New perspective: the multi-targets mechanism of hydroxychloroquine in the treatment of rheumatoid arthritis based on network pharmacology

Network pharmacology is a new method of bioinformatics in exploring drug targets in recent 3 years. Hydroxychloroquine (HCQ) is a multi-targets drug that are clinically effective in rheumatoid arthritis (RA) but whose mechanism is not well understood. The predicted targets of HCQ and the proteins related to RA were returned from databases. Followed by protein-protein interaction (PPI) network, the intersection of the two group of proteins was conducted. Furthermore, gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment was used to analyse these proteins in a macro perspective. Finally, the candidate targets were veried by molecular docking. and with their network which regulate MAPK, cell and Molecular docking analysis shows that and stacking are main of chemical force.


Background
Rheumatoid arthritis (RA) is de ned as one of the chronic autoimmune diseases, characterized by destruction of joints and connective tissues with associated metabolic, vascular, bone and psychological comorbidities [1][2]. The dysregulated innate and adaptive immunity characterized by immune responses against autoantigens, disordered cytokine secretion, osteoclast and chondrocyte activation mediated by immune complex-complement pathway [2][3]. About 0.5-1.0% of adults affected by RA in developed countries while the ratio was around 0.4% in South East Asia and Eastern Mediterranean region. The prevalence of RA is higher in female than that in male and the ration also rises with age. The quality of life is severely affected by persistent and progressive joint in ammation and damage which could result in disability nally [4].
Different combinations of NSAIDs and glucocorticoids are mostly used to mitigate the pain and in ammatory effects. In addition, DMARDs such as hydroxychloroquine (HCQ), methotrexate, sulfasalazine and le unomide are also widely used to protect joints by slowing down the in ammatory arthritis. In recent years, biologics that aim to relieve in ammation through depleting B lymphocytes or inhibiting in ammatory mediators such as interleukin 6 or tumor necrosis factor α pathways. It has been demonstrated that the early intervention of DMARDs and the availability of timely medications could greatly improve the prognosis in a large proportion of RA patients [6]. When combined with other DMARDs, HCQ could provide moderate clinical bene t to patients in terms of the control in RA activity [7].
Recently, HCQ has been revealed to bene t the metabolic pro le and to a lesser extent cardiovascular events in RA patients [8].
However, the mechanism of HCQ in the treatment of RA is still unknown. With the rapid progress of Bioinformatics, Systematic Biology and Polypharmacology, network-based drug discovery and evaluation is considered a promising approach toward more cost-effective drug development. Network pharmacology introduces a paradigm shift from the current "one research-based target, one drug" strategy to a novel version of the "network multi-targets" strategy [9][10][11][12].
In our research, we evaluated potential targets of HCQ on RA patients by using the approach of network pharmacology. Firstly, we predicted potential molecular targets of HCQ. Then we investigated the intersection of these targets with RA-related genes. Protein-protein interaction (PPI) network was built to enlarge the amount of proteins which are closely related to the mutual genes. Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment was conducted on the enlarged amount of proteins. Finally, we performed docking studies to verify the chemical force that allow HCQ binding to its predicted targets. Our results may be helpful to further nd how HCQ can be effective against RA and and facilitate the development of novel drugs.

2.1.Predicted Target Proteins of HCQ
The chemical structure (SMILES) of HCQ was searched on PubChem website and made target prediction on different databases (SwissTargetPrediction, DrugBank and PharmMapper) based on it. The species was limited to "Homo sapiens". A total of 100 human proteins that possibly targeted by HCQ were returned [13].
2.2.Collection of Related Genes associated with RA Targets related to RA were returned from databases of OMIM and Genecards by searching with the key word of "rheumatoid arthritis" and "Homo sapiens". A total of 4329 human genes related to RA were returned [14].

2.3.Screening of Pivotal Target Proteins as well as GO and KEGG Analysis
The plug-in "Bisogenet" in Cytoscape software was used to conduct the PPI network of the mutual targets between HCQ and RA. The key targets were screened according to the parameters of "degree", "betweenness" and "closeness" calculated by Bisogenet. GO and KEGG enrichment analysis was conducted on the database of DAVID and Metascape to make macroscopic evaluation of target genes about their molecular function and systemic involvement [15].

