Abstract

Background

   Rheumatoid arthritis (RA) is a chronic inflammatory autoimmune disease
   in which macrophages produce cytokines that enhance inflammation and
   contribute to the destruction of cartilage and bone. Additive Sishen
   decoction (ASSD) is a widely used traditional Chinese medicine for the
   treatment of RA; however, its active ingredients and the mechanism of
   its therapeutic effects remain unclear.

Methods

   To predict the ingredients and key targets of ASSD, we constructed
   “drug-ingredient-target-disease” and protein–protein interaction
   networks. Gene ontology and Kyoto Encyclopedia of Genes and Genomes
   enrichment analyses were performed to explore the potential mechanism.
   The activity of the predicted key ingredients was verified in
   lipopolysaccharide-stimulated macrophages. The binding mode between the
   key ingredients and key targets was elucidated using molecular docking
   and molecular dynamics simulation.

Results

   In all, 75 ASSD active ingredients and 1258 RA targets were analyzed,
   of which kaempferol, luteolin, and quercetin were considered key
   components that mainly act through inflammation-related pathways, such
   as the PI3K-AKT, TNF, and IL-17 signaling pathways, to ameliorate RA.
   Transcriptome sequencing suggested that kaempferol-, luteolin-, and
   quercetin-mediated inhibition of glycolysis reduced the
   lipopolysaccharide-induced production of proinflammatory factors. In
   vitro experiments indicated that kaempferol, luteolin, and quercetin
   decreased Glut1 and LDHA expression by diminishing PI3K-AKT signaling
   to inhibit glycolysis. Molecular dynamic simulation revealed that
   kaempferol, luteolin, and quercetin stably occupied the hydrophobic
   pocket of PI3Kδ.

Conclusion

   Our results show that the PI3Kδ-mediated anti-inflammatory responses
   elicited by kaempferol, luteolin, and quercetin are crucial for the
   therapeutic efficacy of ASSD against RA.

   Keywords: Autoimmune disease, traditional Chinese medicine, glycolysis,
   inflammation, PI3K-AKT

Introduction

   Rheumatoid arthritis (RA), a chronic inflammatory autoimmune disease
   that affects approximately 1% of the population worldwide, is
   characterized by painful, swollen joints that can severely impair
   physical function and quality of life.[40]1,[41]2 The
   immunopathogenesis of RA spans decades, beginning with the loss of
   self-tolerance, which is marked by the production of autoantibodies
   against post-translationally modified protein antigens. After a
   prolonged period of asymptomatic autoimmunity, tissue tolerance fails
   and synovial inflammation ensues as defective DNA repair and rewired
   metabolic networks drive T-cell maldifferentiation toward highly
   proliferative, tissue-invasive, proinflammatory effector T cells. The
   transition of synovial stromal cells into autoaggressive effector cells
   converts synovitis from acute to chronic destructive, which erodes the
   cartilage and bone.[42]3

   Macrophages play an important role in RA pathogenesis. Compared with
   the normal synovium, biopsy specimens from patients with RA contain
   more macrophages.[43]4 These cells make up almost the entire synovial
   lining, and their numbers considerably increase in the sublining
   tissue.[44]4 The degree of synovial macrophage infiltration is
   positively correlated with the progression of joint
   destruction.[45]5,[46]6 During the course of RA, synovial macrophages
   function as antigen-presenting cells, driving pathogenic T-cell
   infiltration, and produce cytokines (eg, IL1B, IL6, TNF-α, and IL23) to
   perpetuate inflammation by recruitment and activation of additional
   immune cells. Synovial macrophages also directly differentiate to
   mature osteoclasts under specific microenvironment conditions. In
   addition, inflammatory macrophages can produce angiogenic factors (eg,
   vascular endothelial growth factor (VEGF), transforming growth factor,
   and TNF-α) and degrading enzymes (eg, MMPs, namely MMP1, 3, 7, 10, 12,
   14, and 25), which participate in the hypervascularization and
   articular surface erosion observed in RA, respectively.[47]5,[48]7
   Importantly, all agents currently used in the clinics for the treatment
   of RA, including conventional synthetic (cs)-, biologic (b)-, and
   targeted synthetic (ts)-disease-modifying antirheumatic drugs (DMARDs),
   and even corticosteroids, quantitatively or qualitatively affect the
   synovial macrophage components, directly or indirectly.[49]7 The
   decrease in the number of synovial macrophages is a validated biomarker
   of the clinical efficacy of antirheumatic treatment.[50]8,[51]9
   Therefore, synovial macrophages are clearly linked to the initiation,
   progression, severity, and response to RA treatment.

   Sishen decoction (SSD), a well-known traditional Chinese medicine
   prescription, was first recorded in “YanFang Xin Bian” of the Qing
   Dynasty and has been used clinically to treat arthritis for hundreds of
   years. SSD comprises five traditional herbs: Astragalus membranaceus
   (AM), Lonicera japonica (LJ), Polygala tenuifolia (PT), Dendrobium
   nobile (DN), and Cyathula officinalis (CO). Additive SSD (ASSD),
   additionally contains Polygonum cuspidatum (PC) and Hedyotis diffusa
   (HD), and has a good therapeutic effect against RA. However, its main
   active ingredients and the molecular mechanism underlying its effect
   against rheumatic diseases remain poorly understood. In this study,
   using network pharmacology, we constructed
   “drug-ingredient-target-disease” and protein–protein interaction (PPI)
   networks to predict key ingredients and key targets of ASSD against RA,
   and then used databases to performed Gene Ontology (GO) and Kyoto
   Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses to
   predict the potential mechanisms underlying the therapeutic effect of
   ASSD against RA. Next, we verified the predicted results using
   lipopolysaccharide (LPS)-stimulated macrophages. Finally, molecular
   docking and molecular dynamics simulation were conducted to elucidate
   the binding mode of key ingredients and targets. A bioinformatics
   analysis of GSE datasets was performed to evaluate the role of key
   targets in RA patients. A flowchart of the study is shown in [52]Figure
   1.

