2.1 Network pharmacology study

2.2 Chemicals and reagents

Coeloglossum viride var. bracteatum extract (CE) was prepared and its composition was validated as previously described (Li et al., 2021). Lipopolysaccharides (LPS) was purchased from Solarbio (#L8880). WST-1 cell proliferation and cytotoxicity assay kit was purchased from Beyotime (#C0036L). In situ cell death detection kit for terminal deoxynucleotidly transferase-mediated dUTP nick-end labelling (TUNEL) detection was purchased from Roche (#11684795910). The total RNA isolation kit was purchased from Beijing Zoman Biotechnology Co., Ltd (#ZP404).

2.3 Cell culture

To generate primary astrocytes and primary hippocampal neurons, the brains of neonatal Sprague-Dawley (SD) rats (provided by the Department of laboratory animal science of Peking University Health Science Center, license number SCXK-2011–0012) were dissected and the prefrontal-cortex and hippocampus were isolated. The isolated tissues were then digested (0.25% trypsin), filterted and centrifuged (1,100 rpm, 5 min) as previously described (Pan et al., 2017, Cai et al., 2021). Finally, astrocytes and neurons were cultured in DMEM medium containing 10% fetal bovine serum (FBS), 100 units/mL penicillin and 100 μg/mL streptomycin at 37 °C with 5% CO₂ for 7–10 days.

BV2 cells were provided by the Brain Science Center, Beijing Institute of Basic Medical Sciences and cultured in DMEM medium supplemented with 10% FBS, 100 units/mL penicillin and 100 μg/mL streptomycin at 37 °C with 5% CO₂ for 12 days.

2.4 Measurement of cell viability and apoptosis

The WST-1 assay was applied to measure cell viability. Briefly, after culturing and treating cells in 96-well plates with different conditions, 10 μL of WST-1 solution was added and incubated at 37 °C and 5% CO₂ for 1–2 h. Finally, the absorbance was recorded at 450 nm and the cell viability was calculated compared to the control absorbance.

Apoptosis was measured by the TUNEL assay according to the instructions of the in situ cell death assay kit. Briefly, cells were stained and observed under a fluorescent microscope. The rate of cell apoptosis was assessed by calculating the ratio of TUNEL/DAPI-positive cells.

2.5 Determination of inflammatory factors

The gene expression of inflammatory factors, including interleukin-6 (IL-6), interleukin-1β (IL-1β), tumor necrosis factor-α (TNF-α), and induced nitric oxide synthase 2 (Nos2, also known as iNOS), was examined by real-time quantitative polymerase chain reaction (RT-qPCR) experiments according to the instructions of the RNA kit. The sequences of the primers are listed as follows: β-actin, 5′-TGTCCACCTTCCAGCAGA-3′ (forward) and 5′-GCTCAGTAACAGTCCGCCTA-3′ (reverse); Il6, 5′-ATTCTGTCTCGAGCCCACCA-3′ (forward) and 5′-AGGCAACTGGCTGGAAGTCT-3′ (reverse); Il1β, 5′-AATGCCTCGTGCTGTCTGA-3′ (forward) and 5′-TGTCGTTGCTTGTCTCTCCT-3′ (reverse); Tnfα, 5′-TGACCCCCATTACTCTGACC-3′ (forward) and 5′-GGCCACTACTTCAGCGTCTC-3′ (reverse); Nos2, 5′-AAGAGACGCACAGGCAGAG-3′ (forward) and 5′-CAGGCACACGCAATGATGG-3′ (reverse). Parameters for RT-qPCR cycles were set as follows. 120 s at 95 °C, 10 s at 95 °C (45 cycles), 95 °C for 10 s, 37 °C for 30 s. The relative expression levels of genes were calculated by the 2^−ΔΔCt method and normalized using β-actin as an internal control.

