QKL injection ameliorates Alzheimer ’ s disease-like pathology by regulating expression of RAGE

The onset of Alzheimer ’ s disease is related to neuron damage caused by massive deposition of A β in the brain. Recent studies suggest that excessive A β in the brain mainly comes from peripheral blood, and BBB is the key to regulate A β in and out of the brain. In this study, we explored the pathogenesis of AD from the perspective of A β transport through the BBB and the effect of QKL injection in AD mice. The results showed that QKL could improve the cognitive dysfunction of AD mice, decrease the level of A β and A β transporter — RAGE, which was supported by the results of network pharmacology, molecular docking and molecular dynamics simulation. In conclusion, RAGE is a potential target for QKL ’ s therapeutic effect on AD.


Introduction
Alzheimer's disease (AD) is the most common cause of dementia in older adults, characterized by central nervous system degeneration (Chhatwal et al., 2020).Its symptoms include progressive memory loss and multidisciplinary cognitive impairment (Gaugler et al., 2019).The pathogenesis of AD remains poorly understood.However, extensive amyloid-β (Aβ) deposition in the brain can be seen in the most patients with AD, which eventually leads to neuronal apoptosis and dementia (Corriveau et al., 2017).The blood-brain barrier (BBB) is an important structure to maintain the stability of the internal environment of the brain, preventing peripheral circulation substances from entering the brain at will (SM, M. and Z. N, 2021).Recent studies have suggested that soluble Aβ in the brain is mainly derived from the blood (Sweeney et al., 2018a;Starling, 2018).There are two main pathways of Aβ for peripheral blood to enter brain (Shang et al., 2016).One is the damage in the continuity of the brain microvascular wall (Liu et al., 2020a), and the other is an imbalance in the expression of Aβ transport proteins in brain microvascular endothelial cells (Kim et al., 2014).Therefore, we speculated that the key point was in BBB, and wanted to observe the changes of BBB transport proteins and tight junctions as well as the cognitive functions.
Qingkailing (QKL) injection is a famous Chinese medicine that is approved by the China Food and Drug Administration, which was originally prepared by a group of scientists at the Beijing University of Chinese Medicine in the 1970s and has been clinically used throughout China for over 30 years.It is derived from a popular traditional Chinese prescription drug, the Angong Niuhuang pill (BUoC, 1975), whose main ingredients include geniposide, baicalin, cholic acid and hyodeoxycholic acid, which has protective effects on the brain (Ma et al., 2019;Ma et al., 2020).Our research and that of others have shown that QKL can effectively work on the BBB (Ma et al., 2019;Zhang et al., 2020;Zhu et al., 2011) .In our previous study, we found that QKL significantly improves learning and memory in AD mice (Qiu et al., 2010).However, the underlying mechanisms of QKL in treating AD remain to be established.Network pharmacology reveals new ways of applying drugs by elucidating the complex relationships among targets, drugs, diseases and pathways.Molecular docking technology predicts and verifies the pharmacodynamic elements of traditional Chinese medicine.Molecular dynamics simulation (MDS) verifies the binding abilities between small molecule compounds and key targets proteins.All of them are widely used in the research of traditional Chinese medicine.This study aims to assess the effect of QKL intervention on Aβ transport across the BBB of AD mice based on in vivo experiment, network pharmacology analysis, molecular docking and MDS.

Animals
Male senescence-accelerated mouse resistant (SAMR) 1 and senescence-accelerated mouse prone (SAMP) 8 mice (12 weeks old) were provided by Beijing Vital River Laboratory Animal Technology Co., Ltd.(License number: SCXK (Beijing) 2012-0001).The mice were raised in the Experimental Animal Center of Dongzhimen Hospital of Beijing University of Chinese Medicine.They were placed in separate cages (5 mice per cage) and maintained at a constant temperature (23

Grouping and administration
After a week of adaptive feeding, thirty SAMP8 mice were numbered and divided into the following two groups, with 15 mice each: AD model group and QKL group; fifteen SAMR1 mice were used as the control group.
The control group was treated with an intraperitoneal injection of saline and intragastric administration of pure water.The AD model group was treated with the same administration as the control group.The QKL group was intraperitoneally injected with 3 mL/kg 20 % QKL (diluted in normal saline to a 20 % injection) and intragastrically administered pure water.All the groups were administered at 9:00 am from Monday to Friday for 4 weeks.

