Firstly, chemical compounds in five herbs constituting WTD were obtained from TCM Database @ Taiwan¹, which is the biggest TCM database around the world so far (Chen, 2011). Human protein targets of the above chemical compounds were retrieved in PubChem². Besides, RA-related human genes were searched in GenBank databases³.

Human protein targets of WTD and RA-related human genes obtained from the first step were uploaded to IPA software. Subsequently, a set of networks were generated by IPA based on different bio-functions, in which nodes with various shapes were represented as molecules with different functions, and biological relationship between two nodes was expressed as an edge (line). In addition, the networks were ranked by scores calculated by IPA, which thereby represented the significance of molecules to networks. Furthermore, canonical pathways were obtained by means of core analysis in IPA. The significance of molecules to canonical pathways was reflected by P-value calculated with Fisher’s exact test. A smaller P-value suggested greater significance. Finally, comparison analysis was accomplished by IPA, so as to further study the action mechanism of WTD in RA.

Wu-Tou decoction is composed of 6 g Aconiti Radix (Aconitum carmichaeli Debx.), 9 g Ephedrae Herba (Ephedra sinica Stapf), 9 g Astragali Radix (Astragalus membranaceus Bunge var. mongholicus (Bunge) P.K. Hsiao), 9 g Paeoniae Radix Alba (Paeonia lactiflora Pall.) and 9 g Glycyrrhizae Radix EtRhizoma (Glycyrrhiza uralensis Fisch.). All crude drugs were purchased from Beijing Tongrentang. WTD was prepared as previously described (Zhang et al., 2015). Firstly, all crude drugs were soaked in 2 L water for 2 h, then they were decocted to boiling for 1 h. The decoction was filtered through a four-layer gauze. Later, the drugs were boiled once again for 0.5 h with 2 L water and the decoction was filtered in accordance with the above method. The filtrates were merged and concentrated to a concentration of 0.75 g crude drug/mL. Subsequently, the decoction was further filtered with a 0.22 μm pore-size filter (Millipore, Billerica, MA, USA) for tests in vitro, and stored at 4°C for no more than 1 week. Finally, the decoction was diluted with saline for tests in vivo and PBS for tests in vitro to certain concentrations before use.

Human leukemic U937 cells were purchased from American Type Culture Collection (Manassa, VA, USA), and cultured in RPMI 1640 medium (GIBCO, Gaithersburg, MD, USA) supplemented with 10% fetal bovine serum (GIBCO, Gaithersburg, MD, USA) and 1% penicillin-streptomycin (PS) at 37°C in a humidified atmosphere with 5% CO₂.

U937 cells were planted into a 96-well cell culture plate at a density of 1.0 × 10⁴ cells/well and incubated with 10 ng/mL phorbol 12-myristate 13-acetate (PMA) (Sigma, St. Louis, MO, USA) at 37°C for 48 h in a 5% CO₂ incubator. Subsequently, the cells were washed with PBS twice, followed by incubation with WTD (2.5–160 μg/mL) for up to 45 h at 37°C in a 5% CO₂ incubator. 10 μL CCK-8 reagent (Dojindo, Tokyo, Japan) was then added into each well and the cells were incubated for another 3 h. Eventually, the absorbance of each well was measured at 450 nm (test wavelength) with an enzyme-linked immunosorbent assay (ELISA) reader (Bio-Tek Instruments, Winooski, VT, USA).

U937 cells were seeded in a 6-well cell culture plate at a density of 1 × 10⁶ cells/well and incubated with PMA for 48 h at 37°C. Then cells were washed for three times and treated with or without WTD at 20 μg/mL or 40 μg/mL for 2 h, followed by stimulation with 25 nM MIP-1β (PeproTech, Rocky Hill, NJ, USA) for the duration suggested in previous studies (Tomkowicz et al., 2006; Fantuzzi et al., 2008; Cheung et al., 2009). Maraviroc (MVC) (MCE, Monmouth Junction, NJ, USA), a CCR5 inhibitor, was used as the positive control at a concentration of 1 μM (Rossi et al., 2011). Cell supernatants were collected to detect levels of TNF-α, MIP-1α, and RANTES by ELISA, and the cells were harvested to detect levels of total and phosphorylated CCR5, PKC δ and p38 with western blotting.

