The deidentified nondiseased (NL) and RA synovial tissues (RASTs) were procured under a protocol approved by the Washington State University IRB (IRB#14696) from the Cooperative Human Tissue Network (Columbus, OH) and National Disease Research Interchange (Philadelphia, PA). STs from RA patients were obtained from joint surgery or synovectomy according to an Institutional Review Board (IRB)-approved protocol and in compliance with the Declaration of Helsinki. NLSTs were obtained from autopsy or amputation. Disease RA tissue was digested in dispase, collagenase, and DNase before being seeded in 72 cm² flasks.

Cells were grown in RPMI 1640 medium supplemented with 15% fetal bovine serum (FBS, Invitrogen, Waltham, MA, USA), 5,000 U/ml penicillin, 5 mg/ml streptomycin from Sigma (St. Louis, MO, USA), and 10 µg/ml gentamicin (Invitrogen, Waltham, MA, USA). Upon confluency (>85%), cells were passaged with brief trypsinization. Experiments were done using cells passed at least four to five times to ensure a pure fibroblast population. Similarly, we also established primary human synovial fibroblasts from nondiseased (NLSFs) tissues under an IRB-approved protocol.

Human primary NLSF and RASF cells were maintained in RPMI 1640 culture media, with 10% FBS and antibiotics. For chronic exposure of cytokines ranging 12 days of exposure, 2 ng/ml M-CSF+10 ng/ml RANKL were used with or without a combination of 100 ng/ml IL-6 along with 100 ng/ml IL-6Rα. For acute exposure, cells were starved overnight in plain RPMI 1640, followed by stimulation cytokines. Acute treatment of cytokines ranging between 30 min and 4 h using 25 ng/ml MCSF+50 ng/ml RANKL was used with or without combination 100 ng/ml IL-6 along with 100 ng/ml IL-6R experiments. For chronic exposure of cells, RPMI 1640 with 10% FBS in combination with cytokines was changed every 3 days. On day 11 of differentiation, media without FBS was replenished for an additional 24 h before termination of the experiment. Supernatant and whole-cell extracts were prepared accordingly for further experiments. Likewise, we isolated bone marrow cells from 8-week-old C57BL/6 and allowed them to differentiate using 25 ng/ml M-CSF+10 ng/ml RANKL for 14 days with or without 100 ng/ml IL-6 or 100 ng/ml IL-6+100 ng/ml IL-6R (IL-6/IL-6R) for bone marrow macrophage-derived osteoclast studies. Human PBMCs were obtained from a commercial source (Zen-Bio, Inc., Research Triangle, NC, USA).

For the inhibitor experiments, overnight-starved RASF cells were subjected to 2 h pretreatment with 5 µM tofacitinib or 3 µg/ml anti-IL-6 antibody followed by M-CSF+RANKL and/or IL-6/IL-6R stimulation for 30 min for Western blotting. TRAP staining was performed on RASFs based on manufacturer instructions from PMC-AK04F-COS (CosmoBio, Carlsbad, CA, USA).

Chromatin immunoprecipitation (ChIP) was performed for Ets2 as described previously (19). Human RASFs from three different donors were subjected to differentiation by M-CSF+RANKL with or without IL-6+IL-6R combination for 12 days. Detailed protocol is provided in the SI Materials and Methods. The list of primer pairs used for the ChIP assay has been provided in SI ( Supplementary Table S1 ).

The untargeted proteomics analysis was conducted on an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) equipped with a Nano-Acquity UPLC system (Waters, Milford, MA, United States) and in-house developed nanospray ionization source at the University of Washington Proteomics Center, Seattle, WA. A detailed method is provided in the SI Material and Methods file. Candidates from IL-6+IL-6R treatment cohorts were also subjected to Metascape and or Ingenuity Pathways Analysis-based (IPA-Qiagen, Hilden, Germany) gene ontology prediction. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD028374.

Human RASFs (2 × 10⁴) were seeded in calcium phosphate-coated 24-well plates (CSR-BRA-24KIT, CosmoBio, Carlsbad, CA, USA) and allowed to reach ∼85% confluency. Cells were differentiated under 2 ng/ml M-CSF+10 ng/ml RANKL with and without combination of 100 ng/ml IL-6+100 ng/ml IL-6R for 12 days. Every 3 days, media were replaced with fresh cytokines. After 12 days of differentiation, plates were washed with 10% sodium hypochlorite for 45 min, followed by an extensive wash using deionized water. Plates were air-dried, followed by staining with 1% toluidine blue staining for an additional 5 min. Plates were rewashed and air-dried. Each well of toluidine blue stain plates were photographed using a Zeiss microscope. Pit areas (three independent pits from each triplicate well) were calculated using ImageJ software (NIH, Bethesda, MD, United States). The total cumulative pit area was divided by the whole field and reported as a percentage of pit area per field.