The Morris water maze experiment
We performed behavioral testing in the Morris water maze (MWM) to assess spatial learning and memory ability for six days.The water maze consisted of a black circular tank (120 cm diameter, 40 cm high) and a transparent platform (9 cm diameter, 23 cm high) under water.The pool was filled with warm water (23 ± 1 • C) about 24 cm deep and divided into four equal quadrants, designated as I II III and IV.After 4 weeks of drug administration, all experimental mice were subjected to the MWM experiment, including the navigation training for 5 days and probe trial for 1 day.Navigation training was first conducted with four trials each day and the drug administration continued during the task.The transparent platform was located in center of the target quadrant (III), approximately 1.5 cm below the water.The mice were allowed to swim randomly of those quadrants in water until they find the submerged platform during the maximum swimming time (90 s).The escape latency (the time that mice swim from the point of entry to the platform) was determined.The mice were guided to stay on the target platform for 10 s if they did not find the platform within 90 s.One day after navigation training, probe trial was conducted.The mice were allowed to swim from the relative quadrant(I) to identify the transparent platform, which has removed in the target quadrant(III) before.The number of crossings, the time spent in the target quadrant and strategy to exploring were measured.

Sample collection
Serum and brain tissue samples from six mice of each group were collected on the next day of the MWM.Blood was drawn from eyeballs removed from the mice.Then serum was separated (3000 rpm for min) and stored at − 80 • C. The head was quickly severed, and the brain was dissected on an ice tray; the olfactory bulb and cerebellum were discarded.The remaining brain tissue was washed with ice-cold saline, dried using a filter paper, and weighed.A 10 % homogenate of the tissue was prepared and centrifuged at a low temperature and high speed (3000 rpm for 20 min).The supernatant was then separated and stored at − 80 • C.

ELISA
Aβ levels in the brains of mice were measured using a human Aβ Colorimetric solid phase sandwich Enzyme Linked Immuno-Sorbent Assay (ELISA) kit (BioSource, Camarillo, CA, USA) according to the manufacturer's instructions.Dilute Reconstituted Standard and Anti-Rabbit IgG HRP solution.Add 50 μL of standards, controls, or samples, 50 μL of Hu Aβ 40 Detection Antibody solution into the appropriate wells.
Cover the plate with a plate cover and incubate 3 h at room temperature.
Wash wells 4 times and add 100 μL Anti-Rabbit IgG HRP into each well except the chromogen blanks.Cover the plate with plate cover and incubate for 30 min at room temperature.Wash wells 4 times and add 100 μL Stabilized Chromogen to each well.The substrate solution begins to turn blue.Incubate for 30 min at room temperature in the dark.Add 100 μL Stop Solution to each well.The solution in the wells changes from blue to yellow.Read the absorbance at 450 nm with a plate reader (THERMO Multiskan FC, MK3).Use ELISACalc software to generate the standard curve and read the concentrations.

Light microscopy
Six mice per group were treated with cardiac perfusion of 4 % paraformaldehyde; their brains were fixed in 4 % paraformaldehyde for 24 h.After gradient alcohol dehydration and xylene transparent paraffin embedding, the tissue was sectioned by coronal slicing (5 μm), placed on polylysine anti-stripping glass slides, and stored in a cool dark place.Nissl staining was performed using a kit (Solarbio, G1430) containing a Cresyl violet Stain and a Nissl 's differentiation solution.After conventional dewaxing, the sections were placed in the Cresyl violet Stain in a 56 • C incubator for 1 h.After rinsed with deionized water, the sections were differentiated in Nissl 's differentiation for 2 min.The sections were sealed after treatment with anhydrous ethanol and xylene.The neuronal structure of the hippocampal CA1 region was observed under a light microscope (Olympus BX60 and SPOT-(Diagnostic instruments ens.Inc.)) after Nissl staining.

