A Novel Drug Combination of Mangiferin and Cinnamic Acid Alleviates Rheumatoid Arthritis by Inhibiting TLR4/NFκB/NLRP3 Activation-Induced Pyroptosis

Growing evidence shows that Baihu-Guizhi decoction (BHGZD), a traditional Chinese medicine (TCM)-originated disease-modifying anti-rheumatic prescription, may exert a satisfying clinical efficacy for rheumatoid arthritis (RA) therapy. In our previous studies, we verified its immunomodulatory and anti-inflammatory activities. However, bioactive compounds (BACs) of BHGZD and the underlying mechanisms remain unclear. Herein, an integrative research strategy combining UFLC-Q-TOF-MS/MS, gene expression profiling, network calculation, pharmacokinetic profiling, surface plasmon resonance, microscale thermophoresis, and pharmacological experiments was carried out to identify the putative targets of BHGZD and underlying BACs. After that, both in vitro and in vivo experiments were performed to determine the drug effects and pharmacological mechanisms. As a result, the calculation and functional modularization based on the interaction network of the “RA-related gene–BHGZD effective gene” screened the TLR4/PI3K/AKT/NFκB/NLRP3 signaling-mediated pyroptosis to be one of the candidate effective targets of BHGZD for reversing the imbalance network of “immune-inflammation” during RA progression. In addition, both mangiferin (MG) and cinnamic acid (CA) were identified as representative BACs acting on that target, for the strong binding affinities between compounds and target proteins, good pharmacokinetic features, and similar pharmacological effects to BHGZD. Notably, both BHGZD and the two-BAC combination of MG and CA effectively alleviated the disease severity of the adjuvant-induced arthritis-modified rat model, including elevating pain thresholds, relieving joint inflammation and bone erosion via inhibiting NF-κB via TLR4/PI3K/AKT signaling to suppress the activation of the NLRP3 inflammasome, leading to the downregulation of downstream caspase-1, the reduced release of IL-1β and IL-18, and the modulation of GSDMD-mediated pyroptosis. Consistent data were obtained based on the in vitro pyroptosis cellular models of RAW264.7 and MH7A cells induced by LPS/ATP. In conclusion, these findings offer an evidence that the MG and CA combination identified from BHGZD may interact with TLR4/PI3K/AKT/NFκB signaling to inhibit NLRP3 inflammasome activation and modulate pyroptosis, which provides the novel representative BACs and pharmacological mechanisms of BHGZD against active RA. Our data may shed new light on the mechanisms of the TCM formulas and promote the modernization development of TCM and drug discovery.


INTRODUCTION
Rheumatoid arthritis (RA), characterized by hyperplasia and inflammation of the synovial joints, is an intractable and highly prevalent autoimmune disease with unknown pathogenic triggers (1,2). Currently, the key therapeutic approaches to RA are mainly based on drugs and surgeries (3), including disease-modifying anti-rheumatic drugs (DMARDs), non-steroidal antiinflammatory drugs (NSAIDs), glucocorticoids, and biological response modifiers, which may reduce synovitis and systemic inflammation (4). However, there are no specific drugs for RA therapy due to its various comorbidities, such as cardiovascular diseases (5)(6)(7)(8), osteoporosis (9,10), interstitial lung disease (9,11), malignancies (11,12), and hypertension (13), as well as the heterogeneous and complex clinical manifestations with wide individual differences. Traditional Chinese medicine (TCM), as a complementary and alternative medicine, has its unique advantages in the clinical treatment of RA, especially at the active stage with moist heat arthralgia spasm syndrome (active RA), which is characterized by the hyperactive immune response and excessive inflammatory cytokines, subsequently leading to osteoclast differentiation and active joint inflammation (14,15). Among TCM-originated DMARDs, Baihu-Guizhi decoction (BHGZD) has been indicated to efficiently reverse the aggressive progression of active RA with moist heat arthralgia spasm syndrome and to reach a remarkable response rate (90%) (16)(17)(18)(19). Our previous studies identified the chemical profiling of BHGZD in vitro and verified its immunomodulatory and antiinflammatory activities (20)(21)(22). However, the bioactive compounds (BACs) and the underlying mechanisms of BHGZD against active RA have not been fully elucidated.
In the current study, the candidate BACs of BHGZD were identified by the ultra-fast liquid chromatography-quadrupoletime-of-flight tandem mass spectrometry (UFLC-Q-TOF-MS/ MS) system in vivo. Then, we carried out transcriptomic profiling, target prediction, and network calculation to screen the candidate targets of BHGZD for reversing the imbalance network of "immune-inflammation" during the progression of active RA. After that, the binding affinities between candidate BACs and candidate targets of BHGZD were calculated using the molecular docking simulation, the results of which were subsequently verified by surface plasmon resonance (SPR) assay and microscale thermophoresis (MST) analysis.
Moreover, the in vivo biodistribution and pharmacokinetic characteristics of candidate BACs were also determined. Furthermore, the pharmacological effects and the molecular mechanisms of BHGZD, and its representative BACs against active RA were further validated by a series of experiments based on the adjuvant-induced arthritis-modified rat model (AIA-M) as well as LPS/ATP-induced pyroptosis cellular models on both RAW264.7 macrophage and MH7A cells.

