A Transwell coculture system was utilized to construct an in vitro BBB model. The hCMEC/D3 and HBVP cells were seeded into the upper and lower chambers, respectively, of a six-well Transwell plate at a 1:1 ratio and cultured until they reached more than 85% confluence. After digestion and counting, the hCMEC/D3 and HBVP cells were added to their respective chambers at a 1:1 volume ratio, and the culture medium was added to the appropriate level. After 48 h of coculture, a 4-h liquid surface leakage test was performed to evaluate the BBB model’s establishment. After changing the culture medium, if the liquid surface difference was greater than 0.5 cm, the culture continued for 4 h. If the liquid surface difference remained unchanged, the in vitro BBB model was considered to be preliminarily established.

Lipopolysaccharide (LPS), a constituent of the outer membrane of Gram-negative bacteria, is commonly used to induce neuroinflammation (Yu et al., 2024). After successfully establishing the in vitro BBB model, LPS was used to induce BBB damage. Logarithmically growing hCMEC/D3 and HBVP cells were seeded into 96-well plates. After 8 h of incubation, 100 μL of a diluted LPS solution (initial concentration of 100 mg/mL) was added to each well at final concentrations of 0.5, 1, 1.5, and 2 μg/mL. Cell viability was evaluated at 24, 48, and 72 h to determine the optimal culture time and concentration.

The LPS-induced hCMEC/D3 and HBVP cells were treated with 5%, 10%, and 20% XXD-medicated serum, and cell viability was calculated to screen for the optimal concentration. The successfully established in vitro BBB models were subdivided into the following groups: control group, LPS group, 24-h XXD-medicated serum group, 48-h XXD-medicated serum group, 72-h XXD-medicated serum group, and blank serum group.

Logarithmically growing hCMEC/D3 and HBVP cells were modeled, grouped, administered, and incubated as described above. Following this, Western blotting experiments were performed. The cells were collected, centrifuged, lysed, and centrifuged again to collect the supernatant. Protein levels were determined via the BCA assay. Each well was loaded with 20 μg of total protein. Proteins were separated by SDS-PAGE (6% stacking gel at 60 V, 10% separating gel at 90 V), followed by wet transfer at 300 mA for 90 min. Subsequently, the samples were blocked using 5% skim milk on a shaker for 1 h and subsequently incubated with primary antibodies (antibodies against P-glycoprotein [P-gp], cannabinoid receptor types 1 and 2 [CB1 and CB2], and major facilitator superfamily domain-containing protein 2a [Mfsd2a]) at a 1:5000 dilution at 4°C overnight with shaking. Following three washes in TBST, secondary antibodies were applied at room temperature for 90 min. Subsequently, TBST washing was performed three times, ECL reagent was used for darkroom exposure, and band gray values were analyzed using ImageJ software (v 1.8.0).

The in vitro BBB model was incubated, then removed, digested, and centrifuged. RNA extraction was performed with TRIzol, followed by cDNA synthesis using a reverse transcription kit. RT-PCR was conducted according to the manufacturer’s instructions. The amplified products were separated via 1% agarose gel electrophoresis and imaged using an image analyzer. Expression levels of mRNA were determined using the 2^−ΔΔCT method. Primer details are provided in Table 4.

The incubated in vitro BBB model was extracted post-incubation, PBS-washed, and fixed with 4% paraformaldehyde. Membrane permeabilization was achieved using Triton X-100, followed by BSA blocking (37°C, 1 h). Sequential antibody incubations were performed: primary (overnight) and secondary (1 h). Nuclear counterstaining employed DAPI, and the samples were subsequently imaged using a fluorescence microscope at a magnification of ×20.