CCR5 siRNA was obtained from Dharmacon (Lafayette, CO, USA), and control siRNA was purchased from Santa Cruz Biotechnology Inc. (Delaware, CA, USA). Transfection was performed using Lipofectamine 2000 (Invitrogen Life Technologies, Carlsbad, CA, USA) in strict accordance with instructions from manufacturers. Briefly, PMA-induced U937 cells were grown to 30–50% confluence in a 6-well plate before they were transfected with Lipofectamine 2000. SiRNA and Lipofectamine 2000 were premixed for 20 min before they were added into the cells. After 24 h of transfection, the cells were incubated with or without WTD for 2 h, followed by exposure to MIP-1β for another 24 h. After that, cell supernatants were collected to detect levels of TNF-α and RANTES by ELISA.

Male Sprague Dawley (SD) rats aged 8–10 weeks with a mean weight of 180–200 g were purchased from Research Institute of Experimental Animals, Chinese Academy of Medical Science. The rats were fed with food and water ad libitum, and then were allowed to acclimatize themselves for 1 week before the initiation of experiment. The rats were housed in a temperature-, humidity-, and light-controlled environment. The light-dark cycle was 12 h:12 h with the light phase being set from 06:00 to 18:00. The rodent license of the laboratory (NO.SYXK 11-00-0039) was issued by National Science and Technology Ministry of China. This study was approved by the Research Ethics Committee of Institute of Basic Theory of Chinese Medicine, China Academy of Chinese Medical Sciences.

Collagen induced arthritis (CIA) animal model was constructed as previously described (Hsu et al., 2012). Male SD rats were immunized using an emulsion consisted of equivalent parts of incomplete Freund’s adjuvant (Chondrex, Redmond, WA, USA) and bovine type II collagen (C II) (Chondrex, Redmond, WA, USA). The rats received an intradermal injection of 200 μL emulsion (200 μg of bovine CII) into the base of tail on day 0. Booster doses were administered on day 7 with 100 μL injection of same emulsion being given into the base of tail. The onset of CIA could be observed between days 11 and 13 after the first immunization. The following clinical parameters were measured, including body weight and arthritis index (AI) score. AI score for each hind ankle joint was recorded by the same observer, who was blind to the treatment received by animals. Scoring was performed with a 0–4 scale (Shahrara et al., 2003), where 0 = no swelling or erythema; 1 = slight swelling and/or erythema; 2 = low-to-moderate edema; 3 = pronounced edema with limited joint usage; and 4 = excess edema with joint rigidity. The maximum score of each rat was eight.

Methotrexate with a purity of over 99.0% was purchased from Sigma (St. Louis, MO, USA). The rats were randomly divided into five groups after the successful induction of CIA model with eight rats in each group, which were normal group (N), model group (M), WTD high dose group (WH, 7.5 g/kg/d), WTD low dose group (WL, 3.75 g/kg/d) and methotrexate group (MTX, 1 mg/kg/w). All agents were intragastrically administered at a volume of 1 mL/100 g for 4 weeks. The dosage of WTD in WL group (3.75 g crude drug/kg/d) was nearly equal to that for RA patients (42 g crude drug/person/day). The dosage of MTX was set as previously described (Le Goff et al., 2009).

Cell supernatants were collected after 24 h stimulation with MIP-1β. Levels of TNF-α, MIP-1α, and RANTES in U937 cell supernatants, as well as those of IL-1β, IL-2, IL-6, TNF-α, macrophage inflammatory protein (MIP)-1α, MIP-2, regulated upon activation normal T cell expressed and secreted (RANTES) and interferon-inducible protein (IP10) in rat serum were detected with commercially available enzyme-linked immunosorbent assay (ELISA) kits (eBioscience, San Diego, CA, USA) according to manufacturer’s instructions.

