Abstract

   Background: Dianbaizhu (Gaultheria leucocarpa var. yunnanensis), a
   traditional Chinese/ethnic medicine (TC/EM), has been used to treat
   rheumatoid arthritis (RA) for a long time. The anti–rheumatic arthritis
   fraction (ARF) of G. yunnanensis has significant anti-inflammatory and
   analgesic activities and is mainly composed of methyl salicylate
   glycosides, flavonoids, organic acids, and others. The effective
   ingredients and rudimentary mechanism of ARF remedying RA have not been
   elucidated to date.

   Purpose: The aim of the present study is to give an insight into the
   effective components and mechanisms of Dianbaizhu in ameliorating RA,
   based on the estimation of the absorption, distribution, metabolism,
   and excretion (ADME) properties, analysis of network pharmacology, and
   in vivo and in vitro validations.

   Study design and methods: The IL-1β–induced human fibroblast-like
   synoviocytes of RA (HFLS-RA) model and adjuvant-induced arthritis in
   the rat model were adopted to assess the anti-RA effect of ARF. The
   components in ARF were identified by using UHPLC-LTQ-Orbitrap-MS^n. The
   quantitative structure–activity relationship (QSAR) models were
   developed by using five machine learning algorithms, alone or in
   combination with genetic algorithms for predicting the ADME properties
   of ARF. The molecular networks and pathways presumably referring to the
   therapy of ARF on RA were yielded by using common databases and visible
   software, and the experimental validations of the key targets conducted
   in vitro.

   Results: ARF effectively relieved RA in vivo and in vitro. The five
   optimized QSAR models that were developed showed robustness and
   predictive ability. The characterized 48 components in ARF had good
   biological potency. Four key signaling pathways were obtained, which
   were related to both cytokine signaling and cell immune response. ARF
   suppressed IL-1β–induced expression of EGFR, MMP 9, IL2, MAPK14, and
   KDR in the HFLS-RA .

   Conclusions: ARF has good druggability and high exploitation potential.
   Methyl salicylate glycosides and flavonoids play essential roles in
   attuning RA. ARF may partially attenuate RA by regulating the
   expression of multi-targets in the inflammation–immune system. These
   provide valuable information to rationalize ARF and other TC/EMs in the
   treatment of RA.

   Keywords: rheumatoid arthritis, Dianbaizhu, anti–rheumatic arthritis
   fraction (ARF), quantitative structure–activity relationship (QSAR),
   ADME, network pharmacology

Introduction

   Rheumatoid arthritis (RA) is a chronic autoimmune disease with a higher
   prevalence in women, which according to [57]Oliveira and Fierro (2018)
   is characterized by an inflammatory process, with a global prevalence
   ranging from 0.3 to 1%. The current drugs for RA are mainly divided
   into the following categories: nonsteroidal anti-inflammatory drugs,
   disease modifying anti–rheumatic drugs, glucocorticoids, and biological
   response modifiers ([58]Zhang et al., 2015). Limitations with these
   treatments have been associated with side effects and dosing
   inconvenience and have been observed in a proportion of patients
   ([59]Zhang et al., 2015). An increasing number of complementary and
   alternative drug therapy have been developed to alleviate the severity
   of RA and to bring improvement in physical conditions of patients. The
   development of traditional Chinese/ethnic medicines (TC/EMs) featured
   by multicomponent therapy provides a representative approach for the
   treatment of RA.

   Gaultheria leucocarpa var. yunnanensis (Franch.), known as “Dianbaizhu”
   in TC/EMs, belongs to the Ericaceae family and is mainly distributed in
   Southwest China. It has been widely used as a folk medicine for the
   treatment of inflammatory diseases, such as RA and chronic tracheitis
   in the Yi nationality ([60]Liu et al., 2013). Many of the
   over-the-counter TC/EMs containing Dianbaizhu are used for the
   treatment of RA, for example, the Touguxiang ointment, Jingulian
   capsule/tablet, Fenghechubi tincture, and Dianbaizhu syrup. The main
   constituents of these include salicylate derivatives, lignans,
   flavonoids, and organic acids that have been isolated and identified
   from G. yunnanensis ([61]Liu et al., 2015; [62]Xu et al., 2016).
   Isolates obtained from Gaultheria, such as MSTG-B, MSTG-A, gaultherin,
   and chlorogenic acid, have anti-inflammatory and analgesic activities,
   which have been all verified using in vivo and in vitro models
   ([63]Zhang et al., 2011; [64]Xie et al., 2014; [65]Nabavi et al.,
   2017). The anti–rheumatic arthritis fraction (ARF) of G. yunnanensis
   has been ascertained by screening activity and found to exhibit better
   activity than the above-mentioned singular effective constituents. The
   research group of this study demonstrated by simulating the
   gastrointestinal fluid and human gut bacteria models in vitro that ARF
   is relatively stable in the gastrointestinal tract ([66]Wang et al.,
   2019). The bioactive constituents and molecular mechanisms that
   underlie the effects of ARF against RA progression remain unclear.