SAMR1 mice were designated as the control group (Control), while SAMP8 mice were randomly divided into the model group (Model), the probiotics group (Probiotics), the high-dose Xixin Decoction granules group (H-XXD), the medium-dose Xixin Decoction granules group (M-XXD), and the low-dose Xixin Decoction granules group (L-XXD), with eight mice in each group. The dosages were calculated based on the equivalent clinical dosages and the body surface area of the animals (Huang et al., 2004). The probiotics group received a dose of 0.39 g·kg^-1·d^-1, the M-XXD group received a dose of 2.535 g·kg^-1·d^-1, the H-XXD group received twice the equivalent clinical dose at 5.07 g·kg^-1·d^-1, and the L-XXD group received half the equivalent clinical dose at 1.2675 g·kg^-1·d^-1. The treatment was administered continuously for 8 weeks. The drugs were dissolved in distilled water to prepare solutions of the appropriate concentrations. Each treatment group received the drugs by gavage daily at the designated dose, with a volume of 10 mL·kg^-1·d^-1. Equivalent volumes of distilled water were administered to both the control and model groups via gavage.

The Morris water maze test was performed for six consecutive days after administration. The initial 5 days consisted of place navigation trials. Mice were released into the water pool from four different quadrants each day, facing the wall. The escape latency, defined as the time required to find the submerged platform, was recorded with a limit of 60 s per trial. Upon reaching the platform, mice were permitted a 10-s rest period. If mice were unable to find the platform within the allotted time, they were gently guided to it, allowed a 10-s rest, and assigned a maximum escape latency of 60 s. On the sixth day, a spatial probe test was conducted by removing the hidden platform and releasing mice from the first quadrant, facing the pool wall. They were allowed to swim freely for 60 s. During this trial, the time spent in the target quadrant and frequency of crossing the former platform site were recorded.

Following the final behavioral test, four mice per group were randomly chosen and injected with a 2% Evans blue solution (4 mL/kg) through the tail vein. Specific methods referred to the published literature (Schoch et al., 2014). Upon turning blue in limbs and eyes, the mice were anesthetized via intraperitoneal injection of 0.5% sodium pentobarbital (0.1 mL/g). The specific methods were performed with reference to the literature (Levin-Arama et al., 2016). Cardiac perfusion was performed through the left ventricle with physiological saline until the liver and kidneys turned white and clear fluid flowed from the right atrium. This was followed by rapid perfusion with 4% paraformaldehyde precooled to 4 °C, which was maintained at a slow and steady rate for 10–15 min. The brain tissue was subsequently collected, fixed in 4% paraformaldehyde (4°C, 12 h), embedded in OCT, rapidly frozen at −20°C, and sectioned coronally at 20 μm. Evans blue extravasation was observed under a laser confocal microscope and quantified using ImageJ 1.8.0 software.

Hippocampal tissue was precooled on ice and homogenized in nine times the volume of the PBS solution, and the supernatant was collected. Following the instructions provided with the ELISA kit, the levels of Aβ_1-42, NF-κB p65, NLRP3, TNF-α, and IL-1β were measured. Absorbance at 450 nm was assessed using an enzyme-labeled instrument.

Anesthesia was administered to the mice via intraperitoneal injection of 0.5% sodium pentobarbital. A thoracic incision exposed the heart, allowing for left ventricular needle insertion and right atrial appendage excision. Perfusion commenced with physiological saline until clear fluid emerged from the oronasal cavities, followed by 4% paraformaldehyde perfusion until complete body rigidity. The intact brain was extracted and underwent a series of preparatory steps: 24-h fixation in 4% paraformaldehyde, ethanol dehydration, xylene clearing, and paraffin embedding. Five-micrometer sections were prepared and subsequently washed thrice with PBS before blocking with 3% bovine serum albumin (37°C, 30 min). Primary antibody incubation (targeting P-gp, CB1, CB2, RAGE, IBA1, TREM2, TLR1, and TLR2) was conducted overnight at 4°C. Following PBS washing, sections were incubated with fluorescent secondary antibodies (37°C, 1 h). After additional PBS washes, DAPI staining was performed, and sections were mounted using an anti-fluorescence quenching medium. Imaging was performed using confocal microscopy at a magnification of ×40, and relative fluorescence intensity was quantified with ImageJ software (version 1.8.0).

