggest that blood-derived SAP leads to BBB dysfunction at an early stage of aging in SAMP8 mice.

SAMP8 serum-derived serum amyloid P component induces neuronal cell death

Our results indicated that SAP co-localized with Tuj1⁺ neurons in the hippocampus of SAMP8 mice (Fig. S3E, F), and it has been reported that SAP induces neuronal apoptosis using primary rat neuron culture [44]. These results suggest that blood-derived SAP in SAMP8 mice induces neuronal apoptosis. To elucidate this, we prepared primary mouse cortical neuron cultures and added the serum protein from 6-month-old SAMR1 and SAMP8 mice for apoptotic analysis (Fig. 5A). Treatment with the serum protein of SAMP8 mice increased the ratio of cleaved-caspase 3⁺ apoptotic neurons (Fig. 5B, C). ELISA confirmed that the serum in SAMP8 mice contained more SAP than that from SAMR1 mice (Fig. 5D), consistent with the results of the proteomic analysis (Fig. 4A). Next, we added recombinant mouse SAP to primary cortical neurons for 6 h to determine whether SAP induces neuronal apoptosis. Treatment with 60 nM SAP significantly increased the ratio of cleaved-caspase 3⁺ apoptotic neurons (Fig. 5B, E). In addition, to confirm whether SAP in the serum protein of SAMP8 mice induces neuronal apoptosis, we added a neutralizing antibody against SAP with the serum protein of SAMP8 mice, which revealed that the inhibition of SAP binding to neurons mitigated serum-induced neuronal apoptosis in SAMP8 mice (Fig. 5B, F). Therefore, SAP, which is more abundant in the serum of SAMP8 mice, promotes neuronal cell death.

Fig. 5.

Blood-derived SAP promotes neuronal apoptosis and synaptic density decline in a primary cortical neuron culture. (A) Schematic diagram showing the study design to quantify apoptosis and synaptic density in E15.5 mouse-derived primary cortical neurons treated with the sera of SAMR1 and SAMP8 mice for 48 h, recombinant SAP for 6 h, and the neutralizing anti-SAP antibody for 2 h before sera treatment. (B) Confocal images of cleaved-caspase 3 (red) and MAP2 (green) stained cortical neurons treated with vehicle, 1000 µg/mL total serum protein of SAMR1 and SAMP8 mice, 60 nM recombinant SAP, or 500 µg/mL total serum protein of SAMP8 mice and the neutralizing anti-SAP antibody, showing that apoptotic cells with cleaved-caspase 3⁺ granules (arrowheads). Cleaved-caspase 3; MAP2 double positive neurons (arrows) are counted as cleaved-caspase 3⁺ neurons. (C) Ratio of cleaved-caspase 3⁺ apoptotic neurons in primary cortical neurons treated with 100, 250, 500, or 1000 µg/mL serum protein of SAMR1 and SAMP8 mice. Data represent mean ± SEM from at least three independent cultures. Statistic uses two-way ANOVA. (D) ELISA for SAP concentration in the sera of 6-month-old SAMR1 and SAMP8 mice. The SAP volume is normalized with fluid volume of sera. The dashed line indicates the median, dot lines indicate quartiles, and dots indicate the results of each mouse. Statistic uses unpaired and two-tailed t test. (E) Ratio of cleaved-caspase 3⁺ apoptotic neurons in primary cortical neuron cultures treated with vehicle and 60 nM recombinant SAP. Data represent mean ± SEM from four independent cultures, and dots indicate the results from each experimental culture. Statistic uses unpaired and two-tailed t test. (F) Ratio of cleaved-caspase 3⁺ apoptotic neurons in primary cortical neurons treated with control rabbit IgG and neutralizing anti-SAP antibody 2 h before the treatment with 500 µg/mL serum protein of SAMR1 and SAMP8 mice. Data represent mean ± SEM from four independent cultures, and dots indicate the results from each experimental culture. Statistic uses two-way ANOVA. (G) Confocal images of synaptophysin (magenta) and MAP2 (green)-, and PSD95 (magenta) and MAP2 (green)-stained cortical neurons treated with vehicle, 500 µg/mL serum protein of SAMR1 and SAMP8 mice, 60 nM recombinant SAP, or 500 µg/mL serum protein of SAMP8 mice and the neutralizing anti-SAP antibody. (H, I) Colocalized area of Synaptophysin⁺ presynapses or PSD95⁺ postsynapses with MAP2⁺ cytoplasmic and neurite area in primary cortical neurons treated with vehicle and 500 µg/mL serum protein of SAMR1 and SAMP8 mice. Data represent mean ± SEM from four independent cultures. Statistics use two-way ANOVA. (J, K) Colocalized area of Synaptophysin + presynapses or PSD95⁺ postsynapses with MAP2⁺ neuronal area in primary cortical neurons treated with vehicle and 60 nM recombinant SAP. Data represent mean ± SEM from five independent cultures, and dots indicate the results from each experimental culture. Statistic uses unpaired and two-tailed t test. (L, M) Colocalized area of Synaptophysin⁺ presynapses or PSD95⁺ postsynapses with MAP2⁺ neuronal area in primary cortical neurons treated with control rabbit IgG and neutralizing anti-SAP antibody 2 h before the treatment with 500 µg/mL sera of SAMR1 and SAMP8 mice. Data represent mean ± SEM from four independent cultures, and dots indicate the results from each experimental culture. Statistics uses two-way ANOVA. ns, not significant

