Inference of RA-related pathological processes affected by GSZD
Because drug indications are often determined by the functions of their corresponding targets and drugs with similar chemical structures generally exert similar therapeutic effects, we predicted the putative targets of nine herbs contained in GSZD based on the similarities in drug structure and function as described by our previous studies [34, 35]. A total of 77 putative targets were identified out of 165 chemical components containing in GSZD (Additional file 1: Table S7).
Then, we compared known drugs with similar structures to the chemical components in the nine GSZD herbs. As a result, the nine GSZD herbs shared 15 putative targets with known drugs for the treatment of RA, autoimmune diseases, inflammatory diseases, and pain, as well as for the provision of anesthesia (Table 1).
The putative targets of Herba Ephedrae, Radix Aconiti Lateralis Preparata, Radix Glycytthizae, Raidix Saposhnikoviae, Ramulus Cinnamomi, Rhizoma Atractylodis Macrocephalae and Rhizoma Zingiberis Recens included nitric oxide synthase 2 (NOS2); nuclear receptor subfamily 3, group C, member 1 (NR3C1); phospholipase A2, group IB (PLA2G1B); and serpin peptidase inhibitor, clade A, member 6 (SERPINA6), which are the targets of several glucocorticoids and FDA-approved anti-inflammatory and immunosuppressive agents for the treatment of RA, such as Alclometasone, Amcinonide, Betamethasone, etc. based on DrugBank (Version 4.3, http://www.drugbank.ca/), suggesting that these putative targets might be involved in the anti-inflammatory and immunosuppressive effects exerted by GSZD on RA.
The management of pain is an important component of RA patient care, and cholinergic receptor, nicotinic, alpha 2 (CHRNA2), gamma-aminobutyric acid A receptor (GABRA1), 5-hydroxytryptamine receptor 1B (HTR1B), 5-hydroxytryptamine receptor 1D (HTR1D), NOS2; opioid receptor, delta 1 (OPRD1) and prostaglandin-endoperoxide synthase 2 (PTGS2) have been identified as therapeutic targets for severe pathologic pain. The current study predicted that Herba Ephedrae, Radix Aconiti Lateralis Preparata, Radix Glycytthizae, Raidix Saposhnikoviae, Ramulus Cinnamomi, Rhizoma Atractylodis Macrocephalae and Rhizoma Zingiberis Recens might target these molecules.
Providing anesthesia to patients with osteoarticular disorders during RA progression involves a number of risks not only due to the mechanical deformations caused by the disease but also in relation to the cardiovascular, respiratory, renal, and digestive systems . Thus, to benefit RA patients, it is of great clinical significance to control anesthesia effectively. Here, seven GSZD herbs, including Herba Ephedrae, Radix Aconiti Lateralis Preparata, Radix Glycytthizae, Raidix Saposhnikoviae, Paeonia lactiflora, Ramulus Cinnamomi and Rhizoma Zingiberis Recens, shared targets [cholinergic receptor muscarinic 1 (CHRM1), CHRNA2, fatty acid amide hydrolase (FAAH), farnesyltransferase CAAX box alpha (FNTA), GABRA1 and OPRM1] with known anesthetic drugs, including glycopyrrolate, methohexital, metocurine, mivacurium, naloxone, propofol and thiamylal.
Collectively, the putative targets of GSZD mainly have roles in the progression of inflammation, joint destruction and pathological pain. As such, GSZD’s therapeutic efficacy in the treatment of RA may arise from its regulation of the expression or activities of these targets.
