The potential indications enriched in DO combined with the pharmacological effects of approved drugs screened by CMap were used for drug repurposing of QR, and the most representative indication was selected for subsequent research. The representative indication targets were retrieved based on MalaCards (https://www.malacards.org/), TTD, PharmgKB (https://www.pharmgkb.org/), OMIM and DisGeNET (https://www.disgenet.org/). In addition, to collect the QR targets, CTD (https://ctdbase.com), DrugBank (https://go.drugbank.com/), PubChem (https://pubchem.ncbi.nlm.nih.gov/), HIT (http://www.badd-cao.net:2345/), PharmMapper (http://www.lilab-ecust.cn/pharmmapper/), and SwissTargetPrediction (http://www.swisstargetprediction.ch/) databases were used. The QR targets were matched with the targets of the relevant indication to obtain the common targets of QR treatment indication.

The common targets were put into the STRING database (https://cn.string-db.org/), where the organism was set as “homo sapiens” and the data file was imported into Cytoscape for interaction network visualization. Then, the CytoHubba plugin (http://apps.cytoscape.org/apps/cytohubba) was 103 used to analyze the PPI network and calculate the core targes.

All the common targets were imported into Metascape (https://metascape.org/) for functional enrichment analysis (Duan et al., 2019; Duan et al., 2021). The Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment results with significant statistical differences (p < 0.05) were visualized using the OmicShare tools (https://www.omicshare.com/tools).

C57BL/6 female mice, 8–10 weeks of age (18–20 g), were obtained from SPF Biotechnology Co., Ltd (Beijing, China), and housed in specific pathogen-free conditions at the First Affiliated Hospital of Henan University of Chinese Medicine (animal ethics committee approval No. YFYDW2023025), Henan, China. All efforts were made to reduce animal suffering, and the Institutional Committee on Care and Use of Research animals approved the procedures used in this study. The inactivated Mycobacterium tuberculosis H37Ra (No. 231141, Difco) was added to Freund’s incomplete adjuvant (No. F5506, Sigma) to prepare the 10 mg ml^-1 solution, and the same volume of myelin oligodendrocyte glycoprotein peptide (MOG35-55, MEVGWYRSPFS-RVVHLYRNGK; L2880; GenScript) solution (2 mg ml^-1) was dissolved in sterile PBS to prepare oil antigen emulsion.

The C57BL/6 mice were randomly divided into the normal control (NC) group, the EAE group and EAE + QR group (n = 6 in each group). EAE was induced by MOG 35-55, as previously described (M. L. Zhang et al., 2017). Briefly, each mouse was anesthetized with 1% phenobarbital sodium and then immunized subcutaneously. On days 0 and 2 post-immunization (p.i.), 200 ng pertussis toxin (No. 180, List Biological) was intraperitoneally injected into mice. The treatment by oral gavage started from day 10 p.i. The EAE + QR group mice were treated with 50 mg (kgd)⁻¹ QR (Hendriks et al., 2004) (No. C10798568, Macklin) suspension prepared by using 0.5% sodium carboxymethyl cellulose (CMC-Na) as the vehicle. NC and EAE groups mice were treated with 0.5% CMC-Na in equal volumes. The EAE model was scored on a five-point scale for neurological function (Glenn et al., 2017): 1-animal tail weakness; 2-weakness of hind limbs; 3-hind limb paralysis; 4-anterior and posterior limbs were paralyzed; 5-dying or dying. In addition, NC group mice were used as naive control. The above trials were repeated twice to ensure the reproducibility of clinical efficacy.

To observe mice’s central nervous system (CNS) inflammation and demyelination. The mice were anesthetized with 1% pentobarbital sodium and perfused with 0.9% sodium chloride solution on day 21 p.i. The left brain was quickly removed and fixed with 4% paraformaldehyde, embedded in paraffin, sectioned, and stained with HE inflammatory infiltration staining and LFB myelin staining. The scoring criteria for inflammatory infiltration (Zhao et al., 2018) were as follows: 0-no sign of inflammation; 1-scattered inflammatory cells; 2-some inflammatory cells and karyopyknosis; 3-perivascular inflammatory cell infiltrate; 4-marked inflammatory cell infiltration into the parenchyma. Demyelination level scoring criteria (Dou et al., 2021) were as follows: 0-no demyelination; 1-rare demyelination; 2-a few demyelinated areas; 3-a large area of demyelination.

