Male C57BL/6J mice (6–8 weeks of age, 20–25 g) were purchased from the Tengxin Biotechnology Company (Chongqing, China). All mice were fed with standard food and sufficient water in cages under a controllable circumstance (temperature 20°C–23°C, humidity 55%–65%, and a 12h light/dark cycle). The animal program was authorized by the Institutional Animal Care and Use Committee of Guizhou Province People’s Hospital (Guizhou, China).

SAE was established by CLP as previously described (21, 22). Briefly, the mice were anesthetized with isoflurane, and a 2 cm-wide midline laparotomy was performed to expose the cecum. The cecum was ligated with a 4.0 silk suture at 1/4 distance to the end below the ileocecal flap and then punctured with a 22G sterilized needle. Intestinal contents were pushed out through the perforation site. The cecum was then returned to the peritoneal cavity, and the laparotomy was closed with 4.0 silk sutures. All animals received a dose of antibiotic (primaxin, 0.5 mg/mouse in 200 uL of sterile saline) subcutaneously immediately at the end of surgery. All animals were fed in single cages with free access to food and clean water. In the sham surgery group, the mice were subjected to all surgical procedures, except for ligation or puncture.

Following anesthetization, the mice were placed in the stereotaxic apparatus (RWD life science Co. Ltd., China). In accordance with previous reports (10), guide cannulas (RWD life science Co. Ltd., China) were planted into the right lateral ventricle (coordinates: 0.6 mm posterior to the bregma, 1.5mm lateral, 2mm depth from the dura) and secured to the skull with dental cement. All the mice were allowed to recover in clean cages for 7 days. And handled daily to check the guide cannula and familiarize them with the investigators.

A MC stabilizer, comolyn was dissolved in sterile saline with a final concentration of 10 mg/mL. For the cromolyn treatment group, 2uL of cromolyn was administered intracerebroventricular (i.c.v) at 30 min before the CLP surgery and every 12h after surgery. The other group received 2uL of sterile saline.

For IHC analyses, brain tissue sections were incubated for 1 h in 10% bovine serum albumin (BSA) with 0.3% Triton X-100 in 0.01 M PBS and then overnight with anti-tryptase (Abcam, UK) at 4°C. Tissue sections were washed and incubated with anti-rabbit secondary antibody for 1 h at room temperature. Immunostaining was visualized with DAB and counterstained with hematoxylin (Solarbio, China). The slides were scanned using a digital camera (Olympus, Japan). For IF analysis, brain sections or cell slides were incubated overnight with primary antibodies: ZO-1, occludin and Ki67 (Abcam, UK) in blocking solution at 4°C, followed by goat anti-mouse Alexa Fluor Plus 488 secondary antibody (Invitrogen, USA), and then covered with mounting medium containing DAPI (Invitrogen, USA) in a dark environment. Images were captured under a fluorescent microscope (Carl Zeiss, Germany). Images were calibrated for the optical density and then measured the areas and integrated densities. Changes in immunofluorescent intensity were shown (mean fluorescence intensity = integrated densities/areas)

Brain paraffin sections and cell smears were stained with 1% TB solution (Solarbio, China) following the manufacturer’s protocol. MCs were detected as violet color or different levels of blue color as observed under a microscope. The number of MCs and their degranulation status were determined by their abnormal morphology and the presence of extracellular cytoplasmic granules under a light microscope using 40x or 100x magnification (Olympus, Japan).

The mice were euthanized at 24 h after surgery. The whole brain was instantly collected and weighed as the wet weight, and then dried at 100°C for 24 h to acquire the dry weight. The BWC was calculated using the formula: BWC = [(wet weight−dry weight)/(wet weight)]∗100%.

FS and EB were applied to evaluate BBB permeability as previously described (23). Briefly, 2% FS (Sigma, Germany) and 2% EB (Sigma, Germany) were respectively injected through the femoral vein at 4 mL/kg under anesthesia at 24 h after modelling. After 1 h, all animals were euthanized and transcardially perfused with 0.9% saline for 20 min to remove the intravascular dye. The hippocampus was then dissected from the brain, weighed separately, homogenized in 0.5 ml of 60% trichloroacetic acid and incubated at 4°C for 30 min. After centrifugation (12,000 rpm, 15 min), the FS concentration in the supernatant was measured at OD 440nm and the EB concentration was measured at OD 620nm by a microplate reader (BioTek, USA).