Combinatorial effects of herbs contained in GSZD acting on RA
The compatibility of a TCM herbal formula emphasizes the “Jun (emperor)–Chen (minister)–Zuo (adjuvant)–Shi (messenger)” rule with proper herbs to synergize the therapeutic efficacies and minimize adverse effects integrally [38, 39]. According to the co-module analysis , the herb-putative target network was divided into three modules, which were respectively centered on Ramulus Cinnamomi, Paeonia lactiflora and Rhizoma Anemarrhenae (Fig. 2). In TCM theory, Ramulus Cinnamomi, Rhizoma Atractylodis Macrocephalae and Herba Ephedrae are considered as the “Jun” herbs and play the leading roles in GSZD . These herbs were linked to anti-inflammatory and anti-allergy activities in previous study  and were computationally confirmed here. Their putative targets were significantly associated with the regulation of inflammatory process, cytokine stimulus response and cytokine production, and complement and coagulation cascades, which are all involved into the main pathological changes during RA progression, such as inflammation, synovial pnnus formation and angiogenesis. Radix Aconiti Lateralis Preparata and Raidix Saposhnikoviae function as “Chen” herbs which enhance the pharmacological actions of the “Jun” herbs. Radix Glycytthizae and Rhizoma Zingiberis Recens are considered as “Shi” herbs and harmonize the actions of other herbs in GSZD . Moreover, Paeonia lactiflora and Rhizoma Anemarrhenae serve as “Zuo” herbs which dispel toxins and guide other drugs to their corresponding meridian channels . Similarly, we found that the biological functions and pathways of the Paeonia lactiflora and Rhizoma Anemarrhenae-centered modules respectively focused on the regulation of drug metabolism and cell surface receptor-mediated signal transduction, and G-protein-coupled receptor protein signaling.
GSZD has a reverse effect on inflammation-immune regulatory network imbalance during RA progression
To elucidate the function of herb putative targets of GSZD, pathway enrichment analysis were performed and found that the top 6 pathways that the GSZD putative targets were significantly associated with neuroactive ligand-receptor interaction, toll-like receptor signaling, osteoclast differentiation, calcium signaling pathway, complement and coagulation cascades and VEGF signaling (all P < 0.001, Additional file 1: Table S8).
Performing molecular network-based analysis by mapping disease-related genes and drug target genes into an interaction network can efficiently illustrate underlying links between drugs and disease. Thus, we constructed a network based on interactions between putative GSZD targets, known RA-related targets and other human proteins. A node may function as a hub if its degree is more than two-fold of the median degree of all nodes in a network . As a result, 135 hubs were identified, and our pathway enrichment analysis showed that these hubs were frequently implicated in T and B cell receptor signaling, Toll-like receptor signaling, osteoclast differentiation, NF-kappa B signaling, TNF signaling, chemokine signaling, VEGF signaling, and neuroactive ligand-receptor interactions. All of these actions play crucial roles in the main pathological events that comprise RA progression, such as inflammation, synovial pannus formation, inflammatory cell infiltration, angiogenesis, joint destruction and pain  (Fig. 3).
Subsequently, a network of hubs based on the direct interactions of the above was constructed (please see the interaction network data in Additional file 1: Table S9). The major hubs of this network were identified by calculating four topological features of each hub in the network: ‘Degree,’ ‘Node betweenness’, ‘Closeness’ and ‘K value’. The median values of ‘Degree’, ‘Node betweenness’, ‘Closeness’ and ‘K value’ were 8.00, 0.45, 41.31 and 6.00, respectively. Therefore, we determined that hubs with ‘Degree’ >8.00, ‘Node betweenness’ >0.45, ‘Closeness’ >41.31, and ‘K value’ >6.00 were major hubs. As a result, 40 major hubs were identified (the detailed information on the topological features of the 40 major hubs is provided in Additional file 1: Table S10). After assessing the intersection of the above with the putative GSZD targets (Additional file 1: Table S7), 10 major hubs were identified as candidate targets for this formula, including albumin (ALB); androgen receptor (AR); cyclin-dependent kinase 1 (CDK1); estrogen receptor 1 (ESR1); histone deacetylase 1 (HDAC1); heat shock protein 90 kDa alpha, class A member 1 (HSP90AA1); NR3C1; retinoic acid receptor alpha (RARA); signal transducer and activator of transcription 3 (STAT3); and vitamin D receptor (VDR).