Preparation of BHGZD
Crude drugs of BHGZD formula were purchased from Beijing Tongrentang Co., Ltd. (Beijing, China). BHGZD was prepared according to the original composition of the formula recorded in Chinese pharmacopoeia 2020 edition and our previous study (22) (Supplementary Materials Section 1, and Table S1).

RA Modeling, Grouping, and Treatment
A total of 49 male Lewis rats (200 ± 20 g in weight, 6-8 weeks old) were obtained from Beijing Vital River Laboratory Animal Technology Co., Ltd. (SCXK 2019-0003, Beijing, China) and randomly divided into two independent clusters: a discovery cluster (n = 24) and a validation cluster (n = 25). All animals were maintained in specific pathogen-free conditions with a constant temperature of 24 ± 1°C with a 12-h light/dark cycle, and had free access to standard rodent diet and water. The discovery and validation clusters were respectively used for gene expression profiling and experimental validation. In the discovery cluster, 24 rats were randomly divided into three groups [3 per group for microarray and 5 per group for mRNA sequencing (mRNA-Seq)] (1): the normal control group, (2) the AIA-M model group, and (3) the BHGZD treatment group. In the validation cluster, 25 rats were randomly divided into five groups (5 per group): (1) the normal control group, (2) the AIA-M model group, (3) the BHGZD treatment group, (4) the MG+CA treatment group, and (5) the methotrexate (MTX) treatment group.
The AIA-M rat model was established simulating the pathological changes and characteristics of active RA with moist heat arthralgia spasm syndrome based on our previous studies (20)(21)(22)(23). Please see the detailed information on the AIA-M rat model establishment in Supplementary Materials Section 2. The dosage of BHGZD for the corresponding treatment group was 21.4 g/kg, equivalent to two times the clinical dose, which has been proved to exert the most prominent therapeutic efficacy in our previous studies (21,22). The dosages for the two-BAC combination of MG and CA for the corresponding treatment group were respectively 600.912 mg/kg and 46.652 mg/kg, equivalent to their content in 21.4 g/kg BHGZD. The dosage for MTX was 0.2 mg/kg. All treatments were performed for 30 days via oral administration from the day of the primary immunization (20)(21)(22)(23).
On the 31st day, all rats were anesthetized by intraperitoneal injection of pentobarbital sodium (50 mg/kg). The blood samples were collected using one-time anticoagulant negative pressure blood collection tubes, and then placed at room temperature for 20 min and centrifuged at 3,000 rpm for 15 min. The sera were collected, vortexed, and centrifuged again at high speed at low temperature (4°C) for 15 min, and stored at −20°C for subsequent assays. Whole blood cells were freshly isolated from blood, then frozen in liquid nitrogen overnight, and stored at −80°C for mRNA-Seq. The synovium tissues were isolated and immediately frozen in liquid nitrogen for microarray analysis. The right joints of rats were removed and preserved in 4% paraformaldehyde (PFA) combined with phosphate-buffered saline (PBS) for 72 h for subsequent assays. The left joints and organs of rats were kept at −80°C for subsequent assays.

Viscera Index and Histology
The viscera indexes were calculated by the weight of the thymus, spleen, liver, and kidney relative to total brain weight. Histological changes were examined using hematoxylin and eosin (H&E), safranin O-fast green (Solarbio, Beijing, China), and Masson trichrome staining (Baso, Wuhan, China) according to routine protocols. Please see detailed information on the protocol in Supplementary Materials Section 5.

Micro-Computed Tomography Analysis
To quantitatively evaluate bone formation within the defects, the specimens were scanned using a micro-CT instrument (GE healthcare, USA) at a resolution of 45 mm. The x-ray settings and cylinder region of interest (ROI) are provided in Supplementary Materials Section 6.

