Inhibitory Effect of Methotrexate on Rheumatoid Arthritis Inflammation and Comprehensive Metabolomics Analysis Using Ultra-Performance Liquid Chromatography-Quadrupole Time of Flight-Mass Spectrometry (UPLC-Q/TOF-MS)

Rheumatoid arthritis (RA) is a common autoimmune disease. The inflammation in joint tissue and system endanger the human health seriously. Methotrexate have exhibited a satisfactory therapeutic effect in clinical practice. The aim of this research was to establish the pharmacological mechanism of methotrexate on RA therapy. Collagen induced arthritic rats were used to identify how methotrexate alleviates inflammation in vivo. Lipopolysaccharide-induced inflammatory proliferation in macrophages was also be detected in vitro. The activation level of Nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) and Nucleotide binding domain and leucine-rich repeat pyrin 3 domain (NLRP3)/Caspase-1 and related cytokines were examined by real-time PCR and western blotting or quantified with the enzyme-linked immunosorbent assay. Comprehensive metabolomics analysis was performed to identify the alteration of metabolites. Results showed that treating with methotrexate could alleviate the inflammatory condition, downregulate the activation of NF-κB and NLRP3/Caspase-1 inflammatory pathways and reduce the level of related cytokines. Docking interaction between methotrexate and caspase-1 was visualized as six H-bonds indicating a potential inhibitory effect. Metabolomics analysis reported three perturbed metabolic inflammation related pathways including arachidonic acid, linoleic acid and sphingolipid metabolism. These findings indicated that methotrexate could inhibit the onset of inflammation in joint tissue by suppressing the activation of NF-κB and NLRP3/Caspase-1 pathways and regulating the inflammation related metabolic networks.


Introduction
Rheumatoid arthritis (RA) is a chronic inflammatory joint disease that is characterized by synovial proliferation, joint destruction, and systemic inflammation [1,2]. It affects about 1-2% adult population globally [3]. RA begins with an autoimmune phase and is followed by an inflammation, which is

Results
How methotrexate alleviate inflammation was studied in collagen induced arthritic (CIA) murine model and Lipopolysaccharide (LPS)-induced macrophages in present study.

Inhibitory Effect of Methotrexate on Murine CIA Paw Swelling
Swelling of CIA paws showed a statistical significance at 11th day according to the visual scores after the booster injection. Paw swelling volume was examined for all experimental animals then. As shown in Figure 1A, the experimental arthritis groups that were injected with collagen emulsion showed a significant increase in paw swelling volume when compared with the control (Control) group (p < 0.05). Drugs administration began at day 11 and continued until the visual scores of the CIA model (Model) group came to a platform stage. As expected, the endpoint of the administration came at day 25. The low dosage (Low) group showed a trend of inhibitory effect on paw swelling when compared with the model group (p = 0.14), as shown in Figure 1B

Downregulation of Methotrexate on NF-κB/NLRP3 Pathways in Murine Model
NF-κB and NLRP3/Caspase-1 pathways mRNA expression was remarkably higher in CIA ankle joint tissue in Model compared with Control indicating a significant activation ( Figure 2). In detail, NF-κB p65 and TNF-α gene expression in all the drug administrated groups decreased significantly (p < 0.05). Level of NLRP3 gene expression in Middle and High group is different from Model group significantly (p < 0.01). The mRNA level of Caspase-1 in drug-administrated groups showed an obvious decrease compared with the model animals (p < 0.01) in a dose-independent manner. However, IL-1β and IL-18 gene expression only decreased significantly in Middle, High, and Positive groups (p < 0.01). The IL-1β and IL-18 gene expression in the Low group simply showed a trend of decrease.
NLRP3 protein in Model rats increased significantly (p < 0.01). So, as the protein expression level of caspase-1 and NF-κB p65 in Model (p < 0.05), as displayed in Figure 3. All of the drug administrated groups showed a downregulated expression of NLRP3, caspase-1, and NF-κB p65 protein when compared with Model. Especially, they were downregulated remarkably in Middle, High and Positive group compared with Model (p < 0.05 or p < 0.01). Moreover, NF-κB p65 protein in Low group also showed a statistical significance (p < 0.05).
Besides, the quantification assay results of IL-1β, IL-18, and TNF-α in plasma were represented in Figure 4. The NF-κB and NLRP3/Caspase-1 pathways related cytokines in plasma of Model increased significantly (p < 0.01 vs. Control) and decreased in all the drug-administrated groups (p < 0.05 for Low vs. Model and p < 0.01 for Middle, High, and Positive groups vs. Model, respectively).