The Y maze consisted of three arms (regions I–III, 30 × 5 × 20cm) that converged to an equilateral triangular central area (region 0) (RWD life science Co. Ltd., China). One arm was chosen as the “start arm”, another as the “novelty arm”, and the third as the “other arm”. Firstly, the animal was placed in the “start arm” with its back to the center, and the “novel arm” was closed. The animal was allowed to explore freely between the two arms for 15 min (training). One hour after the first trial, the animal was placed in the “start arm” with its back to the center, and the “novelty arm” was opened. The animal was allowed to explore freely for 5 min, and its movement trace was recorded. The time spent in each arm of the mouse and the movement distance of the animal in each arm were used as indicators to evaluate spatial recognition and memory. The maze was cleaned with 75% ethanol between tests.

The locomotor and exploratory activities of mice were measured in an open-field apparatus. A mouse was gently placed in the center of a plastic chamber (50 × 50 × 40 cm) for 5 min, and its activities were automatically recorded using a video tracking system (RWD life science Co. Ltd., China). The total distance travelled and time spent in the center of the arena were recorded to assess the locomotor activity and spatial exploration of the mouse. The apparatus was cleaned with 75% alcohol between tests to remove any odor cues.

P815 and bEnd.3 cell lines were purchased from the Chinese Academy of Sciences Shanghai Cell Bank (Shanghai, China). The cells were cultured in Dulbecco’s modified Eagle’s essential medium (Gibco, USA) containing 10% fetal bovine serum (Gibco, USA) and 1% penicillin/streptomycin (Beyotime, China) under 37°C and 5% CO₂. In brief, over-expression plasmid for H1R was generated by polymerase chain reaction (PCR) and cloning into LentiORF pLEX-MCS vector using a forward primer (5′-ATAAGAATGCGGCCGCGCCACCATGAGCCTCCCCAATTCCTC-3′) and a reverse primer (5′-CCGCTCGAGTTAGGAGCGAATATGCAGAATTCTC-3′). Recombinant lentiviral particles were obtained from the transient transfection of 293T cells following a standard protocol. For transduction, bEnd.3 cells were cultured in six-well plates to 60% confluence and added with viral solutions containing 8 µg/mL polybrene (Solarbio, China). Over-expressing cells were selected under puromycin (Beyotime, China). H1R siRNA was purchased from Obio Technology (Shanghai, China) and transfected into bEnd.3 cells by using RNAi max (Invitrogen, USA) following the manufacturer’s protocol. After transfection for 6 h, the lipid and siRNA complexes were removed, fresh 10% serum medium was added and incubation was performed for 72 h.

For co-culture experiments, P815 cells (1 × 10⁵ cells) were planted in the upper chamber of a 12-well Transwell™ plate (Corning, USA) with complete medium and treated with cromolyn for 30 min. The P815 cell chambers were then transferred to a 12-well plate with bEnd.3 cells incubated in the lower chamber, The co-cultured cells were stimulated with LPS (1 μg/mL) for 2 or 24 h.

Expression levels of TNF-α, IL-6, IL-1β and IL-10 were quantified using a commercial ELISA kit from Boster Biological Technology (Wuhan, China). Histamine content was tested with an ELISA kit from Elabsecience (Wuhan, China). Tryptase content was tested with a detection ELISA kit from Meimian Biological Technology (Jiangsu, China). All procedures were carried out following the manufacturer’s instructions.

Proteins were extracted from homogenized cells or tissues in a lysis buffer (Solarbio, China) containing a protease inhibitor cocktail (Solarbio, China) and boiled in a SDS sample buffer (Beyotime, China) for 15 min. Equal amounts of protein per sample were separated by SDS-PAGE (8% or 10%) and electrotransferred to a polyvinylidene difluoride membrane (Milipore, USA). Membranes were blocked with 5% BSA (Beyotime, China) for 1h at room temperature, and rinsed three times (10 min/time) with 0.5% PBST (0.5mL Tween-20 in 1L 0.01 M PBS). Next, incubated with the specified primary antibodies at 4°C overnight, rinsed three times (10 min/time) with 0.5% PBST, and then 1 h with HRP-conjugated goat anti-rabbit IgG (detailed antibodies were shown in Supplementary Table 1 ), rinsed three times (10 min/time) with 0.5% PBST, followed by detection using an enhanced chemiluminescence kit (Milipore, USA).