Gene Expression Profiling
Whole blood cells and the synovium tissues representing the pathological characteristics of systemic disease and target organs during active RA progression were respectively used for whole rat genome microarray analysis (Agilent Technologies, design ID: 014879, Santa Clara, CA, USA) and mRNA-Seq Illumina NovaSeq 6000 (Illumina, CA, USA) to screen RA-related genes [the significant differentially expressed genes (DEGs) between the AIA-M model group and the normal control group] and BHGZD therapeutic effect-related genes (DEGs between the BHGZD treatment group and the AIA-M model group). The gene expression data have been uploaded and are publicly available in the NCBI GEO database (GEO No. GSE189942, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc= GSE189942, December 3, 2021, and GEO No. GSE190523, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc= GSE190523, December 11, 2021), and the DEGs were identified by referring to the criteria of t-test p-value < 0.05 and fold change (FC)>2/<0.5. Please see the detailed information in Supplementary Materials Section 7.

Network Construction and Analysis
The RA-related gene interaction network was constructed using the links among the DEGs between the AIA-M model and normal control groups to identify the key RA-related genes. Then, the interaction network of the "key RA-related gene-BHGZD-effective gene" was constructed using the links between the key RA-related genes and BHGZD effective genes. The genegene interaction data were collected from the String database (version 10.0, http://string-db.org/) (27), and the interaction networks were all visualized by the Cytoscape platform (version 3.9.0, https://cytoscape.org/) (28). The network topological properties of nodes, including "Degree", "Betweenness", and "Closeness", were calculated for screening the key network targets according to our previous descriptions (20)(21)(22)(23)(24)(25)29) and Supplementary Materials Section 8.

Determining the Chemical Compounds of BHGZD and Absorption Distribution Metabolism Excretion Evaluation
A rapid, sensitive, and reliable method by the UFLC-Q-TOF-MS/ MS system was used to identify the chemical profiling of BHGZD as described in Supplementary Materials Section 9 (30). The reference standards of these compounds (Supplementary  (26). In addition, in silico ADME models were used to identify the candidate BACs of BHGZD by calculating the passive intestinal permeability of the Caco-2 module, the oral bioavailability, and the apparent permeability coefficient (Papp) of each BACs as described in Supplementary Materials Section 10.

Molecular Docking Simulation
The molecular docking and virtual screening program was carried out to investigate the direct binding efficiencies of the two BACs contained in BHGZD and the corresponding putative targets. The structures of human TLR4 (PDB ID: 2z65), AKT1 (PDB ID: 3o96), and NFkB (PDB ID: 1nfi) were obtained from the PDB website (

SPR Assay
To confirm the binding affinity of recombinant AKT1 protein and MG, SPR assay was performed with Biacore 8K (Biacore, Cytiva), and the K D value was calculated using the Biacore 8K evaluation software 2.0 (GE Healthcare). Detailed information on SPR assay is provided in Supplementary Materials Section 11.

MST Assay
The binding affinity between recombinant TLR4 protein and CA was measured by a NanoTemper Monolith NT.115 instrument (NanoTemper Technologies, Germany). The K D value was fitted by NanoTemper Monolith affinity software (NanoTemper Technologies, Germany) using 1:1 binding mode. Detailed information on the MST assay is provided in Supplementary Materials Section 12.

In Vivo Pharmacokinetic Analysis
To detect the pharmacokinetic parameters of MG and CA in vivo, quantification analysis was performed by an ultra-highperformance liquid chromatography (Shimadzu Corp., Japan) tandem ion trap quadrupole QTRAP 6500 plus mass spectrometry (AB Sciex, USA). The quantitative analyses for MG and CA were approved by the Guidance for For the drug treatment, both RAW264.7 and MH7A cells were divided into the following groups: (1) Normal control group: with no stimulation and treatment; (2) LPS/ATPinduced cellular model group: cells were stimulated with the corresponding concentrations of LPS and ATP; (3) BHGZD treatment group: after stimulation, cells were treated with 28.54 mg/ml BHGZD for 24 h, which was determined to be the most effective as in our previous study (22); (4) MG+CA treatment group: after stimulation, cells were treated with the two-BAC combination (1.69 ng/ml MG and 0.13 ng/ml CA, with the same content as that in the 5 mg/ml BHGZD formula) for 24 h; (5) MG treatment group: after stimulation, cells were treated with 1.69 ng/ml MG for 24 h; and (6) CA treatment group: after stimulation, cells were treated with 0.13 ng/ml CA for 24 h. The concentration of DMSO was less than 1‰ of the solution for the in vitro experiment.

