Higher PGD2 production by synovial mast cells from rheumatoid arthritis patients compared with osteoarthritis patients via miR-199a-3p/prostaglandin synthetase 2 axis

We previously reported that synovial mast cells (MCs) from patients with rheumatoid arthritis (RA) produced TNF-α in response to immune complexes via FcγRI and FcγRIIA. However, the specific functions of synovial MCs in RA remain unclear. This study aimed to elucidate those functions. Synovial tissues and fluid were obtained from RA and osteoarthritis (OA) patients undergoing joint replacement surgery. Synovium-derived, cultured MCs were generated by culturing dispersed synovial cells with stem cell factor. We performed microarray-based screening of mRNA and microRNA (miRNA), followed by quantitative RT-PCR-based verification. Synovial MCs from RA patients showed significantly higher prostaglandin systhetase (PTGS)1 and PTGS2 expression compared with OA patients’ MCs, and they produced significantly more prostaglandin D2 (PGD2) following aggregation of FcγRI. PGD2 induced IL-8 production by human group 2 innate lymphoid cells, suggesting that PGD2-producing MCs induce neutrophil recruitment into the synovium of RA patients. PTGS2 mRNA expression in RA patients’ MCs correlated inversely with miRNA-199a-3p expression, which down-regulated PTGS2. RA patients’ synovial fluid contained significantly more PGD2 compared with OA patients’ fluid. Synovial MCs might regulate inflammation in RA through hyper-production of PGD2 following FcRγ aggregation. Our findings indicate functional heterogeneity of human MCs among diseases.


Results
Comparison of gene expression profiles between synovial MCs from OA and RA patients. We first compared the gene expression profiles between synovium-derived, cultured MCs from OA patients (n = 3 donors) and those from RA patients (n = 3 donors) using DNA chips. A total of 419 genes among approximately 42,545 full-length genes and expressed sequence tags were significantly more than two fold higher in synoviumderived, cultured MCs from RA patients than in OA patients. Figure 1A shows a part of hierarchical clustering on the basis of those gene expression data. Upregulation of the cyclooxygenase and lipoxygenase pathways of arachidonic acid is thought to be involved in the development of rheumatic diseases, and targeting these pathways might lead to improved treatment strategies 11 . For that reason, from among those upregulated genes, we focused on prostanoid synthetases, including prostaglandin synthetase (PTGS)1, PTGS2, thromboxane synthetase 1 (TBXAS1) and leukotriene C 4 synthetase (LTC4S) (Fig. 1A). DNA chip analysis showed that the normalized expression level of HPGDS was high compared with the other prostanoid synthetases and did not differ significantly between MCs from OA and RA patients (Fig. 1B). The expression levels of PTGES, PTGES2 and PTGIS were quite low. The expression levels of prostanoid receptors (Fig. 1C) and MC-related genes (Fig. 1D) did not appear to differ significantly between MCs from OA and RA patients. We confirmed the above DNA chip results using quantitative RT-PCR (Fig. 1E,F). The expression levels of mRNA for PTGS1, PTGS2 and TBXAS1 were significantly higher in RA patients' MCs than in OA patients' MCs (Fig. 1E). The expression levels of mRNA for LTC4S showed the tendency of being high in RA patients' MCs compared with OA patients' MCs. Furthermore, we confirmed those findings in synovium-derived, cultured MCs by using freshly isolated synovial MCs from OA and RA patients (Fig. 1F). www.nature.com/scientificreports/ of RA patients 23,24 . PGD 2 induced production of Th2 cytokines such as IL-5 and IL-13 by ILC2s 27 . Thus, we investigated the effects of PGD 2 on production of IL-6, IL-8, IL-9 and TNF-α by ILC2s. We found that PGD 2 induced production of IL-8, but not IL-6 and IL-9, by ILC2s ( Fig. 3A,B,D). TNF-α was spontaneously released by ILC2s and PGD 2 did not enhance the production (Fig. 3C). This suggests that higher amounts of IL-8 produced by ILC2s in RA patients' synovial tissues in response to IgG-mediated MC-induced PGD 2 might cause neutrophil recruitment into the synovium of those patients.
Effects of co-culture of OA patients' synovial MCs with RA patients' synovial fibroblasts on PTGS1, PTGS2, TBXAS1 and LTC4S expression in OA patients' MCs. The SCF-Kit system alone is insufficient to fully drive the maturation of MCs, since culture of immature MCs with fibroblasts, but not with  Relationship between expression levels of miR-199a-3p and PTGS2 mRNA in synovium-derived, cultured MCs from OA and RA patients. micro RNAs (miRNAs) are noncoding RNAs implicated in the regulation of gene expression underlying many relevant physiological processes, including cell activation. We investigated one mechanism causing different mRNA expression levels for prostanoid synthetases between RA   30,31 , cultured human fetal lung epithelial cells 32 , cultured human myometrial cells 33 and human endometrial surface epithelial cells 34 . Thus, we performed quantitative RT-PCR to compare the miR-199a-3p expression levels between synovial MCs from RA and OA patients. The results showed that the miR-199a-3p expression level was significantly higher in the OA patients than in the RA patients (P < 0.05, Fig. 5A). Moreover, that expression level correlated inversely with the PTGS2 mRNA expression level in the RA patients' MCs (r = − 0.698, P = 0.010; Fig. 5C), but not in the OA patients' MCs (Fig. 5B). Tables 2 and 3 shows the characteristics of the patients with OA and RA, respectively.  www.nature.com/scientificreports/ To analyze the effect of miR-199a-3p on PTGS2 expression, we overexpressed miR-199a-3p in synoviumderived, cultured MCs obtained from RA patients by a transduction of miRNA mimics. Figure 5D depicts the method for transduction of miRNA mimic to synovium-derived, cultured MCs obtained from RA patients and activation of the cells. Briefly, miR199a-3p or control miRNA mimics were transduced to the cells, and then the cells were activated with TNF-α. In Fig. 5E, the miR199a-3p expression level was significantly increased in synovium-derived, cultured MCs obtained from RA patients transduced with an miR199a-3p mimic compared with control miRNA mimic (P < 0.05, Fig. 5E). We stimulated these MCs with TNF-α (10 ng/mL) for 2 h and analyzed PTGS2 expression in control miRNA mimic-transduced and miR199a-3p mimic-transduced cells. The PTGS2 expression in miR199a-3p mimic-transduced cells was significantly lower than in control miRNA mimic-transduced cells after 2 h of stimulation with TNF-α (P < 0.05, Fig. 5F).