Statistical analyses were performed with GraphPad Prism 9. All data were presented as mean ± SD and tested for normal distribution by Shapiro–Wilk test. For normally distributed data, differences between the two groups were compared by independent-sample t-tests. Meanwhile, differences among multiple means were assessed by one-way, two-way or repeated-measure of ANOVA, followed by a Bonferroni test whenever appropriate. The survival rate was analyzed by the Kaplan–Meier method and compared by the log-rank test. Statistical significance was set at p<0.05.

Singleton et al. suggested that investigators must rigorously control the distance of the cecum ligated to generate consistent mortality and inflammation data when utilizing the CLP model (24). Therefore, we first performed a pre-experiment with 25%, 50% and 75% lengths of the ligated caecum. As shown in Supplementary Figure 1A , the mice with 75% of ligated caecum have over 50% mortality rate within 7 days. In body weight, 75% of ligated caecum contribute to a significant drop than of other ligated lengths ( Supplementary Figure 1B ). In addition, expression levels of inflammation cytokines (TNF-α, IL-6, IL-1β and IL-10) in the hippocampus homogenate were also detected using ELISA, and the results showed that 75% of ligated caecum can also cause more inflammatory cytokines production than that of others ( Supplementary Figures 1C–F ). Therefore, 75% of ligated caecum was selected as the most appropriate ligated length to induce SAE for further studies.

To evaluate the neuroinflammatory response during SAE, we detected the expression of IL-6, IL-1β, TNF-α and IL-10 in the serum and hippocampus homogenate following CLP. As shown in Figure 1B , all cytokines, either pro-inflammatory cytokines (IL-6, IL-1β, TNF-α) or anti-inflammatory cytokines (IL-10), were substantially increased in the hippocampus tissues at 24 h after surgery (p<0.05). The expression levels of IL-6, IL-1β, TNF-α and IL-10 in serum were also significantly increased at 24 h after surgery (p<0.05, Supplementary Figure 2A ). In addition to the severe inflammatory response, the body weight of the CLP mice dropped seriously after surgery ( Figure 1C ). Moreover, several CLP mice did not survive from severe inflammatory response. The 7-day survival rates of the CLP mice decreased significantly compared with those of the sham mice (p<0.05, Figure 1D ). Behavioral manifestations in sepsis survivors were determined using open field trials and the Y maze. In the open field trial, the movement distance for a mouse to explore a new environment significantly decreased for the CLP mice (p<0.01, Figure 1E ). In terms of the time spent in the central region of the open field, the CLP mice spent a significantly shorter time than the sham mice (p<0.05, Figure 1F ). In the Y maze, the CLP mice showed a similarly poor performance because their percentage of total distance traveled in Zone C ( Figure 1G ), time spent in Zone C were remarkably decreased compared with those of the sham surgery mice (p<0.001, Figure 1H ). Additionally, the movement trace of mice has been shown in Supplementary Figure 2B . These findings demonstrated that the CLP mice showed notably cognitive deficits.

To check the activation of brain MCs in CLP-induced neuroinflammation, we evaluated MCs in the hippocampus sections at 24 h after surgery. These brain sections were incubated with 0.1% TB and tryptase antibody for MCs detection. The positive cells were stained purple or deep blue (black arrows) after TB staining. The number of positive cells was increased in the CA1 area of the SAE mice hippocampus compared with that of the sham control mice ( Figure 1I ). Quantification data for TB- and tryptase-positive cells are shown in Figure 1J . The expression of tryptase in hippocampus tissues was further determined using ELISA at 24 h after CLP surgery. The results showed that the expression level of tryptase increased significantly in the CLP mice (p<0.01, Figure 1K ). These data supported the overactivation of MCs in the hippocampus of mice with SAE.

To explore the specific role of MCs in SAE progression, we i.c.v. injected cromolyn, the most characteristic MCs stabilizer, at 30 min before the CLP surgery. The results showed that treatment with cromolyn alone did not affect the activation of MCs in the brain. However, under CLP surgery, cromolyn significantly prohibited the MCs activation in the hippocampus as indicated by TB and tryptase antibody staining ( Figure 2A with quantification in Figure 2B ). With the stabilization of MCs, pre-treatment with cromolyn mitigated the expression of intracerebral inflammatory cytokines (IL-6, IL-1β, TNF-α and IL-10) ( Figure 2C ) and improved the 7-day survival of the CLP mice ( Figure 2D ). In addition to the inflammation inhibition, pre-treatment with cromolyn remarkedly increased the movement distance and the time in center in open field trial ( Figures 2E, F ) and improved the performance of the CLP mice in the Y maze ( Figures 2G, H and Supplementary Figure 2C ), suggesting that this drug alleviated CLP-induced cognitive impairment. All these results implied that MCs play a crucial role in neuroinflammation and cognitive deficits in the CLP-induced SAE model, and cromolyn can limit the adverse neuroinflammation and cognitive outcomes caused by sepsis.