Western blotting was conducted following previous method (Shahrara et al., 2003). Briefly, U937 cells were harvested after 10 min stimulation with MIP-1β. The cells and ankle joints of rats were lysed in RIPA Lysis Buffer (Beyotime, Shanghai, China) containing PMSF (Amresco, Houston, TX, USA). Ankle joint homogenization was completed on ice with a motorized homogenizer. Homogenates were centrifuged at 13000 rpm for 20 min at 4°C, filtered through a 0.45 μm pore-size Millipore filter, and stored at -80°C until use. Protein concentration in each lysate was determined using a bicinchoninic acid assay (Pierce, Rockford, IL, USA) with bovine serum albumin (Sigma, St. Louis, MO, USA) as the standard.

Equal amount of each sample was subjected to 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to nitrocellulose membranes utilizing a wet transblotting apparatus. Nitrocellulose membranes were blocked with 3% bovine serum albumin in Tris buffered saline-Tween 20 (TBST) buffer (20 mM Tris, 137 mM NaCl, pH 7.6, with 0.1% Tween 20) for 30 min at room temperature. Blots were incubated for 10 min at room temperature followed by overnight at 4°C and 30 min at room temperature on the 2nd day with anti-p-CCR5, anti-p-PKC δ, anti-p-p38 antibody at 1:1000 and anti-CCR5, anti-PKC δ at 1:5000, anti-p38 (Abcam, Cambridge, UK) at 1:4,000 in TBST containing 3% bovine sera albumin. After that, blots were washed five times and then incubated in horseradish peroxidase–conjugated antibody (1:10,000 dilution in TBST containing 5% non-fat milk) for 40 min at room temperature. All blots were developed with enhanced chemiluminescence reagents based on instructions from manufacturers. Blots were scanned and analyzed for measurement of the band intensities with UN-SCAN-IT version 5.1 software. Band intensity was calculated as follows: band intensity = sum of all pixel values in the segment selected - background pixel value in that segment.

Ankle joints were dissected from rats, snap-frozen in liquid nitrogen, grounded into powder, and homogenized under RNase-free conditions. RNA isolation and real-time PCR assay were carried out following the protocol in previous study (Liu et al., 2013). Briefly, total RNA was extracted from the tissue homogenate with TRIzol reagent (Invitrogen, Carlsbad, CA, USA) according to manufacturer’s instructions. The total RNA (1 μg) was reverse transcribed to cDNA using PrimeScript^TM RT reagent Kit with gDNA Eraser (TaKaRa, Kusatsu, Japan) in accordance with instruction manual. The specific transcripts were quantified through the method of quantitative real-time PCR using SYBR^®Premix Ex Taq^TM II (Tli RNaseH Plus) with ROX plus (TaKaRa, Kusatsu, Japan). And they were analyzed by ABI 7500 real-time PCR system (AppliedBiosystems, Foster, CA, USA). Gene-specific primers were synthesized by Invitrogen and the following primer sequences were used: TGCTGCTTCTCCTATGGACG (forward) and GCTCAGTGATGTATTCTTGGACC (reverse) for MIP-1α, CAGAGCTTGAGTGTGACG (forward) and TCGTACCTGATGTGCCTC (reverse) for MIP-2, CCCTGCTGCTTTGCCTACCT (forward) and ACACTTGGCGGTTCCTTCG (reverse) for RANTES, GCACCTGCATCGACTTCCAT (forward) and TTCTTTGGCTCACCGCTTTC (reverse) for IP-10, and TGGAGTCTACTGGCGTCTT (forward) and TGTCATATTTCTCGTGGTTCA (reverse) for GAPDH. The mRNA levels were normalized to GAPDH mRNA level. PCR conditions were shown as below: at 95°C for 30 s, 45 cycles at 95°C for 5 s and at 60°C for 40 s. Relative mRNA expression was calculated by comparative CT method.

Ankle joints of rats were fixed in 4% paraformaldehyde, decalcified in 10% EDTA bone decalcifier, and embedded in paraffin. Four-micrometer sections were prepared and stained with H&E for histological examination. Histopathological characteristics were evaluated blindly as described (Kok et al., 2011), where grade 1 = hyperplasia of the synovial membrane and presence of polymorphonuclear infiltrates; grade 2 = pannus and fibrous tissue formation as well as focal subchondral bone erosion; grade 3 = articular cartilage destruction and bone erosion; and grade 4 = extensive articular cartilage destruction and bone erosion.