Growing evidence has shown that an interaction with a high ‘edge-betweenness’ may function as a bottleneck with many ‘shortest paths’ going through it and may thus control the rate of information flow . Here, we further calculated the ‘edge-betweenness’ of each interaction in the network of direct interactions among hubs to select important interactions. Among the candidate GSZD targets, the HDAC1–HSP90AA1 interaction had the highest edge-betweenness value (128.25, Additional file 1: Table S11), suggesting that it functions as a bottleneck in the network. As shown in the interaction network of GSZD herbs and hubs (Fig. 4), a signal axis containing interactions between HDAC1, HSP90AA1 and three known RA-related targets, including nuclear factor of kappa light polypeptide gene enhancer in B-cells 2 (NFKB2), inhibitor of kappa light polypeptide gene enhancer in B-cells, kinase beta (IKBKB) and tumor necrosis factor-alpha (TNF-α), was found to play a crucial role in connecting different modules. These modules were significantly associated with antigen processing and presentation, T and B cell receptor signaling, Toll-like receptor signaling, natural killer cell-mediated cytotoxicity, osteoclast differentiation, NF-kappa B signaling and TNF signaling, implying that GSZD might reverse the inflammation-immune regulatory network imbalance that occurs during RA progression partially by regulating the HDAC1–HSP90AA1–NFKB2–IKBKB–TNF-α axis. To validate this hypothesis, an adjuvant-induced arthritis (AIA) rat model was constructed and used to demonstrate the preventive effects of GSZD on inflammation and joint destruction. Following this, its regulatory effects on the HDAC1–HSP90AA1–NFKB2–IKBKB–TNF-α axis were also assessed both in vitro and in vivo.
GSZD treatment ameliorates the development and severity of arthritis in AIA rats
Through HPLC–DAD, Ephedrine from Herba Ephedrae, mangiferin from Rhizoma Anemarrhenae, paeoniflorin from Paeonia lactiflora, liquiritin from Radix Glycytthizae, 4′-O-beta-Glucopyranosyl-5-O-Methylvisamminol from Raidix Saposhnikoviae, aconitine from Radix Aconiti Lateralis Preparata, and cinnamic aldehyde from Ramulus Cinnamomi were identified in the water extract of GSZD (Fig. 5).
Macroscopic changes of arthritis, such as redness and swelling, were clearly observed in the AIA rats (Fig. 6a), but were attenuated by the treatment of GSZD [18.6 g/(kg day)] and MTX [0.2 mg/(kg day)]. Statistically, the mean arthritis score (all P < 0.05, Fig. 6b), arthritis incidence (all P < 0.05, Fig. 6c), percentage of arthritic limbs (all P < 0.05, Fig. 6d) and time of first appearance of arthritis (for doses of 9.3 and 18.6 g/(kg day), P < 0.05, Fig. 6e) were markedly lower in the GSZD-treated rats, especially in the middle- and high-dosage groups, and in the MTX-treated rats compared to the untreated AIA rats.
GSZD treatment protects against synovitis and joint destruction in AIA rats
Histopathological evaluation of ankle joint sections from the AIA rats showed inflammatory cell infiltration, synovial hyperplasia and articular tissue destruction, which all could be attenuated by the oral administration of GSZD (Fig. 7). In brief, synovial edema and extensive infiltration of inflammatory cells occurred in the AIA rats, but were repaired by the treatment of GSZD, which promoted the proliferation and maturation of fibrovascular granulation tissues and reduced the number of inflammatory cells (Fig. 7a). Cartilage tissue thinning, dissolution and disappearance, as well as extensive inflammatory cell infiltration with plasma cells and lymphocytes, was observed in the articular cartilage of the ankles of the untreated AIA rats. In contrast, GSZD treatment prevented cartilage degeneration and markedly reduced inflammation by promoting cartilage cell proliferation and calcification and reducing inflammatory cell infiltration (Fig. 7b). Similarly, GSZD treatment typically preserved articular cartilage matrix integrity in markedly inflamed joints, as indicated by the retention of toluidine blue staining in the matrix (Fig. 7c). Moreover, the AIA rats showed severe bone destruction with inflammatory cell infiltration and phagocytosis of osteoclasts, which were reversed by the oral administration of GSZD mainly via the promotion of osteoblast proliferation and the acceleration of the calcification and ossification of regenerated cartilage tissues (Fig. 7d).