Downregulation of Methotrexate on NF-κB/NLRP3 Pathways in Murine Model
NF-κB and NLRP3/Caspase-1 pathways mRNA expression was remarkably higher in CIA ankle joint tissue in Model compared with Control indicating a significant activation ( Figure 2). In detail, NF-κB p65 and TNF-α gene expression in all the drug administrated groups decreased significantly (p < 0.05). Level of NLRP3 gene expression in Middle and High group is different from Model group significantly (p < 0.01). The mRNA level of Caspase-1 in drug-administrated groups showed an obvious decrease compared with the model animals (p < 0.01) in a dose-independent manner. However, IL-1β and IL-18 gene expression only decreased significantly in Middle, High, and Positive groups (p < 0.01). The IL-1β and IL-18 gene expression in the Low group simply showed a trend of decrease.
NLRP3 protein in Model rats increased significantly (p < 0.01). So, as the protein expression level of caspase-1 and NF-κB p65 in Model (p < 0.05), as displayed in Figure 3. All of the drug administrated groups showed a downregulated expression of NLRP3, caspase-1, and NF-κB p65 protein when compared with Model. Especially, they were downregulated remarkably in Middle, High and Positive group compared with Model (p < 0.05 or p < 0.01). Moreover, NF-κB p65 protein in Low group also showed a statistical significance (p < 0.05).
Besides, the quantification assay results of IL-1β, IL-18, and TNF-α in plasma were represented in Figure 4. The NF-κB and NLRP3/Caspase-1 pathways related cytokines in plasma of Model increased significantly (p < 0.01 vs. Control) and decreased in all the drug-administrated groups (p < 0.05 for Low vs. Model and p < 0.01 for Middle, High, and Positive groups vs. Model, respectively).     panels of all protein membranes. DCS, Diclofenac sodium. ## p < 0.01 vs. Control. * p < 0.05 vs. Model. ** p < 0.01 vs. Model.

Methotrexate Suppressed the Inflammatory Proliferation of Macrophages
The cytotoxicity of both methotrexate and DCS on RAW 264.7 was monitored with Real Time Cellular Analysis (RTCA) system and displayed as Figure 5. After a continuous incubation with different drug concentration, the cytotoxicity was evaluated at 72 h for methotrexate and at 96 h for DCS in case of proliferation recession. As shown in Figure 5A1,A2, methotrexate significantly inhibited the cell proliferation over the range of 1 μM to 1 mM indicating a cytotoxic effect (p < 0.01). DCS did not exhibit any suppression on the proliferation with all drug concentrations ( Figure 5B, p > 0.05). The non-cytotoxic concentrations were set as 50, 20, 10, 5, 2 nM for methotrexate and 1 mM for DCS to test their anti-inflammatory effects. As a result, compared to the Model, the Control cells exhibited a similar inflammatory proliferative status (p > 0.05). DCS and methotrexate in 50, 20 and 10 nM significantly reduced the inflammatory proliferation (p < 0.01). However, there did not exist any difference on proliferative characteristics compared to the Model after treating with methotrexate of 5 and 2 nM (p > 0.05).

Methotrexate Suppressed the Inflammatory Proliferation of Macrophages
The cytotoxicity of both methotrexate and DCS on RAW 264.7 was monitored with Real Time Cellular Analysis (RTCA) system and displayed as Figure 5. After a continuous incubation with different drug concentration, the cytotoxicity was evaluated at 72 h for methotrexate and at 96 h for DCS in case of proliferation recession. As shown in Figure 5A1,A2, methotrexate significantly inhibited the cell proliferation over the range of 1 µM to 1 mM indicating a cytotoxic effect (p < 0.01). DCS did not exhibit any suppression on the proliferation with all drug concentrations ( Figure 5B, p > 0.05). The non-cytotoxic concentrations were set as 50, 20, 10, 5, 2 nM for methotrexate and 1 mM for DCS to test their anti-inflammatory effects. As a result, compared to the Model, the Control cells exhibited a similar inflammatory proliferative status (p > 0.05). DCS and methotrexate in 50, 20 and 10 nM significantly reduced the inflammatory proliferation (p < 0.01). However, there did not exist any difference on proliferative characteristics compared to the Model after treating with methotrexate of 5 and 2 nM (p > 0.05).

Methotrexate Inhibited NF-κB and NLRP3/Caspase-1 Pathways in Macrophages
The mRNA expression level of all related genes in RAW 264.7 inflammatory cellular model after the treatment was shown in Figure 6. All genes including were remarkably upregulated in inflammatory cellular models (p < 0.01). After the treatment with methotrexate, the expression of all investigated genes was significantly inhibited in dose-dependent manner (p < 0.01 vs. Model). However, methotrexate with 2 and 5 nM only displayed a potential to downregulate the expression of NF-κB p65 and IL-18 genes (p > 0.05 vs. Model).