Hippocampal tissue was extracted, lysed, and homogenized ultrasonically on ice. Post-lysis centrifugation at 12,000 × g for 15 min at 4°C, the supernatant was collected for BCA protein quantification. Each well was loaded with 20 μg of total protein. Proteins underwent SDS-PAGE separation (6% stacking gel, 60V; 10% separating gel, 90V) and wet transfer (300mA, 90min). Membranes were blocked with 5% skim milk (1h, shaking) and incubated overnight with primary antibodies (anti-P-gp, CB1, CB2, RAGE, IBA1, TLR1, TLR2, TREM2, Mfsd2a, MRP2, LRP1, and CMPK2) at 4°C under agitation. After TBST washing, secondary antibody incubation was performed (90min, room temperature). ECL-based chemiluminescence detection was conducted in darkness. Band intensities were analyzed using ImageJ (v1.8.0).

LPS can disrupt the BBB through various pathways, and damage to the BBB can contribute to the progression of a range of diseases, with AD being a notable example (Peng et al., 2021). Therefore, this study utilized LPS to establish a BBB inflammation model. Optimal LPS concentration was determined through cell viability assessment using the MTT assay. The results showed that, after treating hCMEC/D3 and HBVP cells with 0.5–2 μg/mL LPS, cell viability exhibited a notable decline in comparison to the control group and exhibited a negative correlation with the LPS concentration. As high concentrations of LPS could cause irreversible cell damage, 1 μg/mL LPS was selected for subsequent experiments. On comparing the cell viability of hCMEC/D3 and HBVP cells exposed to 1 μg/mL LPS for 24, 48, and 72 h, it was found that cell viability in the 48-h and 72-h groups exhibited no statistically significant difference from that in the 24-h group (Figures 2A, B). Therefore, a 24-h treatment with 1 μg/mL LPS was used as the modeling condition.

After the administration of 1 μg/mL LPS to hCMEC/D3 and HBVP cells for 24 h, different concentrations of XXD-medicated serum were added for co-incubation. The MTT assay results indicated that 1 μg/mL LPS significantly inhibited cell proliferation, while 10% XXD-medicated serum significantly reversed this effect and increased cell viability (Figures 2C, D). Additionally, the 15% concentration of XXD-medicated serum did not show any further improvement in cell viability. This observation suggests that the 15% concentration may have reached a saturation point of the pharmacological effect or potentially introduced additional adverse effects, thereby failing to further enhance cell viability. Therefore, 10% XXD-medicated serum was chosen for subsequent experiments, as it exhibited the best efficacy in restoring cell viability while avoiding the potential drawbacks associated with higher concentrations.

The BBB can be described as a highly evolved CNS microvascular functional unit, comprising CNS endothelial cells, neural progenitor cells, pericytes, astrocytes, and other neural and immune cells, in addition to various functional protein structures (Obermeier et al., 2013). These components form the neurovascular unit (NVU) that maintains BBB function (Huang et al., 2020). Pericytes envelop the abluminal surface of cerebral vascular walls, including capillaries, precapillary arterioles, and postcapillary venules. The highest pericyte coverage rate is seen in the neural tissue among the capillary beds of different organs (Allt and Lawrenson, 2001). Pericytes are crucial in promoting vascular stability and maintaining BBB integrity (Winkler et al., 2011). Therefore, this study employed hCMEC/D3 and HBVP cells to construct an in vitro BBB model.

P-gp, an ABC transporter encoded by the human MDR1 gene, serves as a crucial efflux mechanism in the BBB. It inhibits substrate drug entry into the brain while promoting the elimination of endogenous molecules, including Aβ—a hallmark pathological feature of AD (Pyun et al., 2022). This study revealed that, relative to the control group, LPS treatment significantly downregulated both mRNA and protein expression of P-gp, indicating that P-gp expression is reduced when the BBB is damaged. Relative to the LPS group, the XXD-medicated serum upregulated P-gp mRNA and protein expression after 24, 48, and 72 h of culture, with the most significant effect observed at 48 h (Figure 3), suggesting that XXD can improve BBB transport dysfunction by upregulating P-gp expression.

