Inhibition of NLRP3 and Golph3 ameliorates diabetes-induced neuroinflammation in vitro and in vivo

NLPR3 involves HG induced GA stress

To observe whether HG could cause disturbances in the GA, we examined the expression and distribution of GM130, a Golgi marker. The Golgi stress inducer nocodazole was used as a positive control. The immunofluorescence results showed that in the control groups, the GA was compressed around the nucleus, while in the nocodazole and HG groups, the GA was dispersed around the cytoplasm, whereas NLRP3 KO ameliorated the dispersed GA in the HG group, indicating that NLRP3 KO could block HG-induced GA dispersion (Figure 1A, 1B).

HG promotes the activation of NLRP3/VPS35/Golph3 pathway

NLRP3 was reported to be recruited to the dispersed Golgi apparatus before activation [16] and Golph3 was reported to regulate Golgi stress [10] whereas VPS35 is responsible for protein transportation [17]. We found that HG promoted the interaction of NLRP3 and VPS35 by co-immunoprecipitation and immune-fluorescence methods (Figure 2A, 2B), this indicated that VPS35 was interacted with NLRP3. Next, we found that HG promoted the interaction of VPS35 and Golph3 (Figure 2C, 2D), knockout of NLRP3 suppressed the expression of VPS35 and Golph3 and knockout of VPS35 reduced the expression of Golph3 but not NLRP3, these results demonstrated that VPS35 was activated downstream of NLRP3, Golph3 was activated downstream of VPS35 under HG condition (Figure 2E). Additionally, knockout of Golph3 alleviated HG induced Golgi stress (Figure 2F). Collectively, our results suggested that HG promoted the activation of GA stress in a NLRP3/VPS35/Golph3 pathway.

Golph3 regulates GA stress by interacting with vimentin

To clarify the molecular pathway of HG induced Golph3 activation, we performed the pulldown assay with GST-tagged Golph3 mixed with BV2 cell lysate, and the isolated proteins were analyzed by mass spectrometry and immunoblot. The mass spectrometry results showed that vimentin was pulled down by GST-tagged Golph3 protein (Figure 3A). The immunoblot results also showed that vimentin interacted with Golph3 (Figure 3B). Next, we performed an immunoprecipitation assay to evaluate the cellular interaction between Golph3 and vimentin, and the results showed that vimentin could be precipitated by Golph3 (Figure 3C). The confocal microscopy results further confirmed that Golph3 colocalized with vimentin after treatment with HG (Figure 3D). Vimentin has been reported to regulate Golgi function, and increased vimentin expression could lead to dispersion of the GA [18]. Knockout of Golph3 reduced the expression of vimentin (Figure 3E), while knockout of vimentin could not influence HG induced expression of Golph3, indicating that vimentin was stimulated downstream of Golph3 (Figure 3F). In addition, knockout of vimentin also alleviated HG induced GA stress as indicated by reduced GA dispersal (Figure 3G). These results indicated that Golph3 regulates GA stress by interacting with vimentin. Collectively, we concluded that HG induced GA stress was mediated by NLRP3/VPS35/Golph3/vimentin pathway.

Combination of NLRP3 inhibitor and Golph3 inhibitor ameliorates HG-caused microglia neuroinflammation

We used virtual molecular docking to select specific inhibitors targeting NLRP3 and Golph3, hoping to develop new drug combinations to alleviate neuroinflammation caused by HG. The molecular modeling analysis performed by Autodock vina and Autodock 4.0 showed that zafirlukast is the potential inhibitors of NLRP3 and bromocriptine is the potential inhibitors of Golph3 (Table 1). The binding mode of zafirlukast and NLRP3 is shown in Figure 4A. Obviously, the hydrogen bonding and hydrophobic interactions were observed for the binding of zafirlukast to NLRP3. In detail, six residues, Ala223, Arg603, Glu180, Arg576, Phe579, and Phe406, made a strong total binding energy contribution (ΔE total ≤ -1.0 kcal/mol), suggesting that these six residues are key residues for zafirlukast binding. From the Table 1, we can see bromocriptine with the smallest binding scores resulted from both Autodock vina and Autodock 4.0. The interactions between bromocriptine and Golph3 and the residues of the binding sites are shown in Figure 4B. The seven residues Ala243, Val248, Leu242, Leu239, Leu213, Cys92, and Glu96 forming the hydrogen bonding and hydrophobic interactions with the total binding energy contribution (ΔE total ≤ -1.0 kcal/mol) made a strong binding mode of bromocriptine and Golph3.