Ultrastructure of the hippocampus
Three mice per group were anesthetized with 1 % pentobarbital sodium (0.05 mL/kg, i.p. injection) and perfused transcardially with cold 0.01 M phosphate-buffered saline (PBS) and then with 2.5 % glutaraldehyde.The hippocampus of the brain tissue sampled from mice was used to postfix in 2 % paraformaldehyde and 2.5 % glutaraldehyde (0.1 M, pH 7.2) for 2 h at 4 • C, then the tissue were placed in PBS 10 min for three times, stored at 4 • C. The samples were processed by cell dehydration, drying, and metal spraying in the electron microscope room of the Basic Institute of Chinese Academy of Medical Sciences.They were observed and imaged using a transmission electron microscope (HITACHI; No: H-7650).

Network pharmacology analysis
QKL's potential target genes were predicted using the Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP), Swiss Target Prediction, PubChem and SymMap database, and uniformly transformed by the UniPort.The screening criteria were oral bioavailability (OB) > 30, drug-likeness (DL) > 0.1.To find targets related to AD, we searched the GeneCards, CTD and OMIM using the key word " Alzheimer's disease ".Intersect the data of QKL and AD disease targets to obtain potential targets for QKL treatment of AD.Using cross genes to establish a PPI network with a confidence score of ≥0.9 in the STRING database.GO (Gene Ontology biological process) function and KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway enrichment analysis are performed through the ClusterProfiler software package in R software (v3.6.3).

Molecular docking
The three-dimensional structures of QKL Injection's bioactive components were downloaded as ligands in TCMSP, and the threedimensional structures of the RAGE protein were downloaded as receptors in the RCSB Protein Data Bank (PDB).Open Bable 3.1.1software was used to convert the pdbqt format of key proteins into PDB format, and Autodock Tools 1.5.7 software was used for docking verification.PyMOL software was used to visualize the docking results.The binding energy is usually based on − 5 kcal/mol to evaluate the effectiveness of ligand-receptor interaction.The lower value indicates the more favorable molecular docking results.

Molecular dynamics simulation
The initial conformation for MDS was obtained using the results from molecular docking.MDS was carried out with Desmond/Maestro noncommercial version 2022.1 as a molecular dynamic's software.TIP3P water molecules were added to the systems, which were then neutralized by 0.15 M NaCl solution.After minimization and relaxation of the system, the production simulation was performed for 100 ns in an isothermal-isobaric ensemble at 300 K and 1 bar.Trajectory coordinates were recorded every 100 ps.MDS was performed using Simulation Interaction Diagram from Desmond.

QKL improves the spatial learning and memory abilities in AD mice
On the first day, those groups showed no significant differences.From the second to the fifth day, the escape latency of the AD model group was significantly longer than that of the control group.Compared to AD model group, the escape latency time of the QKL was significantly shorter (P < 0.05) (Fig. 1A).Furthermore, the concurrent recorded average swimming speed reflected no significant differences among the three groups during the navigation training, excluding the effects of exercise ability and vision (Fig. 1B).
During the navigation training, the numbers of mice crossing the target platform were dramatically higher than that in the AD model group (P < 0.05) (Fig. 1C).Compared to control group, the time spent in the target quadrant of the AD model group was significantly shorter (Fig. 1D), whereas, in the QKL group, the time was greatly longer than that in the AD model group(P < 0.05).In the control and QKL groups, the "towards" trend was the main type of search strategy, whereas in the AD model group, "random" and "edge" type strategies were observed in the exploration process (Fig. 1E).

QKL decreases the Aβ level of brain in AD mice
Compared with the control group, the Aβ level in the brain tissue of the AD model group was markedly higher.Compared to AD model group, the Aβ level in the brain tissue of the QKL group was lower.QKL significantly decreased the Aβ level in the brain (Fig. 1F).