2.6 Immunofluorescence staining

Treated cells were seeded on 24-well plates, fixed with 4% paraformaldehyde for 15 min, and then incubated with 0.2% Triton X-100 for 10 min and blocked with PBS containing 10% goat serum. Cells were then incubated with the corresponding antibodies overnight and further immunofluorescence staining was performed at 4 °C. The primary antibodies were GFAP (#3670, Cell Signaling Technology), IBA1 (#019–19741, Wako, Japan) and MAP2 (#4542, Cell Signaling Technology). The secondary antibody used was Alexa Fluor 594-conjugated goat anti-rabbit IgG (#ZF-0513, ZSBG-Bio). DAPI was used to stain cell nuclei. Images of immunofluorescence staining were obtained using a Leica TCS SP8 confocal microscope.

2.7 ATP level measurements

The level of adenosine 5′-triphosphate (ATP) production in microglia reflects the metabolic pattern of energy (Cheng et al., 2014). Here, we added 100 μL of lysis buffer containing luciferase reagents (#G7570, Promega) to plates with cultured microglia for incubation. And then the plates were measured in a microplate reader and the luminescent signal was normalized.

2.8 Seahorse extracellular flux assay

The extracellular acidification rate (ECAR) and oxygen consumption rate (OCR) assays are based on the Seahorse XFe 96 Extracellular Flux Analyzer (Seahorse Bioscience), which reflects the real-time energy metabolism of the cells. All procedures conform to the instructions of the Seahorse XF Glycolysis Stress Test Kit (Agilent Technologies) and the Seahorse XF Cell Mito Stress Test Kit (Agilent Technologies). Briefly, microglia were seeded and treated in Seahorse XF-96-well plates under different conditions for 24 h. Microglia were then washed and maintained with XF assay medium for measurements. We first measured baselines. To measure ECAR, glucose (10 mM), oligomycin (oligo, 1 μM) and 2-deoxyglucose (2-DG, 50 mM) solutions were added sequentially to the Seahorse XF-96-well plates. To measure OCR, oligo (1 μM), p-trifluoromethoxy carbonyl cyanide phenylhydrazone (FCCP, 2 μM) and Rotenone/antimycin A (Rote/A, 0.5 μM) solutions were added sequentially. The results were normalized and analyzed by Seahorse XF 96 Wave software.

2.9 Western blotting

BV2 cells were harvested after treatment and subjected to protein extraction using RIPA buffer. Equal amounts of proteins were loaded onto SDS-PAGE and separated. Proteins were probed using specific primary antibodies for HIF-1α (#bs-0737R) and PKM2 (#bs-0102 M), respectively. After incubation with HRP-conjugated secondary antibodies, protein signals were visualized by chemiluminescence in infrared fluorescence imaging system (Odyssey Clx, LI-COR Biosciences).

2.10 Data analysis

Experimental results were expressed as mean ± s.e.m. Student’s t-test and one-way analysis of variance (ANOVA) were applied to determine statistical differences between groups. Statistical analyses and plots were carried out in the software GraphPad Prism 8.0.

3.1 Network pharmacological analysis of bioactive molecules in CE and potential AD-related targets.

Combining our previous study (Cai et al., 2021) and literature review (Huang et al., 2004, Zhang et al., 2006a, 2006b), we retrieved and finalized nine bioactive molecules that exhibit neuroprotective capacity as major ingredients of CE (Table 1). Among these nine molecules, Dactylorhin A, Dactylorhin B, Loroglossin, Militarine and Coelovirin A were shown to protect the nervous system from brain damage and cognitive impairment in animal models of AD and Parkinson’s disease (PD) (Zhang et al., 2006a, 2006b, Pan et al., 2017, Li et al., 2021). Meanwhile, β-sitosterol was reported to improve memory and reduce amyloid-beta protein (Aβ) formation in APP/PSI mice (Wang et al., 2013, Ye et al., 2020), and Quercetin 3,7-diglucoside, Daucosterol and 4-Hydroxybenzaldehyde were shown to be antioxidants with potential neuroprotective capacity (Jiang et al., 2015, Oh et al., 2017, Sut et al., 2019). However, the manner and mechanism by which these nine molecules act together remains unknown. Here, we constructed a CE-molecules-targets network to demonstrate the linkage between these nine major components (Fig. 1A). Based on the molecules prediction targets in the Swiss Target Prediction database (Suppl. Table S1), the CE-molecule-target network illustrates the distribution of targets, where 482 nodes represent targets and their network topology parameters represent relationships (Suppl. Table S2).