Figure 1.

   [53]Figure 1
   [54]Open in a new tab

   Flow chart of the present study.

Materials and Methods

Reagents and Chemicals

   RAW 264.7 cells were purchased from the Cell Culture Centre, Chinese
   Academy of Medical Sciences & Peking Union Medical College (Beijing,
   China). Kaempferol, luteolin, and quercetin were purchased from Must
   Bio-technology (Chengdu, China). CAL101 was purchased from Beyotime
   (Shanghai, China). Antibodies against pAkt (T-308) and Akt were
   purchased from Cell Signaling Technology (Danvers, MA, USA). Antibodies
   against Glut 1, c-Myc, and Ldha were purchased from Abcam (Cambridge,
   MA, USA).

Data Preparation

   The Traditional Chinese Medicine Systems Pharmacology Database (TCMSP;
   [55]http://ibts.hkbu.edu.hk/LSP/tcmsp.php)[56]10 and the Bioinformatics
   Analysis Tool for Molecular Mechanism of Traditional Chinese Medicine
   (BATMAN-TCM; [57]http://bionet.ncpsb.org/batman-tcm/)[58]11 were used
   to identify the chemical ingredients of ASSD. Based on absorption,
   distribution, metabolism, and excretion (ADME), the ADME-related
   properties, including oral bioavailability (OB) ≥ 30% and drug likeness
   (DL) ≥ 0.18, were employed to screen the active ingredients of ASSD
   with favorable pharmacokinetic properties. Potential targets of the
   active ingredients were predicted by integrating genes collected from
   TCMSP, BATMAN-TCM, and SwissTargetPrediction
   ([59]http://www.swisstargetprediction.cn/).[60]12

   Genecards ([61]http://www.genecards.org/)[62]13 and Therapeutic Target
   Database (TTD, [63]http://database.idrb.cqu.edu.cn/TTD/)[64]14 were
   used to identify RA-related targets in Homo sapiens. The target
   database of ASSD-treated RA was established by comparing and analyzing
   the overlapping targets between the RA-related and active
   ingredient-related targets.

Network Construction

   The ASSD-ingredients-targets-rheumatoid arthritis network was created
   by inputting the bioactive ingredients and ASSD-treated RA targets into
   Cytoscape ([65]http://cytoscape.org/).[66]15 Nodes represent the
   ingredients and targets, whereas edges indicate the interactions of the
   ingredients and targets.

   The PPI network was derived based on STRING database ([67]http://
   string-db.org/).[68]16 The results of the analysis using the STRING
   database were imported into the Cytoscape software to draw the
   visualized network. Clusters were obtained by analyzing the PPI network
   using the MCODE and cytoHubba plug-ins of Cytoscape.

GO and KEGG Pathway Enrichment Analysis

   The Database for Annotation, Visualization and Integrated Discovery
   (DAVID) ([69]http://david.abcc.ncifcrf.gov)[70]17 and the KEGG database
   ([71]http://www.kegg.jp/)[72]18 were used to perform GO and KEGG
   enrichment analyses of SSD-treated RA targets, respectively.
   Statistical significance was set at P < 0.05.

Cell Culture

   RAW 264.7 cells were cultured in Dulbecco’s modified Eagle’s medium
   (DMEM; Gibco, NY, USA) supplemented with 10% fetal bovine serum (Gibco,
   NY, USA). The cells, after reaching approximately 80% confluence, were
   pretreated with kaempferol, luteolin, or quercetin for 30 min, and then
   incubated with LPS (100 ng/mL) and interferon-γ (IFN-γ, 20 ng/mL) for
   24 h.

Cell Viability Assay

   The cytotoxicity of kaempferol, luteolin, and quercetin on RAW 264.7
   cells was assessed using the Cell Counting kit-8 (CCK-8; Boster, Wuhan,
   China). The cells were seeded in 96-well plates at a density of 2×10^4
   cells/well. The next day, cells were treated with kaempferol, luteolin,
   or quercetin at the indicated concentrations. After 24 h of incubation,
   the CCK-8 solution was added to each well and the plate was incubated
   for 4 h. The absorbance at 450 nm was measured using a SpectraMax i3x
   spectrophotometer (Molecular Devices, CA, USA).

Measurement of Nitric Oxide (NO) Levels

   NO levels were measured using the Griess reagent based on the
   accumulation of nitrite and nitrite reductase. Equal volumes of Griess
   reagent I and II (Beyotime) were mixed with the culture supernatant in
   enzyme-labeled plates. The absorbance at 540 nm was measured using the
   SpectraMax i3x spectrophotometer (Molecular Devices).

Measurement of Il1b and Il6 Levels Using Quantitative Real-Time PCR

   Total RNA was extracted from the cells using the TRIzol reagent
   (Takara, Tokyo, Japan) according to the manufacturer’s instructions.
   cDNA was synthesized using a reverse-transcription kit in a 20 μL
   reaction mixture (Takara, Tokyo, Japan). Gene expression levels were
   quantified using SYBR Premix Ex Taq (Takara) on a 7900HT Real-Time PCR
   system (Applied Biosystems, MA, USA). The mRNA levels of target genes
   were normalized against those of GAPDH. The primers for the target
   genes were synthesized at Sangon Biotech Co. Ltd. (Shanghai, China).
   The sequences of the primers were as follows: Il1b forward:
   5′-TGCCACCTTTTGACAGTGATG-3′ and Il1b reverse:
   5′-TGATGTGCTGCTGCGAGATT-3′; Il6 forward: 5′-GGGACTGATGCTGGTGACAA-3′ and
   Il6 reverse: 5′- TGCCATTGCACAACTCTTTTC-3′.

Measurement of Cd86 Levels Using Flow Cytometry

   Cells were blocked with anti-mouse CD16/32 antibody (BioLegend, San
   Diego, CA, USA) before staining with an anti-mouse CD86-PerCP/Cy5.5
   antibody (BioLegend, CA, USA). After washing, the cells were analyzed
   using a CytoFLEX S cytometer (Beckman Coulter, CA, USA). All data were
   analyzed using CytExpert according to the manufacturer’s instructions.