   The development of systems biology and bioinformatics has created an
   opportunity for the discovery of the mechanisms of action of Chinese
   herbal medicine that are used to treat RA. Network pharmacology is a
   suitable tool to clarify and interpret the synergistic effects and the
   underlying mechanisms of multicomponent and multi-target agents from a
   holistic perspective ([67]Guo et al., 2015; [68]Kibble et al., 2015).
   In most studies, network pharmacology considers drug-like ingredients
   in herb databases, while the inclusiveness and coverage ratio of these
   ingredients are often neglected. Thus, especially for minority drugs,
   such as ethnic medicine Dianbaizhu, the primary ingredients and targets
   predicted by network pharmacology may deviate from the truth.
   Therefore, it becomes necessary to establish a more comprehensive
   evaluation system of drug-likeness of TC/EMs.

   The pharmaceutical industry at present undergoes tremendous pressure,
   especially in reducing healthcare costs and screening the amount of new
   compounds ([69]Yang et al., 2019). The molecular properties for
   absorption, distribution, metabolism, and excretion (ADME) are crucial
   in the evaluation system for drug design and development. The
   development of many potential drugs has been discontinued because of
   their poor absorption. Several screening paradigms, including the ADME
   properties, have been used to enhance the probability of success
   through the drug development stage. Traditional research methods and
   models have been undergoing changes that cannot meet the requirements
   of rapid advances of new drug research and development, together with
   operation of massive data. It has become a trend to develop ADME
   prediction models using machine learning (ML) to process high
   throughput information, particularly the quantitative
   structure–activity relationships (QSARs), as a statistical model, which
   quantitatively correlates chemical structural information (described as
   molecular descriptors) to the response end points (biological activity,
   property, toxicity, etc.) ([70]Khan and Roy, 2018). This occupies a
   crucial position in the establishment and prediction of the ADME data
   model. Currently, ADME predictions have been also applied to design new
   compounds against the novel coronavirus disease 2019 (COVID-19)
   ([71]Hage-Melim et al., 2020). As one of the basic and hot topics in
   chemometrics research, this has attracted much attention and been
   widely used, for instance, in treatment technology for organic
   micropollutants ([72]Huang et al., 2020), migration and transformation
   of organic pollutants, development and design of drugs, graph signal
   processing ([73]Matsushita et al., 2019), and environmental-related
   research ([74]Song et al., 2020). A QSAR model is developed by
   classification and/or regression analysis of select descriptors
   contributing toward targeted properties ([75]Gaikwad et al., 2016). The
   related analyses implemented by ML mainly include the partial least
   square (PLS), random forest (RF), K-nearest neighbors (KNN), error back
   propagation training (EBPT), discrimination analysis (DA), PLS-DA,
   support vector machine (SVM), and other single classifier algorithms
   ([76]Jiang et al., 2020; [77]Maharao et al., 2020; [78]Spiegel and
   Senderowitz, 2020). Some impediments in using the reported QSAR models
   have long existed, including variable selection, data redundancy, and a
   lack of consistent and homogenous data in the public domain ([79]Lee et
   al., 2020). The accelerated pace of drug discovery has heightened the
   need for efficient prediction methods. The genetic algorithm (GA) has
   great advantages in variable selection, model optimization, and high
   efficiency. Although the application of the GA alone or in combination
   with other algorithms in QSAR model building is a crucial end point,
   little or no data exists in the public domain ([80]Ermondi and Caron,
   2019). Therefore, an attempt was made to combine the GA with several
   single classifiers to build a QSAR model in order to get efficient ADME
   prediction models with a good prediction performance.