In this study, we assumed that MC activation mediates SAE progression through BBB disruption. To test our hypothesis, we assessed brain breakage by firstly measuring BWC. At 24 h, BWC was significantly increased in the CLP mice compared with that in the sham mice (p<0.0001, Figure 3A ). Cromolyn significantly reduced BWC in the CLP mice (p<0.05, Figure 3A ). FS and EB extravasation are widely used markers for BBB breakage. As shown in Figures 3B, C , the FS levels in the hippocampus and global brain tissue samples were similar in the sham and cromolyn groups. A marked increase in FS levels was observed in the hippocampus and brain tissue in the CLP mice compared with those in the sham mice, and this increase was notably reversed by cromolyn treatment. Additionally, EB permeability tests were performed to assess BBB damage. The results were similar to those from FS extravasation ( Figures 3D, E ), and representative sections for EB staining are shown in Figure 3F . These data suggested that MC activation increases BBB permeability in SAE mice, which is partially alleviated by cromolyn treatment.

The barrier function of BBB is mainly performed by BMVECs and their TJs. ZO-1, occludin and claudin-5 are integral membrane proteins composed of TJ strands and contribute to BBB integrity. Previous studies reported that matrix metalloproteinases (MMPs) can degrade the extracellular matrix (ECM) and destroy the TJ protein (25). Thus, we firstly investigated the expression of MMP and TJ proteins, including MMP2, MMP9, ZO-1, occludin and claudin-5, in the hippocampus using Western blot. As shown in Figures 4A, B , the protein expression of MMP2 and MMP9 increased significantly in the CLP group compared with that in the sham group (p<0.05). Meanwhile, the protein levels of ZO-1, occludin and claudin-5 were sharply reduced in the CLP group (p<0.05). Cromolyn treatment significantly decreased MMP expression and increased TJ protein levels in the CLP group. We then assessed the distribution of TJs between ECs in the hippocampus by immunofluorescence staining. The results showed that SAE-induced neuroinflammation disrupted the distribution of ZO-1 and occludin, spreading them in the form of a dotted appearance and weakening the fluorescence intensity. This finding demonstrates the possibility of losing the protein–protein interaction between TJ proteins might. Meanwhile, the sham group showed uniform staining of TJs ( Figures 4C, D ). Fortunately, Cromolyn spread in the form of a line and enhanced the fluorescence intensity. All these data revealed that MCs are involved in maintaining the integrity of TJs and their activation can lead to TJ cleavage, resulting in BBB disruption. Cromolyn treatment can ameliorate BBB dysfunction by protecting the TJs between BMVECs.

To specifically explore the relationships between MCs activation and sepsis, we performed in vitro experiments with P815 cells, a mouse MC line, and quantified the MC degranulation under LPS stimulation. Compound 48/80 (C48/80) is one of the most characteristic MCs activators. C48/80 treatment 007Agroup (10mg/mL) was used as the positive control. As shown in Figure 5A , 1mg/mL LPS significantly promoted the β-hexanosidase release rate in P815 cells compared with that in the control group as early as 2 h, the rate was continuously increased at 24 h. Additionally, cromolyn (10mg/mL) treatment significantly alleviated the LPS-induced excessive β-hexanosidase release rate. The morphological properties of degranulated MCs were investigated using the TB staining of the cell smears. As shown in Figures 5B, C , the number of degranulated P815 cells was significantly higher in the LPS and C48/80 groups than that in the control group at 2 h. A few “ghostlike cells” appeared in the LPS and C48/80 groups at 2 h (red arrows). Nearly all P815 cells were activated in the LPS and C48/80 groups at 24 h, and many “ghostlike cells” were present. Cromolyn significantly reversed the LPS-induced degranulation of P815 cells. Considering that the release of inflammatory factors is also an important marker of MC activation (26), we additionally evaluated the production of histamine, tryptase, TNF-α, IL-1β, IL-6 and IL-10 in P815 cells after LPS treatment for 24 h. The results showed that the production of histamine, tryptase, TNF-α, IL-1β, IL-6 and IL-10 significantly increased in the LPS group. This trend was partially reversed by cromolyn treatment ( Figures 5D–I ). These data suggested that LPS can successfully induce the degranulation of MCs and that cromolyn treatment can effectively stabilize MCs under LPS stimulation.