Comparison of PGD 2 and PGE 2 concentrations in SF obtained from OA and RA patients.
We performed EIA to measure the PGD 2 and PGE 2 concentrations in SF obtained from OA (n = 14) and RA (n = 10) patients. Tables 2 and 3 show the characteristics of the patients. The concentration of PGD 2 (Fig. 6A), but not PGE 2 (Fig. 6B), was significantly higher in the RA patients' SF than in the OA patients' SF.

Discussion
We present the first evidence that synovial MCs from RA and OA patients have different phenotypes and function. FcRγ aggregation induced significantly higher PGD 2 production by the synovium-derived, cultured MCs from RA patients via miR-199a-3p/PTGS2 axis compared with the OA patients' MCs.
Since the LTC4S expression level showed the tendency of being higher in the RA patients' MCs than in the OA patients' MCs ( Fig. 1E), LTC 4 production by the RA patients' MCs following FcεRI aggregation was significantly higher (Fig. 2G). In contrast, LTB 4 production by the RA patients' MCs following FcεRI aggregation was significantly lower than that by the OA patients' MCs ( Fig. 2H). Since LTB 4 and LTC 4 are metabolites of LTA 4 , LTB 4 synthesis in RA patients' MCs may be relatively lower. LTB 4 and LTC 4 by MCs from both RA and OA patients after FcγRI aggregation was markedly lower than after FcεRI aggregation (Fig. 2C,D,G,H). FcεRI consists of heterotetramer αβγ 2 -chains. On the other hand, FcγRI consists of heterotrimer αγ 2 -chains. Since FcεRI β-chain reportedly amplified LT production by mouse MCs 35 , FcγRI aggregation might cause no or lesser amounts of LT production by human MCs as well. Thus, immune complexes might induce large amounts of only PGD 2 among the arachidonic acid metabolites in RA patients' synovial MCs.
We previously reported that co-culture of IL-3-dependent mouse bone marrow-derived MCs with mouse fibroblasts resulted in 1.4-fold and 387-fold up-regulation of PTGS1 and PTGS2 mRNA expression, respectively, in the MCs 29 . Thus, we hypothesized that fibroblasts from RA patients might influence PTGS1 and PTGS2 mRNA expression in OA patients' MCs, but our co-culture experiment did not find that to be true. The reason is not Table 1. List of genes whose expression levels of miRNAs were significantly more than three times higher in MCs from OA patients than in MCs from RA patients. Raw data from miRNA microarrays, normalized normalized data obtained from miRNA microarrays (please see "Materials and methods" section), MCs mast cells, OA osteoarthritis, RA rheumatoid arthritis. www.nature.com/scientificreports/  www.nature.com/scientificreports/  36 . However, NSAIDs do not affect PTGS1 and PTGS2 mRNA expression. Synovium-derived, cultured MCs from both OA and RA patients had been cultured under the same conditions for more than 12 weeks. Therefore, it would be difficult to conclude that NSAID treatment of patients might affect the gene expression profiles of the cultured MCs derived from synovia of OA and RA patients. We thus hypothesized that epigenetic modification might be the cause of the difference in the gene expression profiles of MCs derived from synovia of OA and RA patients.
Microarray-based screening found that the expression levels of thirty miRNAs were significantly more than three times higher in OA patients' MCs than in RA patients' MCs (Table 1). Among those 30 miRNAs, miR-199a-3p is a direct regulator of PTGS2 expression in OA chondrocytes 30,31 , cultured human fetal lung epithelial cells 32 , cultured human myometrial cells 33 and human endometrial surface epithelial cells 34 . The miR-199a-3p/ PTGS2 axis plays different roles in different cells and diseases. In human OA chondrocytes, miR-199a-3p directly suppressed the luciferase activity of a PGST2 3′UTR reporter construct and inhibited IL-β-induced PTGS2 protein, suggesting that miR-199a-3p may be an important regulator of human cartilage hemostasis and a new target for OA therapy 30 . In fact, epigallocatechin-3-O-gallate, the most abundant and active polyphenol in green tea, which has been reported to have anti-arthritic effects, inhibited PTGS2 mRNA/protein expression and PGE 2 production by up-regulating miR-199a-3p expression in IL-1β-stimulated human OA chondrocytes 31 . The developmental decline in miR-199a/miR-214 expression in the fetal lung led to increased expression of critical targets, including PTGS2, NF-κB p50/p65, CREB1 and C/EBPβ that enhance surfactant protein-A expression and alveolar type II cell differentiation 32 . The levels of the clustered miRNAs, i.e., miR-199a-3p and miR-214, were significantly decreased in the myometrium of pregnant mice and humans, whereas the miR-199a-3p/miR-214 target, PTGS2, which induced synthesis of contracting