Methotrexate Inhibited NF-κB and NLRP3/Caspase-1 Pathways in Macrophages
The mRNA expression level of all related genes in RAW 264.7 inflammatory cellular model after the treatment was shown in Figure 6. All genes including were remarkably upregulated in inflammatory cellular models (p < 0.01). After the treatment with methotrexate, the expression of all investigated genes was significantly inhibited in dose-dependent manner (p < 0.01 vs. Model). However, methotrexate with 2 and 5 nM only displayed a potential to downregulate the expression of NF-κB p65 and IL-18 genes (p > 0.05 vs. Model). Consistent with the results in CIA murine model, proteins including NLRP3, Caspase-1 and NF-κB p65 in inflammation cellular Model rose sharply compared with Control (p < 0.01). Their levels decreased significantly after the therapy with methotrexate and DCS in a dose-independent manner for NLRP3 and Caspase-1 or in a dose-dependent manner for NF-κB p65 (p < 0.01 vs. Model) as shown in Figure 7. However, the pathways related cytokines including IL-1β, IL-18 and TNF-α increased significantly in LPS induced cellular inflammation Model (p < 0.01 vs. Control) and decreased in a dose-dependent manner (p < 0.01 vs. Model, Figure 8). Consistent with the results in CIA murine model, proteins including NLRP3, Caspase-1 and NF-κB p65 in inflammation cellular Model rose sharply compared with Control (p < 0.01). Their levels decreased significantly after the therapy with methotrexate and DCS in a dose-independent manner for NLRP3 and Caspase-1 or in a dose-dependent manner for NF-κB p65 (p < 0.01 vs. Model) as shown in Figure 7. However, the pathways related cytokines including IL-1β, IL-18 and TNF-α increased significantly in LPS induced cellular inflammation Model (p < 0.01 vs. Control) and decreased in a dose-dependent manner (p < 0.01 vs. Model, Figure 8).

Multivariate Analysis and Differentiated Metabolic Patterns
Metabolomics analysis on the metabolic perturbation after the treatment was performed with Ultra-Performance Liquid Chromatography-Quadrupole Time of Flight-Mass Spectrometry (UPLC-Q/TOF-MS) systems. The stability and suitability of spectrum and chromatogram was evaluated based on the 10 ions in both ESI + and ESI − modes from six quality control (QC) sample injections respectively. All relative standard deviation (RSD) values including repeatability and intermediate precision were calculated as less than 6.0% (Table S1)

Multivariate Analysis and Differentiated Metabolic Patterns
Metabolomics analysis on the metabolic perturbation after the treatment was performed with Ultra-Performance Liquid Chromatography-Quadrupole Time of Flight-Mass Spectrometry (UPLC-Q/TOF-MS) systems. The stability and suitability of spectrum and chromatogram was evaluated based on the 10 ions in both ESI + and ESI − modes from six quality control (QC) sample injections respectively. All relative standard deviation (RSD) values including repeatability and intermediate precision were calculated as less than 6.0% (Table S1)

Multivariate Analysis and Differentiated Metabolic Patterns
Metabolomics analysis on the metabolic perturbation after the treatment was performed with Ultra-Performance Liquid Chromatography-Quadrupole Time of Flight-Mass Spectrometry (UPLC-Q/TOF-MS) systems. The stability and suitability of spectrum and chromatogram was evaluated based on the 10 ions in both ESI + and ESI − modes from six quality control (QC) sample injections respectively. All relative standard deviation (RSD) values including repeatability and intermediate precision were calculated as less than 6.0% (Table S1) 4 orthogonal projections to latent structures discriminant analysis (OPLS-DA) models were also established with a supervise based on the experimental treatment groups. As shown in OPLS-DA plots of Figure 9, all samples in different models were appreciably separated from each other indicating there exists no extremely abnormal sample. As a result, OPLS-DA models established can be used to identify the differentiated metabolites and potential biomarkers (R2 = 99.99%, Q2 = 78.32% for C vs. M in ESI+, R2 = 99.32%, Q2 = 82.46% for M vs. T in ESI+; R2 = 99.99%, Q2 = 45.02% for C vs. M in ESI−, R2 = 99.56%, Q2 = 97.10% for M vs. T in ESI−). Besides, the validation of all OPLS-DA models for differentiated metabolites analysis have been confirmed by cross validation ANONA analysis (p < 0.0001). In PCA, t [1] are the first variables summarized from the original data and t [2] are the second ones. In OPLS-DA, t [1] is the predictive X-scores and to [1] is the Orthogonal X-scores. All p values were estimated with CV-ANOVA test. * in A3 means the multiple magnification to the coordinate axes.