We statistically evaluated the anti-inflammatory and bone protective effects of GSZD with semi-quantitative grading scales (on a scale of 0–3)  and assessed articular cartilage matrix integrity in different groups based on the loss of toluidine blue staining . As shown in Fig. 7e and f, the inflammation score and degree of cartilage damage in the GSZD-treated AIA rats were significantly decreased in a dose-dependent manner compared to the untreated AIA rats (all P < 0.05). Treatment with GSZD also significantly and dose-dependently reduced bone destruction in inflamed joints (all P < 0.05, Fig. 7g). More interestingly, the therapeutic effects produced by high-dosage GSZD treatment on inflammation score, degree of cartilage damage and bone destruction score in inflamed joints in AIA rats did not significantly differ from those produced by MTX treatment (Fig. 7).
GSZD treatment partially reverses RA progression by targeting the HDAC1–HSP90AA1–NFKB2–IKBKB–TNF-α axis in vitro and in vivo
To reveal the pharmacological mechanisms of GSZD’s action on AIA, the expression levels of HDAC1, HSP90AA1, NFKB2, IKBKB and TNF-α proteins in the inflamed joints of AIA rats and in the human fibroblast-like synoviocytes-rheumatoid arthritis (HFLS-RA) cell line were detected by western blot analysis following different treatment protocols. Compared to normal controls, HDAC1, HSP90AA1, NFKB2, IKBKB and TNF-α protein expression were markedly increased in the inflamed joints of AIA rats (all P < 0.05, Fig. 8) but were efficiently reduced by GSZD treatment. Compared with untreated AIA rats, GSZD treatment at doses of 9.3 and 18.6 g/(kg day) significantly reduced the expression of HDAC1 (all P < 0.05, Fig. 8a) and HSP90AA1 (all P < 0.05, Fig. 8b). Notably, the administration of GSZD markedly and dose-dependently decreased the expression levels of NFKB2, IKBKB and TNF-α proteins (all P < 0.05, Fig. 8c–e). More importantly, these findings were consistent with the results from in vitro experiments performed on cultured HFLS-RA, as shown in Fig. 9.
Innate immune responses in the rheumatoid synovium contribute to inflammation and joint destruction in RA . NFKB2, IKBKB and TNF-α have recently been identified to play crucial roles in this chronic inflammation of synovial joint linings, which has initiated the development of a series of targeted and highly effective therapeutics for RA. Mammalian HDACs can be divided into two classes: class I HDACs (HDACs 1, 2, 3, 8), which are homologues of yeast PRD3, and class II HDACs (HDACs 4–7 and 9), which are homologues of yeast Hda1 . It has been reported that HDAC1 activity and expression are dramatically increased in RA synovial tissues compared to normal tissues and are upregulated by TNF-α stimulation in RASFs, suggesting the need to develop HDAC1 inhibitors for the treatment of RA . HSP90AA1, a chaperone family member, functions to guide the late-stage tertiary folding of numerous proteins . HSP90AA1 guides the folding of NF-kappa B signaling pathway members, such as receptor-interacting protein and IKK, which can be degraded following HSP90AA1 inhibition, blocking NF-kappa B signaling pathway activation and causing a subsequent loss of cytokine production in macrophages and other cell types . Thus, accumulating evidence suggests that an HSP90AA1-targeted agent would be useful in the treatment of inflammatory diseases, including RA. Here, we employed in vivo and in vitro experimental validation to demonstrate that GSZD ameliorates the upregulation of HDAC1, HSP90AA1, NFKB2, IKBKB, and TNF-α, in line with its role in reducing synovial inflammation and preventing cartilage destruction during RA progression.