SLAMF8 and NINJ2 promote neuroinflammation and oxidative stress through TLR4 NF kappa B pathway in Alzheimer’s disease

The mRNA expression patterns of human Alzheimer’s disease (AD) were obtained from the Gene Expression Omnibus (GEO) database. In this study, we utilized GSE122063↗ to build co-expression networks and pinpoint hub genes associated with Alzheimer’s disease (AD). The microarray dataset included gene expression profiles captured in percutaneous allograft biopsies from 12 patients with Alzheimer’s disease (AD) and 11 controls without dementia. Normalization of the data was performed using the Robust Multiarray Average (RMA) technique, followed by log2 transformation and quantile normalization.

After applying background correction and normalization, differentially expressed genes (DEGs) were detected using the “limma” R package (v3.56.2). The co-expression network was built using the “WGCNA” R package (v1.72-5), with a threshold of 0.85 for the correlation coefficient and a soft-thresholding power equivalent to 12. A cut height of 0.2 was established to combine modules with comparable characteristics, and the pink module was shown to be most strongly linked to AD.

The elevated genes inside the pink module were subjected to Gene Ontology (GO) enrichment analysis using the “enrichR” package. The obtained findings were then shown using the “ggplot2” R package. The Cytohubba plugin in Cytoscape v3.10.2 was used to identify the top 20 hub genes. Protein-protein interaction networks were then constructed using the STRING database.

The SLAMF8 and NINJ2 structures were acquired from the AlphaFold Protein Structure Database. The interaction between SLAMF8 and NINJ2 was investigated through molecular docking using the HDOCK server¹⁷. Visualization, analysis, and mapping of the best predicted binding mode (lowest score) were carried out using the Pymol program (version 4.3.0).

APP/PS1 transgenic mice were acquired from HFK Bioscience Cooperation (14002a, Beijing, China) using the C57BL/6 genetic background. The SLAMF8 knockout (KO) mice were produced by Cyagen (Jiangsu, China). Over 10 generations of backcrossing between SLAMF8 knockout mice and C57BL/6 mice were conducted to guarantee a genetic background of over 99% C57BL/6, which was verified by microsatellite analysis. Backcrossed SLAMF8 knockout (KO) mice were subsequently crossed with APP/PS1 mice to produce the following groups: wild-type (SLAMF8+/+), APP/PS1 (SLAMF8+/+), APP/PS1/SLAMF8 knockout (SLAMF8−/−). All mice were kept in controlled environments at a temperature of 22 °C, with a fixed 12-hour light-dark cycle, and were provided with unrestricted access to food and water. Under permit No. 2024PS149K, the Animal Ethics Committee of China Medical University granted approval for all animal experiments, which were carried out in accordance with the ARRIVE principles. Mice were anesthetized using isoflurane (1–3% for induction and maintenance) for all surgical procedures and sacrificed using carbon dioxide (CO₂) inhalation followed by cervical dislocation to ensure humane treatment. Furthermore, all experimental procedures were performed in strict accordance with relevant guidelines and regulations.

Spatial memory development was evaluated in age-matched 16-month-old wild-type (SLAMF8+/+), APP/PS1 (SLAMF8+/+), and APP/PS1/SLAMF8 knockout (SLAMF8−/−) mice using the Morris Water Maze test (Morris, 1984). A conventional black circular tank measuring 150 cm in diameter, 70 cm in height, and 45 cm in water depth, filled to a depth of 30 cm with water kept at a temperature of 22 ± 1 °C, constitutes the Morris water maze. The tank is partitioned into four quadrants, including northeast (NE), northwest (NW), southeast (SE), and southwest (SW), utilizing the EthoVision tracking system software (Noldus Information Technology, Wageningen, The Netherlands)¹⁸. Within the SE quadrant, a concealed platform is positioned at the centre, submerged at a depth of around 2 cm below the water surface. Diverse images function as supplementary visual indicators within the maze and are strategically placed around the tank.

