Abstract

   Qing-Luo-Yin (QLY) is a traditional Chinese medicine (TCM) formula used
   to treat Hot Syndrome-related rheumatoid arthritis (RA). Previously, we
   uncovered partial mechanisms involved in the therapeutic actions of QLY
   on RA. In this study, we further elucidated its anti-rheumatic
   mechanisms and investigated its possible interactions with methotrexate
   (MTX) in vivo using an integrating strategy coupled with network
   pharmacology and metabolomics techniques. Chemical composition of QLY
   was characterized by HPLC analysis. Collagen induced arthritis (CIA)
   was developed in male SD rats. The CIA rats were then assigned into
   different groups, and received QLY, MTX or QLY+MTX treatments according
   to the pre-arrangement. Therapeutic effects of QLY and its possible
   interactions with MTX in vivo were evaluated by clinical parameters,
   digital radiography assessment, histological/immunohistochemical
   examination, and serological biomarkers. Mechanisms underlying these
   actions were deciphered with network pharmacology methods, and further
   validated by metabolomics clues based on UPLC-Q-TOF/MS analysis of
   urines. Experimental evidences demonstrated that QLY notably alleviated
   the severity of CIA and protected joints from destruction. Re-balanced
   levels of hemoglobin and alanine transaminase in serum indicated
   reduced MTX-induced hepatic injury and myelosuppression under the
   co-treatment of QLY. Network-based target prediction found dozens of RA
   related proteins as potential targets of QLY. Upon the further
   biological function enrichment analysis, we found that a large amount
   of them were involved in nucleotide metabolism and immune functions.
   Metabolomics analysis showed that QLY restored amino acids, fatty
   acids, and energy metabolisms in CIA rats, which solidly supported its
   therapeutic effects on CIA. Consistently to findings from network
   pharmacology analysis, metabolomics study also found altered purine,
   pyrimidine, and pentose phosphate metabolisms in CIA rats receiving QLY
   treatment. All these clues suggested that inhibition on nucleic acid
   synthesis was essential to the immunosuppressive activity of QLY in
   vivo, and could contribute great importance to its therapeutic effects
   on CIA. Additionally, QLY induced significant antifolate resistance in
   rats, which would prevent folate from depletion during long-term MTX
   treatment, and should account for reduced side effects in combination
   regimen with MTX and QLY.

   Keywords: Qing-Luo-Yin, rheumatoid arthritis, network pharmacology,
   metabolomics, nucleotide metabolism

Introduction

   Rheumatoid arthritis (RA) is a prevalent systematic autoimmune disease,
   which is characterized by severe joint destruction and chronic local
   inflammation ([35]Smolen and Aletaha, 2009). Conventional treatment
   approaches for RA patients are mainly dependent on disease-modifying
   anti-rheumatic drugs (DMARDs) ([36]Tanaka, 2016). In the past decades,
   methotrexate (MTX) has become the anchor DMARD, and is extensively
   adopted as first-line regimen for the superior efficacy and economical
   merits ([37]Cronstein, 2005). It significantly slows the progress of
   disease, and minimizes the detrimental effects on joints. However,
   there are a number of patients with inadequate response to it.
   Furthermore, despite the fact that low dose of MTX is well tolerated,
   the treatment usually accompanies with remarkable side effects
   ([38]Dubey et al., 2016).

   Traditional Chinese medicine (TCM) as one of the oldest continuously
   surviving traditional medicines is derived from clinical practices in
   ancient China. Much different from Western medicine, TCM emphasizes the
   integrality of body ([39]Feng et al., 2006), and aims at multiple
   pathogenesis factors simultaneously upon the exact diagnosis of Zheng
   (a term in TCM, generally encompasses etiology, pathology and disease
   location) ([40]Yu et al., 2006; [41]Su et al., 2012). Contrarily,
   Western medicine usually modulates one single defined target or symptom
   based on the methodology of reductionism. By the development of system
   biology, limits of this strategy are gradually disclosed. Western
   treatments usually fail to achieve effective and sustainable outcomes
   in systematic and chronic diseases, and accompany with high risk of
   side effects ([42]Feng et al., 2006; [43]Yu et al., 2006). By
   comparison, TCM exhibits notable merits of efficacy and safety in many
   cases, and provides us an efficient alternative for treatments of some
   complicated diseases, such as RA. Objectively speaking, both the
   medical systems have their own advantages and shortcomings, and the
   integrated therapeutic approach would be more effective and safer when
   well optimized.