Measurement of Mitochondrial Reactive Oxygen Species (ROS)

   Mitochondrial ROS levels were assessed using a MitoSox probe (Yeasen,
   Shanghai, China). The cells were incubated with the MitoSox dye at 37
   °C for 30 min. Thereafter, the medium was aspirated and the cells were
   quickly washed twice with phosphate-buffered saline to remove the
   unbound dye. Images were captured at 580 nm using an ImageXpress Micro
   Four High-Content Imaging System (Molecular Devices). For quantitation
   of intensity, images from 27 sites in three wells per experiment were
   analyzed using the MetaXpress analysis software.

RNA Sequencing

   Total RNA was isolated and purified using the TRIzol reagent
   (Invitrogen) following the manufacturer’s instructions. Poly (A) RNA
   was purified from total RNA using Dynabeads Oligo (dT) 25–61005 (Thermo
   Fisher, CA, USA) and fragmented into small pieces using the Magnesium
   RNA Fragmentation Module (catalog# e6150; NEB, USA). The cleaved RNA
   fragments were reverse-transcribed into cDNA using SuperScript II
   Reverse Transcriptase (catalog# 1896649; Invitrogen, USA), which were
   used to synthesize U-labeled second-stranded DNA with Escherichia coli
   DNA polymerase I (catalog# m0209; NEB), RNase H (catalog# m0297; NEB),
   and dUTP Solution (catalog# R0133; Thermo Fisher). Single- or
   dual-index adapters were ligated to the fragments and size selection
   was performed using AMPureXP beads. After treatment of the U-labeled
   second-stranded DNA with heat-labile UDG enzyme (catalog# m0280; NEB),
   the ligated products were amplified using PCR. Finally, 2×150 bp
   paired-end sequencing (PE150) was performed on an Illumina NovaSeq 6000
   (LC-Bio Technology Co., Ltd., Hangzhou, China) following the
   manufacturer’s protocol.

Measurement of NAD^+/NADH and ATP/ADP Ratios

   The ATP/ADP ratio was measured using an ADP/ATP Ratio Assay Kit
   (Dojindo Laboratories, Kumamoto, Japan). The cells were plated in white
   96-well plates and treated as required. ATP and ADP working solutions
   were added to each well, according to the manufacturer’s instructions.
   Absorbance was detected using a SpectraMax i3x spectrophotometer
   (Molecular Devices).

   The NAD^+/NADH ratio was measured using an NAD^+/NADH Assay Kit
   (Beyotime). The cells were lysed in a lysis solution. The lysates were
   centrifuged at 12,000 rpm for 5 min, and the supernatants were
   collected. Subsequently, the NAD^+ and NADH concentrations were
   measured according to the manufacturer’s instructions.

Measurement of Glucose Consumption

   To measure glucose consumption after treatment with LPS and kaempferol,
   luteolin, or quercetin, the culture medium was replaced with fresh
   medium before treatment and collected 24 h later. The glucose levels in
   the culture medium were measured using a glucose assay kit (Beyotime).

Measurement of the Expression Levels of Glut1, Ldha, and c-Myc and
Phosphorylation of AKT Using Western Blot Analysis

   Cells were lysed in RIPA buffer (50 mm Tris-HCl, pH 7.4, 150 mm NaCl,
   1% NP-40, and 0.5% sodium deoxycholate) containing a Complete Protease
   Inhibitor Cocktail and Phosphatase Inhibitor Cocktail (Roche, Basel,
   Switzerland). The lysates were centrifuged at 12,000 rpm for 15 min and
   the proteins were quantified using a BCA protein assay kit (Boster,
   Wuhan, China). Total proteins were resolved using sodium dodecyl
   sulfate-polyacrylamide gel electrophoresis and subsequently transferred
   onto a polyvinylidene difluoride membrane. The membranes were incubated
   with specific primary and secondary antibodies. Images were acquired
   using the Amersham Imager 600 (Cytiva, DC, USA).

Molecular Docking and Molecular Dynamics Simulation

   The crystal structure of PI3Kδ was obtained from the RCSB Protein Data
   Bank (PDB ID 7LQ1). The 3D structures of kaempferol, luteolin, and
   quercetin were downloaded from the PubChem database. Autodock Vina was
   used to analyze molecular docking between PI3Kδ and the small
   molecules. The free binding energy was obtained as an indicator to
   reflect the binding, and the docking conformation with the highest
   output score was considered to be the binding conformation for
   subsequent molecular dynamics simulation.

   Molecular dynamics simulation was performed by Gromacs 2022. Before
   simulation, the small molecules were described by the GAFF2 small
   molecule force field, whereas the PI3Kδ protein was described by the
   AMBER14SB force field, along with the TIP3P solvent box. Molecular
   dynamics simulation commenced with constant number, volume, and
   temperature (NVT) and constant number, pressure, and temperature (NPT)
   equilibration runs, each lasting 100 ps at 298 K. Subsequently, a 100
   ns molecular dynamics simulation of the complex system was executed,
   with configurations saved every 10 ps. The binding free energy of the
   small molecules with PI3Kδ was calculated using the molecular mechanics
   Poisson–Boltzmann surface area (MM/PBSA).

Statistical Analysis

   Data were analyzed using Prism (GraphPad Software) and are presented as
   the mean values ± standard deviation (SD). Differences between groups
   were statistically analyzed using two-tailed Student’s t-tests.
   Statistical significance was set at P < 0.05.