Our multi-omics analysis revealed the downregulation of the synaptic vesicle membrane peptide Sypl2 in the hippocampus of SAMP8 mice (Fig. 2C, F, H), suggesting that blood-derived SAP induces the loss of synapses as well as apoptosis. We analyzed the synaptophysin⁺ presynaptic and PSD95⁺ postsynaptic area in primary cortical neurons treated with the serum protein of SAMR1 and SAMP8 mice, showing that the control serum treatment in SAMR1 mice had no effect or enhanced the synaptophysin⁺ presynaptic or PSD95⁺ synaptic density, respectively, because the sera contained various neurotrophic factors. However, the serum protein from SAMP8 mice reduced the presynaptic and postsynaptic areas compared with those from SAMR1 mice (Fig. 5G-I). In addition, recombinant mouse SAP treatment reduced synaptophysin⁺ or PSD95⁺ synaptic area in primary cortical neuron cultures (Fig. 5G, J, K). Furthermore, treatment with a neutralizing antibody against SAP blocked SAMP8 serum-induced reduction in synaptophysin⁺ or PSD95⁺ synaptic area (Fig. 5G, L, M). These results suggest that SAP, which is upregulated in the sera of SAMP8 mice, is transported to the hippocampus, where it promotes neuronal cell death and synaptic loss.

Discussion

In this study, to reveal how age-related disruptions in brain homeostasis induce neurodegeneration, we clarified the mechanism of neurodegeneration in SAMP8 mice, which showed accelerated senescence approximately twice as fast. First, in 6-month-old SAMP8 mice, learning and memory abilities declined drastically compared to those of control SAMR1 mice (Fig. 1A, B). Since SA-β-Gal⁺ senescent cells were detected particularly in the hippocampi of the SAMP8 mice (Fig. 1C-E), it is considered that the decline in learning and memory abilities at the early stage of aging in SAMP8 mice is caused by the accelerated senescence of hippocampal neurons. To comprehensively understand the state of the hippocampi of 6-month-old SAMP8 mice, bulk transcriptomic and proteomic profiling was performed, showing that BBB maintenance-, immune response-, and neuronal loss-related changes occurred in SAMP8 mice (Fig. 2). Consistent with the multi-omics profiling, pathological analyses indicated that neuronal apoptosis and synaptic decline occurred in the hippocampus of SAMP8 mice (Fig. 3A-D). Moreover, the occluding-positive tight junctions were disrupted in the hippocampi of SAMP8 mice (Fig. 4F, G). Previous studies have reported that aged SAMP8 mice show drastic extravasation of blood components into the brain [9]; therefore, we hypothesized that neuronal cell death was induced by the toxicity of blood component factors transferred into the brain through the BBB. Using the serum proteome of SAMP8 mice, we identified Apcs, Ces1b, and Orm1 as blood-derived peptides that might be transported into the hippocampus (Fig. 4A, B). We confirmed that SAP was translocated into the brain through the BBB (Fig. 4D, E). Interestingly, BBB permeability was regulated by SAP (Fig. 4H-L), suggesting that an increase in SAP in the blood contributes to BBB dysfunction and transport of SAP into the brain parenchyma. Furthermore, blood-derived SAP promoted neuronal apoptosis and pre- and post-synaptic decline in primary cortical neurons (Fig. 5). Overall, our results indicate that an increase in the blood component, SAP, disrupts BBB homeostasis, and the translocation of SAP into the brain leads to neurodegeneration during aging in SAMP8 mice (Fig. 6).