Statistical analysis of multiple comparisons between control versus treatment group means was performed using a one-way ANOVA followed by Dunnett’s test or using Tukey’s multiple comparison test when comparing multiple treatment groups to each other. All tests assumed normal distribution where p < 0.05 was considered significant.

To understand the development of TRAP-positive RASFs that indicate osteoclast-like phenotype, human RASFs treated under different conditions for 12 days were washed with ice-cold PBS, fixed immediately, and stained using a TRAP substrate. Analysis of the stained area showed that IL-6/IL-6R increased the number of TRAP-positive cells in cultured conditions more than M-CSF/RANKL, the known drivers of osteoclastogenesis ( Figures 1A, B ). To confirm this observation, we differentiated bone marrow-derived macrophage (BMDM) using M-CSF/RANKL or treated in combination with IL-6 for 14 days. Upon differentiation, IL-6 markedly increased in the number of TRAP-positive BMDMs compared with M-CSF/RANKL treatment ( Supplementary Figure S1 ). To confirm this phenotypic switch of RASFs to osteoclast-like cells also influenced their function, we performed the in vitro pit formation assay using calcium phosphate-coated plates. Evaluation of the pit areas showed that RASFs treated with chronic IL-6/IL-6R caused a significantly higher erosion of calcium phosphate from the plates independent of the osteoclastic cytokines M-CSF/RANKL ( Figure 1C ). Quantitative analysis showed that the percentage of pit area formed were significantly increased by ~2-fold in IL-6/IL-6R-treated samples compared with M-CSF/RANKL, and no further increase in the percentage of pit area formed was observed with M-CSF/RANKL plus IL-6/IL-6R treatment ( Figure 1D ; p < 0.01; n = 3), suggesting that IL-6 trans-signaling can independently be a major factor in RASF-induced bone loss. Furthermore, our FACS analysis identified the lack of CD14+ cell population in RASF culture preparations used in the study, suggesting the pure synovial fibroblast lineage of the cells used ( Supplementary Figure S2 ).

Human RASFs grown for 12 days under the differentiation with M-CSF/RANKL and/or IL-6/IL-6R were evaluated for osteoclast-specific proteins. Western blot analysis of osteoclast-specific signature proteins confirms upregulation of cathepsin K, cathepsin B, p-STAT-3, RANKL, OPG, and RANKL expression by both M-CSF/RANKL and IL-6/IL-6R treatment ( Figures 2A, B ). Interestingly, while M-CSF/RANKL were able to preferentially activate traditional osteoclast markers such as OSCAR, OPG, and RANKL, IL-6/IL-6R was unique in inducing Ets2 transcription factor, in addition to p-STAT3 activation, to induce cathepsins K and B expression in human RASFs ( Figure 2A ). Furthermore, IL-6 trans-signaling in RASF-induced dose-dependent Ets2 expression by 70%–85% (p < 0.01; n = 4) and consequently induced cathepsins K and B expression by ~70% and 100%–125%, respectively, compared with untreated RASFs ( Figures 2C, D ; p < 0.05 or p < 0.01; n = 3). To compare these changes to a classical osteoclast model, primary human PBMC-derived macrophages were differentiated into osteoclasts by M-CSF/RANKL or IL-6/IL-6R for 12 days. Western blot analysis showed that IL-6/IL-6R induced the expression of cathepsin B, Ets2, and p-STAT-3 (Ser⁷⁰⁵) that were independent of and comparable with M-CSF/RANKL treatment ( Figure 2E ). Furthermore, BMDM in response to M-CSF/RANKL and IL-6/IL-6R combination showed the marked upregulation of osteoclast-specific cathepsin K, cathepsin B, MITF, and Ets2 expression when compared with M-CSF/RANKL alone treatment ( Figure 2F ). Western blot analysis of joint homogenates prepared from naïve and AIA rats showed a marked increase in the expression of Ets2 along with other bone resorption markers including cathepsin K, cathepsin B, and RANKL ( Figure 2G ), suggesting a potential role of Ets2 in bone resorption in RA.