Results

Data Preparation

   To identify the active ingredients of ASSD using TCMSP and BATMAN-TCM,
   the classical ADME parameters, OB and DL, were used for screening. A
   total of 75 active ingredients in the formula were screened, 20 from A.
   membranaceus, 23 from L. japonica, 5 from P. tenuifolia, 6 from D.
   nobile, 4 from C. officinalis, 10 from P. cuspidatum, and 7 from H.
   diffusa. Among these, five active ingredients (kaempferol, quercetin,
   β-sitosterol, stigmasterol, and luteolin) were duplicated. Detailed
   information on the active ingredients is provided in [73]Supplementary
   Tables 1 and [74]2. The active ingredient-related targets were
   retrieved from TCMSP, BATMAN-TCM, and SwissTargetPrediction, and 873
   targets were obtained after overlapping. The data of active
   ingredient-related targets are shown in [75]Supplementary Tables 1 and
   [76]2.

   A total of 1258 RA-related targets were obtained from Genecards and
   TTD. Using a Venn diagram of RA-related and active ingredient-related
   targets, 276 common targets were filtered as the target database of
   ASSD-treated RA ([77]Figure 2).

Figure 2.

   [78]Figure 2
   [79]Open in a new tab

   Venn diagram of 873 additive Sishen decoction (ASSD) and 1258
   rheumatoid arthritis (RA) targets showing an overlap of 276 targets.

ASSD-Ingredients-Targets-RA Network and Key Ingredients

   To evaluate which ASSD ingredients are critical to RA treatment, the
   ASSD-ingredients-targets-RA network was built using the Cytoscape
   software by converting the 75 active ingredients and 276 common targets
   to nodes. As shown in [80]Figure 3, the network consists of 365 nodes
   and 1795 edges. Among them, kaempferol (edges =84), luteolin (edges
   =96), and quercetin (edges =155) were closely linked to the targets,
   suggesting that these compounds are the key ingredients for the
   therapeutic effect of ASSD against RA.

Figure 3.

   [81]Figure 3
   [82]Open in a new tab

   The ASSD-ingredients-targets-RA network. The green rhombuses represent
   7 traditional herbs in additive Sishen decoction (ASSD), blue
   rectangles represent the active ingredients in 7 traditional herbs, and
   yellow ellipses represent 276 potential targets in ASSD-treated
   rheumatoid arthritis (RA).

PPI Network and Key Targets

   To further explore the underlying targets of ASSD protection against
   RA, a PPI network for the 276 common targets, with 7794 edges was
   constructed using the STRING database and Cytoscape software. By
   analyzing the network using MCODE, a cluster containing 70 targets was
   constructed. The cluster was further assessed with cytoHubba, and the
   top-10 targets were selected according to their degree values; these
   included MAPK1, MAPK3, JUN, IL6, AKT1, TNF, PTGS2, MMP9, STAT3 and
   CXCL8 ([83]Figure 4).

Figure 4.

   [84]Figure 4
   [85]Open in a new tab

   Protein–protein interaction (PPI) network of the 276 potential targets.
   The nodes represent proteins, and edges represent protein–protein
   interactions.

GO and KEGG Pathway Enrichment Analysis

   GO analysis of the 276 common targets using the DAVID system revealed
   that the functions of these potential targets were related to many
   biological processes, molecular functions, and cellular components
   ([86]Figure 5).

Figure 5.

   [87]Figure 5
   [88]Open in a new tab

   Gene ontology (GO) enrichment with the top-10 P-values for each item.

   Based on the P value, the top 20 significantly enriched KEGG pathways
   for the 276 common targets was identified to decipher the putative
   mechanism underlying the effect of ASSD on RA. As shown in [89]Figure
   6, inflammation-related pathways, such as the PI3K-AKT, TNF-, and IL-17
   signaling pathways, were involved in ASSD-mediated RA treatment.

Figure 6.

   [90]Figure 6
   [91]Open in a new tab

   Kyoto encyclopedia of genes and genomes (KEGG) enrichment with the
   top-20 P-values.

Kaempferol, Luteolin, and Quercetin Inhibited LPS+IFN-γ-Induced Inflammatory
Responses in RAW 264.7 cells

   Using the network and pathway analyses, kaempferol, luteolin, and
   quercetin were found to be the key ingredients of ASSD responsible for
   its effect against RA. The anti-inflammatory properties of these
   ingredients could provide the major mechanism for ASSD-mediated RA
   treatment. Therefore, the anti-inflammatory effects of kaempferol,
   luteolin, and quercetin were investigated in LPS-induced RAW 264.7
   cells.

   The cytotoxicity of kaempferol, luteolin, and quercetin was evaluated
   by treating the RAW 264.7 cells with various concentrations of these
   compounds for 24 h using the CCK-8 assay. Kaempferol, luteolin, and
   quercetin did not affect the viability of RAW 264.7 cells at
   concentrations of 60, 40, and 60 μmol/L, respectively ([92]Figure 7a).
   Therefore, concentrations between 1 and 40 μmol/L were selected for
   subsequent experiments.

Figure 7.

   [93]Figure 7
   [94]Open in a new tab

   Inhibition of lipopolysaccharide (LPS)+interferon gamma (IFN-γ)-induced
   inflammatory responses in RAW 264.7 cells by kaempferol, luteolin, and
   quercetin. (a) Cytotoxicity of kaempferol, luteolin, and quercetin in
   RAW 264.7 cells after 24 h of treatment. (b) Nitric oxide (NO) levels
   in the culture medium were measured using the Griess reagent. (c)
   Quantitative RT-PCR analysis of the mRNA levels of Il1b and Il6 in RAW
   264.7 cells stimulated with LPS+IFN-γ in the presence of the indicated
   doses of kaempferol, luteolin, or quercetin for 24 h. (d) Flow
   cytometry analysis of Cd86 expression in RAW 264.7 cells stimulated
   with LPS+IFN-γ in the presence of kaempferol, luteolin, or quercetin
   (20 μmol/L) for 24 h. Three samples were analyzed for each group. The
   data were analyzed using an unpaired t-test. ^## P<0.01 compared with
   the control group, * P<0.05 and ** P<0.01 compared with the LPS+IFN-γ
   group. Bar graphs show mean ± SD.