QKL alleviates damage to neurons, mitochondria and endoplasmic reticulum in AD mice
In the control group, neurons in the CA1 of the hippocampus had a complete structure and the pyramidal cells had a normal morphology and uniform distribution, lined in an orderly manner.In the AD model group, the neurons in CA1 were reduced, degeneration and necrosis were observed in the neurons, the cell volume was reduced, and the pyramidal cells were disorganized and reduced in number.Moreover, the nucleus was pyknotic, staining was deepened, and the shape was irregular with neurotropism.In the QKL group, neuron damage was reduced compared to that in the AD model group.Partial neuronal necrosis, nuclear shrinkage, irregular shape, and deep staining were observed.Pyramidal cells were numerous and lined in an orderly way (Fig. 2A and B).
The changes in brain neurons and tight junctions were observed under the electron microscope.In the control group, the structure of the neurons was intact and the morphologies of the mitochondria and endoplasmic reticulum were normal.In the AD model group, some neurons and nuclei indicated pyknosis and the mitochondria and endoplasmic reticulum were swollen.In the QKL group, the neurons were intact, but the mitochondria and endoplasmic reticulum were slightly swollen (Fig. 2C-E).However, there was no difference in tight junctions between the groups (Fig. 3A).

QKL decreases RAGE expressions in AD mice
A certain expression of the Aβ transport proteins RAGE, LRP1, and Pgp was observed in the control group.Compared with control group, the expression of RAGE in AD model group was significantly higher (P < 0.01); however, there was no significant difference in the expression of LRP1 and P-gp.Compared to the AD model group, the expression of LRP1 and P-gp in QKL group was not significantly different.The QKL group exhibited a significant reduction in the expression of RAGE (P < 0.05) (Fig. 3B-E).

Network pharmacology uncovered the potential targets and signaling pathways regulated by QKL in AD
A total of 237 potential QKL targets and 2343 potential AD targets were gained.As the Veen diagram showed, there were 117 common targets between QKL and AD (Fig. 4A).All targets were listed in the supplementary materials (Tables 1-3).These targets were imported into STRING database to obtain a PPI network, in which the top 8 targets were TP53, IL6, TNF, ESR1, IL1B, MAPK8, EGFR and MAPK1 (Fig. 4B).In order to further clarify the potential mechanism of the effect of QKL on AD, 117 common targets were analyzed by GO and KEGG pathway enrichment analysis.As shown in Fig. 4C, GO analysis might be related to regulation of inflammatory response, reactive oxygen species metabolic process and response to oxidative stress.KEGG analysis mainly involved AGE-RAGE signaling pathway, TNF signaling pathway, PI3K-Akt signaling pathway and Il-17 signaling pathway (Fig. 4D).

Molecular docking
Considering the importance of the AGE-RAGE pathway, molecular docking was performed between RAGE and six bioactive components of QKL injection (quercetin, beta-sitosterol, Stigmasterol, kaempferol, luteolin, and Cholic acid).The results showed that all the binding free energies between RAGE and the six components were less than − 5 kcal/ mol, indicating that the six main components of QKL Injection had good binding with RAGE.The molecular docking mode of Quercetin, kaempferol, luteolin and Cholic acid with RAGE is shown in Fig. 5A-D.The amino acid residues SER2, KPI4, GLN80 of RAGE formed hydrogen bonds with Quercetin, with good affinity and binding energy of − 5.6 kcal/mol (Fig. 5A).The amino acid residues KPI4 and GLN80 of RAGE form hydrogen bonds with kaempferol, and the binding energy is − 5.2 kcal/mol (Fig. 5B).The amino acid residues GLU88, KPI4 and GLN80 of RAGE form hydrogen bonds with luteolin, with good affinity and binding energy of − 5.5 kcal/mol (Fig. 5C).The amino acid residues ASP6, PHE3, GLU88, KPI4 and GLN80 of RAGE formed hydrogen bonds with Cholic acid, with good affinity and binding energy of − 5.4 kcal/ mol (Fig. 5D).