Cell Viability Assay
Cell viability was analyzed with the cell counting kit-8 (CCK-8) assay kit (Beyotime Biotechnology, Shanghai, China) as described in Supplementary Materials Section 16.

Flow Cytometry
To identify and quantify the pyroptosis population of RAW264.7 cells, FITC Apoptosis Detection kit I staining with both the Annexin V-fluorescein isothiocyanate and propidium iodide (PI, BD Biosciences, San Jose, CA, USA) was used according to the manufacturer's guidelines. Flow cytometry was performed using NovoCyte 2040R (ACEA Bioscience, San Diego, CA, USA), and analyzed using the NovoExpress 1.4.1 software (ACEA Bioscience, San Diego, CA, USA). Detailed information of this experiment is provided in Supplementary Materials Section 17.

Terminal Deoxynucleotidyl Transferase-Mediated dUTP Biotin Nick End Labeling Staining
The occurrence of pyroptosis in both MH7A cells and synovium tissues in AIA-M rats was determined by in situ TUNEL staining

Immunofluorescence Staining
Double fluorescence staining was performed as described previously (22)

Immunohistochemical Staining
To detect the expression levels of NLPR3/ASC of the knee joints in AIA-M rats and corresponding treatment groups, immunohistochemical staining was carried out using a DAB kit (Cat No. AR1027, Boster Biological Technology Co., Ltd., Wuhan, China) and a rabbit/mouse two-step detection kit (Cat No. SV0002/SV0001, Boster Biological Technology Co., Ltd., Wuhan, China) according to the routine protocols. Immunohistochemistry quantification was performed using ImageJ (Image Progressing and Analysis in Java, version 1.42q, https://imagej.nih.gov/ij/), following the ImageJ User Guide.

Cytokine Analysis and Lactate Dehydrogenase Measurement
To evaluate the therapeutic effects of two BACs from BHGZD on the regulation of the "immune-inflammation" system, the levels of cytokines, proteins, and LDH release in RAW264.7 and MH7A cellular supernatant, as well as the sera of AIA-M rats were determined using enzyme-linked immunosorbent assay (ELISA) to assess the integrity of membranes in accordance with the manufacturer's guidelines.

Statistical Analyses
Statistical analyses were performed using GraphPad Prism 8.0 Software (San Diego, CA, USA). Data are expressed as the mean ± SD and analyzed by one-way ANOVA with Bonferroni's or Dunnett's post-hoc test for comparison of multiple columns. Differences were considered statistically significant when the pvalue was less than 0.05.

RESULTS
BHGZD Reverses the Imbalance of the "Immune-Inflammation" System in Active RA Rats The AIA-M rat model was successfully established (the achievement ratio of this model was 100%) and verified according to its reliable and rapid-onset characteristics, and severe disease progression by external stimulus of wind, dampness, and heat, which may be the major pathological changes and features of active RA with moist heat arthralgia spasm syndrome (22). In contrast, the administration of BHGZD effectively alleviated the disease severity in arthritis of active RA rats. Detailed information on the characterizations of the active RA rat model and the therapeutic effects of BHGZD in active RA severity based on this model was reported in our previous study (22).
To determine the candidate targets of BHGZD against active RA, the gene expression profiles of whole blood cells and synovium tissues were carried out. A total of 473 RA-related genes (275 upregulated and 178 downregulated genes) and 1,802 BHGZD therapeutic effect-related genes (456 upregulated and 1,346 downregulated genes) were identified. The heatmap and volcano plot of DEGs are shown in Figures 1A-H, suggesting a distinct separation between the AIA-M model and normal control groups, as well as the BHGZD treatment and AIA-M model groups. Among them, there were 135 RA-related genes reversely regulated by the treatment of BHGZD ( Table 1). Detailed information on the RA-related genes and BHGZD therapeutic effect-related genes is respectively provided in Supplementary Tables S4 and S5.
Then, the interaction network of the "RA-related gene-BHGZD effective gene" was constructed using the links among RA-related genes and BHGZD effective genes. Following the calculation of the nodes' topological features in the network (the median values of "Degree", "Betweenness", and "Closeness" were 7.0000, 0.0009, and 0.2860, respectively), four BHGZD candidate targets (PI3K, AKT1, NFkB, and IL-1b) against active RA ( Figure 1I, and Supplementary Table S6) were identified due to their topological importance. In addition, previous data obtained from our research group and from other research groups indicated that BHGZD may restore the imbalance of the "immuneinflammation" system via inhibiting TLR4-induced NLRP3 inflammasome signaling (21,22), and TLR4/NFkB signaling may play a vital role in the regulation of the inflammatory response (32,33). Moreover, PI3K/AKT signaling is involved in the process and release of pro-inflammatory cytokines (34,35), and the activation of this signaling leads to autoimmunity, showing the increased activity in some autoimmune diseases, including RA (36). On this basis, we hypothesize that BHGZD might reverse the main pathological changes of active RA, including synovial inflammation, cartilage damage, and bone erosion, via regulating TLR4/PI3K/AKT/NFkB/NLRP3 signaling.