The examination consists of two stages: location navigation (skill acquisition) and probe. The location navigation phase consists of four repeated trials conducted daily over a span of three consecutive days. During each trial, the animal is immersed in the water, positioned towards the pool wall in one of the quadrants, and given a maximum of 60 s to locate the concealed platform. Failure of an animal to find the platform within the designated time is met with manual guidance and a 20-second period of time spent there to enhance learning. The investigation phase takes place on the fourth day, precisely 24 h following the last learning trial, when the educational platform is no longer available. It is anticipated that each rat will swim for a duration of 60 s, with the greatest amount of time spent in the quadrant where the platform was previously positioned, known as the target quadrant. Behavioral parameters, including distance travelled, escape latency (time to locate the platform), swimming speed, time spent in the target quadrant, and number of platform crossings during the probe phase, are captured by a video camera and analysed using the EthoVision program. After completing the behavioral assessments, the mice were thoroughly sedated and administered 15 ml of phosphate-buffered saline through transcardial perfusion. Subsequently, the brains were removed from the skull. The experimental procedure is illustrated schematically in Fig. 1.

The preparation of Aβ₁₋₄₂ oligomers was based on the method established by Klein¹⁹. The lyophilized synthetic Aβ₁₋₄₂ peptides (107761-42-2, Abcam Co., Ltd., Cambridge, UK) were dissolved in a solution of 1,1,1,3,3,3-hexafluoro-2-propanol (105228, Sigma-Aldrich Corporation, St. Louis, MO, USA), dried by evaporation, and then stored at a temperature of − 80 °C. In preparation for use, the desiccated peptides were dissolved in anhydrous DMSO to a concentration of 5 mM. To prepare Aβ1–42 oligomers, 5 mM Aβ1–42 in DMSO was diluted to 100 µM with ice-cold, phenol red-free Ham’s F-12 medium. The mixture was vortexed for 30 s and incubated at 4 °C for 24 h (Supplementary Fig. 1).

The SH-SY5Y cells (CL-0208) were purchased from Procell Life Science & Technology Co., Ltd. (Wuhan, China), and cultivated in a culture medium prepared specifically for SH-SY5Y (CM-0208, Procell). The cells were incubated at 37 °C in a humidified environment with 5% CO2 for 24 h. Next, the cells were exposed to 5 µM Aβ₁₋₄₂ oligomers for a duration of 24 h.

The HMC3 cells (CL-0620) were purchased from Procell (Wuhan, China), and cultivated in a culture medium specifically designed for HMC3 cells (CM-0620, Procell). The cells were incubated at 37 °C in a humidified environment with 5% CO2 for 24 h. Furthermore, the cells were exposed to a concentration of 0.1 µM LPS (L5293, Sigma-Aldrich) for a duration of 24 h.

Total RNA was extracted from cells using TRIzol™ Reagent (Thermo Fisher Scientific) according to the manufacturer’s protocol. The isolated RNA was quantified and assessed for purity using a spectrophotometer. cDNA was synthesized from 1 µg of RNA using the HiScript^® II Q RT SuperMix for qPCR (+ gDNA wiper) (R223-01, VAZYME). Quantitative PCR (qPCR) was performed using ChamQ Universal SYBR qPCR Master Mix (Q711-02, VAZYME) on a 7500HT Fast Real-Time PCR System (Applied Biosystems; Thermo Fisher Scientific). The specific primers utilized for each gene are listed in Table 1. Gene expression levels were normalized to GAPDH as an internal control and quantified using the 2^−ΔΔCt method.

Cells were transduced with PLV-puro vectors (OriGene) containing the full-length open reading frame (ORF) of SLAMF8. Stable SLAMF8-expressing cell lines were generated after a minimum of two weeks of puromycin selection (58-58-2, Sigma-Aldrich). For gene knockdown, SLAMF8 expression was suppressed using SLAMF8-specific shRNA, while a non-targeting shRNA was used as a control (Generalbiol, Anhui, China). Additionally, siRNA targeting NINJ2 (siNINJ) was introduced into cells using Lipofectamine 2000 (Life Technologies). The specific target sequences are provided in Table 1. After 48 h of transfection or shRNA expression, cells were harvested for downstream analyses, and transfection efficiency was assessed using RT-qPCR and Western blot analysis.