Fig. 6.

Proposed model of the blood-derived neurotoxic SAP property in SAMP8 mice at early aging stages. Based on transcriptomic analysis of the hippocampus of SAMP8 mice, changes in the expression of BBB maintenance-associated genes were observed, suggesting that SAMP8 mice exhibit dysregulation of BBB function. Consistent with this, the hippocampus of SAMP8 mice showed the vulnerability of tight junctions between endothelial cells. The integration of proteomic and transcriptomic analyses of the hippocampi and sera from SAMP8 mice suggests transport of SAP, carboxylesterase 1, and orosomucoid 1 from the blood into the brain parenchyma at 6 months of age. Particularly, SAP is abundantly detected around the hippocampal BBB components, including astrocytes. In addition, SAP has multiple functions, including attenuating the tight junction strength and inducing neuronal apoptosis and synaptic decline. Taken together, we propose a model: at early stages of aging in SAMP8 mice, upregulated serum SAP is transported to the brain parenchyma through vulnerable tight junctions or accelerated endocytosis in endothelial cells, leading to neurodegeneration

It is possible that the cognitive decline in SAMP8 mice is caused by hippocampal neurodegeneration. The present study indicated that 6-month-old SAMP8 mice displayed a decline in recent memory using the passive avoidance test and a hippocampus-specific increase in SA-β-Gal⁺ senescent cells (Fig. 1). Senescence has three consequences: neuroinflammation, impaired neurogenesis, and synaptic dysfunction [48]. In the hippocampi of 6-month-old SAMP8 mice, pro-inflammatory factors did not increase in the proteomic analysis (Fig. 2F, G), and drastic glial activation was not observed (Fig. 3E-I), suggesting that neuroinflammation did not occur in the early aging stage of SAMP8 mice. It has been reported that the microglia of 10-month-old SAMP8 mice have characteristics to express more Il-1b than that of SAMR1 mice [7]. Therefore, it is possible that older SAMP8 mice exhibit senescent cell-induced neuroinflammation. Although we did not study neurogenesis in SAMP8 mice, a previous study showed that the number of doublecortin⁺ cells is decreased in the hippocampus of SAMP8 mice [49]. Moreover, the inhibition of neurogenesis causes only a decline in remote, but not recent, memory [50], suggesting that the decrease in hippocampal neurogenesis in SAMP8 mice induces a decline in learning and memory abilities; however, further investigation is required. Furthermore, since SAMP8 mice showed a decrease in PSD95⁺ synapse intensity (Fig. 3C, D), it is suggested that the synaptic decline in senescent cells induces cognitive decline, which is supported by a report showing that rats with cognitive decline exhibit a reduction in PSD95⁺ post-synapses in hippocampal neurons [51]. Therefore, it is possible that the observed reduction in excitatory synapses observed in senescent cells contributes to cognitive decline during the early aging stage in SAMP8 mice.