MG and CA May Be the Representative BACs Contained in BHGZD Against Active RA
To screen the potential BACs contained in BHGZD against active RA, herein we identified the chemical profiling of this herbal formula in sera using the UFLC-Q-TOF-MS/MS system. As a result, a total of 31 chemical compounds were identified in the serum samples 2 h after the treatment of BHGZD. Among them, 14, 16, and 1 chemical compound were from Anemarrhenae Rhizoma, Glycyrrhizae Radix et Rhizoma, and Cinnamomi Ramulus, respectively, and belong to glycosides, flavonoids, organic acid, triterpenoids, and saponins. Quantitatively, timosaponin B II, mangiferin, glycyrrhizic acid, neomangiferin, 7-O-methyl mangiferin, anemarrhensaponin I, vitexin, formononetin, liquiritigenin, isoliquiritin, isoliquiritin apioside, isoschaftoside, and isovitexin were all detected at 30 min after the treatment of BHGZD. Detailed information on chemical profiling is provided in Supplementary Table S7, and the quantitative detection data are provided in Supplementary  Table S8. In addition, the drug-like properties of the chemical compounds contained in BHGZD, including the intestinal absorption rate and oral bioavailability, were evaluated using ADME models in silico, and a total of 8 candidate BACs with good drug-likeness were identified (Supplementary Table S9).
After that, molecular docking was performed to determine the binding affinities of candidate BACs contained in BHGZD to corresponding proteins of TLR4/PI3K/AKT/NFkB/NLRP3 signaling. As a result, a total of 9 chemical compounds were    found to bind these proteins with binding affinities of more than −4.0 kcal/mol (Supplementary Table S10). In particular, MG ( Moreover, the content of MG and CA was determined to be 336.96 ng and 26.14 ng in 1 mg of BHGZD lyophilized powder, respectively, with a conversion ratio of 8.33% (calculated by the weight of BHGZD lyophilized powder divided by the weight of crude herbs contained in BHGZD; 1 g of BHGZD lyophilized powder is equivalent to the amount of 12 g of BHGZD).
The UHPLC-QTRAP-MS/MS system was further used to detect the pharmacokinetic properties of MG and CA in sera after the treatment of BHGZD (Supplementary Figure S1). A summary of MG and CA pharmacokinetic parameters using a non-compartment model analysis is presented in Table 2, and the mean plasma concentration-time profiles from time 0 to 24 h after