The sections were dewaxed and hydrated using xylene and a graded series of alcohols. Activity of endogenous peroxidase was quenched with 3% H₂O₂. Then the tissues were incubated with anti-p-CCR5, anti-p-PKC δ, and anti-p-p38 (Abcam, Cambridge, UK) overnight at 4°C, followed by incubation with Polink-2 plus^® Polymer HRP Detection System (ZSGB-BIO, Beijing, China) according to instructions from manufacturers. Final color product was developed with DAB Kit (ZSGB-BIO, Beijing, China). Later, sections were counterstained with hematoxylin (Leagene, Beijing, China). Meanwhile, PBS was used for control staining instead of primary antibodies. Images were captured using a LEICA DFC300 FX (Leica microsystems Ltd) attached to LEICA DM6000B at a magnification of 200×.

Immunohistochemical semi-quantitative analysis was performed as previously described (Zhang et al., 2015). Uncounterstained specimens were examined utilizing a Leica image analyzer and analyzed by computer image analysis (Leica QWin V3) in a blind manner. Ten digital images were recorded for each specimen from ankle joints so as to localize and identify areas with positively stained cells, and quantitative analysis was performed according to the color cell separation. The results were expressed as the mean region of interest, which represented the percentage of area covered with positively stained cells per image at a magnification of 200×.

To further detect phosphorylation levels of CCR5, PKC δ, and p38 in macrophages in ankle joints, immunofluorescent staining was performed according to previous study (Ma et al., 2011). Sections were treated with 10% bovine serum albumin for 1 h at 24°C, and then incubated with each primary antibody for 18 h at 4°C. The primary antibodies used were rabbit polyclonal anti-p-CCR5, anti-p-PKCδ and anti-p-p38 mixed with mouse monoclonal anti-CD68. After being washed with PBS, the sections were incubated with a mixture of FITC-conjugated and rhodamine-conjugated secondary antibodies for 1 h at 24°C, followed by incubation with DAPI for 10 min at 24°C. Sections were mounted in the anti-fading medium Vectashield after being washed. Samples were examined using a confocal laser scanning microscope, LSM510MET (Carl Zeiss, Jena, Germany).

Statistical analysis was performed with SPSS 18.0 software. All data were expressed as mean ± SD. Comparisons of numerical data between two groups were calculated by Student t-tests. Differences in mean values of various groups were analyzed by ANOVA. Difference with P-value < 0.05 was considered as statistically significant.

One hundred and seventy-four chemical compounds (Supplementary Table 1) and 186 human target proteins (Supplementary Table 2) of WTD were obtained in this study. Molecular networks constructed by RA-related genes or targets of WTD could be seen in Supplementary Tables 3, 4. The top 3 of them were shown in Supplementary Figures 2, 3, respectively. The top 12 bio-functions regulated by both targets of WTD and RA-related genes were shown in Supplementary Figure 4. The results indicated that the action mechanism of WTD in RA might be related to its regulation effect on inflammatory response, hematological system development and function, cell-to-cell signaling and interaction, and so on. Furthermore, 346 and 394 signaling pathways of WTD and RA were obtained, respectively. Among which, 314 signaling pathways were shared by WTD and RA. The top 10 shared signaling pathways associated with cell immune response and cytokine signaling were shown in Figure 1A. The top 1 shared signaling pathway, CCR5 signaling pathway in macrophages, was focused for further study. It could be seen from Figure 1B that ligands of CCR5, such as HIV1, MIP-1α, MIP-1β, and RANTES, could interact with CCR5 and activate CCR5 signaling pathway. PKC δ and p38 (yellow molecules) in this signaling pathway were shown as targets of WTD in RA. Besides, the signaling pathway could ultimately regulate the expression of chemoattractants and proinflammatory mediators.