Based on the GEO database, a gene expression dataset numbered GSE13899 was screened (Boomgaarden et al., 2010). The dataset contains gene expression profiles of CD14⁺ monocytes before and after 2 weeks of QR (150 mg d^-1) supplementation in three healthy volunteers. The results show that a total of 156 DEGs of QR have been screened, including 135 upregulated DEGs and 21 downregulated DEGs with p-value <0.05 and |log FC| > 1 as the threshold: (Figure 2).

Under the condition of Norm CS ≥ 1.3, we screened 2312 compounds with similarity to QR gene expression signatures. According to the DrugBank database, the top 15 approved drugs of the above compounds were screened for the potential pharmacological effects and clinical indications of QR (Table 2; Figure 3A). The results revealed that the clinical pharmacological effects of the top 15 approved drugs are mainly involved in immune, nervous, endocrine, respiratory, digestive, and other systems, including anti-inflammatory, immunosuppressive, hypoglycemic, anti-tumor, anti-asthma, antimalarial, suggesting that QR may have the similar pharmacological effects and clinical indications.

Based on Coexpedia platform, the QR-regulated DEGs are mainly enriched in 84 significant disease entries (p < 0.05), and the top 10 disease entries were screened for display according to the p-value. As shown in Figure 3B, the primary potential indications for QR include neuromuscular diseases (muscular dystrophy), autoimmune diseases (multiple sclerosis), mouth diseases (gingivitis), joint diseases (arthritis), neoplasms (liposarcoma, leukemia, and lymphoma), metabolic diseases (diabetes mellitus and hemochromatosis), and CNS diseases (adrenoleukodystrophy). Integrating analysis of the above clinical indications cured by the top 15 approved drugs and the top 10 enriched DO diseases, the result found that the potential indications of QR are primarily related to multiple sclerosis (MS), osteoarthritis, type 2 diabetes mellitus, and acute leukemia. Notably, 7 of the above 15 top drugs have been confirmed to have better therapeutic effects on MS, including five drugs already used clinically and two drugs proven by animal experiments. Therefore, MS was selected as a representative disease for the subsequent study to verify the feasibility of the drug repurposing method for QR.

After identifying MS as a potential indication for QR therapy, targets were collected in the disease database using the keywords “multiple sclerosis”, “relapsing-remitting multiple sclerosis”, “secondary progressive multiple sclerosis”, “primary progressive multiple sclerosis”, and “progressive recurrent multiple sclerosis” respectively. As shown in the Venn diagram (Figure 4A), 370 QR and 271 MS targets were collected, with 44 common targets. Through the analysis of the PPI network (Figure 4B) by CytoHubba, six core targets were screened out, which were tumor necrosis factor (TNF-α), interleukin 6 (IL-6), interleukin 1 β (IL-1β), interferon γ (IFN-γ), interleukin 17A (IL-17A), and interleukin 2 (IL-2), respectively. These targets are closely related to the inflammatory immune response of autoimmune diseases (Ge et al., 2018), suggesting that QR may treat MS mainly by affecting these targets.

The results of the Gene Ontology (GO) enrichment analysis are shown in Figure 4C. As for the biological process, QR affected cellular response to cytokine stimulus, inflammatory response, cellular response to an organic substance, response to external stimulus, and immune response. For the cellular component, QR has mainly located in the exterior of the cell plasma membrane, membrane, and axon. For the molecular function, the effects of QR on MS were related to the binding of a cytokine receptor, growth factor receptor, TNF receptor superfamily, G protein-coupled receptor, and chemokine receptor.

The KEGG enrichment indicated that 44 common targets were enriched in 87 pathways (p < 0.05). The top 15 main pathways are TNF signaling pathway, IL-17 signaling pathway, Toll-like receptor signaling pathway, NOD-like receptor signaling pathway, and NF-κB signaling pathway (Figure 4D). KEGG functional annotation (Figure 4E) indicates that QR may mainly affect the immune system by regulating cell signaling pathways, thereby interfering with immune and neurodegenerative diseases, such as MS.

The symptoms such as loss of tail tonicity, staggering gait, hind-limb paralysis, four-limb paralysis, and even death appeared sequentially in EAE mice. The first clinical score of the CMC-Na-treated EAE group mice appeared around day 9 p.i. and increased in the following days. The disease peaked on day 19 p.i., with reduced body weight (Figure 5A). Notably, the EAE + QR group mice were heavier (p < 0.01), and the mean clinical scores and cumulative clinical scores were lower (p < 0.01) than the EAE group mice (Figures 5B, C).