Global Profiles of Distinct Metabolites and Perturbed Metabolic Pathways
Distinct metabolites were selected according to the Variable Importance for the Projection (VIP) indexes that were calculated based on the OPLS-DA models above. Metabolites with VIP values more than 1.0 and p values were less than 0.05 were chosen for tandem mass spectrometry (MS/MS) tandem identification and metabolic pathway analysis. A total of 24 remarkably changed metabolites in plasma were determined with the details listed in Table 1. Comparison details were provided in Figures S1-S24.
All of the identified metabolites ions were also displayed in S-plots ( Figure 10), which combined correlation and covariation values of the OPLS-DA models. The edging distribution of all the distinct metabolites in the loading plot revealed enough significance and confidence to illustrate the group difference within the OPLS-DA models. All of the potential biomarkers were evaluated with receiver operating characteristic curve (ROC), as shown in Figure S25. Metabolites with AUC > 0.9 were defined as high diagnostic accuracy. As a result, 19 differentiated metabolites were identified as potential biomarkers with high sensitivity and specialty. is the predictive X-scores and to [1] is the Orthogonal X-scores. All p values were estimated with CV-ANOVA test. * in A3 means the multiple magnification to the coordinate axes.

Global Profiles of Distinct Metabolites and Perturbed Metabolic Pathways
Distinct metabolites were selected according to the Variable Importance for the Projection (VIP) indexes that were calculated based on the OPLS-DA models above. Metabolites with VIP values more than 1.0 and p values were less than 0.05 were chosen for tandem mass spectrometry (MS/MS) tandem identification and metabolic pathway analysis. A total of 24 remarkably changed metabolites in plasma were determined with the details listed in Table 1. Comparison details were provided in Figures S1-S24.
All of the identified metabolites ions were also displayed in S-plots ( Figure 10), which combined correlation and covariation values of the OPLS-DA models. The edging distribution of all the distinct metabolites in the loading plot revealed enough significance and confidence to illustrate the group difference within the OPLS-DA models. All of the potential biomarkers were evaluated with receiver operating characteristic curve (ROC), as shown in Figure S25. Metabolites with AUC > 0.9 were defined as high diagnostic accuracy. As a result, 19 differentiated metabolites were identified as potential biomarkers with high sensitivity and specialty. , include not only C14768, but also C14769, C14771, C04742 and C04805; , include not only C02198, but also C00696. Besides, the relative levels of the distinct metabolites were displayed with a heatmap after the mathematic LOG transformation ( Figure 11A). Samples were naturally clustered into three major branches, which is consistent with the biological groups. It is noted that the drug-administrated samples share a large similarity on their metabolic patterns.
Eight metabolic pathways were identified to be involved (Table S3, Figure 11B) according the results from MetaboAnalyst 4.0 platform. Three of them, including arachidonic acid, linoleic acid, and sphingolipid metabolisms were significantly perturbed (Impacts > 0.1 and −log(P) > 5). The comprehensive interaction between inflammatory pathways and metabolic pathways were summarized and displayed in Figure 10C. Besides, the relative levels of the distinct metabolites were displayed with a heatmap after the mathematic LOG transformation ( Figure 11A). Samples were naturally clustered into three major branches, which is consistent with the biological groups. It is noted that the drug-administrated samples share a large similarity on their metabolic patterns.

Potential Docking Mode of Methotrexate within Caspase-1 Protein
Molecular docking performed to investigate the potential interaction between methotrexate and the critical inflammatory cytokines catalyzation related enzyme, caspase-1 revealed an accurate docking site and the appropriate conformation. The precise modes of interaction and relative binding energy were measured as Table 2. Moreover, the number of H-bonds and the docking scores was calculated with the Schrödinger demo. As shown in Figure 12A,B, both the molecular conformation of methotrexate and the polar contacts (seven contacts, including six H-bonds and one other polar contacts) were displayed. The residue ARG-341 has the most contacts with the ligand.

Potential Docking Mode of Methotrexate within Caspase-1 Protein
Molecular docking performed to investigate the potential interaction between methotrexate and the critical inflammatory cytokines catalyzation related enzyme, caspase-1 revealed an accurate docking site and the appropriate conformation. The precise modes of interaction and relative binding energy were measured as Table 2. Moreover, the number of H-bonds and the docking scores was calculated with the Schrödinger demo. As shown in Figure 12A,B, both the molecular conformation of methotrexate and the polar contacts (seven contacts, including six H-bonds and one other polar contacts) were displayed. The residue ARG-341 has the most contacts with the ligand.