2.4.Molecular Docking
AutoDock 4.2 software was used to analyze the chemical interactions between proteins and small molecules. The 3D crystal structures of potential targets of HCQ were found from the database of RCSB-PDB. Their structures were modi ed with AutoDock software including ligand and water removal, hydrogen addition, amino acid optimization and patching. ChemBioDraw 3D software was used to visualize the 3D chemical structures and minimize their energy. Results were saved in MOL.2 format. Molegro Virtual Docker predicted docking partners by comparing the predicted conformation with the observed crystal structure. A model was considered accurate if its root mean square deviation from the crystal structure was ≤ 2 Å; reliable if the deviation was ≤ 4 Å, and reliable or accurate if the deviation was less than 3 Å [16].

3.1.The Predicted Targets of HCQ based on databases
The potential targets of HCQ were predicted by databases according to the 2-dimensional and 3dimensional chemical structure (Fig. 1A). The top 100 targets (Fig. 1B) of them with high index of possibility were chose. Most of these targets are G protein-coupled receptors (29%), kinase (26%) and surface antigen (13%) (Fig. 1C).

3.2.Genes associated with RA and Topological Network Analysis
The intersected 64 of the total 4329 genes involved in RA with potential targets of HCQ were conducted with PPI network by cytoscape software through the plug-in of "Bisogenet". Then enlarged result of 3316 more proteins associated with the 64 proteins were returned. A total of 82340 edges (interactions) of the 3316 nodes (targets) could be seen in Fig. 2A. In order to show the most important nodes of these 3316 proteins, the index of degree 56 (the median) was used as a criterion to screen them preliminarily (Fig. 2B). Further more, 913 nodes and 39807 edges were screened by the index of betweenness 335 and closeness 0.515. The returned 147 related proteins and their 3767 interrelationships which may play important roles in the treatment of RA with HCQ could be seen in Fig. 2C. Finally, top 20 of them was screened (Table.1).

Gene Ontology and Pathways Enrichment of the 3316 Related Proteins
A total of 3316 human genes which participate in the mechanism of HCQ in curing RA, were conducted with GO and KEGG enrichment (Fig. 3). Macrobiological evaluation of these proteins was performed. According to GO enrichment: these proteins were mainly located in colin-1-rich granule lumen, colin-1rich granule, vesicle lumen and so forth (Fig. 3A); as to molecular functions, these proteins mainly take part in ATP binding, adenyl nucleotide binding, RNA binding and so forth (Fig. 3B); the biological process of HCQ acted on the network mainly in inhibiting cell activation, myeloid leukocyte activation and regulating exocytosis (Fig. 3C); KEGG pathway analysis further showed that these proteins were mainly involved ErbB, HIF-1, NF-κB, FoxO, Chemokine, MAPK, PI3K/Akt pathway and so forth (Fig. 3D).

3.4.Molecular docking
A total of 20 candidate potential targets of HCQ (Table.1) were performed with molecular docking analysis which provided a visual explanation of the interaction between HCQ and its potential protein targets associated with RA (Table.2 Fig. 4). The score below − 20 was considered to have better docking power. Smoothened homolog (SMO), sphingosine kinase (SPHK) 1, SPHK2 and gatty-acid amide hydrolase 1 (FAAH1) were the most possible target of HCQ in curing RA which had top 5 highest binding force and spatial t with HCQ. We found that hydrogen bond, ionic bond and π-π stacking were the main forms of interaction. For instance, the hydroxyl, amino and carbonyl groups of HCQ formed hydrogen bonds with the proteins, while with the benzene ring and aromatic ring of HCQ engaged in π-π stacking ( Fig. 4).