Close

Fig. 1

Workflow of the PPI network used to construct the CE-molecules-targets network and extract the key targets. (A) In the CE-molecules-targets network, the aquamarine circles represent the putative targets of the corresponding molecules, the wheat colored diamonds represent the molecules in CE, the light pink triangles represents CE, and the edges represent the connections. (B) PPI network of AD-related targets. (C) PPI network of CE-related targets. (D) Crossover network after merging the AD-related PPI network and the CE-related PPI network. (E, F) PPI networks of key targets after filtering by topological parameters.

Export to PPT

1017 CE-related targets were obtained from the GeneCards database, DisGeNET database and DrugBank database (Suppl. Table S3). To explore the key targets and protein interactions during CE treatment of AD, we built a PPI network of CE-related targets and AD-related targets using the BisoGenet program (Fig. 1B and 1C). In addition, a PPI network with a total of 178 nodes and 2317 edges was eventually screened (Fig. 1F), yielding the hub targets that may play an important role in the treatment of AD.

All topological parameters of the hub target PPI network were calculated in Cytoscape, by which we can rank the importance of the targets (Fig. 2A). The color and size of the nodes are proportional to their degrees in the network. For example, NPM1, HSPA5, HSPA8, EEF1A1 and HDAC5 may be critical nodes of great degree (Fig. 2A; Table S4). However, the application of network pharmacology emphasizes the synergistic effects of multiple targets at the overall system level (Yuan et al., 2017, Li 2021). Therefore, we applied the MCODE algorithm, an automated method for large protein interaction networks, to cluster and find molecular complexes (Bader and Hogue 2003). Seven clusters named “MCODE_” and their corresponding GO annotations were generated in Metascape (Fig. 2B and 2C). The GO terms of MCODE_5 and MCODE_7 are consistent with earlier studies (Pan et al., 2017, Li et al., 2021), describing that CE may treat AD by regulating oxidative stress and inflammation. In addition, the GO terms of MCODE_6 provide a new perspective focusing on energy metabolism.

Close

Fig. 2

Key target networks and gene clusters. (A) Key target network. The size and color of the diamonds are proportional to their degree centrality. (B, C) Modular description of the gene clusters and Metascape generated by the MCODE algorithm.

Export to PPT

Other than GO annotation of the clustered targets, we also performed GO enrichment analysis and KEGG pathway enrichment analysis to gain a comprehensive understanding of the biological functions of all 178 key targets (Fig. 3). As a result, pathways involved in inflammation were enriched, such as Toll-like receptor signaling pathway (hsa04620), TNF signaling pathway (hsa04668), NF-kappa B signaling pathway (hsa04064), MAPK signaling pathway (hsa04010) and IL-17 signaling pathway (hsa04657). Intriguingly, glycolysis/gluconeogenesis (hsa00010) was enriched in KEGG pathway analysis, while the GO terms ATP binding (GO: 0005524) and ATPase activity (GO: 0016887) were also found to be consistent with the clustering targets. Recent studies have proposed that AD progression can be accelerated by pro-inflammatory activation and aberrant glycolysis in different brain cell types, and targeting energy metabolism as well as inflammation may be a therapeutic strategy for AD (Tejera et al., 2019, Pan et al., 2022). Therefore, the results of the network pharmacology analysis suggest that CE may regulate inflammation and energy metabolism in AD, which is a promising finding.