To characterize the composition of the secretory proteins in response to IL-6/IL-6R stimulation and determine if that reflects the signature of osteoclast-like proteins, the conditioned media obtained from human RASFs untreated or treated with M-CSF/RANKL, IL-6/IL-6R, or the combination of M-CSF/RANKL and IL-6/IL-6R (All) were used for the untargeted proteomic analysis using MS. Initial data analysis identified roughly 107 common secreted proteins from RASFs under various treatments; these are presented with differential spectral count in the heatmap ( Figure 3A ). Further analysis of proteomic data identified 113 proteins unique to the secretome of RASFs treated with IL-6/IL-6R, in contrast to only 73 proteins specific for M-CSF/RANKL stimulation ( Figure 3B ). An additional 27 proteins were identified in both M-CSF/RANKL- and IL-6/IL-6R-stimulated supernatants. Western blot analysis on the concentrated supernatants from these differentiated RASFs confirmed increased production of cathepsin B, cathepsin K, and RANKL in IL-6/IL-6R-activated secretome ( Figure 3C ). Overall, gene signatures of all secretome in response to M-CSF+RANKL, IL-6+IL-6R, and M-CSF+RANKL+IL-6+IL-6R are listed in Supplementary Table S2 ; followed by Gene Ontology studies of M-CSF+RANKL, IL-6+IL-6R, and MCSF+RANKL+IL-6+IL-6R which are also described in Supplementary Figure S3 . Focusing on the 113 proteins secreted explicitly by IL-6/IL-6R treatment of RASFs, we performed a Gene Ontology study using Metascape and Ingenuity Pathway Analysis ( Figures 3D, E ), revealing the metabolic changes in RASFs. Metascape analysis of the proteomics data identified the metabolic reprogramming of the RASFs as evident from the changes in metabolic pathway-related candidates such as carboxylic acid metabolism, oxidation, glycolysis, phagocytosis, stem cell differentiation, and ammonia metabolism ( Figure 3D ). In addition, the Ingenuity Pathway Analysis predicted the impact on tryptophan, isoleucine, and valine pathways, hematopoiesis from pluripotent stem cells, hepatic fibrosis, retinoid X receptor activation, endocytosis, and the complement system by IL-6/IL-6R treatment ( Figure 3E ). Details of the top 10 molecular networks predicted by IPA in response to IL-6/IL-6R treatment are mentioned in Supplementary Table S3 .

IL-6/IL-6R-initiated trans-signaling activates the JAK/STAT pathway (20), whereas M-CSF/RANKL drives MAPK-dependent signaling for cell survival and differentiation (21). To understand how early signaling events play a role in RASF reprogramming, we stimulated cells with M-CSF/RANKL and/or IL-6/IL-6R for 30 min. Western blot analysis of the cell lysates showed that IL-6/IL-6R selectively induced the activation of p-JAK1 (Tyr^1022/1023), p-JAK1 (Tyr^1007/1008), p-STAT3 (Tyr⁷⁰⁵), and p-STAT3 (Ser⁷²⁷), without activating NF-κBp65 in RASFs ( Figure 4A ). Interestingly, IL-6/IL-6R induced the expression of p-Akt, a potent cell survival and proliferation marker protein. While M-CSF/RANKL combination was ineffective in activating the JAK/STAT3 pathway itself, the combination partially suppressed IL-6/IL-6R-induced activation of JAK/STAT3 and Akt pathways ( Figure 4A ). Pretreatment of RASFs with tofacitinib or anti-IL-6 antibody showed not only the reduction in JAK/STAT signaling but also suppressed IL-6/IL-6R-induced activation of p-Akt in human RASFs ( Figure 4A ). Interestingly, anti-IL-6 antibody was more effective in reducing the phosphorylation of JAK1 while tofacitinib seems to be more effective on suppressing the phosphorylation of JAK2.

To understand if IL-6 trans-signaling influences M-CSF/RANKL-induced MAPK activation, we analyzed the same samples for proteins upstream of the MAPK pathway (IRAK1/IRAK4/TAK1) in RASF signaling (22). Western blot analysis showed that IL-6/IL-6R markedly induced phosphorylation of p-IRAK4 (Thr³⁴⁵/Ser³⁴⁶) and p-TAK1 (Thr^184/187) expression alone or in combination with M-CSF/RANKL in RASFs, which was not effectively inhibited by tofacitinib or anti-IL-6R antibody ( Figure 4B ), suggesting that IL-6/IL-6R and M-CSF/RANKL activation of the MAPK pathways are independent of its JAK-STAT signaling and may remain activated even with IL-6-targeted treatments. Quantitative analysis of TAK1 phosphorylation and STAT3 phosphorylation suggesting TAK1 is unaffected by IL-6 inhibitors ( Figures 4C, D ).