PGs, was coordinately increased 33,37 . Overexpression of miR-199a-3p and miR-214 in cultured human myometrial cells inhibited PTGS2 protein and blocked TNF-α-induced myometrial cell contractility, suggesting their physiological relevance 33 . Epithelial sodium channel-dependent CREB activation led to suppression of miR-199a-3p and miR101, which in turn augmented PTGS2 up-regulation during embryo implantation 34 . We found that miR-199a-3p correlated inversely with PTGS2 expression in RA patients' MCs (r = − 0.698, P = 0.010; Fig. 5C), suggesting that miR-199a-3p may be a regulator of PTGS2 mRNA expression in RA patients' synovial MCs. Furthermore, we confirmed that PTGS2 mRNA expression by miR199a-3p mimic-transduced synovium-derived, cultured MCs obtained from RA patients was significantly lower than that by control miRNA mimic-transduced cells, when the cells were activated with TNF-α (P < 0.05, Fig. 5F). The 3′ UTR of the mTOR gene is reportedly targeted by miR-498, which was included in the list of thirty genes. Consequently, silencing of mTOR reduced PTGS2 expression in OA chondrocytes 38 . Thus, other miRNAs might also affect PTGS2 mRNA expression in RA patients' synovial MCs. PGD 2 was reported to be detected in SF obtained from RA patients 11,22,39 . PGD 2 is synthesized by various cells, including MCs 16 , antigen-presenting cells 17 , T helper 2 (Th2) lymphocytes 18 and synovial fibroblasts 40 . The cell sources of PGD 2 in RA patients' SF have not been identified. In our study, OA and RA patients' fibroblasts showed no significant differences in expression of PTGS1 and PTGS2 mRNAs. Thus, increased PGD 2 synthesis would not be due to production by synovial fibroblasts. Our in vitro study found that synovial MCs (10 5 cells) from RA patients produced 4000 pg/mL PGD 2 following FcγRI aggregation. Furthermore, PGD 2 metabolites have been reported to be biomarkers of in vivo MC activation in RA patients 41 . Therefore, MCs might be one cell source of PGD 2 in RA patients' SF.
Upregulation of the PTGS1 and PTGS2 pathways of arachidonic acid (AA) is thought to be involved in the development of rheumatic diseases, and targeting these pathways might lead to improved treatment strategies 11 . Thus, to clarify the quantitative and qualitative changes in lipid mediators in the synovium of severe RA patients, we recently compared the profiles of lipid mediators in SF obtained from RA and OA patients using liquid chromatography-tandem mass spectrometry/mass spectrometry 42 . The concentrations (levels based on the area-under-the-curve/mL) of the majority of PTGS-1/2 products of AA appeared to be higher in SF from RA patients compared with OA patients. We determined the absolute concentrations (pmol/mL) of representative eicosanoids, including 6-keto PGF 1α (a stable metabolite of PGI 2 ), PGF 2α , PGE 2 , PGD 2 and 12-hydroxyheptadecatrienoic acid (HHT), in the SF from RA and OA patients. The PGF 2α and PGE 2 concentrations were significantly higher in the RA patients' SF. Thus, although the PGD 2 concentration in the SF did not differ significantly between Figure 6. Comparison of PGD 2 and PGE 2 concentrations in SF obtained from OA and RA patients. SF was obtained from OA (n = 14) and RA (n = 10) patients. Each point represents one donor. Significance was determined using the Mann-Whitney U test (***P < .0005). www.nature.com/scientificreports/ the RA and OA patients in our previous study 42 , we confirmed our earlier findings regarding upregulation of the PTGS pathways in RA compared with in OA 42 . This study has a number of limitations. First, all the samples used in this study were obtained from hospitalized patients who had undergone total knee replacement surgery. It was reported that RA patients in remission had significantly reduced synovial MC density compared with patients with clinically active RA 43 . Thus, the enrollees were limited to patients with severe clinical disease, hence limiting extrapolation of the findings to patients with milder clinical disease. Second, we cannot rule out the possibility that use of a wide range of antirheumatic drugs-including methotrexate, oral glucocorticoids, anti-TNF-α therapy and anti-IL-6 therapy that potently suppress specific inflammation, might be responsible for the differences in gene expression profiles in OA and RA patients' MCs. However, the RA patients' anti-TNF-α therapy and anti-IL-6 therapy had been discontinued 2-4 weeks before the total knee arthroplasty. Since expression of PTGS2 mRNA and protein is reportedly enhanced in various human cell types by such inflammatory cytokines as IL-1β and TNF-α 44 , anti-TNF-α therapy and anti-IL-6 therapy would not enhance PTGS2 mRNA expression levels.