Both BHGZD and the Two-BAC Combination of MG and CA Improve Arthritis Severity in Active RA Rats
To verify the pharmacological effects of the two-BAC combination of MG and CA against active RA, the in vivo experiments were performed based on the AIA-M rat model (22,24). Similar pathological characteristics and changes in AIA-M rats were observed from the validation cluster to the discovery cluster ( Figure 3). As a result, the ankles and knuckle joints of AIA-M rats with severe redness and swelling were remarkably improved by the treatment of the two-BAC combination (equivalent to BHGZD at a dose of 21.4 g/kg) and BHGZD (dose of 21.4 g/kg, Figures 3A and B). Importantly, both the treatment of BHGZD and the two-BAC combination significantly alleviated disease severity, including reducing arthritis incidence, the diameter of the limb, and arthritis score (all p < 0.05, on the 27th day after immunization), and simultaneously elevated pain thresholds (mechanical-, acetone-, and thermal-induced hyperalgesia), all of which were similar to the pharmacological effects of MTX (positive drug, dose of 0.2 mg/kg, Figures 3C and D).
In order to investigate the joint destruction and synovial inflammation, both the ankle and knee joints of AIA-M rats in different groups were scanned using Micro-CT analysis on the 31st day after immunization. The quantified data revealed that BMD, TMD, BV/TV ratio, and Tb.Th were dramatically decreased, and BS/BV ratio and Tb.Sp was observably increased in the AIA-M model group (all p < 0.05, Figure 3E and Supplementary Figure S2B) with rough bone surfaces and severe bone erosion ( Figure 3F, and Supplementary Figure  S2A). As shown in Figure 3F, both BHGZD and the two-BAC combination treatment groups showed smooth bone surface with a significantly increased BMD, TMD, BV/TV ratio, and Tb.Th, as well as a significantly decreased BS/BV ratio and Tb.Sp (all p < 0.05, Figure 3E and Supplementary Figure S2B), suggesting that the two-BAC combination may efficiently reverse bone erosion, which was similar to the pharmacological effects of MTX.
Moreover, H&E, safranin O fast green, and Masson trichrome staining were carried out to evaluate the degree of joint lesions in AIA-M rats. As shown in Figures 3G and H, the treatment of the two-BAC combination, BHGZD, and MTX all apparently reversed the histopathological changes of knee joints in AIA-M rats, including the decreased inflammatory cell infiltration, the prevention of cartilage and bone destruction, and synovial hyperplasia (all p < 0.05). Consistently, a significant reduction of the safranin O fast green-positive area and an elevation of the Masson-positive area were observed on the cartilage surface of AIA-M rats, reflecting a loss of articular cartilage, but reversed by the two-BAC combination, similar to the pharmacological effects of BHGZD and MTX (all p < 0.05, Figures 3G and H). These findings revealed that BHGZD and the two-BAC combination of MG and CA may effectively alleviate the progression of bone damages, repair bone erosion, and simultaneously improve pathological changes of articular cartilage and synovial inflammation in the arthritic joints of AIA-M rats.
In terms of the response to inflammation, pathological changes of the thymus and spleen and viscera indexes in different groups were examined. As shown in Figure S2C, the lower proportion of white pulp in the spleen with decreased cell density lymphatic sheath, lymphoid follicular hyperplasia, marginal zone, red pulp, and germinal center was observed in AIA-M rats, which were significantly improved by the treatment of BHGZD and the two-BAC combination (all p < 0.05, Supplementary Figure S2D). A thinner thymic cortex, less thymic lobule, an unclear boundary, and more vacuoles in the cytoplasm of epithelial reticular cells were also observed in AIA-M rats, which were remarkably improved by the treatment of BHGZD and the two-BAC combination (Supplementary Figure  S2C). The treatment with BHGZD and the two-BAC combination protected the morphological structure of the spleen and thymus, and simultaneously decreased spleen and thymus indexes significantly (all p < 0.05, Supplementary Figure S2E).  Furthermore, H&E-stained tissues of liver and kidney, and the liver and kidney indexes showed no toxic damages ( Supplementary  Figures S2C and F).
Both BHGZD and the Two-BAC Combination of MG and CA Suppress NLRP3 Inflammasome-Induced Pyroptosis Via Regulating TLR4/PI3K/AKT/NFkB Signaling The above network-based data imply that BHGZD might reverse the imbalance of the "immune-inflammation" system during active RA progression via regulating TLR4/PI3K/AKT/NFkB/NLRP3 signaling. To the best of our knowledge, pyroptosis is a kind of NLRP3 inflammasome-induced inflammatory cell death, characterized by cell swelling and release of pro-inflammatory cytokines depending on the activation of caspase-1, contributing to inflammation in arthritis (37). The formation of the NLRP3 inflammasome complex requires the interaction of NLRP3 with ASC and caspase-1, which are vital for the assembly and activation of the NLRP3 inflammasome. In our previous