We hypothesized that Ets2 activity and binding in osteoclast-specific CTSK and CTSB gene promoters are mechanistically linked to the differentiation of human RASFs to an osteoclast-like phenotype. To address this hypothesis, we first verified Ets2 nuclear translocation in response to M-CSF/RANKL and/or IL-6/IL-6R combination in the nuclear extract prepared as previously described (23). We observed that M-CSF/RANKL increased the nuclear translocation of Ets2 by twofold (p < 0.05; n = 3) compared with the untreated samples ( Figure 5A ). Interestingly, IL-6/IL-6R increased Ets2 nuclear content by almost threefold, which was higher than M-CSF/RANKL-stimulated samples and did not increase further in M-CSF/RANKL and IL-6/IL-6R combination (All) treatment ( Figure 5A ; p < 0.01; n = 3). Ets2 nuclear localization loading control tubulin and lamin B shown in Supplementary Figure S4 . Aligned to this finding, pretreatment of RASFs with tofacitinib significantly inhibited the combination-triggered nuclear translocation of Ets2 to the level seen in unstimulated samples ( Figure 5A ; p < 0.01; n = 3), suggesting that JAK/STAT is a pivotal pathway to interfere in this process. We further utilized bioinformatic and ChIP-seq analysis to determine the core-binding motif (GGAA/T) for Ets2 and the transcriptionally active regions in CTSK and CTSB gene promoters. We observed four Ets2-binding sites in CTSK and five in CTSB within −500 bps upstream of respective promoters ( Figures 5B, C ). ChIP-seq analysis of previously published GSE31621 datasets further demonstrated that the genomic regions around −500 bps upstream of both the CTSK and CTSB gene are highly occupied with histone H3K27 acetylation (H3K27ac) modifications ( Figure 5B ). These high levels of the active chromatin marks around Ets2 binding indicate that Ets2 is directly involved in activating CTSK and CTSB gene expression and promoting differentiation. To further define the involvement of Ets2, we performed ChIP assay in RASFs treated with M-CSF/RANKL, IL-6/IL-6R, or M-CSF/RANKL and IL-6/IL-6R combination for 12 days. The results of the ChIP analysis showed that Ets2-immunoprecipitated DNA showed a significant increase in the occupancy of Ets2 protein in the CTSK promoter after treatment with M-CSF/RANKL, IL-6/IL-6R, and their combination when compared with control IgG ( Figure 5C ; n = 3). We also found similar enrichment in the CTSB promoter. Taken together, our promoter analysis and ChIP assay findings indicate that Ets2 binding is implicated in the osteoclast-like differentiation of RASFs. Furthermore, we performed the repeated knockdown of Ets2 on every 4th day for 12 days in IL-6/IL-6R-treated RASFs, followed by Western blot analysis to observe a reduced expression of IL-6/IL-6R-induced cathepsin B and cathepsin K, which confirms Ets2 transactivation of cathepsin genes ( Figure 5D ; p < 0.01, n = 3). Our observation of no change in expression of osteoclastic transcription factor MITF suggests that it is not directly involved in IL-6/IL-6R-induced reprogramming of RASFs to osteoclast-like phenotype. Interestingly, the reduced TRAP staining of RASFs on 12 days of IL-6/IL-6R treatment upon Ets2 knockdown further validated its role in the osteoclast-like phenotype of RASFs ( Figure 5E ).

Synovial tissue-derived RASFs possess a unique capability to dedifferentiate and have recently been identified as synovial mesenchymal stem cells (SMCs) (24). To understand how human RASFs respond to IL-6/IL-6R-induced osteoclast-like phenotype and functions compared with nondiseased human SFs (NLSFs), we treated both cell types with IL-6/IL-6R for 5 or 10 days in culture. Western blot analysis of the cell lysates showed selectively higher expression of osteoclast-specific markers (cathepsin K, cathepsin B, RANKL, and Ets2) by RASFs compared with the NLSFs, with no difference in p-STAT3 activation between two groups ( Figure 6A ). To our surprise, we observed a temporal decrease in the endogenous levels of gp130 and JAK1 in both NLSFs and RASFs treated with IL-6/IL-6R. When probed for the stem cell markers, RASF lysates showed markedly higher basal expression level of KLF4, Nanog, Oct4, and Sox-2, along with consistently higher response of these markers to IL-6/IL-6R stimulation compared with NLSFs ( Figure 6B ; p < 0.05). This suggests that RASFs are more susceptible than NLSFs to cytokine-driven metabolic reprogramming. Knockdown of Ets2 in RASFs stimulated with IL-6/IL-6R revealed that Ets2 might play a critical role in its invasive phenotype or osteoclast-like phenotype reprogramming by targeting Thy1, podoplanin, and RANKL expression ( Figures 6C, D ). Although IL-6/IL-6R did not induce the expression of Thy1, the knockdown of Ets2 abrogated the expression of Thy1 in untreated and IL-6/IL-6R-treated samples, suggesting its involvement in RASF heterogeneity beyond the regulatory role of IL-6 trans-signaling.