Close

Fig. 3

GO enrichment analysis and KEGG pathway enrichment analysis of the hub targets. (A) GO enrichment analysis. The top ten GO terms were selected to describe the biological process (BP), cellular component (CC) and molecular function (MF) of the hub target. (B) KEGG pathway enrichment analysis. The pathways enriched in environmental information processing, genetic information processing, human disease, cellular processes, organismal systems, and metabolism were displayed.

Export to PPT

3.2 CE attenuates LPS-induced inflammation in astrocytes and apoptosis in neurons

To confirm the results of the network pharmacology analysis that CE may regulate inflammation, we treated primary astrocytes and hippocampal neurons with LPS and CE, respectively, while LPS is widely used to stimulate inflammation in cells (Calsolaro and Edison 2016). Cell viability was assessed with the WST-1 assay, and the expression levels of inflammatory factors were detected with RT-qPCR. The results showed that the cell viability of astrocytes decreased with increasing LPS concentration and the expression levels of TNF-α and IL-6 became higher (Fig. 4A and B), indicating that LPS managed to activate inflammation in astrocytes. Furthermore, since the 10 μg/mL LPS-treated group had relatively low cell viability and higher expression levels of TNF-α and IL-6, we selected this concentration of LPS to activate inflammation in these cell models in the next experiments. We next determined the anti-inflammatory effect of different concentration gradients of CE. The results demonstrated that treatment with CE, especially at 10 mg/mL, inhibited the LPS-induced decrease in cell viability and down-regulated the expression levels of inflammatory factor in astrocytes (Fig. 4A and B).

Close

Fig. 4

CE attenuates LPS-induced inflammation in astrocytes and apoptosis in neurons. (A) CE rescued the cell viability of astrocytes from LPS treatment. Left, Student’s t test was performed, n = 5, * p < 0.05, *** p < 0.001. Right, Student’s t test was performed, n = 5, *** p < 0.001 versus control; # p < 0.01, ## p < 0.005, ### p < 0.001 versus 10 μg/mL LPS. (B) CE down-regulated the expression levels of inflammatory factors in LPS-treated astrocytes. One-way ANOVA was performed, n = 5, * p < 0.05, ** p < 0.01, *** p < 0.001. (C) Immunofluorescence staining experiments showed morphological protection of CE against LPS on astrocytes. (D-E) Immunofluorescence staining experiments and TUNEL assays showed that CE protected the neurons from LPS-induced apoptosis. Student’s t test was performed, n = 9, *** p < 0.001 (LPS vs Control), ### p < 0.001 (CE + LPS vs Control), n.s denotes not significant (CE vs Control).

Export to PPT

We also performed immunofluorescence to evaluate the morphological changes in cells treated with LPS and CE. Astrocytes activated by LPS for inflammation (labeled by GFAP) experienced a decrease in cell number and cell bodies, while astrocytes treated with CE, morphologically maintained their activity and size of the control (Fig. 4C). Similarly, CE improved the performance of neurons (labeled by MAP2) treated with LPS, showing greater cell numbers and neuronal connections (Fig. 4D). Consistent with this improvement, the results of the TUNEL assay showed an inhibitory effect of CE on LPS-induced neuronal apoptosis (Fig. 4E). Collectively, consistent with the results of the network pharmacology analysis, CE attenuated LPS-induced inflammation in astrocytes and neurons.

3.3 CE alleviates LPS-induced inflammation and enhances energy metabolism in microglia

Microglia-mediated neuroinflammation might contribute to the progression of AD, where the metabolic switching from OXPHOS to aerobic glycolysis plays an important role (Leng and Edison 2021, Pan et al., 2022). Our network pharmacology analysis also provides evidence that CE may regulate energy metabolism against AD. Here, we used BV2 cells in various models to assay inflammatory factors and energy metabolism markers to elucidate whether CE could shift the metabolic patterns of activated microglia. In line with previous findings in astrocytes and neurons (Fig. 4), LPS activated inflammation in microglia (labeled by IBA1), while CE reduced activation (Fig. 5A). Quantitative data of activated cells were consistent with morphological findings (Fig. 5B), where activated microglia displayed larger size as well as abnormal shape (Fig. 5A). We also found that ATP levels were increased in the CE-treated group compared to the control or LPS-treated group (Fig. 5C). In addition, CE reduced the gene expression levels of inflammatory factors, including TNF-α, NOS2, IL-1β, and IL-6 in LPS-treated microglia (Fig. 5D). Together, CE alleviated LPS-induced inflammation in microglia and promoted ATP expression, indicating that regulation of energy metabolism is involved in the anti-inflammatory effects of CE.