Discussion
Collagen-induced arthritis (CIA) in rats is the most widely used model for RA because it shares many similar immunological and pathological features with RA in patients [17]. Innate immune components, especially macrophage is a prime source of pro-inflammatory cytokines and critically contribute to RA pathology [3,18]. In present study, we established the CIA murine model and the inflammatory macrophage cellular model for pharmacological research on the anti-rheumatic effect of methotrexate. We discovered that methotrexate could alleviate inflammatory swelling in CIA rats and inhibit the inflammatory proliferation of macrophages by suppressing NF-κB and NLRP3/Caspase-1 pathway and downregulate the level of pathway related cytokines. Besides, methotrexate has the potential to inhibit catalytic activity of caspase-1 directly. Three inflammation related metabolic pathways including arachidonic acid, linoleic acid and sphingolipid metabolism also participated in the pathogenesis of RA and were regulated significantly after the treatment with methotrexate.
NF-κB is responsible for the regulation of inflammation and immune responses and it is involved in the pathogenesis of RA. The damage-associated molecular patterns binding to Toll-like receptors, as the priming signal activate the upregulation of the NF-κB nucleus transfer [3,19]. The translocated free NF-κB then activates its target genes to transcribe various kinds of inflammatory mediators, such as proIL-1β, proIL-18, and NLRP3 [19,20]. NLRP3 protein was constructed as NLRP3 inflammasome assembly with other components including caspase recruitment domain containing protein 8, pro-caspase-1 and so on together. The activation of caspase1, which is mediated by the inflammasome further induces maturation of IL-1β and IL-18 in RA [21]. The similar activation of NF-κB and NLRP3/Caspase-1 pathways were also observed in CIA ankle joint tissues and LPSinduced macrophages in present study. A great deal of evidence has been accumulated and revealed the beneficial effect of direct inhibition of NF-κB on inflammatory disease [22,23]. Herein, methotrexate inhibited the activation of NF-κB and NLRP3/Caspase-1 pathways in joint tissue and