Molecular dynamics simulation
MDS was performed to verify the binding abilities between small molecule compounds and key targets proteins with optimal binding abilities in molecular docking.RAGE, along with quercetin and luteolin, was selected for MDS due to their stronger binding forces exhibited during the molecular docking process.A trajectory of 100 ns was computed using high-performance computing, and the binding stability of RAGE with small molecules was tested based on the root-mean-square deviation (RMSD) system, while the conformational changes at specific sites in the complex were tested based on the root-mean-square fluctuation (RMSF).
In Fig. 6A, the RMSD of the protein stabilized around 5.6 Å, while the ligand's RMSD fluctuated and eventually stabilized at approximately 12.5 Å.In Fig. 6C, the RMSD of the protein stabilized around 5.6 Å, while the ligand's RMSD fluctuated and eventually stabilized at approximately 10.5 Å.This indicates that both the protein and ligand experienced significant fluctuations relative to their initial conformations, suggesting that their initial conformations had moderate stability.After undergoing MDS, a more stable binding conformation was formed based on the original structure.In Fig. 6B and D, the amino acid residues involved in forming interactions with small molecules (indicated by green lines) exhibited lower RMSF values, indicating minimal conformational changes during binding.In contrast, larger RMSF values were Fig. 3. Tight injunction and Aβ transport proteins of cerebrovascular endothelial cells.Tight injunction of cerebrovascular endothelial cells didn't change much over those groups (A).The expression of RAGE in the AD model group was significantly higher than that in the control group, and the QKL group exhibited a significant reduction compared with the AD model group (B and C).The expression of LRP1 and P-gp showed no significant difference over the three groups (D and E).Values are mean ± S.E.M. n = 6 statistical analysis was performed using one-way ANOVA.##P < 0.01 compares with control group.*P < 0.05 compares with AD model group.
observed for the atoms of the small molecules themselves, suggesting that they experienced greater fluctuations during MDS.During the binding process of RAGE with both quercetin and luteolin, amino acids GLU74 and ASN92 played crucial roles, with hydrogen bond formation frequencies of 39 % and 36 %, respectively.In addition to these interactions, multiple hydrophobic bonds and water bridges facilitated binding.Furthermore, intramolecular hydrogen bonds were formed within the ligands to stabilize their binding conformations.