   Traditional Chinese medicine formulas are usually composed of many
   herbs, which results in the complex chemical composition. Hence, it is
   difficult to fully understand their therapeutic mechanisms. High
   throughput techniques give us helpful tools to screen out potential
   bioactive ingredients from them. However, the dilemma is still hard to
   be resolved, for the formula should be taken as a whole but not the sum
   of some well investigated natural compounds. This situation makes
   network-oriented approaches more preferable ([44]Li, 2015). To achieve
   a comprehensive and systematic understanding of their therapeutic
   mechanisms, new research strategies based on computational simulation
   and prediction flourish recent years ([45]Guo et al., 2017). As one of
   these newly developed disciplines, network pharmacology possesses
   obvious advantages over conventional methods in elucidation of
   comprehensive mechanisms ([46]Li et al., 2007; [47]Hopkins, 2008). The
   network target-based concept could largely reflect complicated
   interactions between biomacromolecules and chemical ingredients. The
   research technique constructed under this guideline is regarded as a
   representative method of emerging network pharmacology ([48]Barabasi et
   al., 2011), and has been successfully applied in many TCM related
   research fields, such as therapeutic mechanisms elucidation ([49]Zhang
   Y. et al., 2015; [50]Zhang et al., 2016), new pharmacological actions
   prediction ([51]Zhang et al., 2018) and potential toxic ingredients
   screening ([52]Zhang B. et al., 2015). Meanwhile, metabolomics focuses
   on systematic analysis of metabolites from drugs treated objects. It
   gives us an opportunity to obtain the panoramic view of network effects
   of formulas on bodies and testify results generated by computational
   modeling analysis.

   Xin’an medical family is an important TCM academic school originated in
   South Anhui district in Song dynasty, and Zhang-Yi-Tie is one of its
   main branches still flourishing nowadays. Qing-Luo-Yin (QLY) is widely
   regarded as the representative formula of this sect, which was created
   by its 14th generation Jiren Li, a famous contemporary TCM master based
   on Xin’an medical theory and his medical experiences. It is composed of
   four components: Kushen (radix of Sophora flavescens Ait.),
   Qingfengteng [caulis of Sinomenium acutum (Thunb.) Rehder and E. H.
   Wilson], Huangbai (cortex of Phellodendron chinense C. K. Schneid.),
   and Bixie [rhizome of Dioscorea collettii var. hypoglauca (Palib.) S.
   J. Pei and C. T. Ting]. As a Cold natured formula, it is mainly used to
   treat the Hot Syndrome-related RA. Previously, we revealed that targets
   of QLY against RA-related key biological processes mainly involved in
   angiogenesis, inflammatory response and immune functions by using a
   network target-based research technique, but critical upstream factors
   leading to these changes were still unknown ([53]Zhang et al., 2013).
   In the present study, we evaluated therapeutic effects of QLY on
   collagen induced arthritis (CIA) in rats and investigated its possible
   interactions with MTX in an integrated regimen. A computational
   workflow based network pharmacology study and a metabolomics analysis
   were carried out to further elucidate mechanisms underlying these
   actions.

Materials and Methods

Ingredients Preparation From QLY and Target Prediction

   Firstly, we collected information about chemical composition of each
   herb in QLY formula from literatures. Those meeting certain ADME
   properties and drug-likeness standard (wQED > 0.3) were chosen for
   further target prediction analysis (Supplementary File [54]S1)
   ([55]Bickerton et al., 2012). After filtering redundant information, we
   obtained 234 ingredients, including 124, 47, 61, and 13 compounds from
   Kushen, Qingfengteng, Huangbai, and Bixie, respectively. Their
   structures were retrieved from the PubChem database^[56]1. All the
   chemical information was then used as data source for target
   prediction.

   To achieve in silico prediction, potential targets of these ingredients
   were analyzed by using the drugCIPHER method, a state-of-art
   network-based algorithm for global prediction of compound targets
   ([57]Zhao and Li, 2010). In principle, this technique predicts
   relationships between bioactive ingredients and candidate targets based
   on network-based integration of multiple pharmacological similarities
   by using FDA covered agents, corresponding targets and protein–protein
   interactions as references. In order to obtain high precision results,