Discussion
Collagen-induced arthritis (CIA) in rats is the most widely used model for RA because it shares many similar immunological and pathological features with RA in patients [17]. Innate immune components, especially macrophage is a prime source of pro-inflammatory cytokines and critically contribute to RA pathology [3,18]. In present study, we established the CIA murine model and the inflammatory macrophage cellular model for pharmacological research on the anti-rheumatic effect of methotrexate. We discovered that methotrexate could alleviate inflammatory swelling in CIA rats and inhibit the inflammatory proliferation of macrophages by suppressing NF-κB and NLRP3/Caspase-1 pathway and downregulate the level of pathway related cytokines. Besides, methotrexate has the potential to inhibit catalytic activity of caspase-1 directly. Three inflammation related metabolic pathways including arachidonic acid, linoleic acid and sphingolipid metabolism also participated in the pathogenesis of RA and were regulated significantly after the treatment with methotrexate.
NF-κB is responsible for the regulation of inflammation and immune responses and it is involved in the pathogenesis of RA. The damage-associated molecular patterns binding to Toll-like receptors, as the priming signal activate the upregulation of the NF-κB nucleus transfer [3,19]. The translocated free NF-κB then activates its target genes to transcribe various kinds of inflammatory mediators, such as proIL-1β, proIL-18, and NLRP3 [19,20]. NLRP3 protein was constructed as NLRP3 inflammasome assembly with other components including caspase recruitment domain containing protein 8, pro-caspase-1 and so on together. The activation of caspase1, which is mediated by the inflammasome further induces maturation of IL-1β and IL-18 in RA [21]. The similar activation of NF-κB and NLRP3/Caspase-1 pathways were also observed in CIA ankle joint tissues and LPS-induced macrophages in present study. A great deal of evidence has been accumulated and revealed the beneficial effect of direct inhibition of NF-κB on inflammatory disease [22,23]. Herein, methotrexate inhibited the activation of NF-κB and NLRP3/Caspase-1 pathways in joint tissue and decreased the level of systematic cytokines in a dose dependent manner. The activity-inhibiting binding of methotrexate within caspase-1 may have the potential to partly explain the therapeutic effect, but the detailed mechanism is not clear.
Metabolomics techniques as a high-throughput approach have been extensively utilized to investigate the pathogenesis of RA and therapeutic targets of anti-RA drugs [11,24]. In the present study, we discovered that arachidonic acid, linoleic acid, and sphingolipid metabolism perturbed and they were significantly regulated after administrating with methotrexate.
Arachidonic acid, a component that is released from cellular membrane would be metabolized as a variety of metabolites, including EETs, HETEs, prostaglandins, thromboxane, and leukotrienes by cyclooxygenases or lipoxygenases and promote the nuclear translocation of NF-κB, which is at least partly induced by eicosanoids [25,26]. At the same time, the NF-κB pathway is the most critical activator for NLRP3 inflammasome and the massive production of lipid mediators [27]. EETs are a series of lipid mediators with critical physiological functions that include anti-inflammation, anti-hypertension, and organ protection effects [28]. The anti-inflammation effect is associated with the regulation of NF-κB translocation, as described previously [29]. Methotrexate also increased the production of EETs and inhibited the activation of NF-κB in present study. Prostaglandins and leukotrienes, which was believed to cause pain and activate the formation of cytokine were reduced after the treatment of methotrexate. So did the level of thromboxane A2. HETEs as a class of lipids with distinct inflammation bioactive roles were also been influenced by methotrexate. In brief, methotrexate modulate the NF-κB and NLRP3/caspase-1 inflammasome-mediated arachidonic acid metabolic lipid profiles, inhibit the release of pro-inflammatory lipids and promote the production of anti-inflammatory mediators in CIA murine rats.
The precursor of arachidonic acid, linoleic acid can be converted to arachidonic acid and epoxides of linoleic acid (including 9,10-EpOME and 12,13-EpOME) [30]. The EpOMEs that are produced by neutrophils and macrophages are considered as leukotoxins and they contribute to the pathogenesis of inflammatory disease, such as acute respiratory distress syndrome [31]. We observed a significant increase in CIA rats and a decrease of EpOMEs after the treatment of methotrexate. Linoleic acid itself seems to have potent anti-inflammatory effect, which is associated with the diminished NF-κB binding activity and delayed translocation of NF-κB [32,33]. A lower level of linoleic acid reported before in RA serum is consistent with our results in CIA [34]. Linoleic acid also has the potential to ameliorate the high expression level of NLRP3 genes and reduced the release of mature IL-1β and other cytokines [35,36]. Despite that methotrexate could damage the dietary absorption of linoleic acid [37], it had already been reported to increase linoleic acid in carcinoma cell and was demonstrated in present study in CIA rats to play its anti-inflammation role [38].
Sphingolipids as a class of bioactive lipids, including ceramides, sphingomyelins, and lactosylceramides play a key modulating role in cell cycle, apoptosis, and inflammatory responses [39]. They could also participate in the osteoblast crosstalk on bone homeostasis [40]. Sphingomyelin, a critical structural component of biological membranes is tightly associated with the activation of NF-κB [39,41]. The hydrolyzed metabolites of sphingomyelin, ceramides have already been linked to the inflammation process and act as a mediator inducing the activation of NF-κB and the sensitivity of NLRP3 inflammasome [42]. Another pro-inflammatory key branching point of sphingolipid biosynthesis, lactosylceramides were also reported to involve the activation of Ras/NF-κB and inflammatory gene expression [43]. Similar to our results, the level of these sphingolipids increased significantly in RA synovial fluid [44]. Therapeutic effect of methotrexate on CIA murine sphingolipid metabolism was reported for the first time in the present study. We demonstrated that methotrexate might have downregulated the NF-κB and NLRP3/caspase-1 pathways at least partly by intervening the sphingolipid metabolism.
RA was accompanied with a differentiated energy metabolism and expenditure [45]. Trehalose was reported to inhibit the inflammation by regulating NF-κB [46]. Herein, we observed a mild alteration of energy metabolism in CIA rats after the administration of methotrexate, indicating a potent mechanism that is unestablished. Glycerophospholipids are a critical series compounds for participating in many biological regulatory processes, including the pathogenesis of RA [47]. In present study, we identified large series phosphatidylcholines, most of which had long polyunsaturated side chains. The detailed role of every phosphatidylcholines metabolites was unknown, but their potential predictive effects for diagnosis and therapy will benefit for the development and investigation of biomarkers in clinical practice.
Although the subjects with RA mainly manifests a pathological change in synovial joint and cartilage, recent studies reported that the autoimmune response in patients with RA was initiated outside the joint, affected different parts of the body, and resulted in multi-organ disorders inevitably [48]. The inflammation related pathways and cytokines that we detected in the joint ankle tissue could reflect the pathophysiological changes more sufficiently. Given that many immune components such as innate B lymphocyte and adaptive T lymphocyte have been reported to participate the pathogenesis of RA [49], whether methotrexate could potentially regulate the inflammatory mechanism of these cells need further elucidation. Untargeted metabolomics in present study was utilized to analyze all the measurable molecules in a sample, including chemical unknowns. But, the extremely limited ability to quantify with the untargeted metabolomics require us to perform the targeted lipidomic research on the RA patients treated with methotrexate. Small size for metabolomics study is the other limitation of this study. Besides, as a classical chemical therapeutic drug, the unknown side effects after inhibiting the potential target also need our further evaluation. Nevertheless, this pilot study will be a pioneer for further determination on the discovered metabolism pathways.