Cell lysates were prepared using RIPA buffer (P0013B, Beyotime, Shanghai, China), followed by centrifugation at 12,000 x g for 15 min at 4 °C to clear the debris. The protein content in the supernatants was quantified using a BCA protein assay. Equal protein amounts were then subjected to SDS-PAGE and subsequently transferred onto PVDF membranes. The membranes were blocked with 5% non-fat milk in TBST for 1 h at room temperature, followed by an overnight incubation at 4 °C with the primary antibodies. After washing, HRP-conjugated secondary antibodies were applied for 1 h at room temperature. Detection of protein bands was carried out using a Bio-Rad imaging system (Hercules, CA, USA), and ImageJ software was used for densitometric analysis, with normalization to β-actin levels. The specific primary antibodies used are listed in Table 2.

HMC3 cells were treated with 1 µg/mL of lipopolysaccharide (LPS) for 24 h. After incubation, the supernatants were collected and stored at − 80 °C until analysis. The concentrations of IL-1β, IL-6, and TNF-α in the supernatants were measured using commercial ELISA kits specific for IL-1β (EK0394, Boster, Wuhan, China), IL-6 (EK0411, Boster, Wuhan, China), and TNF-α (EK0527, Boster, Wuhan, China), following the manufacturer’s protocols.

Intracellular reactive oxygen species (ROS) levels were measured using DCFH-DA (D6883, Sigma-Aldrich), while mitochondrial ROS levels were assessed using MitoSOXTM Red reagent (M36008↗, Thermo Fisher Scientific), following the method described by Kauffman et al. (2016). SH-SY5Y cells treated with 5 µM Aβ₁₋₄₂ were incubated with 10 µM DCFH-DA for 30 min at 37 °C in the dark to detect intracellular ROS. For mitochondrial ROS detection, cells were incubated with 5 µM MitoSOX™ Red reagent under the same conditions. After incubation, cells were washed with PBS and imaged using a confocal laser-scanning microscope (Leica, Wetzlar, Germany). DCFH-DA was excited at 488 nm, and emission was collected at 525 nm. MitoSOX™ Red was excited at 510 nm, with emission collected at 580 nm.

Cells were lysed in cell lysis buffer (Beyotime, Shanghai, China) on ice for 30 min, followed by centrifugation to clear debris. The supernatant was precleared with rProtein A/G Plus MagPoly beads (RM09008, ABclonal, Wuhan, China) and incubated overnight at 4 °C with antibodies against SLAMF8 or NINJ2. The immune complexes were captured using fresh rProtein A/G Plus MagPoly beads, washed, and eluted for analysis by SDS-PAGE and Western blotting.

SH-SY5Y and HMC3 cells were fixed with 4% paraformaldehyde (MA0192, Meilunbio, Dalian, China) for 20 min at room temperature, followed by permeabilization with 0.1% Triton X-100 (P0096, Beyotime, Shanghai, China) for 10 min. After blocking with 5% BSA in PBS for 1 h, cells were incubated with primary antibodies against SLAMF8 overnight at 4 °C. The next day, cells were thoroughly washed with PBS to remove unbound primary antibodies. They were then incubated with fluorescently labeled secondary antibodies for 1 h at room temperature in the dark to prevent photobleaching. Finally, cells were mounted with antifade medium and imaged using a fluorescence microscope.

Statistical analyses were performed using GraphPad Prism 9.0 (GraphPad Software, CA, USA) and SPSS 19.0 (SPSS Inc., Chicago, USA). Measurements between two groups (mean ± standard deviation) were compared using Student’s t-test. Comparisons among multiple groups were conducted using one-way analysis of variance (ANOVA). Two-way repeated measures ANOVA (Two-way RM-ANOVA) was used to compare different groups across multiple time points. A p-value of < 0.05 was considered statistically significant.

Transcriptome data from GSE122063↗, including frontal and temporal cortex tissue samples from 12 Alzheimer’s disease (AD) patients (3 males and 9 females) and 11 non-demented controls (5 males and 6 females) (Fig. 2a), were utilized for further analysis. Gene co-expression networks were constructed using the WGCNA package. (Fig. 2b), through which multiple co-expression modules were identified via gene clustering (Fig. 2c). Correlation analysis between gene modules and AD status revealed that the pink module (MEpink, p < 0.001, r = 0.53) was significantly positively correlated with AD (Fig. 2d). Using gene module membership (GM > 0.8) and gene significance (GS > 0.4) as criteria, we identified key genes within the pink module (Fig. 2e). From these genes, 52 upregulated differentially expressed genes were identified (Fig. 2f), and their interactions were predicted using the STING database (Fig. 2g). Further analysis using cytoHubba extracted the top 20 genes with the strongest interactions (Fig. 2h). GO enrichment analysis of these 20 genes revealed their predominant involvement in immune and inflammatory response pathways (Fig. 2i).