Then, we tested the anti-neuroinflammatory effect of ZAF and BRO on HG induced BV2 cells. The cell viability tests showed that ZAF at 5 μM and BRO at 2.5 μM have no effect on BV2 cell viability (Figure 4C, 4D). ZAF, BRO alone and combination of ZAF and BRO could reduce the transcription level of proinflammatory cytokines IL-1β and IL-6 in microglial cells (Figure 4E, 4F) and the expression level of proinflammatory cytokines IL1-β and IL-6 in BV2 cells treated with HG (Figure 4G, 4H). Moreover, ZAF, BRO alone and combination of ZAF and BRO could reduce the expression level of NLRP3, caspase1, IL-1β in NLRP3 inflammasome pathway (Figure 4I) and the expression level of Golph3, GM130, and vimentin in GA stress pathway (Figure 4J).

Combination of NLRP3 inhibitor and Golph3 inhibitor ameliorates STZ-induced diabetes neuroinflammation

To test the effect of ZAF, BRO and the combination of ZAF and BRO on diabetes neuroinflammation in vivo, we built STZ-induced diabetes mice (Figure 5A, 5B) and evaluated the therapeutical effect of ZAF, BRO and the combination of ZAF and BRO. First, we investigated the learning behavior of mice treated with ZAF or BRO, or ZAF combined with BRO by the Morris maze test. We found that diabetes mice treated with combination of ZAF and BRO demonstrated significantly increased associative and spatial learning compared with the diabetes untreated group as indicated by the latency of mice successfully finding the targets (Figure 5C). The motor performance as indicated by time spent in the course showed that STZ induced diabetes mice spent longer time in finding the targets in the acquisition course, while, mice in ZAF, BRO, or combination of ZAF and BRO treatment group showed shorter time in finding the targets (Figure 5D). The swimming speed of mice in ZAF, BRO, or combination of ZAF and BRO were significantly faster than diabetes untreated group (Figure 5E). Moreover, ZAF, BRO and ZAF combined with BRO treatment mice showed better performance as measured by the number of crossing over the platform area (Figure 5F, 5G).

Next, we found that ZAF, BRO and the combination of ZAF and BRO reduced the transcription and expression level of proinflammatory cytokine IL-1β and IL-6 of the hippocampus of diabetes mice, respectively (Figure 5H–5K). ZAF, BRO and the combination of ZAF and BRO respectively reduced the expression level of NLRP3 pathway related protein NLRP3, caspase1 and IL-1β (Figure 5L), also suppressed the expression of Golph3 pathway related proteins Golph3, GM130 and vimentin of the hippocampus of diabetes mice, respectively (Figure 5M).

Moreover, we explored the effect of NLRP3 inhibitor combined with GOLPH3 inhibitor on the pathological injury of brain tissue in diabetic mice. We used H&E staining to detect and analyze the pathological changes in hippocampus DG region, CA3 region and cerebral cortex in control group and diabetic groups (Figure 6A, 6B). We found that zafirlukast combined with bromocriptine could alleviate the pathological changes of brain neurons such as vacuolation and swelling caused by diabetes, and the pathological score collected from three pathologists showed that the pathological score was significantly decreased in zafirlukast combined with bromocriptine treatment group in hippocampus and cortex sections, suggesting that zafirlukast combined bromocriptine played a protective role in diabetic induced neuro-damage and neuro-inflammation.