Assessment on RA Swelling
The experimenter recorded the visual scores according to the recommended scoring system from manufacturer of Chondrex every two days. Paw swelling volume was also recorded before and after drug administration with paw volume measuring instrument (PV-200, Chengdu Tai Meng Software Co., Ltd., Chengdu City, Sichuan Province, China). The end-line of the paw volume measurement was defined as 0.3 cm above the ankle. Experimenter continued to score the swelling level until the animals were sacrificed. The observer was blind to the biological groups of experimental animals.

Sacrifice and Sample Collection
All of the animals were sacrificed when the average visual score of Model measured nearly and stably up to 4. The animals were anesthetized with sodium pentobarbital and sacrificed by collecting whole blood from abdominal aorta directly into the heparin sodium treated vacuum tube. Plasma samples was prepared by centrifuging the blood at 3000 rpm for 15 min at 4 • C and stored at −80 • C. The preparation of ankle joints tissue was performed as described previously [50]. All of the ankle joints tissues were stored at liquid nitrogen for RNA and protein extraction.

Cell Viability Assay with Real Time Cellular Analysis
Proliferation of RAW 264.7 cells, as obtained from American Type Culture Collection (ATCC TIB-71) was monitored with Real Time Cellular Analysis system (xCELLigence RTCA S16, ACEA Biosciences, San Diego, USA) in Dulbecco's modified Eagle's medium (DMEM, Caisson, Smithfield, UT, USA) containing 10% fetal bovine serum (FBS, Genedirex, Las Vegas, NV, USA) with 1% penicillin/streptomycin in a humidified atmosphere with 5% CO 2 in a 37 • C incubator. The toxicity assays of both methotrexate and diclofenac sodium were performed with RTCA system at a series of geometric concentrations (1 mM, 100 µM, 10 µM, 1 µM, 100 nM, 10 nM, 1 nM, 0 fM). The non-toxic dosages of methotrexate and diclofenac sodium were selected to investigate their anti-inflammation effect. 3500 cells in every well were cultured in the E-plates of the RTCA system and stimulated with lipopolysaccharide (LPS, 1 µg/mL, Sigma, Darmstadt, Germany. Cat. L2630) at 12th h for 4 h. After stimulating with LPS, the medium of Model group was continued to be under the stimulation of LPS. The anti-inflammation groups were treated with methotrexate in different concentrations or diclofenac sodium respectively. Control was cultured with 10% DMEM+PBS after the four hours' stimulation with LPS.

Cell Culture and Induction of Inflammation
RAW 264.7 cells were seeded in the six-well plates with 1 × 10 5 cell/well and cultured in the medium described above. The stimulation and treatment steps are consistent with the RTCA. All cells were harvested at the endpoint which was determined according to the highest peak of RTCA monitor for RNA and protein extraction. The supernatants were collected after removing cell debris by centrifugation prior to use for enzyme-linked immunosorbent assay (ELISA).

Metabolomic Procedure and Data Processing
All of the plasma samples, 200 µL for each sample, were added methanol (600 µL, Methanol (HPLC), Fisher Chemical, Cat. A452-1) to make free of protein. Then the supernatant (500 µL) was lyophilized at −60 • C and 10.0 pa. The residue was re-dissolved in 100 µL of methanol-water (4:1, v/v). An aliquot of 2 µL was injected for UPLC-MS/MS analysis with Waters ACQUITY UPLC system (Waters Corporation, Milford, MA, USA), which had been equipped with a BEH C18 column (2.1 mm × 100 mm, 1.7 mm, Waters Corporation). A Waters Xevo G2-S Quadrupole Time-Of-Flight (QTOF) mass spectrometer (Waters Corporation) coupled to the UPLC system was utilized to carry out the mass spectrometry with an electrospray ionization in both and positive (ESI + ) and negative (ESI − ) ion modes. The chromatographic conditions and spectrometric parameters were set and optimized, as described previously [51].
To ensure the stability and suitability of MS analysis, a QC sample was prepared by pooling the same volume (20 µL) from every plasma samples. Ten chromatographic peaks of ions from the QC sequencing datasheet were selected to evaluate the validation of systematic method. Repeatability and intermediate precision on spectrum and chromatogram were also estimated with six successive QC injections or 6 replicates of a plasma in both ESI + and ESI − modes before the injection of the plasma samples. Moreover, four QC injections were performed randomly through the whole worklist.
The pre-processing of raw data was finished with MarkerLynx XS V4.1 software for alignment, deconvolution, and data reduction, so as to pair the mass and retention time with the corresponding intensities of all detected peaks. The main parameters were set similarly as before [51]. The processed files from ESI + and ESI − modes were exported for multivariate analysis.