Discussion
The SAMP8 rapid-aging AD mice, an ideal AD model showing agerelated learning and memory impairment, as well as Aβ deposition (Liu et al., 2020b), was utilized in this study.SAMR1 has normal aging characteristics and is often used as a normal control for SAMP8.The apposition of Aβ is believed to play an important role in the pathogenesis of AD.Earlier studies have suggested that the occurrence of AD can be attributed to neurotoxicity of Aβ accumulation in brain, which was closely related to the mutations in Aβ precursor protein (APP), presenilin-1 (PS-1), and presenilin-2 (PS-2) in the brain, resulting in the overexpression of APP, increase of APP or abnormal processing of APP (Li et al., 2012).However, with the study continues, there was no evidence that the production of Aβ increased in the brain in typical and lateonset AD (Sagare et al., 2012).
The latest research demonstrated that soluble Aβ in peripheral tissues can enter the brain through the BBB and lead to AD finally (T et al., 2021).However, only a small amount of Aβ can be transported to the brain through microvessels under normal conditions, owing to the unique structure of the BBB in a healthy brain (Wang et al., 2021).The BBB effectively prevents Aβ in the circulatory system from entering the brain (Sweeney et al., 2018b;van Gerresheim et al., 2021).Aβ enters the brain mainly by disrupting the continuity of cerebral microvascular wall and changing the expression of Aβ transporter (Shang et al., 2016;Liu et al., 2020a;Kim et al., 2014).Therefore, our research focused on the BBB.In the first, we observed the hippocampus under the electron microscope, but found no destruction of tight junctions in AD mice.Therefore, we speculate that soluble Aβ in the blood enters the brain through the transport proteins of cerebral microvascular endothelial Then we estimated the expression of Aβ transportation proteins including RAGE, LRP1 and P-gp.RAGE promotes Aβ transport from the blood to the brain (Cai et al., 2016).Previous studies have shown that RAGE expression in the brain of AD population was 2.5 times higher than that in the normal population (Han et al., 2011).LRP1 and P-gp on BBB were associated with efflux of Aβ in the brain (Zlokovic et al., 2010;Sita et al., 2017).In our study, we observed that RAGE expression in the AD model group was significantly higher than that in the control group, which is consistent with literature reports (Kim et al., 2016).However, we observed no change in LRP1 or P-gp expression in the brains of AD mice.Therefore, increased expression of RAGE on BBB made more soluble Aβ into brain, playing an important role in AD pathogenesis.
According to the above results, we suspect that a treatment aiming at restoring the BBB can improve symptoms of AD.QKL can effectively work on the BBB (Ma et al., 2019;Zhang et al., 2020;Zhu et al., 2011).However, it's unclear whether QKL could interfere with the transport of Aβ.Therefore, this study investigates the intervention of QKL in the  Cognitive dysfunction and memory impairment is one of the most important clinical pictures of AD, so we first tested the effect of QKL on the cognitive function of AD mice.The result showed that the spatial learning and memory abilities were impaired in AD mice; however, QKL improved these abilities evidently.This finding is consistent with our reports that QKL can ameliorate memory impairment (Qiu et al., 2010).
Decreased spatial learning and memory was associated with Aβ level in the brain of AD mice.To confirm this, we tested the Aβ level in the brain of each group.In contrast with the control group, the Aβ level in the brain of the AD model group obviously was higher, which suggests that the increase of Aβ level in the brain may relate to the change in the structure and function of the BBB.QKL can significantly reduce Aβ level in the brain in AD mice, which is consistent with the efficacy of QKL in improving the spatial memory of AD mice.
Meanwhile, we observed the hippocampus under the light and electron microscope.The destruction of brain neurons in the AD model group was obviously observed, especially mitochondria and endoplasmic reticulum, and the destruction of neuronal structures in the QKL group was alleviated.Aβ has been observed to accumulate in the mitochondria of posthumous and alive AD patients (M, A., M. F, and W. B, 2010).Research indicates that mitochondrial structural changes caused by Aβ resulting in increased mitochondrial fragmentation, decreased mitochondrial fusion, disruption of electron transport chain, and synaptic damage (PH et al., 2010;PH and MF, 2008).Recent evidence supports that accumulation of Aβ-peptides may upset the ER homeostasis resulting in ER stress and imbalance of calcium homeostasis, which involve in AD (C, H. and M. B, 2014;AC et al., 2011;P et al., 2016).Increased accumulation of Aβ aggravated the destruction of mitochondria and endoplasmic reticulum, causing damage to neurons and cognitive impairment, while Aβ level decreased by QKL has certain reduction to these damages.
Then we studied the effect of QKL on the BBB.There are two types of Aβ Transport proteins on the BBB.RAGE is responsible inward transportation, while LRP1 and P-gp are responsible for outward transportation.QKL significantly reduced RAGE expression in AD mice.However, QKL had no effect on the tight junctions, the expression of LRP1 or P-gp on BBB in AD mice.This also confirmed our previous conjecture that QKL greatly reduced the possibility of AD by reducing the entry of Aβ from the periphery, not increasing the outflow or closing tight junctions.
Further Network pharmacology analysis indicated that the mechanism of QKL affects AD included AGE-RAGE signaling pathway, TNF signaling pathway, Il-17 signaling pathway, inflammatory response and oxidative stress, which is consistent to the change of RAGE in our experiment.RAGE mainly exerts pro-inflammatory effects, whose important ligand is AGE.RAGE could cause activate NF-κB, inducing a series of inflammatory cascade related to inflammatory reaction and oxidative stress, further damaging the structure and function of brain neuron, hippocampus and vascular endothelium.Activation of NF-κB also induces further synthesis of RAGE, leading to increased inflammatory response (Wu et al., 2021).Abnormal activation of the TNF signaling pathway promotes neuroinflammation and neuronal apoptosis, and contributes to the pathological processes of diseases such as MCI and AD (Ding et al., 2021).Il-17 is mainly derived from helper T cell 17, which could destroy the BBB, promote neuronal apoptosis and inflammatory response, by activating neutrophils and stimulating the production of pro-inflammatory cytokines (Chen et al., 2022).
The results of molecular docking also confirmed the effect of QKL on RAGE.We selected the four main active components of QKL Injection, some of which have been reported to have the ability to cross the bloodbrain barrier and have neuroprotective effects.Quercetin is a plantbased flavonol belonging to the polyphenol family of flavonoids, which has been proven to protect against Aβ-induced pathology (Schiavi et al., 2023).Kaempferol is a natural flavonoid existing in a range of plants and foods, which could improve cognitive impairment in AD through its effects on oxidative stress, neuroinflammation, and Ab induced neurotoxicity (Babaei et al., 2018;Nejabati and Roshangar, 2022).Luteolin is a type of flavonoid compound with anti-inflammatory and neuroprotective activities (Kou et al., 2021;Singh et al., 2022).Recent researches have shown that luteolin can inhibit endoplasmic reticulum stress-dependent neuroinflammation in AD mice, and amyloid protein β induced oxidative stress and mitochondrial damage (Kou et al., 2021;He et al., 2023).Studies have also found that hydrophilic cholic acids had a protective effect on the BBB (Palmela et al., 2015).In this study, we found that quercetin, kaempferol, luteolin and acid could effectively dock with RAGE.During MDS, the binding conformations of RAGE with both quercetin and luteolin were optimized based on their original structures, resulting in increased stability and higher affinity between the interacting partners.These computational data provide some evidence for the reliability of network pharmacology predictions and experimental results, which indicated that RAGE is a key target during QKL treatment of AD.
As shown in Fig. 7, QKL significantly inhibited the transport of Aβ by inhibiting the expression of RAGE.And then it decreased the Aβ level, consequently reduced Aβ's damage to mitochondria and endoplasmic reticulum of neurons, and finally improved cognitive impairment in AD mice.The above findings suggested that RAGE may be the therapeutic target of QKL and we further used network pharmacology, molecular docking, and molecular dynamics simulation to prove this finding.However, there are several limitations to this study.While we have preliminarily revealed the potential benefits of QKL for treating AD, further research is needed to establish a positive control group.And it is unknown whether the increase of soluble Aβ in brain can lead to increased deposition.Whether QKL can enter BBB and directly affect neurons also needs to be further studied.