BBB dysfunction is considered to occur at an early stage of aging in SAMP8 mice. In general, the BBB contributes to a highly controlled selective permeability through strong tight junctions, transporters, and selective endocytosis in endothelial cells, which are dysregulated during aging [52]. Age-dependent loss of BBB integrity has been observed in healthy humans through MRI [53]. Furthermore, aged mice show increased IgG leakage into the brain parenchyma [54]. Similarly, SAMP8 mice showed drastic IgG leakage into the parenchyma at 12 months of age [9]. In addition, we demonstrated alterations in the expression of BBB-related genes, including transporters and tight junction-regulated genes, and in the vulnerability of occludin-positive tight junctions in CD31⁺ endothelial cells in the hippocampi of 6-month-old SAMP8 mice (Figs. 2C and D and 4F and G). These results suggest that the dysregulation of BBB function has already occurred at an early stage of aging, which advances to complete IgG leakage into the parenchyma at 12 months of age. Although the present study did not clarify the detailed mechanism of BBB dysfunction in SAMP8 mice, our results suggested two hypotheses: Our data indicated an increase in the number of GFAP⁺ astrocytes in SAMP8 mice (Fig. 3J, K), which regulate molecular transport between the blood and brain; therefore, it is possible that the increase in GFAP⁺ astrocytes causes structural changes in the BBB and alters molecular transport. Conversely, it is possible that the BBB breakdown causes an increase in the number of GFAP⁺ astrocytes. However, the contribution of astrocytes to BBB dysfunction remains controversial. We must note that we did not detect the change in S100β⁺ astrocyte number (Fig. 3J, L). Though it is known that GFAP is more specific astrocyte marker than S100β [55], further study about the change of astrocyte dynamics in 6-month-old SAMP8 hippocampus is required. Second, because the sera of SAMP8 mice contained more inflammation-related peptides (Fig. S2B), it is possible that blood factors exert toxicity to endothelial cells, which induces the vulnerability of tight junctions and accelerates non-selective endocytosis. We demonstrated that SAP induced vulnerability of tight junctions in primary endothelial cells (Fig. 4K, L). Collectively, it is possible that the increased toxicity of blood factors contributes to BBB dysfunction during aging.

Further studies are needed to demonstrate the mechanisms underlying SAP-induced BBB tight junction vulnerability. The present study clarified that the treatment of recombinant SAP increases the BBB permeability in an in vitro 3D BBB model (Fig. 4I) and induces the disruption of occludin-related tight junctions in primary brain endothelial cells (Fig. 4K, L). It is suggested that the upregulated serum SAP can impair the selective permeability of BBB by disrupting the tight junction between endothelial cells in 6-month-old SAMP8 hippocampus. Meanwhile, our in vitro assessments for BBB function have limitations since in vitro experiments did not mimic the physiological condition completely. In the future, it is necessary to verify that the circulating SAP induces BBB dysfunction using the transgenic mice, which show liver-specific overexpression of SAP.

Further research is required to understand the mechanism by which SAP is transported from the blood to the brain while passing through the BBB in SAMP8 mice. Although our proteomic analyses suggested that several proteins, including SAP, were translocated into the brain parenchyma at an early stage of aging in SAMP8 mice, the number of translocated proteins was unexpectedly low (Fig. 4A, B). We performed proteomic analyses after the removal of high-abundance proteins, such as albumin and IgG; therefore, these abundant blood components might also leak into the brain. In addition, it is possible that the disruption of tight junctions in the BBB allows blood components to be paracellularly transported into the brain in SAMP8 mice, but the shared DEPs between the sera and hippocampi of 6-month-old SAMP8 mice were scarce. Therefore, the dysregulation of intracellular transport in endothelial cell in 6-month-old SAMP8 mice as well as the vulnerability of tight junctions is a possibility. Intracellular protein transport in endothelial cells occurs via receptor-mediated selective transcytosis and non-selective caveolae transcytosis [56]. Proteomic analysis of the hippocampus of SAMP8 mice showed a decrease in FcγR2, one of the receptors for SAP (Fig. 2F), although it was not confirmed which cell types showed a change in FcγR2 expression. This implied that changes in FcγR2 expression in endothelial cells affected receptor-mediated transcytosis. Furthermore, the caveolar transcytosis is enhanced by aging [54], supporting the possibility that SAP translocation is caused by the increase of caveolae transcytosis in SAMP8 mice. However, drastic non-selective leakage of blood components was not detected; therefore, it is necessary to further consider the dynamics of caveolar vesicle transcytosis in SAMP8 mice.