   A systematic study of the multiple components and multiscale mechanisms
   was carried out to investigate the remedial effect of ARF on RA, and
   the following steps were taken: 1) the components of ARF of G.
   yunnanensis were identified using ultra-high-performance liquid
   chromatography coupled with linear ion trap Orbitrap mass spectrometry
   (UHPLC-LTQ-Orbitrap-MS^n); 2) the organ indexes (OIs) in
   adjuvant-induced arthritis (AIA) in rats were measured in vivo,
   histological and pathological changes of the joints were surveyed, and
   the repression capability of ARF on RA, as well as the release of
   inflammatory factors, was analyzed; 3) QSAR models of the ADME features
   based on using RF, SVM, KNN, PLS-DA, and EBPT alone or in combination
   with the GA were proposed. The ADME-related parameters of the compounds
   of ARF were predicted and evaluated; 4) protein–protein interaction
   (PPI) network analyses, the KEGG pathway, and gene ontology (GO)
   enrichment analyses were performed for key targets; 5) the
   compound–common target and RA-related pathway networks were developed
   to investigate the potential mechanisms, and the key targets were
   selected for subsequent experimental verification in vitro.
   Consequently, a better comprehension of the underlying pharmacodynamics
   of ARF could provide new insights for treatment against RA. At present,
   this exploration strategy could present the critical active ingredients
   and potential mechanisms of ARF in relieving RA and bring weighty
   benefits in screening new clinical therapeutic approaches on RA. The
   whole framework diagram of this study is depicted in [81]Figure 1.

FIGURE 1.

   [82]FIGURE 1
   [83]Open in a new tab

   A schematic picture of the whole development strategies in unraveling
   the pharmacological mechanisms of ARF derived from Dianbaizhu in
   treating RA. For the ADME prediction section, the blue solid arrows
   represent the modeling process of the QSAR models, and the black solid
   line and dotted line arrows represent the evaluation and validation
   processes of the models. For the other parts, the black solid arrows
   stand for process transition. The red solid arrows reflect the
   relevance and progressiveness among the three portions.

Materials and Methods

Reagents and Antibodies

   Dianbaizhu was collected from Chuxiong (Yunnan, China). Methotrexate
   (MTX) was bought from SPH Sine Pharmaceutical Laboratories Co. Ltd.
   (H3102067804, Shanghai, China). Aspirin was purchased from Jiangsu
   Pingguang Pharmaceutical Co. Ltd. (H19980197, Jiangsu, China). Freund’s
   complete adjuvant (FCA, 101722747) and β–actin (A5441) were obtained
   from Sigma (USA). Human fibroblast-like synoviocytes of RA (HFLS-RA)
   were purchased from Hunan Fenghui Biotechnology Co. Ltd. (Hunan,
   China). The DMEM/F12 medium (18219003), FBS (190913), and penicillin
   streptomycin solution (30002303) were obtained from Beijing BioDee
   Biotechnology Co. Ltd. (Beijing, China). IL-1β ([84]P01584) was
   supplied by Novoprotein Technology Co. Ltd. (Jiangsu, China). Rat IL-2
   ELISA kit (KA30281115), Rat IL-1β ELISA kit (KA301B90421), Rat IL-6
   ELISA kit (KA30690741) and Rat TNF-α ELISA kit (KA38290222) were bought
   from Beijing Biodragon Immunotechnologies Co., Ltd. (Beijing, China).
   The primary antibody (anti-rabbit) and secondary antibody (anti-mouse)
   were bought from Cell Signaling Technology, Inc. (CST, USA). Antibodies
   against the following proteins were used: EGFR (18986-1-AP), MMP9
   (10375-2-AP), IL2 (Ab92381, Abcam), MAPK14 (8690, CST) and KDR
   ([85]Ab126679, Abcam). These antibodies as well as RIPA Lysis Buffer
   (S1004), protease inhibitor (KGP603), BCA KIT (Thermo, USA) were
   provided from Beijing Bioway Biotechnology Co., Ltd. (Beijing, China).
   The AB-8 resin was obtained from Cangzhou Bon Adsorber Technology Co.,
   Ltd. (Hebei, China).

Preparation of Anti-Rheumatic Arthritis Fraction

   Dianbaizhu was acquired from the dried aerial parts of Gaultheria
   leucocarpa var. yunnanensis (Franch.) TZ Hsu and RC. Fang from
   Gaultheria Kalm ex L. (Ericaceae) ([86]Wang et al., 2019). Samples were
   authenticated by Professor Shengli Wei (Beijing University of Chinese
   Medicine). The voucher specimens and herbs were deposited in the
   Chinese Materia Medica Chemistry laboratory (B417). ARF was prepared as
   previously described ([87]Wang et al., 2019). Briefly, the dried aerial
   parts of G. yunnanensis (17 kg) were extracted using 238-L 30%
   ethanol–water three times for 2 h each time by the thermal recycling
   extract method. The solvent was completely evaporated from the ethanol
   extraction solution to yield a crude extract (approximate 1.7 kg). The
   total ethanol extract was dissolved in water, and then adsorbed by the
   AB-8 resin column. It was eluted with distilled water and 35%
   ethanol–water in order, yielding the 35% elution fraction (approximate
   500 g). This fraction was concentrated under reduced pressure to yield
   ARF powder. After which it was dissolved in distilled water and passed
   through a 0.22-μm filter for UHPLC-LTQ-Orbitrap-MS^n analyses.