study, the immunomodulatory and anti-inflammatory activities of BHGZD were verified, especially in inhibiting pyroptotic death, which may be attributed to the activation of TLR4-NLRP3 inflammasome signaling (22). Currently, our TUNEL data showed a significant proportion of pyroptosis cells in the arthritic joints of AIA-M rats, which was improved by the treatment of BHGZD and the two-BAC combination ( Figure 4A). Similar to the pharmacological effects of MTX, high expression levels of NLRP3 and ASC, high expression and enhanced activity of caspase-1, and increased expression of IL-1b and IL-18 (the signature inflammatory cytokines in pyroptosis) in the AIA-M model group (all p < 0.05) were all significantly decreased by the treatment of BHGZD and the two-BAC combination (all p < 0.05, Figures 4B-E). Pyroptosis was assessed by LDH activity, which may be used to verify cell membrane integrity and the release of intracellular soluble component. Consistently, the enhancing activity of LDH in the sera of AIA-M rats was observed, which was significantly decreased by the treatment of BHGZD, the two-BAC combination, and MTX (all p < 0.05, Figure 4E), suggesting the alleviation of downstream membrane damage.
To verify the in vivo findings based on AIA-M rats, an established method (LPS plus ATP) was applied to induce NLRP3 inflammasome activation in both RAW264.7 and MH7A cells. The cell viability on RAW264.7 cells with the treatment of MG and CA was initially examined using the CCK8 assay, and the results exhibited no cell toxicity under 0.21-6.74 ng/ml MG or 0.02-0.52 ng/ml CA treatment alone, as shown in Supplementary Figure S2G. Thus, 1.69 ng/ml MG and 0.13 ng/ml CA treatment were chosen in the following assays (the same content as that in the 5 mg/ml BHGZD formula). Flow cytometry analysis revealed that MG, CA, and the two-BAC combination treatment prominently reduced the PI [a marker of cells that stains necrotic, dead, and membrane-compromised cells (38)] positive cell rate of RAW264.7 cells induced by LPS/ATP,  which was inferior to that in the BHGZD treatment group (all p < 0.05, Figures 5A and B). As shown in Figures 5C and 6A, LPS/ ATP induced an increase in the number of both FLICA-positive RAW264.7 cells and TUNEL-positive MH7A cells, which was reduced by the treatment of BHGZD, MG, CA, and the two-BAC combination. Intriguingly, the inhibition of MG or CA alone on cell pyroptosis induced by LPS/ATP was weaker than that in the two-BAC combination treatment group (all p < 0.05, Figures 5A-C and 6A). In addition, the expression levels of NLRP3, ASC, and caspase-1 were significantly increased in both RAW264.7 and MH7A cells induced by LPS/ATP, the same as the levels of IL-1b, IL-18, and LDH release in the supernatant of RAW264.7 and MH7A cells, which were all reduced by the treatment of BHGZD, MG, CA, and the two-BAC combination (all p < 0.05, Figures 5D-G and 6B-E). GSDMD, a crucial mediator of pyroptosis downstream of canonical and non-canonical inflammasomes (39)(40)(41), is cleaved by caspase-1 at a specific site (GSDMD-NT) and subsequently causes cell lysis and IL-1b release (42,43). In the current study, the activities of GSDMD-NT in sera and its levels in knee joints were both significantly elevated in AIA-M rats, which was reduced by the treatment of BHGZD, the two-BAC combination, and MTX (all p < 0.05, Figures 4C-E).
Notably, GSDMD-NT occurred in both RAW264.7 and MH7A cells induced by LPS/ATP, which was decreased by the treatment of BHGZD, MG, CA, and the two-BAC combination ( Figures 5E,  F and 6C, D).
After determining the inhibitory effects of BHGZD and the two-BAC combination on NLRP3 inflammasome-induced pyroptosis, we further investigated the changes of its upstream determinant TLR4/PI3K/AKT/NFkB signaling in different groups. Importantly, the expression levels of TLR4 in both sera and knee joints of AIA-M rats were increased, and subsequently were effectively reversed by the treatment of the two-BAC combination, with similar trends to BHGZD and MTX (all p < 0.05, Figures 4C-E). The protein expression ratios of p-PI3K/ PI3K, p-NFkB-p65/NFkB-p65, and p-AKT1/AKT1 were all significantly increased in AIA-M rats, which were distinctively reduced by the treatment of BHGZD and the two-BAC combination (all p < 0.05, Figures 4C, D), indicating their inhibitory effects on TLR4/PI3K/AKT/NFkB signaling activation. In addition to the increased accumulation of inflammatory cells, the expression levels of inflammatory cytokines in AIA-M rats sera, including TNF-a, IL-6, and IL-12, were abnormally elevated in AIA-M rats, and reduced by the treatment of the two-BAC combination, in accordance with the effects of BHGZD and MTX (all p < 0.05, Figure 4E). Interestingly, we also achieved the same findings in the in vitro experiment validations based on both RAW264.7 and MH7A cells induced by LPS/ATP (all p < 0.05, Figures 5E-G and 6C-E).
These findings demonstrated that both BHGZD and the two-BAC combination of MG and CA may suppress TLR4/PI3K/ AKT/NFkB signaling-related protein activation, and subsequently inhibit NLRP3 inflammasome-induced pyroptosis.