   NO is a key factor promoting the development of inflammatory responses.
   The production of NO in LPS+IFN-γ-treated cells was markedly increased,
   as measured using the Griess reagent; however, treatment with
   kaempferol, luteolin, and quercetin alleviated the LPS+IFN-γ-induced
   increase in NO levels in a concentration-dependent manner ([95]Figure
   7b).

   To further explore the anti-inflammatory effects of kaempferol,
   luteolin, and quercetin, the mRNA levels of inflammatory factors were
   measured using RT-PCR. LPS+IFN-γ stimulated the production of
   proinflammatory cytokines, including Il1b and Il6, which was
   significantly reduced by treatment with kaempferol, luteolin, and
   quercetin ([96]Figure 7c). Moreover, kaempferol, luteolin, and
   quercetin decreased the levels of LPS+IFN-γ-simulated proinflammatory
   cytokines in a dose-dependent manner.

   Under the action of LPS+IFN-γ, macrophages manifest the classically
   activated phenotype (M1). To confirm the effect of kaempferol,
   luteolin, and quercetin on M1 macrophages, the expression of M1 marker
   Cd86 was analyzed using flow cytometry. Kaempferol, luteolin, and
   quercetin treatment led to a significantly lower percentage of M1
   macrophages thanLPS+IFN-γ stimulation ([97]Figure 7d).

Inhibition of Glycolysis Involves Kaempferol, Luteolin, and
Quercetin-Mediated Anti-inflammatory Responses

   Because mitochondrial ROS production is central to determining the
   inflammatory phenotype of macrophages,[98]19 we examined the effects of
   kaempferol, luteolin, and quercetin on mitochondrial ROS levels using
   MitoSOX. Addition of LPS+IFN-γ led to an increase in mitochondrial ROS
   levels, which were decreased by kaempferol, luteolin, and quercetin in
   a dose-dependent manner ([99]Figure 8a). Following LPS stimulation,
   macrophages repurpose their mitochondria from ATP production to ROS
   generation, with glycolysis taking on the role of ATP generation to
   promote a proinflammatory phenotype.[100]19 To examine the roles of
   kaempferol, luteolin, and quercetin in glycolysis, we performed RNA
   sequencing. Genes involved in glycolysis were upregulated by LPS+IFN-γ,
   but they were downregulated by kaempferol, luteolin, and quercetin,
   suggesting that glycolysis is crucial for the kaempferol-, luteolin-,
   and quercetin-mediated anti-inflammatory action ([101]Figure 8b–d).

Figure 8.

   [102]Figure 8
   [103]Open in a new tab

   Inhibition of glycolysis involves kaempferol, luteolin, and
   quercetin-mediated anti-inflammatory responses. (a) MitoSox staining,
   indicative of mitochondrial ROS levels, of RAW 264.7 cells stimulated
   with lipopolysaccharide (LPS)+ interferon gamma (IFN-γ) in the presence
   of the indicated doses of kaempferol, luteolin, or quercetin for 24 h.
   (b-d) Heat map of statistically differentially expressed genes by RNA
   sequencing transcriptome analysis in RAW 264.7 cells stimulated with
   LPS+IFN-γ in the presence of kaempferol, luteolin, or quercetin
   (20 μmol/L) for 24 h. Black Bar points to the genes downregulated in
   glycolysis and red bar points to the genes upregulated in oxidative
   phosphorylation. (e) The ATP/ADP ratio in cell lysates was determined
   using an ATP/ADP assay kit. (f) The NAD^+/NADH ratio in cell lysates
   was determined using an NAD^+/NADH assay kit. Three samples were
   analyzed for each group. The data were analyzed using an unpaired
   t-test. ^## P<0.01 compared with the control group, * P<0.05 and **
   P<0.01 compared with the LPS+INF-γ group. Bar graphs show mean ± SD.

   The LPS+IFN-γ treatment also decreased the ATP/ADP ratio, which is
   consistent with the shift in ATP synthesis from oxidative
   phosphorylation to glycolysis; however, kaempferol, luteolin, and
   quercetin restored the ATP/ADP ratio ([104]Figure 8e). The NAD^+/NADH
   ratio was also decreased upon LPS+IFN-γ treatment as the conversion of
   pyruvate to lactate in glycolysis regenerates cytosolic oxidized NAD^+
   to support glycolytic ATP generation, which was enhanced in the
   presence of kaempferol, luteolin, and quercetin ([105]Figure 8f). Taken
   together, these results indicate that kaempferol, luteolin, and
   quercetin prevent the expression of proinflammatory factors via the
   inhibition of glycolysis.

Kaempferol, Luteolin, and Quercetin Suppress Glycolysis Through PI3K-AKT
Signaling

   Macrophages upregulate the glycolytic module, which is linked to excess
   glucose uptake and increased expression of the glucose transporter,
   Glut1.[106]20,[107]21 Consistent with previous reports, LPS+IFN-γ
   induced a time-dependent increase in the expression of Glut1
   ([108]Figure 9a). Kaempferol, luteolin, and quercetin treatment not
   only decreased the expression of Glut1 ([109]Figure 9b), but also
   reduced glucose consumption ([110]Figure 9c).

Figure 9.

   [111]Figure 9
   [112]Open in a new tab

   Kaempferol, luteolin, and quercetin suppressed glycolysis through the
   PI3K-AKT signaling. (a-d) RAW 264.7 cells were stimulated with
   lipopolysaccharide (LPS)+ interferon gamma (IFN-γ) for the indicated
   time points. Immunoblotting and normalized expression of Glut, Ldha,
   and c-Myc against GAPDH. (b-e) RAW 264.7 cells were stimulated with
   LPS+IFN-γ in the presence of kaempferol, luteolin, or quercetin
   (20 μmol/L) for 24 h. Immunoblotting and normalized expression of Glut,
   Ldha, and c-Myc against GAPDH. (c) Glucose levels in the culture medium
   were measured with a glucose assay kit. (f) RAW 264.7 cells were
   stimulated with LPS+IFN-γ for the indicated time points. Immunoblotting
   and normalized expression of p-Akt (T308) against Akt. (g and h) RAW
   264.7 cells were stimulated with LPS+IFN-γ in the presence of the PI3K
   inhibitor CAL-101 (1, 3, or 10 μmol/L) for 24 h. The immunoblot shows
   normalized expression of p-Akt (T308) against Akt as well as of Glut,
   Ldha, and c-Myc against GAPDH. (i) RAW 264.7 cells were stimulated with
   LPS+IFN-γ in the presence of kaempferol, luteolin, or quercetin
   (20 μmol/L) for 24 h. Immunoblotting and normalized expression of p-Akt
   (T308) against Akt. Three samples were analyzed for each group. The
   data were analyzed using an unpaired t-test. ^## P<0.01 compared with
   the control group, * P<0.05 and ** P<0.01 compared with the LPS+IFN-γ
   group. Bar graphs show mean ± SD.