Thus, SAP may be an initial factor that promotes neuronal loss in SAMP8 mice. SAP is an acute-phase response protein produced in the liver with multiple biological functions. It has been reported to activate classical complement activation and phagocytosis, which contribute to bacterial resistance against bacteria [57], indicating that SAP is an essential factor in maintaining homeostasis. By contrast, SAP directly contributes to neuronal toxicity. It is known that the pharmacological removal of SAP from the brain parenchyma attenuates the amyloid β deposition in Alzheimer’s disease models, showing that SAP promotes amyloid fibril formation [58] and neuronal apoptosis [44, 59]. In addition to the early aging stage in SAMP8 mice, even blood-derived SAP promoted neuronal apoptosis. We found that treatment with a neutralizing antibody against SAP attenuated serum-induced neuronal apoptosis and synaptic decline in vitro in SAMP8 mice (Fig. 5). In the present study, we did not elucidate the mechanism by which SAP induces neuronal apoptosis. Interestingly, proteomic and transcriptomic analyses of the hippocampus of SAMP8 mice revealed the expression of FcγR2 (Figs. 2F and 4A), a receptor for SAP [60]. Additionally, FcγR2 induces T cell apoptosis [61]. Therefore, these results suggest that SAP regulates neuronal apoptosis by interacting with FcγR2.

The SAMP8 mouse is a senescence-accelerated mouse model, which is characterized by early onset of neurodegeneration. Our data indicate that the hippocampus of 6-month-old SAMP8 mice already exhibits increased neuronal apoptosis (Fig. 3A, B), whereas in healthy aging, a marked increase in neuronal apoptosis is observed in the hippocampus of 18-month-old wild-type mice [62]. Collectively, the SAMP8 hippocampus shows neuronal apoptosis earlier than that of a healthy mouse during aging. Furthermore, it is necessary to understand the differences in the characteristics of aged microglia between SAMP8 and wild-type mice, since SAMP8 microglia may demonstrate a different transition compared to that of wild-type microglia. Notably, the numbers of Iba1⁺ and P2RY12⁺ microglia are reduced in the hippocampus of 6-month-old SAMP8 mice (Fig. 3E-G), consistent with a previous study showing that the Iba1⁺ area is decreased at this age [63]. In contrast, during healthy aging in 6–18-month-old mice, change in microglial density is not observed in the hippocampus; instead, microglial morphology changes from ramified to ameboid type [64, 65]. Although there is no report on microglial morphology in mice older than 10 months due to the short lifespan of SAMP8 mice, SAMP8 microglial morphology does not change remarkably until 10 months of age [7]. Taken together, the reduction of microglial density is a characteristic phenotype in older SAMP8 mice, compared with that in wild-type mice. Recent single-cell RNA-seq analysis of CD11b⁺ microglia in the cortex of 9-month-old SAMP8 mice showed an expansion of specific microglial types, including interferon response microglia and accelerated aging-associated microglia [66]; therefore, further investigation for SAMP8 microglial functions based on single-cell analysis is warranted. Furthermore, SAMP8 astrocytes may demonstrate a transition similar to those of normal astrocytes during aging. Our data showed no change in astrocyte density in the hippocampus of 6-month-old SAMP8 mice (Fig. 3J-L), consistent with a prior report that in healthy older mice, the changes in astrocyte density are mild [67]. Therefore, though a detailed analysis is required, SAMP8 astrocytes and normal-aged astrocytes show a similar phenotype. In summary, the SAMP8 mouse can be a highly practical aging model for neuronal apoptotic analysis. Besides, given the BBB dysfunction observed in the SAMP8 hippocampus (Fig. 4F, G), this model is also useful for the analysis of aging-associated BBB impairment.