At last, we determined the activation of microglia and the expression of NLRP3 and Golph3 in the treatment group by immunofluorescence method. Our results showed that the microglia were activated in STZ induced diabetes mice in hippocampus DG (Figure 7A, 7B) and CA1 (Supplementary Figures 1, 3) region, but not cortex (Supplementary Figures 2, 4), zafirlukast and bromocriptine treatment could suppress the activation of microglia in DG and CA1 regions indicated by the expression of microglia activation marker IBA1. At the same time, zafirlukast and bromocriptine treatment could suppress the expression of NLRP3 in DG (Figure 7A, 7B) and CA1 (Supplementary Figure 1) regions, and suppressed the expression of GOLPH3 in DG (Figure 8A, 8B) and CA1 (Supplementary Figure 3) regions, indicating that STZ induced diabetes might cause neuroinflammation in hippocampus regions, and zafirlukast combined with bromocriptine could ameliorate diabetes induced neuroinflammation by suppressing the expression of NLRP3 and Golph3.

In summary, these results showed that ZAF and BRO combination could ameliorate diabetes neuroinflammation in vivo.

Cell culture and treatment

The BV2 cell line was kept in our lab. Cells were cultured in low-glucose DMEM (5.5 mmol/L) (HyClone, USA, catalog: SH30021.01) containing 10% FBS (BI, Israel, catalog: 04-001-1ACS) and 1% penicillin-streptomycin (Gibco, USA, catalog: 15140122) at 37° C in an atmosphere of 5% CO₂. The cells (1 × 10⁶) were treated with different concentrations of D-glucose (Sigma, USA, catalog: 154-17-6) for 12h.

Establishment of diabetes mice and drug treatment

Forty male C57BL/6 mice (6 ~ 8 weeks old) were purchased from the Experimental Animal Center of First Hospital of Jilin University. For diabetes mice model, 40 C57BL/6 mice were divided into 5 groups, 8 mice in each group. The mice from diabetic group were administered with streptozotocin (STZ, Sigma, USA) intraperitoneally at the dose of 70 mg/Kg in 0.1 M citrate buffer (CIT), and PH 4.2 ~ 4.4 every other day for 3 times. The mice from control group were injected with CIT alone and served as the vehicle controls. The blood glucose level and body weight were recorded weekly for 4 consecutive weeks. The blood glucose level was determined by a glucometer, and diabetes was confirmed by fasting blood glucose > 16.7 mmol/L. The diabetes cognitive decline was determined by the Morris maze test. Then the control group and diabetes groups were given DMSO only, the zafirlukast (Yuanye, Shanghai, China, catalog: 107753-78-6) groups were given zafirlukast at the dosage of 50 mg/kg by intraperitoneal injection, the bromocriptine groups were given bromocriptine (MedChemExpress, USA, catalog: HY-12705A) at the dosage of 12 mg/kg by intraperitoneal injection, the combination groups were given both zafirlukast and bromocriptine.

Cas9-mediated knock out (KO) cell line construction

CRISPR/Cas9-mediated gene editing was performed using a px459 vector (Addgene, catalog: 48139) targeting murine nlrp3, vps35, golph3, vimentin in BV2 cells, the plasmid construction was described in a previous study [35]. The sgRNAs were displayed in Table 2 and were synthesized by Comate Bioscience Company, China. Plasmids were transfected with Attractene Transfection Reagent (Tiangen, China, catalog: 301004) according to the manufacturer’s instructions. Clonal lines were established by 96-well plate screening. The KO cell lines were certified by western blotting after the clones had formed. For HG treatment experiment, the KO cells were treated with HG at 35 mM for 12 h.

Immunohistochemistry and immunocytochemistry analysis

The immunohistochemistry was described as previous studies. Brain sections (30 μm) were prepared for IHC. First, the sections were made for antigen retrieval using citrate buffer, then washed with PBS for three times. After that, the sections were blocked with block buffer (5% BSA) for 1h and then cultured with primary antibody at 4° C overnight. Then the sections were washed with PBS for three times, incubated in second antibodies for 1 h at 37° C in dark, last, sections were stained with Hoechst for 10 minutes.