   Lactate dehydrogenase A (Ldha) is a glycolytic enzyme that converts
   pyruvate to lactate and is induced under intensified glycolysis.[113]22
   Immunoblot analysis revealed that activation of macrophages with
   LPS+IFN-γ enhanced the expression of Ldha in a time-dependent manner,
   in consonance with the induction of the Ldha transcriptional regulator,
   c-Myc ([114]Figure 9d). Treatment with kaempferol, luteolin, or
   quercetin significantly reduced the expression of Ldha and c-Myc
   ([115]Figure 9e).

   Glut 1 and Ldha are induced in CD8^+ T effector cells via PI3K-AKT
   signaling.[116]22 In macrophages, upon LPS+IFN-γ stimulation with
   increasing time, PI3K-dependent Akt phosphorylation showed a response
   curve similar to that of Ldha and c-Myc ([117]Figure 9f). To
   investigate whether PI3K-AKT signaling was involved in the expression
   of Glut 1 and Ldha, a PI3K-specific inhibitor, CAL-101, was used
   ([118]Figure 9g). The blockade of PI3K signaling with increasing doses
   of CAL-101 progressively repressed the expression levels of Glut1,
   Ldha, and c-Myc ([119]Figure 9h). To further determine the role of
   PI3K-AKT signaling in kaempferol-, luteolin-, and quercetin-mediated
   inhibition of glycolysis, we examined Akt phosphorylation and found
   that it was decreased by kaempferol, luteolin, and quercetin treatments
   ([120]Figure 9i). Taken together, these data suggest that kaempferol,
   luteolin, and quercetin act through the PI3K-AKT pathway to decrease
   the expression of Glut1 and Ldha, thereby inhibiting glycolysis.

Molecular Docking and Molecular Dynamics Simulation

   Kaempferol, luteolin, and quercetin exhibited the same inhibitory
   effect on AKT phosphorylation as CAL101, which is a potent and
   selective ATP-competitive inhibitor of PI3Kδ. Molecular docking
   revealed that kaempferol, luteolin, and quercetin occupied the cleft
   formed between the N- and C-lobes of the kinase domain of PI3Kδ, which
   is the hydrophobic pocket for ATP binding ([121]Figure 10a–c). The
   hydrogen bond, hydrophobic interaction, and van der Waals (VDW) force
   are the main forces that maintain interactions between PI3Kδ and the
   compounds.

Figure 10.

   [122]Figure 10
   [123]Open in a new tab

   Molecular docking of PI3kδ and the compounds. (a) Binding mode of PI3kδ
   and kaempferol. (b) Binding mode of PI3kδ and luteolin. (c) Binding
   mode of PI3kδ and quercetin.

   To further evaluate the structural stability of PI3Kδ and compounds,
   molecular dynamics simulation was performed. Root mean square deviation
   (RMSD) is an essential factor to evaluate the stability of a system.
   The RMSD of the complex structures, proteins, and compounds gradually
   stabilized as the simulation progressed, without significant
   fluctuations, indicating that the complex structure gradually
   stabilized ([124]Figure 11a). Centroid evolution analysis was performed
   by measuring the distance between the initial binding site and
   compounds and the distance between the protein center and compounds to
   evaluate the binding state. [125]Figure 11b shows that the distance
   between the initial binding site and compounds fluctuated slightly, and
   the distance between the protein center and compounds was stable,
   indicating that compounds did not deviate from and stably bound the
   initial binding site and the protein center. Buried SASA evaluated the
   binding state by measuring the size of the binding interface between
   the protein and compound. [126]Figure 11c showed that the embedding
   areas of kaempferol, luteolin, and quercetin in the protein of the
   complex were stable with slight fluctuations, indicating that the
   binding interface between PI3Kδ and the compounds was stable The high
   degree of simulated conformational overlap indicates that small
   molecules always bind to the initial site. [127]Figure 11d shows that
   the superimposed conformation of the compounds in the complex was near
   the initial binding site, and the degree of conformation overlap was
   strong, indicating that compounds bound to the initial binding site of
   the protein during the simulation.

Figure 11.

   [128]Figure 11
   [129]Open in a new tab

   Stability analysis of PI3kδ and compounds in molecular dynamic
   simulations. (a) RMSD of the protein, ligand, and protein-ligand
   complex. (b) Centroid evolution analysis of PI3kδ and compounds. (c)
   Buried SASA analysis of PI3kδ and compounds. (d) The simulated
   conformational congruence of PI3kδ and the compounds.

   Hydrogen bonding is an important interaction force between proteins and
   ligands, which can reflect the strength of electrostatic interactions.
   During the simulation, the number of hydrogen bonds for kaempferol,
   luteolin, and quercetin mainly fluctuated between 2 and 3, 3 and 4, and
   4 and 6, respectively ([130]Figure 12a). To further identify the
   residues that form hydrogen bonds with compounds in the protein and the
   stability of hydrogen bonds, the percentage of hydrogen bonds formed
   between PI3Kδ and the compounds was analyzed. During the simulation,
   the hydrogen bonds, 828VAL:ligand ([131]Figure 12b), 828VAL:ligand
   ([132]Figure 12c), 813TYR:ligand ([133]Figure 12d), 828VAL:ligand
   ([134]Figure 12d) and ligand:826GLU ([135]Figure 12d), were formed with
   high frequency, high occupancy, and strong stability.