Recently, the detailed dynamics and mechanisms of aging-related BBB dysfunction have been revealed with the development of experimental techniques [52, 68–71]. To address the challenges of observing microstructures such as the BBB, novel techniques for visualizing deep cerebrovascular structures are being developed. For example, ultrasound localization microscopy enables the analysis of real-time cerebral blood flow with high resolution [69], which also clarified that the vascular coverage area, vascular diameter, neurovascular coupling, and cerebral blood flow are decreased in the older mouse brain [70]. In addition, reports indicate that aging induces impaired cerebrovascular functions, including reduced functional hyperemia and increased oxidative stress in endothelial cells [71, 72]. Collectively, these findings reveal that aging drives dynamic changes in cerebral blood vessels. Further studies are required to clarify the mechanisms by which aging causes cerebrovascular dysfunction. The present study suggests that a blood circulating factor, SAP, is one of the causes of senescence-associated cerebrovascular dysfunction.

In this study, we propose that blood-derived neurotoxic factors transported due to BBB dysfunction may cause cognitive decline and dementia. Additionally, BBB dysfunction may be caused by an increase in toxic factors in the blood during aging. Although there is still a limitation in the experimental techniques for BBB function analysis, further studies are required to clarify the association between the mechanisms of BBB dysfunction caused by aging and the pathogenesis of neurodegenerative diseases.

Abbreviations

CAN

Acetonitrile

AGC

Auto gain control

AUC

Area under the curve

BBB

Blood-brain barrier

BSA

Bovine serum albumin

DEGs

Differential expressed genes

DEPs

Differential expressed peptides

DIV

Day in vitro

DMEM

Dulbecco’s modified eagle medium

DNB

DNA Nanoball

E

Embryonic day

FBS

Fatal bovine serum

FDRs

False discovery rates

GO

Gene Ontology

P

Postnatal day

PBS

Phosphate buffered saline

PC

Principal component

PLL

Poly-L-lysin

SAMP8

Senescence accelerated mouse- prone 8

SAMR1

Senescence accelerated mouse-resistant 1

SA-β-Gal

Senescence-associated β-Galactosidase

SAP

Serum amyloid P component

SDS

Sodium dodecyl sulfate

TEER

Trans-endothelial electrical resistance

TFA

Trifluoroacetic acid

VCAM-1

Vascular cell adhesion molecule-1

Authors’ contributions

F.S. performed pathological experiments and in vitro experiments. A.F. performed pathological experiments. S.K. performed part of transcriptome analysis. Y.U. performed synaptic analysis in vivo. M.I., D.N., and R.K. performed proteomics analysis. K.N. and T.I. provided mouse serum. K.Y. conducted the transcriptome analysis. Y.M. modified the manuscript. M.G. modified the manuscript and provided funding. K.H. wrote the manuscript, provided funding and supervised this project.

Funding

This work was supported by grants from the Sasakawa Scientific Research Grant 2023–4001, The Mitsubishi Foundation, Astellas Foundation for Research on Metabolic Disorders, Nagase Brothers, Inc. Research Grant, Uehara Memorial Foundation, Brain Science Foundation, Shiseido Female Researcher Science Grant, Life Science Foundation of Japan, Mochida Memorial Foundation for Medical and Pharmaceutical Research to K.H. and Sasakawa Scientific Research Grant 2025–4010 to F.S.

Data availability

The datasets of proteomic data using the hippocampi and sera of 6-month-old male SAMR1 and SAMP8 mice for this study can be found in jPOST REPOSITORY (ID: JPST003961/PXD066582).

Declarations

Competing interests

The authors declare no competing interests.