Additionally, differential expression analysis demonstrated that SLAMF8 is significantly upregulated in AD groups (Fig. 2j, k). To further explore the differential expression of SLAMF8, classic preclinical AD models were utilized, including Aβ₁₋₄₂-treated SH-SY5Y cells²⁰ and LPS-treated HMC3 microglia²¹. SLAMF8 was found to be highly expressed at both the protein (Fig. 2m,m1) and mRNA levels (Fig. 2n) in these models. This finding was further supported by immunohistochemistry (IHC) staining, which demonstrated a significant increase in SLAMF8 protein expression in the hippocampus of APP/PS1 mice compared to WT mice (Fig. 2l).

To explore the role of SLAMF8 in AD, we stably introduced SLAMF8 expression vectors or control vectors into AD cell models (Aβ₁₋₄₂-treated SH-SY5Y cells and LPS-treated HMC3 cells). Western blot and quantitative analysis confirmed that the protein levels of SLAMF8 were elevated in the SLAMF8 overexpression group compared to the control group. (Fig. 3a, a1). Overexpression of SLAMF8 significantly increased mitochondrial superoxide production and oxidative stress in Aβ₁₋₄₂-treated SH-SY5Y cells (Fig. 3b, b1, c) and promoted neuroinflammation in LPS-treated HMC3 cells (Fig. 3d). Conversely, knockdown of SLAMF8 in both Aβ₁₋₄₂-treated SH-SY5Y cells and LPS-treated HMC3 cells (Fig. 3e,e1) significantly inhibited mitochondrial superoxide production and oxidative stress in Aβ₁₋₄₂-treated SH-SY5Y cells (Fig. 3f, f1, g) and decreased neuroinflammation in LPS-treated HMC3 cells (Fig. 3h).

To gain deeper insights into the regulatory mechanisms of SLAMF8 in Alzheimer’s disease (AD), we assessed the expression levels of Toll-like receptor 4 (TLR4), phosphorylated p65 (phospho-p65), total p65, phosphorylated IκBα (phospho-IκBα), and total IκBα across different experimental groups. The findings revealed that SLAMF8 activated the TLR4/NF-κB signaling pathway (Fig. 4a, a1). Overexpressing SLAMF8 led to an upregulation of TLR4, an increased p-IκBα/IκBα ratio, and a higher phospho-p65/p65 ratio. Conversely, silencing SLAMF8 reversed these effects (Fig. 4b, b1). To assess whether SLAMF8’s role in AD is mediated through TLR4/NF-κB activation, we treated Aβ₁₋₄₂-exposed SH-SY5Y cells and LPS-stimulated HMC3 cells, with and without SLAMF8 overexpression, with TAK-242, a selective TLR4 inhibitor that blocks NF-κB activation. TAK-242 effectively reduced TLR4 expression (Fig. 4c, c1) and mitigated the pro-inflammatory and oxidative stress effects induced by SLAMF8, as indicated by reduced ROS levels (Fig. 4d, d1, e) and lower neuroinflammatory markers (Fig. 4f).

To further determine the downstream targets of SLAMF8, we queried the IntAct database (https://www.ebi.ac.uk/intact/home↗)²², which displays the interaction network of SLAMF8 with multiple proteins proteins (Fig. 5a). Among the potential interacting proteins, NINJ2 is a membrane protein involved in nerve regeneration and has been implicated in various neuroinflammatory processes. It plays a role in cellular adhesion and may influence signaling pathways related to inflammation and neurodegeneration. Additionally, NINJ2 was reported to potentially interact with SLAMF8, making it the most interesting candidate for interaction with SLAMF8²³.