The immunocytochemistry study was described previously. Briefly, the cells were fixed with 4% paraformaldehyde, washed with PBS for 3 times, blocked with 5% BSA, then incubated with primary antibodies at 4° C overnight. Then cells were washed with PBS for 3 times, incubated with secondary antibody at 37° C for 1 h in dark, stained with Hoechst for 10 minutes. The cells were mounted on slides using glycerinum visualized by fluorescence microscope (Olympus U-TV0.63XC, Tokyo, Japan). The microglia were labeled with anti-IBA-1 antibody (Abcam, Beijing, China, catalog: ab178846; Diluted with 5% fetal bovine serum, 1:200), anti-NLRP3 (1:1,000), (Cell Signaling Technology, USA, catalog: 15101S); anti-Golph3 (Finetest, Wuhan, China, catalog: FNab03561), anti-GM130 (Finetest, Wuhan, China, catalog: FNab03558), anti-Vimentin (Finetest, Wuhan, China, catalog: FNab09410); Alexa Fluor dye-conjugated secondary antibodies were used for ICC experiments.

Quantitative real-time PCR (qRT-PCR)

Total RNA was extracted from cultured cells (1 × 10⁶) using a total RNA extraction kit (QIAGEN, USA, catalog: 90001), and reverse transcription was performed using an EasyScript^® Reverse Transcriptase kit (TransGen Biotech, Beijing, China, catalog: AE301-02). qRT-PCR was carried out with the Applied Biosystems 7900HT fast real-time PCR system using SYBR Green PCR master mix (ROCHE, USA, catalog: 4913850001). The primers for IL-1βwere 3’-TGCCACCTTTTGACAGTGATG-5’, 5’-AAGGTCCACGGGAAAGACAC-3’; for IL6 were 3’-CCCCAATTTCCAATGCTCTCC-5’, 5’-CGCACTAGGTTTGCCGAGTA-3’. All reactions were performed in triplicate, and each experiment was repeated three times. The relative expression of each target gene was calculated using the 2^-ΔΔCt method.

Western blot

The method was performed as described in a previous study [36]. The antibodies used were as follows: anti-NLRP3 (1:1,000), (Cell Signaling Technology, USA, catalog: 15101S), anti-cleaved-caspase-1 (1:1,000), (Cell Signaling Technology, USA, catalog: 89332S), anti-Cleaved-IL-1β (1:1000) (Cell Signaling Technology, USA, catalog: 63124S), and anti-GAPDH (1:2,000) (Bioworld, China, catalog: AP0066); anti-Golph3 (Finetest, Wuhan, China, catalog: FNab03561), anti-GM130 (Finetest, Wuhan, China, catalog: FNab03558), anti-Vimentin (Finetest, Wuhan, China, catalog: FNab09410); Goat anti-rabbit antibody (Bioss, Beijing, China, catalog: bs-0295G); Goat anti-mouse antibody (Bioss, Beijing, China, catalog: bs-0296G).

ELISA

Cells (1 × 10⁶) were incubated in 6-well plates and treated with or without HG at 35 mM for 12 h. The cell supernatant was collected and the pro-inflammatory cytokine IL1β (Boster, Wuhan, China, catalog: EC0394) and IL6 (Boster, Wuhan, China, catalog: EC0394) was detected using ELISA kit according to manufactures’ instruction.

Co-immunoprecipitation

The BV2 (1 × 10⁸) cells were collected and lysed with RIPA lysis buffer (1% NP-40, 0.25% deoxycholate) (Beyotime, China, catalog: P0013D) with 1 mM phenylmethylsulfonyl fluoride (PMSF) (Beyotime, China, catalog: ST505). 30 μg protein was added into 500 ul RIPA lysis buffer, along with 1μg portions of the corresponding antibodies, 1 μg normal rabbit IgG (Santa Cruz Biotechnology, USA, catalog: sc-2026). Then, 30 μL protein A/G (SMRRT, China, catalog: SA032005) beads were added to the protein-antibody mixture and incubated at 4° C overnight. After the incubation, samples were centrifuged at 2, 500 rpm for 4 min at 4° C and washed three times with RIPA Lysis Buffer. Next, the supernatant was removed and 30 μL of 2 × loading buffer was added boiled for 10 minutes, immunoblot (IB) analysis was performed.