Close

Fig. 5

CE alleviates LPS-induced inflammation and enhances energetic metabolism in microglia. (A-B) CE protected microglia from LPS-induced activation and morphological changes, n = 5. (C) CE rescued LPS-induced decrease of ATP, n = 5. (D) CE alleviated the LPS-induced elevation of gene expression in inflammatory factors, n = 3. (E-F) ECAR measurements of microglia. Oligo, oligomycin; 2-DG, 2-deoxyglucose. (G-H) OCR measurements of microglia. (I) Measurement of lactate production and the activities of HK and PK. (J) RT-qPCR assay of the gene expression levels of Hif1α, Glut1, Hk2, and Pkm2. (K) Western blotting detects the protein levels of HIF-1α and PKM2. The relative expression levels were normalized to β-actin. Student’s t-test was performed in B, C, D, I, J, and K, *** p < 0.001 (LPS vs Control), ### p < 0.001 (CE + LPS vs LPS), n.s denotes not significant. One-way ANOVA was performed in F and H, ** p < 0.01, *** p < 0.001, n.s, not significant.

Export to PPT

Furthermore, based on the results of the network pharmacology analysis and previous experiments, we cultured different microglia models to investigate the role of CE in metabolic regulation. The activation of microglia can be described as two phenotypes: M1 and M2 (Guo et al., 2022). M1 microglia exert pro-inflammatory effect leading to cell death, while M2 microglia protect neuronal cells from inflammation (Colonna and Butovsky 2017, Guo et al., 2022). Here, we treated microglia with LPS (LPS induces M1 polarization) or IL-4 (IL-4 induces M2 polarization) to induce microglia polarization and assessed the role of CE in both models. Energy metabolism, which includes ECAR and OCR related to cellular glycolysis and oxygenation, was examined with the Seahorse extracellular flux assay. The results showed that glycolysis in microglia was successfully enhanced by LPS but could be inhibited by CE (Fig. 5E and 5F), whereas CE had no effects on glycolysis in M2 microglia (IL-4-treated). In addition, OXPHOS was decreased in LPS-treated microglia, whereas CE reversed this phenomenon by promoting OXPHOS as in the IL-4-treated group (Fig. 5G and H). Therefore, CE can reprogram the metabolic state of microglia by regulating glycolysis and OXPHOS in different microglia models and promote the conversion of M1 microglia to M2 microglia.

We further examined the activity of hexokinase (HK) and pyruvate kinase (PK), as well as the levels of lactate. It was found that the activities of HK and PK were significantly increased in LPS-activated BV2 cells, accompanied by increased lactate production, which was reversed by CE (Fig. 5I). In addition, RT-qPCR assays for the expression of glycolysis-related genes, such as glucose transporter 1 (Glut1), hexokinase 2 (Hk2), lactate dehydrogenase A (Ldha), glucose-6-phosphate 1-dehydrogenase X (G6pdx) and M2-type pyruvate kinase (Pkm2), also confirmed that LPS significantly induced elevated expression levels of these genes (Fig. 5J). However, the elevated expression of LPS-activated microglia glycolysis-related genes was also restored by CE (Fig. 5J). Consistently, we also examined the changes in PKM2 and HIF-1α protein levels, again confirming that CE reversed the LPS-induced elevation of PKM2 and HIF-1α. These results suggest that CE inhibited the LPS-induced enhancement of microglia glycolytic activity and improved energy metabolism.