Molecular Docking of Methotrexate within Caspase-1
Molecular docking of methotrexate within caspase-1 was performed with Schrödinger software packages. Maestro (version 2015-2 demo, Schrödinger) was used for protein and ligand preparation, receptor grid generation and docking. The X-ray crystal structure of inhibited interleukin-1β converting enzyme (Protein Data Bank (PDB) code: 1IBC) was retrieved and downloaded from the PDB database (available online: http://www.rcsb.org/pdb). Caspase-1 protein was prepared in the Protein Preparation Wizard to optimize the structure with assigning bond orders and water orientations, removing water molecules, adding hydrogens, creating zero-order bonds to metals, and disulfide bonds. Energy of the protein was minimized with the default parameters. Receptor grid of the protein ligand was generated with the limited size of 20 Å at the active site. The crystal structure of ligand, methotrexate, was downloaded from National Center for Biotechnology Information Open Chemistry Database (NCBI, PubChem CID: 126941), and optimized in the Ligand Preparation Wizard prior to the molecular docking. Docking was performed on the Workspace using GLIDE v.6.7. At least 10 conformations were set for methotrexate. The docking result was plotted with Pymol 1.8 software, which was based on python 2.7.

Statistic and Bioinformatic Analysis
Multivariate analysis for metabolomics data pre-possessed was finished with SIMCA-P software (v14.1, Umetric, Umeå, Sweden). PCA and OPLS-DA models were established. Variable importance of project (VIP) values, which indicated a significant difference between the groups were also estimated statistically. RSD also was calculated for the pooled QC injections to assess the quality and stability of MS data. All the distinct metabolites were identified by matching accurate mass (ppm < 10) to the Human Metabolome Database (HMDB Version 4.0) [52], with confirmation determined by comparing characteristic tandem mass spectrometry (MS/MS) fragmentation patterns with METLIN [53] and HMDB Database [52] or demonstrated by referring the chemical standards (HPLC grade, Sigma). All distinct metabolites were analyzed with MetaboAnalyst 4.0 for metabolic pathways [54]. The comprehensive metabolic network was constructed with Cytoscape Software (v3.6.1) [55] based on the data from Kyoto Encyclopedia of Genes and Genomes (KEGG) database [56].
Statistical significance of the difference was estimated using the F-test. Kolmogorov-Smirnov test was used to ensure the normality of the data. Student's t test for data with homogeneity of variance or Welch's t-test was performed for pairwise two-group analysis. Moreover, multiple comparisons among the groups were performed by one-way analyses of variance (ANOVA). Mann-Whitney-Wilcoxon test was performed for the dataset, which does not follow the normality. All statistical significance was accepted at p < 0.05. The statistical analysis was completed with R (v3.3.3) basic statistical packages. All of the statistical bar charts were finished with R's Basic Graphic and ggplot2 (v2.2.1) packages. The heatmap of all the differentiated metabolites were prepared with ComplexHeatmap (v1.19.1) package. The specificity and sensitivity of biomarkers was displayed by ROC estimated with pROC (v1.12.1) package.

Conclusions
We demonstrated that methotrexate play an inhibitory effect on rheumatoid arthritis swelling and inflammation by downregulating NF-κB and NLRP3/Caspase-1 pathways and potentially affecting the activity of caspase-1. The critical perturbed metabolic pathways, including arachidonic acid, linoleic acid, and sphingolipid interacting with NF-κB and NLRP3/Caspase-1 could contribute to the pathogenesis of RA and were also involved in the pharmacological regulation of methotrexate. Various differentiated mediators provide us with series potential clinical biomarkers for future investigation. The elucidating pharmacological mechanism of methotrexate will benefit its rational administration in clinical practice.

Acknowledgments:
The authors would like to thank Cuizhu Wang from Research Center of Natural Drug, School of Pharmaceutical Sciences, Jilin University for providing her advice and help on the metabolomics researching work.

Conflicts of Interest:
The authors declare no conflict of interest.