Declaration of competing interest
The authors declare that there is no conflict of interests regarding the publication of this article.

Fig. 1 .
Fig. 1. Results of Morris water maze experiment.Mean escape latency time to the visible platform during cued navigation training for five days (A).Average swimming speed during the navigation training (B).Crossing numbers of the target platform (C).Spent time to the target quadrant (D).Swimming route of exploration experiment (E).Aβ levels in brain tissue of mice measured by ELISA (F).Values are means ± S.E.M. n = 60.#P < 0.05 compares to the control group.*P < 0.05 compares to the AD model group.Values are mean ± S.E.M. n = 6.###P < 0.001compares with control group.*** P < 0.001 compares with AD model group.

Fig. 2 .
Fig. 2. The neurons in the hippocampus CA1.Representative photomicrographs of the Nissl staining were shown in the Control, model and QKL groups, scale bar = 100 μm (A) and 25 μm (B).Neurons (C), mitochondria (D) and endoplasmic reticulum (E) of neurons were observed and photographed under the transmission electron microscopes.Scale bar = 5 μm (C) and 200 nm (D and E).

Fig. 4 .
Fig. 4. Network pharmacology analysis of QKL treating AD. (A) (B) The PPI network of common targets.(C) Analysis of GO functional enrichment.(D) Analysis of KEGG enrichment.

Fig. 6 .
Fig. 6.Molecular dynamics simulation.The RMSD (A) and RMSF (B) results of RAGE with quercetin.The RMSD (C) and RMSF (V) results of RAGE with luteolin.

Fig. 7 .
Fig. 7. Diagram illustrating the mechanism of QKL regulating cognitive dysfunction and decreasing Aβ accumulation by depressing the expression of RAGE.
• C ± 2 • C) under a 12 h/12 h light/dark period.The experimental design was approved by the Animal Ethics Committee of Dongzhimen Hospital of Beijing University of Traditional Chinese Medicine.