Molecular docking analysis indicated a possible interaction between SLAMF8 and NINJ2 (Fig. 5b). Immunofluorescence analysis showed that SLAMF8 and NINJ2 were partially colocalized in SH-SY5Y cells, HMC3 cells (Fig. 5c), and the APP/PS1 mouse hippocampus (Supplementary Fig. 2). To confirm the interaction between SLAMF8 and NINJ2, co-immunoprecipitation (Co-IP) experiments were conducted on Aβ₁₋₄₂-treated SH-SY5Y cells and LPS-treated HMC3 cells. More NINJ2 was detected by the anti-SLAMF8 antibody compared to the IgG control, and more SLAMF8 was detected by the anti-NINJ2 antibody (Fig. 5d), indicating a direct interaction between SLAMF8 and NINJ2. Interestingly, although SLAMF8 overexpression upregulated NINJ2 protein levels (Fig. 5e, e1), it did not affect NINJ2 mRNA levels (Fig. 5f). Given this, we next investigated whether SLAMF8 influences the stability of NINJ2. Cells transfected with shSLAMF8 or control shNC were treated with cycloheximide, a protein synthesis inhibitor. The results indicated that the presence of SLAMF8 in these cells enhanced NINJ2 stability (Fig. 5g, g1).

To explore the impact of NINJ2 on SLAMF8-mediated neuroinflammation and oxidative stress, we transfected AD cell models with SLAMF8 or control vectors, followed by co-transfection with siRNA targeting NINJ2. The knockdown of NINJ2 nullified the enhancing effect of SLAMF8 on ROS production in Aβ₁₋₄₂-treated SH-SY5Y cells (Fig. 6a, a1,b), as well as the neuroinflammatory response in LPS-treated HMC3 cells (Fig. 6c). We next investigated whether SLAMF8 activates the TLR4/NF-κB signaling pathway via NINJ2 mediation. Consistent with our expectations, knockdown of NINJ2 suppressed p65 phosphorylation in these cells (Fig. 6d, d1). Taken together, these findings indicate that SLAMF8’s pathogenic role and its activation of the TLR4/NF-κB signaling pathway are at least partly reliant on NINJ2 expression in AD cell models.

To investigate the effect of SLAMF8 on learning tasks and hippocampus-dependent memory, we utilized the Morris Water Maze (MWM) to assess spatial memory in mice after a 14-day adaptation period. During the 3-day cue training, all experimental groups demonstrated a progressive decrease in both escape latency and swimming distance (Fig. 7a, c). However, the APP/PS1 (SLAMF8+/+) group showed a slower and less pronounced reduction in these metrics compared to the other groups. Notably, the APP/PS1 (SLAMF8+/+) mice exhibited significantly longer escape latency compared to control mice, while APP/PS1 (SLAMF8−/−) mice displayed a significant reduction in escape latency (Fig. 7b). Additionally, the learning curve, analyzed by path length, revealed significant effects of both time (indicating learning performance over days) and group differences, as determined by repeated-measures ANOVA (Fig. 7c).

AUC analysis demonstrated that APP/PS1 (SLAMF8+/+) mice traveled significantly longer distances compared to control mice during the 3-day acquisition phase, whereas APP/PS1 mice with SLAMF8 gene deficiency (SLAMF8−/−) significantly reduced this increase in path length (Fig. 7d). In the probe trial, APP/PS1 (SLAMF8+/+) mice spent significantly less time in the target quadrant searching for the platform compared to controls, while APP/PS1 mice with SLAMF8 gene deficiency exhibited an increased duration in this quadrant (Fig. 7e). Additionally, during the probe phase, no significant differences were observed in the number of crossings over the area where the platform was previously hidden among the three groups (Fig. 7f). During the probe phase, APP/PS1 (SLAMF8+/+) mice exhibited a considerable delay in escape latency compared to other groups (Fig. 7g). Despite this delay, one-way ANOVA indicated that there were no significant variations in total swimming distance (Fig. 7h) or swimming speed (Fig. 7i) among the groups on the fourth day of the MWM test.

Furthermore, Western blot analysis of mice brain tissue revealed that both NINJ2 expression and p65 phosphorylation were significantly reduced in the APP/PS1 (SLAMF8−/−) group compared to the APP/PS1 group (Fig. 7j), supporting the mechanisms observed in vitro.