Discussion
HCQ is a multi-targets antimalarial drug, which is widely used in rheumatology. However, the exact pharmacological mechanism is still unclear. As one of the DMARDs, HCQ could relief RA activity and improve the prognosis of it. Antimalarial agents have numerous biological effects that are responsible for their immunomodulatory actions [5][6][7][8]. According to our results of network pharmacology, SMO, SPHK1, SPHK2 and FAAH1 play vital roles of HCQ in the treatment of RA.
Synovitis is the main characteristic of RA. Excessive proliferation of broblast-like synoviocytes (FLSs) and synovial angiogenesis are the most important contributors to the progression of RA synovitis and joint destruction. Sonic hedgehog (SHH) signaling pathway plays a pivotal role in FLSs proliferation in a SMO-dependent manner. Upregulation of SMO promotes proliferation of FLSs [17][18]. Targeting SHH signaling pathway may help control joint damage in patients with RA [19]. According to our analysis by network pharmacology, the binding of HCQ with SMO is involved in the pathological process of synovitis through the SHH pathway (Fig. 5).
SPHK (including SPHK1 and SPHK2) is a key lipid kinase in sphingolipid metabolic pathway, which phosphorylate phingosine into sphingosine-1-phosphate (S1P) [20][21]. The importance of SPHK and S1P in in ammation and angiogenesis has been demonstrated in many hyperproliferative/in ammatory diseases such as RA [22]. The level of S1P exhibits signi cantly higher than those non-in ammatory osteoarthritis counterparts [23]. Furthermore, S1P receptor was found to be expressed in RA synovium, which means that in ammatory cytokines would further promote the progress of synovitis [21]. As mentioned above, excessive proliferation of FLSs was induced mainly through SHH pathway. In addition, in ammatory further accelerate the process through mitogen-activated protein kinases/extracellular signal-regulated kinases (MAPK/ERK) signaling pathway. For instance, interleukin 6, tumor necrosis factor-α, angiopoietin 1, neuropilin 1 and vascular endothelial growth factor regulate the lesion of rheumatoid joint and the proliferation of FLSs through the MAPK/ERK pathway [24]. SphK blockade suppresses cytokines and MMP-9 release in RA peripheral blood mononuclear cells [23]. Targeting SPHK may help control joint damage in aspect of in ammation. According to our analysis by network pharmacology, the binding of HCQ with SPHK1 and SPHK2 plays important role in inhibiting the in ammatory process of synovitis (Fig. 5).
In recent years, the role of the endocannabinoid (EC) system in the pathogenesis of RA attracted more attention of researchers. EC system modulates function of immune cells and mesenchymal cells such as broblasts, which contribute to cartilage destruction in RA [25]. The action of EC system in immune system regulation, via primary cannabinoid receptor (CB) activation, followed by inhibition of production of pro-in ammatory cytokines, auto-antibodies and matrix metalloproteinase (MMPs), FLSs proliferation and T-cell mediated immune response [26]. Since FAAH is a major EC-degrading enzyme, the therapeutic possibility of FAAH inhibition is promising [27]. Thus, due to the result of network pharmacology, the binding of HCQ with FAAH acts as one of the multi-targets mechanism in the treatment of RA (Fig. 5).

Conclusions
Collectively, 4 key targets (SMO, SPHK1, SPHK1 and FAAH) (Fig. 5) involving 3316 proteins become the multi-targets mechanism of HCQ in the treatment of RA due to our research, through the pathway related to proliferation of broblast-like synoviocytes and inhibition of production of pro-in ammatory cytokines. According to GO analysis, they totally enrich in the functions of: regulation of cell activation, myeloid leukocyte activation, regulated exocytosis and so forth. ErbB, HIF-1, NF-κB, FoxO, Chemokine, MAPK, PI3K/Akt pathways and so forth are the main pathways that took part in the multi-targets machanism according to KEGG analysis. This paper introduced the new concept of network pharmacology into the clinical treatment of rheumatic diseases with multi-targets drugs, which is conducive to the exploration and evaluation of multi-targets drugs that are clinically effective in rheumatic treatment but whose particular mechanism is not well understood. In addition, it can also provide an explanation of drug usage and guidance for relevant scienti c research.

Declarations
Ethics approval and consent to participate: None declared.
Consent for publication: The article is approved by all the co-authors for publication.
Availability of data and materials: The data used or analysed during the current study are available from the corresponding author on reasonable request.
Competing interests: The authors declare that they have no con ict of interests.     Molecular models of HCQ binding to its predicted protein targets. Proteins (A) SMO, (B) SPHK1, (C) SPHK2, (D) FAAH are shown interacting with a HCQ molecule, represented by a green stick model. hydrogen bonds and demarcate π-π interactions are shown in these models.