Endocannabinoids, originating from dietary omega-3 and omega-6 polyunsaturated fatty acids, are endogenous bioactive lipids. The endocannabinoid system (ECS) is characterized by two primary receptors: CB1 and CB2. CB1 is predominantly expressed in CNS presynaptic regions (Zou and Kumar, 2018). It modulates cytokine release both within and outside the CNS and immune cell migration, with expression levels influenced by cell activation and stimulus type (Paul et al., 2012). CB2 is predominantly expressed in peripheral immune cells and brain tissue (Nagarkatti et al., 2009). Primarily serving an immunomodulatory function, it regulates immune cell migration and cytokine release both within and beyond the immune system (Massi et al., 2006). During neuroinflammation or BBB dysfunction, P-gp regulates ECS-related receptor protein expression, thereby influencing the inflammatory process. Regulating the P-gp/ECS axis can alleviate BBB dysfunction and neuroinflammation. Mfsd2a, a member of the major facilitator superfamily (MFS), plays a critical role in BBB integrity and DHA transport. Mfsd2a-deficient mice exhibit significantly lower brain DHA levels, resulting in neuronal loss and cognitive impairment due to DHA’s crucial role in brain development and maintenance. Mfsd2a inhibits CNS endothelial cell transport and mitigates BBB damage. Consequently, it has gained prominence in neurological disease research (Huang and Li, 2021).

This study demonstrated that, relative to the control group, LPS treatment upregulated CB1 protein expression while downregulating both CB2 and Mfsd2a expression. indicating that neuroinflammation can lead to an abnormal ECS as well as abnormal Mfsd2a expression, resulting in BBB dysfunction. Compared with the LPS group, XXD-medicated serum downregulated CB1 expression and upregulated CB2 and Mfsd2a expression after incubation for 24, 48, and 72 h (Figure 4), suggesting that XXD can improve BBB dysfunction by modulating the expression of the transport-related protein Mfsd2a and the P-gp/ECS axis.

SAMP8 mice serve as an optimal AD model, demonstrating age-associated learning and memory impairments, in addition to exhibiting the majority of pathological characteristics associated with AD pathogenesis, including oxidative stress, inflammation, and Aβ deposition. Therefore, SAMP8 mice aid in visualizing the effects of AD and provide an effective method for finding new therapeutic targets (Liu et al., 2020).

The Morris water maze is a commonly used behavioral test for assessing spatial cognition in rodents (Vorhees and Williams, 2006). This study investigates, compared to SAMR1 controls, SAMP8 mice exhibited significantly prolonged escape latencies over five consecutive days, along with decreased target quadrant dwell time and crossings, indicating marked spatial cognitive impairment. High-dose XXD treatment, relative to untreated SAMP8 mice, significantly reduced escape latencies throughout the 5-day period and increased target quadrant residence time. Medium- and low-dose XXD groups also demonstrated significantly shortened escape latencies (Figures 5A–C). These results suggest XXD’s potential to ameliorate spatial cognitive deficits in SAMP8 mice.

Evans blue, a widely utilized azo dye, is often used for tracing and observing BBB integrity. Under normal conditions, Evans blue, which is bound to plasma albumin, is unable to pass through the BBB and thus is incapable of staining the nervous system. However, if the BBB is disrupted, Evans blue is able to penetrate and cause staining (Liu et al., 2018). In this study, Evans blue extravasation was markedly increased in the brain tissue of SAMP8 mice, indicating BBB damage and increased permeability. Compared with the SAMP8 group, Evans blue extravasation in the brain tissue of XXD-treated groups significantly decreased (Figures 5D, E). Therefore, XXD may reduce BBB permeability by improving its ultrastructure, although the specific mechanisms require further investigation.