Scientific Reports
PGD 2 was reported to be active in the resolution of inflammation in experimentally induced arthritis in mice 19,21 and in human RA 22 . Increased PGD 2 may trigger an anti-inflammatory/pro-resolution cascade. That is because it spontaneously undergoes non-enzymatic dehydration and is converted into 15-deoxy-Δ12,14prostaglandin J 2 (15d-PGJ 2 ). 15d-PGJ 2 is a cyclopentenone PG that has been shown to be immuno-modulatory and anti-inflammatory due to its ability to inhibit NFκB signaling and cytokine release and to act as an agonist of PPARγ 45 . However, the role of MC-derived PGD 2 in the pathogenesis of RA should be determined in MCdeficient experimental arthritis in mice reconstituted with bone marrow-derived MCs obtained from HPGDSdeficient mice. NSAIDs' inhibition of cyclooxygenase-dependent PG synthesis ameliorates RA manifestation because they inhibit mainly production of PGE 2 , which exacerbates synovial inflammation in RA patients 46 .
Since ILC2s were reportedly involved in the pathogenesis of RA 23,24 , we investigated the effect of PGD 2 on RArelated cytokine production (IL-6, IL-8, IL-9 and TNF-α) by ILC2s. We found that PGD 2 induced IL-8 production by ILC2s, suggesting that PGD 2 -producing MCs induce neutrophil recruitment into the synovium of RA patients.
In conclusion, we demonstrated that human synovial MCs might regulate inflammation through hyperproduction of PGD 2 in RA following FcRγ aggregation. These findings indicate that human MCs show functional heterogeneity among diseases.

Materials and methods
Patient enrollment and processing of SF. We enrolled RA and OA patients. The diagnosis of each patient was established by the treating doctor. SF and synovial tissue samples were obtained during total knee arthroplasty performed at the Department of Orthopeadic Surgery, Nihon University, after receiving informed consent. Two milliliters of SF were treated with hyaluronidase, followed by centrifugation at 860×g for 10 min. The supernatants were collected, and the tubes were filled with N 2 gas. The samples were then frozen at − 80 °C.