DISCUSSION
Growing clinical evidence shows that various immune cells sustainably influx and migrate into the joints via secreting different types of immunomodulatory cytokines, leading to pyroptosis-induced persistent synovitis and cartilage degradation (44,45). Thus, reversing the imbalance of the "immuneinflammation" system, especially alleviating pyroptosis, may be a promising therapeutic strategy for active RA. The current study performed an integrative research combining UFLC-Q-TOF-MS/ MS, gene expression profiling, network calculation, pharmacokinetic profiling, SPR/MST assay, and pharmacological experiment validations, and identified TLR4/PI3K/AKT/NFkB/ NLRP3 signaling-induced pyroptosis as one of the candidate effective targets of BHGZD for reversing the imbalance network of "immune-inflammation" during active RA progression. In addition, both MG and CA were identified as representative BACs acting on that target, for the strong binding affinities between compounds and target proteins, good pharmacokinetic features, and similar pharmacological effects to BHGZD. Notably, both BHGZD and the two-BAC combination of MG and CA effectively improved disease severity of active RA rats including elevating pain thresholds, relieving joint inflammation and bone erosion via inhibiting TLR4/PI3K/AKT/NFkB signaling to suppress the activation of the NLRP3 inflammasome, leading to the downregulation of downstream caspase-1, the reduced release of IL-1b, and the modulation of GSDMD-mediated pyroptosis. Consistent data were obtained based on the in vitro pyroptosis models of RAW264.7 and MH7A cells induced by LPS/ATP. Thus, RAW264.7 macrophages might be the probable immune cells targeted by both BHGZD and the two-BAC combination of MG and CA.
inflammation" system, including distinct redness and swelling, an increase in arthritis surface temperature, arthritis incidence, the diameter of the limb, arthritis score, pathologic changes of the thymus and spleen, as well as severe inflammatory cell infiltration, cartilage and bone destruction, synovial hyperplasia, and the high levels of inflammatory mediators, such as TLR4, IL-6, IL-12, IL-1b, IL-18, and TNF-a, were observed as the distinctive characteristics and biological basis of AIA-M rats, which may be in line with the clinical manifestations in active RA patients. Following the transcriptomic profiling and biomolecular network analyses, a series of active RA-related genes were also identified, and functionally involved in the regulation of the "immune-inflammation" system accordingly.
Considering that TCM contributed to the multi-target interactions of the complex ingredients of its herbal drugs (46)(47)(48), pharmaceutical development has consistently been an urgent challenge. Therefore, a new method that combines BACs contained in herbal formulas has been strongly indicated for new drug discovery. Herein, the two BACs, namely, MG from Anemarrhenae Rhizoma and CA from Cinnamomi Ramulus, were identified as the representative BACs of BHGZD for the strong binding affinities between compounds and target proteins, good pharmacokinetic properties, the high content and the importance of the formula, and similar pharmacological effects to BHGZD. Interestingly, the treatment of MG or CA alone did not exert prominently therapeutic effects compared to that of the two-BAC combination. Similar to the BHGZD formula, the two-BAC combination treatment of MG and CA may exert satisfying therapeutic efficacy on both in vivo and in vitro experiments via regulating TLR4/PI3K/AKT/NFkB signaling, which plays a vital FIGURE 7 | The schematic diagram of the underlying mechanisms of the two-BAC combination of mangiferin (MG) and (CA) against active RA via reversing the imbalance of the "immune-inflammation" system by inhibiting TLR4/NFkB/NLRP3 activation-induced pyroptosis. "⟂": Inhibition. "!": Activation. role in synovial inflammation, cartilage degradation, and bone erosion by regulating inflammation response, immune disorder cells, and pyroptosis (37,(49)(50)(51)(52). More importantly, the two-BAC combination of MG and CA offers key potential advantages over BHGZD for the following points. Firstly, the two-BAC combination may exert similar pharmacological effects in treating AIA-M rats with a definite material basis. Secondly, the two-BAC combination may be flexible in design, easily synthesized on a large scale, easily absorbed, and relatively stable, not dependent on the cultivations of crude herbs. Thirdly, it is better to understand the underlying molecular mechanisms of TCM-based RA therapeutics.
In conclusion, our data offer an evidence that the MG and CA combination from BHGZD may interact with TLR4/PI3K/AKT/ NFkB signaling to inhibit NLRP3 inflammasome activation and modulate pyroptosis, which provides the novel representative BACs and pharmacological mechanisms of BHGZD against active RA (Figure 7). The findings may shed new light on the mechanisms of the TCM formula, and promote the modernization development of TCM and drug discovery.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The name of the repository and accession number (s) can be found below: NCBI Gene Expression Omnibus; GSE190523, GSE189942.