Firstly, the effect of WTD on the viability of U937 cells was evaluated through CCK8 assay, which demonstrated that WTD (2.5–40 μg/mL) had no effect on the viability of U937 cells at 48 h (Figure 2A). Stimulation of U937 with MIP-1β (25 nM) for 24 h resulted in significant increase of TNF-α, MIP-1α, and RANTES, as compared with untreated control. Pretreatment with WTD (20 and 40 μg/mL) led to significant inhibition of MIP-1β-induced TNF-α, MIP-1α and RANTES production (Figures 2B–D).

Then the effects of WTD on the expression of CCR5, PKC δ and p38 as well as their phosphorylation levels in MIP-1β-induced U937 cells were investigated. As could be seen from Figures 3A–C, MIP-1β stimulation significantly increased the phosphorylation levels of CCR5, PKC δ, and p38 (P < 0.05 or P < 0.01), with no effect on their total level. Both WTD and MVC could significantly reduce the phosphorylation levels of CCR5, PKC δ, and p38 in MIP-1β-induced U937 cells (P < 0.05 or P < 0.01). These findings indicated that WTD could modulate CCR5 signaling pathway in MIP-1β-induced U937 cells.

Furthermore, the role of CCR5 signaling pathway in the modulation effects of WTD on MIP-1β-induced production of cytokines and chemokines was further studied. As was shown in Figure 4A, the expression of total and phosphorylated CCR5 was significantly lowered in MIP-1β-induced U937 cells transfected with CCR5 siRNA compared to untransfected cells. CCR5 knockdown inhibited MIP-1β-induced production of TNF-α and RANTES. Much stronger inhibitory effects on the production of TNF-α and RANTES could be seen in MIP-1β-induced U937 cells after combined treatment with CCR5 siRNA and WTD in comparison with treatment with CCR5 siRNA alone (P < 0.05) (Figures 4B,C).

Collagen induced arthritis rats were used in the study to verify the therapeutic effect of WTD on RA. Morphological features of arthritis, including erythema and swelling, could be markedly observed in model group, while WTD treatment relieved arthritis severity in CIA rats (Figure 5A). Consistently, AI scores in WTD-treated CIA rats were significant lower than those in vehicle-treated CIA rats (Figure 5B). Histopathological results (Figure 5C) indicated the presence of massive influx of inflammatory cells, synovial hyperplasia, and severe erosion of cartilage and bone in ankle joints of CIA rats. Whereas, symptoms of inflammation, synovial hyperplasia and joint destruction in rats treated with WTD were remarkably alleviated. Histological score analysis also manifested that WTD could prominently inhibit the histopathological changes in ankle joints of CIA rats (Figure 5D). All these results proved the therapeutic effect of WTD on RA.

Several cytokines and chemokines were detected by ELISA and Real-time PCR, so as to observe the effects of WTD on production of proinflammatory cytokines and chemokines in CIA rats. The results showed that WTD treatment dramatically decreased the serum levels of IL-1β, IL-2, IL-6, TNF-α, MIP-1α, MIP-2, RANTES, and IP-10 in CIA rats (Figures 6A,B). Meanwhile, the mRNA levels of MIP-1α, MIP-2, RANTES, and IP-10 in ankle joints of WTD-treated CIA rats also decreased significantly in comparison with those in vehicle-treated CIA rats (Figure 6C).

The expression of CCR5, PKC δ, and p38, as well as their phosphorylation levels in ankle joints was detected by immunohistochemistry and western blotting, so that the action mechanism of WTD in CIA rats could be investigated. The results could be seen in Figures 7A–D, which showed that WTD significantly decreased the phosphorylation levels of CCR5, PKC δ and p38 in ankle joints of CIA rats (P < 0.05 or P < 0.01). In addition, WTD also decreased total level of CCR5 in CIA rat’s ankle joint, but had no effect on total PKC δ and p38.

Phosphorylation levels of CCR5, PKC δ, and p38 in macrophages in ankle joints were detected by immunofluorescence, so as to further study the effect of WTD on CCR5 signaling pathway in macrophages in CIA rats. It could be observed in Figures 8A–C that the levels of p-CCR5, p-PKC δ, and p-p38 in CD68⁺ cells infiltrated in the ankle joints of vehicle-treated CIA rats were significantly higher than those in normal rats. And higher levels of p-CCR5, p-PKC δ, and p-p38 were inhibited significantly by WTD.