Figure 12.

   [136]Figure 12
   [137]Open in a new tab

   Analysis of hydrogen bond interaction of PI3kδ and compounds in
   molecular dynamic simulations. (a) Hydrogen bond map of PI3kδ and
   compounds. (b) The occupancy of hydrogen bonds for PI3kδ and
   kaempferol. (c) The occupancy of hydrogen bonds for PI3kδ and luteolin.
   (d) The occupancy of hydrogen bonds for PI3kδ and quercetin.

   The VDW and electrostatic (ELE) interaction forces between PI3Kδ and
   the compounds were calculated without considering solvation, and the
   binding forces were analyzed during the simulation. The ELE interaction
   fluctuated considerably compared with the VAW forces, but the ELE
   interaction gradually stabilized, indicating that the binding between
   PI3Kδ and the compounds was stable ([138]Figure 13a). The binding
   energies between PI3Kδ and kaempferol, luteolin, and quercetin were
   −72.364 ± 2.505, −61.419 ± 0.801, and −64.034 ± 2.839 kJ/mol,
   respectively, with considering solvation ([139]Table 1). The binding
   energy was decomposed to obtain the contribution of each amino acid to
   the overall binding energy. [140]Figure 13b shows that the key amino
   acids in the PI3Kδ-kaempferol, PI3Kδ-luteolin, and PI3Kδ-quercetin
   complexes were ILE-910, MET-900, and ILE-825, respectively.

Figure 13.

   [141]Figure 13
   [142]Open in a new tab

   Analysis of the interaction and key amino acid residues of PI3kδ and
   the compounds in molecular dynamic simulations. (a) van der Waals force
   (VDW) and electrostatic (ELE) interaction analysis of PI3kδ and
   compounds. (b) The key amino acid residues between PI3kδ and compounds.

Table 1.

   Binding Free Energy of PI3Kδ with the Compounds (kJ/Mol)
   Complex Inline graphic Inline graphic Inline graphic Inline graphic
   Inline graphic
   PI3Kδ-Kaempferol −139.4±0.925 −63.467±0.988 147.813±1.713 −17.311±0.113
   −72.364±2.505
   PI3Kδ-Luteolin −127.289±2.323 −97.512±4.671 180.68±1.847 −17.298±0.054
   −61.419±0.801
   PI3Kδ-Quercetin −139.114±4.5 −125.144±3.341 217.266±3.974 −17.042±0.111
   −64.034±2.839
   [143]Open in a new tab

   Note: Inline graphic van der Waals energy; Inline graphic ,
   electrostatic energy; Inline graphic , polar solvation energy; Inline
   graphic , nonpolar solvation energy; Inline graphic , binding free
   energy. Inline graphic .

PI3Kδ Was Elevated in Synovial Tissue of RA Patients

   To identify the role of PI3Kδ in RA, PI3Kδ expression in the synovial
   tissue of patients with RA was profiled by bioinformatics analysis of
   two publicly available GSE datasets. In the [144]GSE55235 dataset, the
   mRNA of PI3Kδ (gene name is PIK3CD) was significantly upregulated in
   the synovial tissue of patients with RA (n = 10) compared with those in
   the synovial tissue of healthy people (n = 10) ([145]Figure 14a). PI3Kδ
   was also highly expressed in the [146]GSE55457 dataset (n = 10 for
   health controls and n = 13 for RA) ([147]Figure 14a). Interleukin 7
   (Il7), interleukin 15 (Il15), and interleukin 18 (Il18) were considered
   key biomarkers in RA development and progression that act as mediators
   in the etiology of RA and also serve as a diagnostic tool for RA.
   Specifically, the levels of PI3Kδ showed a positive correlation with
   the levels of Il7, Il15, and Il18 in the synovial tissue of patients
   with RA ([148]Figure 14b–d). This shows that PI3Kδ plays the role of a
   promoter in RA and PI3Kδ may be potentially targeted to treat RA.

Figure 14.

   [149]Figure 14
   [150]Open in a new tab

   PI3kδ was elevated in the synovial tissue of patients with RA. (a) The
   mRNA levels of Pik3cd (gene name of PI3kδ) in synovial specimens of RA
   in the [151]GSE55235 and [152]GSE55457 datasets. Unpaired t-test. *
   P<0.05 and **** P<0.0001 compared with the healthy group. (b-d)
   Pearson’s correlation of Pik3cd with the mRNA levels of Il7, Il15, and
   Il18 in RA in the [153]GSE55235 and [154]GSE55457 datasets. Two-tailed
   Pearson χ^2 test.

Discussion

   In this study, network pharmacology analysis of the 75 active
   ingredients of ASSD and the 1258 RA-related targets revealed that
   kaempferol, luteolin, and quercetin are the key ingredients of ASSD,
   and MAPK1, MAPK3, JUN, IL6, AKT1, TNF, PTGS2, MMP9, STAT3, and CXCL8
   are the key targets of ASSD for RA treatment. The inflammatory-related
   pathways (the PI3K-AKT, TNF, and IL17 signaling pathways) were
   predicted to be involved in ASSD-mediated RA treatment. With the goal
   of ensuring low disease activity or remission, a tight control of
   inflammation is the central focus of the pharmacological treatment of
   RA.