All mice were sacrificed on day 21 p.i., and the brain was collected for pathological analysis. HE staining and LFB staining of the brain showed an extensive infiltration of inflammatory cells, perivascular cuffing and large areas of demyelination in the CMC-Na-treated EAE group, while QR treatment significantly inhibited the infiltration of inflammatory cells into the CNS of EAE mice (p < 0.01, Figure 5D), with decreased demyelination (p < 0.05, Figure 5E). These results indicated that QR had the potential to ameliorate EAE.

TNF-α was the most connected target (degree = 41) to other PPI network protein nodes. Studies have shown that TNF-α binding to Tumor Necrosis Factor Receptor Type 1 (TNFR1) can trigger a series of proinflammatory cytokines and chemokines, thereby increasing the immune cascade of CNS (Hilliard et al., 2020; Taghipour et al., 2022). In addition, combined with the results of drug repurposing of QR, we speculate that QR may play an anti-inflammatory and immunosuppressive pharmacological role by inhibiting TNF-α and TNFR1, thereby alleviating the autoimmune response of MS. Therefore, this study used QR as a docking ligand, TNF-α (PDB ID, 2AZ5; resolution, 2.1 Å) and TNFR1 (PDB ID, 1FT4; resolution, 2.9 Å) as docking target proteins. The reported small molecule inhibitors of these two target proteins were downloaded from the PDB database, and the active sites of the inhibitors were used as the docking pockets of QR. Before QR docking with these two protein targets, all their small molecule inhibitors should be extracted.

The docking protocol was validated by redocking the co-crystallized inhibitors “307” (He et al., 2005) and “703” (Carter et al., 2001) of these two proteins to their active sites, respectively. Root mean square deviation (RMSD) between the docked structures and the initial structures was less than 2.00 Å. The redocking results of “307” and “703” signified that Auto Dock Vina software is reliable (Trott and Olson, 2010). The protocol can be used for docking other compounds. It is generally believed that the binding energy < −4.25 kcal mol^-1 indicates that there is a certain binding activity between the ligand small molecule and the receptor protein; the binding energy < −5.0 kcal mol^-1 indicates that there is a good binding activity between the two; the binding energy < −7.0 kcal mol^-1 indicates that the ligand and receptor have strong binding activity (Hsin et al., 2013). Reassuringly, results showed that QR bound to TNF-α and TNFR1 protein targets through visible hydrogen bonds and strong electrostatic interactions. Moreover, the original inhibitor active sites of these two target proteins were successfully docked by QR, and had low binding energy of −7.1 (2AZ5) and −6.7 (1FT4) kcal·mol^-1, indicating highly stable binding (Table 3; Figure 6). This suggested that QR might play an anti-inflammatory and immunosuppressive effects by inhibiting TNF-α and TNFR1 targets.

Based on molecular dynamics simulation, the data of RMSD and Root Mean Square Fluctuation (RMSF) of TNFα-QR and TNFR1-QR in protein were obtained. Compared with the first frame, the TNFα-QR complex reached stability after 10 ns(RMSD value is 0.30–0.44 nm). The RMSD oscillation amplitude of QR is about 0.14 nm TNFR1-QR complex was stable after 28 ns(RMSD 0.46–0.56 nm). The RMSD oscillation amplitude of QR is about 0.10 nm. It is suggested that during the 100 ns simulation process, the systems of TNFα-QR and TNFR1-QR quickly reached stability without excessive fluctuations (Figures 7A, C).

Another stability factor is the fluctuations of protein during simulation that were evaluated by analyzing RMSF values (Figures 7B, D). On RMSF values, TNF-α and TNFR1 active site residues interacting with QR were all stable with RMSF values below 0.2nm. In particular, most of the RMSF values of the residues that form hydrogen bonds between TNFα-QR (TYR151, SER60, TYR59, and LEU120) and TNFR1-QR (LYS35 and GLU64) complexes are less than 0.15 nm. It can be inferred that QR fits nicely on the active site and maintains interaction with amino acid residues.

Six potential core target genes (TNF-α, IL-6, IL-1β, IFN-γ, IL-17A, and IL-2) were identified by PPI network analysis. To verify the above findings, we determined the mRNA levels of these six target genes in the spinal cord tissues of NC, EAE, and EAE + QR groups by reverse transcription-polymerase chain reaction (RT-PCR). The expressions of TNF, IL6, IL1B, IFNG, IL17A, and IL2 in the CNS of EAE mice were elevated (all p < 0.01) observably compared with that in NC mice, while their expressions were significantly reduced (all p < 0.01) after QR treatment (Figure 8).