Given that the in vivo data indicate a relationship between MCs and BBB dysfunction under the SAE model, we attempted to clarify the mechanism of this correlation. Considering that BMVECs and the TJs between them are the most critical factors for BBB integrity, we further explored the effect of activated MCs on BMVECs in vitro. The mouse MC cell line P815 and mouse BMVEC cell line bEnd.3 were co-cultured in the Transwell system. Tunel assay and Ki-67 immunostaining were applied to investigate the proliferation and apoptosis of BMVECs under various treatments and examine the effect of MCs activation on the cell viability of BMVECs. The results showed an increase in Tunel-positive cells and a decrease in Ki67-positive cells among the bEnd.3 cells under LPS stimulation ( Figure 6A with quantification in Figure 6B ), indicating that LPS could inhibit proliferation and induce apoptosis. In the P815 and bEnd.3 co-culture system, the number of Tunel-positive bEnd.3 cells were further increased, and Ki67-positive bEnd.3 cells were reduced, demonstrating that MCs can promote BMVEC dysfunction under LPS stimulation. To further confirm this conclusion, we pre-treated the co-culture system with the MCs stabilizer, cromolyn, before LPS stimulation and examined the proliferation and apoptosis of bEnd.3 cells. With the treatment of cromolyn, the viability of bEnd.3 cells were partially rescued.

To examine the effect of MC activation on the cell junction of BMVECs, we investigated the expression of MMP and TJ proteins, including MMP2, MMP9, ZO-1, occludin and claudin-5 using Western blot. As shown in Figures 6C, D , the protein expression levels of MMP2 and MMP9 in bEnd.3 cells were up-regulated and those of ZO-1, occludin and claudin-5 were down-regulated by LPS stimulation. In the P815-bEnd.3 co-culture system, these protein expression changes were seriously augmented. Immunofluorescence staining for ZO-1 and occludin showed that their fluorescence intensity was weak in the LPS-treated cells, demonstrating the broken EC junctions ( Figure 6E with quantification in Figure 6F ). Consistent with the in vivo results, cromolyn treatment for the co-culture system decreased the protein level of MMPs and increased the protein expression of TJs. All these results revealed that the activation of LPS-induced MCs can affect the cell viability and cell junctions of BMVECs, thus disrupting the normal functions of BMVECs.

In human gingival fibroblasts, previous studies have confirmed that histamine promotes the expression of the receptors TLR2 and TLR4 and amplifies sensitivity to LPS treatment (27). Approximately 50% of brain histamine is produced by MCs, and H1R is highly expressed in EC. Thus, we hypothesized that MC induce BMVEC dysfunction through the histamine/H1R pathway. To test this hypothesis, we firstly checked the expression of H1R, TLR2, TLR4 and downstream mitogen-activated protein kinases (MAPK) signaling proteins by Western blot. As shown in Figures 7A, B , LPS treatment for 24h up-regulated H1R, TLR2 and TLR4 expression and increased the phosphorylation of P38, P42/44 and JNK in bEnd.3 cells. These protein expression changes were severely augmented in the P815-bEnd.3 co-culture system but ameliorated by cromolyn pre-treatment. To determine whether MCs promote BMVECs inflammation response mediated by histamine/H1R, we overexpressed H1R in bEnd.3 cells ( Figure 7C ). After exposure to LPS, the expression of TLR2, TLR4 and the phosphorylation of P38, P42/44 and JNK markedly increased in the bEnd.3 cells over-expressing HIR ( Figure 7D with quantification in Figure 7E ). To observe the inflammatory effect of P815-bEnd.3 cells co-cultures, we detected IL-6, IL-1β, TNF-α and IL-10 in the supernatant using ELISA and found that the abundance of cytokines was gravely increased in the bEnd.3 cells over-expressing HIR ( Figure 7F ). We used the specific siRNA to silence the expression of H1R in bEnd.3 cells. Adverse effects were found as shown in Figures 8A–C . These data supported that histamine/H1R mediates the interaction between MCs and BMVECs and amplifies the LPS-induced inflammatory response in BMVECs via the TLR2/4-MAPK signaling pathway.