   TNF and IL6 are the underlying targets of ASSD-mediated protection
   against RA. The critical roles of TNF and IL6 in RA are strongly
   supported by the successful clinical benefits of anti-TNF and anti-IL6
   therapies.[155]2 In RA, TNF is mainly produced by monocytes and
   macrophages and also by T cells, B cells, and synovial
   fibroblasts.[156]23,[157]24 TNF primes for or directly mediates
   leukocyte and endothelial cell activation, adhesion, and migration,
   synovial fibroblast activation and survival, cytokine release, and
   angiogenesis, which together represent key pathological features of
   RA.[158]23 Patients with RA who exhibit inadequate response to csDMARD
   often have to receive a first line TNF-α inhibitor as the initial
   bDMARD, unless contraindicated.[159]1 Approximately 70% of patients
   with RA respond to therapeutic inhibition of TNF with antibodies or
   soluble receptors.[160]23 In RA, IL6 is mainly produced by monocytes,
   synovial fibroblasts, B cells, and T cells.[161]23 IL6 regulates innate
   responses similar to TNF in the local synovial environment, but it also
   directly drives the acute-phase response.[162]25,[163]26 In addition,
   IL6 performs an essential immunoregulatory role in adaptive immunity,
   participating in the differentiation of naïve T-cell toward the
   proinflammatory T[H] 17 cell or regulatory T[REG] cell pathway.[164]25
   Targeting IL6 with neutralizing antibodies or Janus kinase (JAK)
   inhibitors has been shown to be effective against RA.[165]27 However,
   there may be some degree of specificity in eliciting inflammatory
   subgroups and synovial changes between TNF and IL6
   inhibition.[166]25,[167]28 Reduction in TNF and IL6 levels is indeed
   involved in the treatment of RA with A. membranaceus,[168]29 L.
   japonica,[169]30 P. cuspidatum,[170]31 and H. diffusa,[171]32 which is
   consistent with the prediction that the therapeutic effects of ASSD
   against RA are mediated, at least in part, via TNF and IL6.

   The PI3K-AKT signaling pathway is assumed to play a crucial role in the
   antirheumatic effects of ASSD. PI3Ks are a large family of lipid
   kinases that phosphorylate phosphoinositides and are divided into
   classes I, II, and III. Class I PI3Ks are further divided into class IA
   (comprising the PI3Kα, PI3Kβ, and PI3Kδ isoforms) and class IB
   (comprising PI3Kγ only), among which PI3Kδ and PI3Kγ are mainly
   expressed in leukocytes.[172]33 AKT, a serine/threonine kinase, is the
   central mediator of PI3Ks and has multiple downstream effectors.[173]34
   The PI3K-AKT signaling pathway, most notably mediated by PI3Kδ and
   PI3Kγ, participates in multiple steps and various stages of RA
   pathogenesis, such as in antigen signaling in B and T cells, signaling
   downstream of FcRs, and cytokine receptors and chemokine receptors in
   mast cells, macrophages, neutrophils, and synoviocytes.[174]33 In vivo
   and in vitro studies have shown that inhibition of the PI3K-AKT
   signaling pathway through pharmacological inhibitors or genetic
   deletion of PI3Ks can reduce inflammatory cell infiltration,
   synoviocyte proliferation and survival, and bone destruction in
   RA.[175]34–37 The PI3K-AKT signaling pathway has attracted considerable
   research interest as a pharmacological target in the treatment of RA.
   Based on the wide range of PI3K-AKT signaling in RA, ASSD may have
   specific roles at all stages of RA progression via this signaling
   pathway.

   According to the results of network and pathway analyses, kaempferol,
   luteolin, and quercetin are the key ingredients of ASSD involved in its
   effect against RA, and the anti-inflammatory effect is the major
   mechanism underlying ASSD-mediated RA treatment. Thus, the role of
   kaempferol, luteolin, and quercetin in the anti-inflammatory effects
   was explored in LPS-stimulated RAW 264.7, and it was found that all
   three ingredients decreased the production of proinflammatory factors
   in macrophages. The inhibition of glycolysis plays a crucial role in
   the anti-inflammatory properties of kaempferol, luteolin, and
   quercetin. The anti-inflammatory effects of kaempferol, luteolin, and
   quercetin have been reported previously;[176]38–40 however,
   investigations on the pharmacological mechanism have mainly focused on
   signaling pathways, such as NF-κB, MAPK, and JAK/STAT, but not on
   glycolysis. Upon LPS stimulation, macrophages shift from ATP production
   via oxidative phosphorylation to glycolysis.[177]19 Glycolysis is less
   efficient than aerobic oxidation, but may prepare macrophages for the
   rapid activation and synthesis of immune mediators required to
   perpetuate inflammation; thus, it is used by inflammatory
   macrophages.[178]41,[179]42 In this study, the inhibition of PI3K-AKT
   signaling by kaempferol, luteolin, and quercetin decreased the glucose
   uptake and pyruvate-to-lactate conversion by diminishing the expression
   of Glut1 and Ldha, which inhibits glycolytic metabolism. PI3K-AKT
   signaling plays a multifaceted role in controlling glucose metabolism
   in T effector cells by promoting the expression of Glut1 and
   LDHA.[180]22 We found that glycolysis in macrophages was also
   controlled by PI3K-AKT signaling via a similar mechanism. In the
   inflamed joints, glucose and lactate concentrations were distinctly low
   and high, respectively. Moreover, macrophages from patients with RA
   commit to a hypermetabolic state with intensified glycolysis.[181]43
   Future studies are required to investigate the role of ASSD in
   glycolysis, inflammation, and disease activity in animal models of RA.
   Targeting metabolic checkpoints may provide new opportunities for
   developing alternative RA treatment strategies.

Conclusion

   In conclusion, using integrated network pharmacology, we identified
   kaempferol, luteolin, and quercetin as the key components of ASSD that
   can be used to treat RA via inhibition of inflammation. We further
   confirmed that kaempferol, luteolin, and quercetin inhibited
   LPS-induced glycolysis in macrophages by blocking PI3K-AKT signaling,
   leading to a decreased expression of proinflammatory factors.
   Kaempferol, luteolin, and quercetin stably occupied the hydrophobic
   pocket of PI3Kδ, which may be the potential target of kaempferol,
   luteolin, and quercetin to treat RA. The various active ingredients of
   ASSD lead to synergistic effects on RA by acting on multiple targets
   and pathways. The findings of this study only partly reveal the complex
   effect of ASSD in RA but can provide new directions that would improve
   the overall understanding underlying the therapeutic value of this
   decoction.

Acknowledgments