As the primary CNS immune cells, microglia are pivotal in Alzheimer’s disease (AD). Upon activation, they can adopt either the neurotoxic pro-inflammatory M1 phenotype or the neuroprotective M2 state (Xie et al., 2021). Studies have shown that Aβ stimulation can upregulate the activity marker CD68 and M1 markers TNF-α, IL-1β, and CD86 in primary microglia (Shi et al., 2017). Injection of Aβ into the mouse hippocampus also increases levels of M1 microglia (Xie et al., 2020). IBA1 is a specific calcium-binding protein participating in microglial remodeling, regulating their migration, membrane ruffling, and phagocytosis, and is commonly used to assess microglial activation (Hendrickx et al., 2017).

TREM2 is specifically expressed in microglia and has anti-inflammatory effects (Qin et al., 2021). Suppression of TREM2-mediated signaling in microglial cells enhances the expression of TNF-α and nitric oxide synthase 2 (NOS2) genes. Conversely, elevated TREM2 levels lead to decreased transcription of TNF-α, IL-1β, and NOS2 (Takahashi et al., 2005). TLRs are crucial components of the neuroimmune system, exerting a pivotal influence on microglial activation and polarization. Activation of TLRs leads to NF-κB upregulation, promoting the synthesis of inflammatory mediators including cytokines (TNF-α, IL-1β, IL-6) and enzymes (iNOS, COX2). TLR2 forms a heterodimer with TLR1, enhancing pro-inflammatory cytokine production via the MyD88-dependent pathway and NF-κB activation. Studies show that increased TREM2 expression decreases specific TLRs (2, 4, 6) and pro-inflammatory mediators (IL-1β, TNF-α, IL-6) in microglia, thereby reducing Aβ_1-42-induced neuroinflammatory responses (Long et al., 2019).

NLRP3 is a core component of the NLRP3 inflammasome, widely distributed in the CNS and highly expressed in microglia. NLRP3 recognizes exogenous or endogenous PAMPs and DAMPs, activates the downstream transcription factor NFκB, phosphorylates and degrades IκBα, facilitates NF-κBp65 nuclear translocation, initiating target gene transcription and upregulating NLRP3, IL-1β, and IL-18 precursor expression. Under PAMP and DAMP stimulation, the NLRP3 inflammasome assembles, releasing mature IL-1β and IL-18 and activating Gasdermin D, which causes microglial membrane rupture and the further release of more inflammatory factors (Hanslik and Ulland, 2020). CMPK2 is a newly discovered upstream factor of the NLRP3 inflammasome. It has been demonstrated that CMPK2 knockdown inhibits NLRP3 inflammasome activation (Chen et al., 2022).

The present research demonstrated that TREM2 expression in the hippocampus of SAMP8 mice was decreased, while IBA1, TLR1, and TLR2 expression significantly increased. Furthermore, the average fluorescence intensity of TMER2 in the mouse hippocampal CA1 region decreased, while the average fluorescence intensities of IBA1, TLR1, and TLR2 increased. Therefore, reduced TMER2 expression in the hippocampal region of SAMP8 mice can upregulate IBA1 expression and promote the overexpression of TLR1 and TLR2, thereby mediating microglia-related neuroinflammation. In addition, CMPK2 protein expression and NLRP3, NF-κB p65, COX-2, TNF-α, and IL-1β levels in hippocampal tissue were significantly increased, suggesting that microglia-related neuroinflammation in the brains of SAMP8 mice may be achieved through the CMPK2/NLRP3 and TLRs/NF-κB pathways. Compared with the SAMP8 group, in all XXD-dose groups, TMER2 expression in the mouse hippocampal region increased (Figure 8D), while IBA1, TLR1, and TLR2 expression significantly decreased (Figures 8A–C). The mean fluorescence intensity of TMER2 in the hippocampal CA1 region increased (Figure 8H), while the average fluorescence intensities of IBA1, TLR1, and TLR2 were reduced (Figures 8E–G). CMPK2 protein expression and NLRP3, NF-κB p65, COX-2, TNF-α, and IL-1β levels in hippocampal tissue were significantly decreased (Figures 9A–G). Thus, XXD may inhibit microglial activation by upregulating TMER2 expression, thereby inhibiting the activation of the CMPK2/NLRP3 and TLRs/NF-κB pathways, ultimately leading to reduced neuroinflammation.