Purification of dispersed synovial MCs and generation of synovium-derived, cultured
MCs. Human synovial MCs were purified with anti-FcεRIα and anti-Kit mAbs using FACS Aria IIu (BD Biosciences). The purities of MCs were > 99%. Human synovium-derived cultured MCs were generated as described previously 8 . Briefly, fresh samples of synovial tissues were obtained after total knee arthroplasty at Nihon University, after obtaining informed consent. Treatment of RA patients with anti-TNF-α therapy and anti-IL-6 therapy had been discontinued 2-4 weeks before total knee arthroplasty, in accordance with the Japanese Guidelines for the use of Infliximab and Etanercept in RA 47 . Briefly, synovial cells were enzymatically dispersed and centrifuged using a density-gradient consisting of 22.5% HistoDenz solution (Sigma-Aldrich; St. Louis, MO, USA) and lymphocyte separation medium (LSM; MP Biomedicals; Santa Ana, CA, USA). Cells at the LSM interface and in the pellet fraction were collected and washed. The cells were then cultured in serum-free Iscove's methylcellulose medium (Stem Cell Technologies Inc.; Vancouver, BC, Canada) and Iscove's Modified Dulbecco's Medium (IMDM; Thermo Fisher Scientific; Waltham, MA, USA) supplemented with 200 ng/mL recombinant human stem cell factor (rhSCF) (PeproTech; Rocky Hill, NJ, USA) and 50 ng/mL rhIL-6 (PeproTech). On day 42, methylcellulose was dissolved in PBS, and the cells were resuspended and cultured in IMDM containing 0.1% BSA, 100 ng/mL rhSCF and 50 ng/mL rhIL-6 (designated as MC medium).

Synovial fibroblasts.
Fresh samples of synovial tissues were obtained after total knee arthroplasty at Nihon University, after obtaining informed consent. Synovial fibroblasts were obtained after culturing enzymaticallydispersed synovial cells 7 .

Isolation and expansion of human group 2 innate lymphoid cells (ILC2s). Peripheral blood mon-
onuclear cells (PBMCs) were isolated from whole blood of healthy volunteers by centrifugation using LSM. Lineage-negative cells were enriched from the isolated PBMCs using magnetic-activated cell sorting (MACS) and Microbeads (Miltenyi Biotec; GladBach, Germany) in accordance with the manufacturer's protocol, with some modification. Briefly, 5 × 10 7 to 1 × 10 8  were determined to be differentially expressed (up-or down-regulated). The false discovery rate is a comparison of the number of times that the real data has a certain P value versus the number of times that randomized data has the same or better P value. The false discovery rate addresses the multiple comparisons problem that occurs when calculating P values for hundreds or thousands of categories, and protects against over-interpreting P values that do not have biological meaning 48 . Hierarchical clustering was performed on the basis of the gene expression data. Of the differentially expressed miRNAs, target miRNAs that interact with target genes were determined by a search of the literature. To further investigate the global molecular network, especially to identify upstream cytokines in pathological signaling cascades, the target genes were imported into IPA (version 27821452) 49  www.nature.com/scientificreports/ (gF(ab′) 2 αmF(ab′) 2 ; Jackson ImmunoResearch Laboratories) for 30 min for the histamine and lipid mediator assays (2 × 10 3 MCs/100 μL). Independent experiments were performed using MCs from different donors.
Co-culture of synovial MCs and fibroblasts. Synovial fibroblasts from RA patients were cultured in IMDM supplemented with 2% FBS for 48 h, and the fibroblasts were grown to confluence. Human synoviumderived, cultured MCs from OA patients were overlaid on the fibroblasts and cultured in IMDM containing 0.1% BSA, 100 ng/mL rhSCF and 50 ng/mL rhIL-6 for 96 h. The MCs were then purified with anti-FcεRIα and anti-Kit mAbs using FACS Aria IIu (BD Biosciences).

Statistical analysis.
To evaluate the quantitative variables, the Mann-Whitney U test was used because of nonparametric distribution of the data. Figures 2 and 3 were analyzed by Sidak's-Bonferroni method. Spearman rank correlation coefficients were calculated to determine the strength of correlations between continuous variables. P values were considered significant at P < 0.05. www.nature.com/scientificreports/