GST pull-down assay

PGEX4T-1 was purchased from PPL (Public Protein/ Plasmid Library, catalog: 27-4580-01) was used to construct p-GEX4T1-λ-Golph3 plasmid. The PCR products target Golph3 gene were amplified from mouse cDNA library using the primers 5’-CG GAATCCATGACCTCGCTGACCCAG-3’ and 3’- CCCTCGAG TTACTTGGTAAACGCAGC-5’. The PCR products and the PGEX4T1 λ were both digested by nuclease EcoR1 (NEB, USA, catalog: R0101S) and Xho1(NEB, USA, catalog: R0146S), constructing the recombinant plasmid using T4 DNA ligase (NEB, USA, catalog: M0202S). The recombinant plasmids were transformed into E.coli prokaryotic expression system to obtain purified GST-tagged Golph3 protein. Purified GST or GST-Golph3 proteins were incubated with glutathione-sepharose beads (PureCube, China, catalog: 32103) at 4° C for more than 2 h. Then the beads were washed for 3 times with 1% Triton 100 (Solarbio, China, catalog: 9002-93-1). After that, the beads were added to BV2 cell lysate at 4° C for 3 h, the cell lysate were extracted from 10⁸ cells. The beads were then washed extensively and the bound proteins were eluted and separated on 10% SDS-PAGE for western blot analysis.

Morris water maze test

The mice were individually placed into a new cage 2 days before the test. The water maze consisted of a circular tank (diameter = 1 m; height = 30 cm) that was filled with tepid water (23 ± 1° C) that was made opaque by the addition of powdered milk. The circular tank was divided into four equal sectors (Target, Opposite, Sector 1, and Sector 2), each with a spatial cue on the tank wall. A white escape platform (diameter = 10 cm, height = 10 cm) was located 1 cm below the water. The water temperature was 24 ± 1° C. The testing lasted for 7 days. First, mice were trained for 6 consecutive days and underwent four training trials per day. In each training trial, a mouse was placed in one of the sectors and allowed to search for the hidden platform for 1 min. If the mouse did not find the platform within 1 min, the experimenter led the animal to it. After the platform was located, the mouse was left on it for 15 s to memorize the spatial cues. After that, the mouse was placed in a cage for 15 s for resting before the next trial. Throughout the experiment, the platform remained at its original position. To assess the learning ability, latency to find the platform in each trial and the total number of successful attempts were registered. Finally, on day 7, a probe trial was administered: the platform was removed, the mouse was placed in the Opposite sector, and the time spent in each sector within 1 min was measured. The Morris maze equipment was kept in college of food science and engineering, Jilin University. The test data were recorded and processed using the ANY-Maze™ (Stoelting Co, Wood Dale, IL, USA).

Virtual screening analysis and molecular docking studies

The compounds from Drugs^@ FDA database (https://www.bindingdb.org/bind/ByFDAdrugs.jsp) were used for virtual screening. The structure of the murine NLRP3 is generated from SWISS-MODEL website (https://swissmodel.expasy.org/interactive) using the nlrp3 sequence (NM_145827.4). The structure of the murine Golph3 is also generated from SWISS-MODEL using the golph3 sequence (NM_025673.2). AutoDock Vina and Autodock 4.0 software was utilized in all the docking experiments, with the optimized model as the docking target. The screening method is restricted to molecular docking, and molecular dynamics simulation has not been carried out. The results were visualized using PyMol [37].

Hematoxylin and eosin (H&E)

Mice were deeply anesthetized by Chloral Hydrate (Yuanye, Shanghai, China) and transcardially perfused with 4% of parafomaldehyde (Beyotime, China). The cortex and hippocampus of mice were carefully removed and fixed in 4% of paraformaldehyde overnight, and were decalcified in 10% of EDTA, and embedded in paraffin. Sections (4 μm) were stained with hematoxylin-eosin. The stained sections were graded by three pathologists in a blinded fashion for the degree of neuron injury and inflammation using a four-point scale from 0 to 5, and 0 represents no damage, 1-2 represents mild damage, 3-4 represents moderate damage and 5 represents severe damage, respectively. Similar results were obtained from three independent experiments.

Statistical analysis

Data were analyzed by GraphPad Prism 8.0. Statistical significance was evaluated using independent sample one-way ANOVA or two-way ANOVA combined with post hoc tests for multiple comparisons. p < 0.05 was considered statistically significant.