Identification of Compounds in Anti-Rheumatic Arthritis Fraction by Using
UHPLC-LTQ-Orbitrap-MS^n

   For the UPLC-ESI-MS^n experiment, most of the analysis conditions were
   consistent with the reported research of [88]Wang et al. (2019), except
   for the gradient program and injection volume. The mobile phase system
   was made up of 0.1% formic acid–aqueous solution (A) and acetonitrile
   (B). The gradient program is as follows: 0–2 min, 10% B; 2–10 min,
   10–20% B; 10–12.5 min, 20% B; 12.5–17.5 min, 20–40% B; 17.5–25 min,
   40–80% B; 25–27.5 min, 80–95% B; and 27.5–33 min, 95% B. The injection
   volume was 5 μL.

Animal Experiments

Animals

   The operations were performed on all the animals in accordance with the
   China Physiological Society's Guiding Principles in the Care and Use of
   Animals, as well as the authorization of the Animal Care Committee of
   Beijing University of Chinese Medicine.

   The 8-week-old SPF Wistar rats with a mean weight of (200 ± 20) g,
   equal numbers of male and female rats, were obtained from Beijing Vital
   River Laboratory Animal Technology Co. Ltd. (certification number SCXK
   (Jing) 2016-0006). The animals were housed in suitable temperature and
   humidity conditions with a 12-h light/dark cycle at the Beijing
   University of Chinese Medicine. All Wistar rats were allowed to
   acclimatize themselves for 7 days, and had free access to tap water and
   food.

Adjuvant Induced Arthritis Model Establishment and Grouping

   ARF was diluted with distilled water for tests in vivo. The lowest
   dosage selection for ARF was twice the daily dose of Dianbaizhu
   followed for patients with RA (25 g/60 kg body weight). After
   acclimation for 1 week, the Wistar rats were randomly divided into
   seven groups (n = 8): the control group (Blank), model group (Model),
   ARF-treated group 1 (608 mg/kg/d, ARF-H), ARF-treated group 2
   (304 mg/kg/d, ARF-M), ARF-treated group 3 (152 mg/kg/d, ARF-L),
   MTX-treated group (0.5 mg/kg/d), and Aspirin-treated group
   (100 mg/kg/d). The RA model was established with subcutaneous
   injections of FCA in the right hind foot. After 7 days of the
   injection, the ARF-, MTX- and Aspirin-treated groups, respectively,
   were orally administered ARF, MTX, and aspirin for a period of 21 days.
   The Blank and Model groups were treated to distilled water.

Severity Assessment of Arthritis

   The rats were observed once every two days after primary immunization.
   The severity of arthritis was evaluated as in previous studies ([89]Guo
   et al., 2016; [90]Slovák et al., 2017), including arthritis score, body
   weight, hind paw volume, percentage of arthritis in limbs, and the time
   arthritis first appeared.

Histopathological Analyses

   Rats were sacrificed by cervical dislocation on the 28th day after the
   first immunization. Both hind limbs, including the paws and ankles,
   were dissected, fixed immediately in 4% paraformaldehyde, and embedded
   in paraffin. Tissue sections (5 μm) were mounted on common slides for
   staining with hematoxylin and eosin (H&E) or Safranin-O–Fast Green. The
   histopathological characteristics were evaluated blindly. The data were
   expressed as mean inflammation scores, and all the sections randomized
   and evaluated by two trained observers who were blinded to the
   treatment groups and the severity of arthritis of each rat ([91]Zhang
   et al., 2015).

Enzyme-Linked Immunosorbent Assay

   The blood samples from all the AIA rats were obtained at day 27 and
   were anticoagulated with sodium heparin. The levels of IL-1β, TNF-α,
   IL-6, and IL-2 in the plasma were detected using commercially available
   rat ELISA kits according to the protocol of the kit, and absorbance was
   determined at 405 nm ([92]Park et al., 2020).

Absorption, Distribution, Metabolism, and Excretion Evaluation

   The ADME properties contribute on vital task in early stages of drug
   discovery process. To evaluate the biological activity of ARF, KNN,
   SVM, RF, PLS-DA, EBPT, and GA were used to establish the QSAR
   prediction model for assessing the pharmacokinetic properties, namely
   oral bioavailability (OB), P-glycoprotein substrate (PGPS),
   P-glycoprotein inhibitor (PGPI), human intestinal absorption (HIA), and
   Caco-2 cell permeability (Caco-2).

   The methods of ML were applied to construct QSAR models according to
   the references previous reported with some changes ([93]Welling et al.,