Positive and negative cooperativity of TNF and Interferon-γ in regulating synovial fibroblast function and B cell survival in fibroblast/B cell co-cultures

Synovial fibroblasts (SF) were reported to produce B cell activating factor (BAFF) in response to stimulation with interferon-γ (IFN-γ) or tumor necrosis factor (TNF). However, the influence of these pro-inflammatory cytokines on other receptors/ligands of the TNF superfamily or associated cytokine receptors in SF has not been investigated yet. Here we show the differential regulation of BAFF (CD257), Fn14 (CD266), TACI (CD267), BAFF-R (CD268), BCMA (CD269), CD40 ligand (CD40L, CD154), IFN-γR (CD119), Leptin receptor (ObR, CD295), VCAM-1 (CD106) and membrane TGF-β in isolated SF and the impact of IFN-γ/TNF co-incubation on proliferation, IL-6 and IL-8 production. In addition, the impact of differentially stimulated SF on B cell survival in co-cultures was assessed. Surface cytokines and cytokine receptors were detected by flow cytometry. Soluble cytokine receptors and cytokines were quantified by ELISA. Proliferation was assessed by cell titer blue. Murine B cell survival in fibroblast/ B cell co-cultures was determined by annexin V/propidium iodide staining and flow cytometry. IFN-γ together with TNF synergistically and significantly increased the cell surface levels of BAFF, Fn14, TACI, BAFF-R, BCMA, CD40L, ObR and IFN-γR in rheumatoid arthritis SF after 72 h incubation. Soluble BAFF was only induced by IFN-γ and inhibited by TNF. Addition of TWEAK had no influence on proliferation or IL-8 production but decreased TNF-induced IL-6 production, whereas APRIL, BAFF and leptin did not modulate TNF or TNF/IFN-γ-induced proliferation or cytokine production. Proliferation was increased by TNF and further enhanced by the addition of IFN-γ. In co-culture experiments, SF stimulated with TNF/IFN but not TNF or IFN-γ alone increased shedding of VCAM-1 and expression of membrane TGFβ, which was associated with reduced survival of murine B cells. IFN-γ and TNF regulate the expression of TNF family member cytokines and associated receptors. Ligation of IFN-γR and Fn14 under pro-inflammatory conditions modulated IL-6/IL-8 production and proliferation. In B cell/SF co-cultures, the combination of TNF/IFN reduced B cell survival possibly via enhanced VCAM-1 shedding and/or increased TGF-β production. IFN-γ is necessary for the observed effects on B cell survival and SF cytokine production and emphasizes its anti-inflammatory role in rheumatoid arthritis.

IFN-γ induces soluble BAFF, which in turn is inhibited by TNF. With the exception of IFN-γR, all TNF family member proteins investigated in this study are also shedded from the cell surface 19,20 , and therefore we assessed the amount of soluble BAFF, APRIL, CD40L, BCMA and TACI. In RASF stimulated with TNF, IFN-γ, or a combination of both cytokines for 72 h, no soluble APRIL, CD40L, BCMA and TACI was detected (data not shown). In contrast, BAFF production was dose-dependently induced by IFN-γ ( Fig. 2A) and inhibited by addition of TNF (p = 0.002; Fig. 2B).

Discussion
In this study, we demonstrated that the combination of TNF and IFN-γ up-regulated the surface expression of the cytokine receptors TACI, BCMA, BAFF-R, Fn14, IFN-γR, ObR, CD40L and also the membrane-bound form of the B cell survival cytokine BAFF in RASF. IFN-γ further enhanced TNF-induced proliferation of RASF while inhibiting IL-6 and IL-8 production. On the other hand, TNF reduced IFN-γ-induced BAFF production. Activation of Fn14 with TWEAK increased TNF-induced proliferation while it decreased TNF-induced IL-6 production by RASF. In RASF/murine B cell co-cultures we found RASF to strongly support B cell survival and this was further enhanced by the addition of BAFF. In addition, RASF stimulated with IFN-γ or TNF increased B cell survival while a combination of both reduced survival. Survival might be linked to the expression and increased shedding of VCAM-1 and membrane-bound TGF-β which are up-regulated by TNF/IFN-γ. To our knowledge, this is the first study that investigated the regulation of surface levels of BAFF, BAFF-R, BCMA, TACI, Fn14, CD40L, ObR and IFN-γR on isolated RASF. In contrast to results from Nakajima et al., we detected BAFF-R on the cell surface of RASF which was increased by TNF/IFN-γ 28 . BCMA and TACI have already been shown to be expressed on RASF and rat arthritic fibroblasts, respectively, but their regulation has not been investigated 14,15 . In line with data from Kamijo et al. we detected Fn14 on RASF and found that TNF alone and TNF/IFN-γ are inducers of its expression 29 . Since we used IFN-γ as one major stimulus, we were also interested in the regulation of its receptor and showed an increase in response to TNF/IFN-γ. This is in line with results from Alvaro-Gracia et al. who demonstrated an increase of IFN-γ binding after TNF stimulation on RASF 30 . Binding of IFN-γ to its receptor is usually associated with a downregulation of IFN-γR, but this does not occur in fibroblasts 31,32 . We also assessed the influence of IFN-γ and TNF on leptin recepter (ObR) expression and found this combination to increase ObR surface levels. Like IFN-γR, ObR signaling is initiated by JAK/STAT activation and some effects of IFN-γ are potentiated by leptin which has been associated with disease activity in RA 18,33 . In addition, TNF has been shown to increase surface levels of ObR by the activation of protein kinase C 34 . Functionally, leptin has already been shown to increase IL-6 levels in SF but we did not observe a potentiation of TNF or TNF/IFN-γ  www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ effects in RASF by leptin. In a next step, we investigated the expression of molecules important for B cell survival and activation in RASF. An indispensable co-stimulatory protein to activate B cells in a T cell dependent manner is CD40L, whose expression by RASF in contrast to CD40, has not been demonstrated directly. The ability of www.nature.com/scientificreports www.nature.com/scientificreports/ RASF to activate B cells and induce antibody class switching, however, suggests CD40L expression [35][36][37] . Indeed, we found CD40L on the surface of RASF and its levels were augmented by TNF/IFN-γ which might support B cell activation by RASF under these conditions. Another important molecule for B cell function is BAFF, which enhances the survival and class switching of (autoreactive) B cells 13,37 . We found an increase of membrane-bound BAFF on RASF after stimulation with TNF/IFN-γ, whereas soluble BAFF was induced only by IFN-γ and inhibited by TNF. This somewhat contrasts results from Reyes et al. that showed increased BAFF levels after TNF stimulation but supports data from Alsaleh et al. who demonstrated reduced BAFF expression in response to TNF 38,39 . However, reported TNF-induced BAFF levels in the study by Reyes were extremely low (below 10 pg/ml) and we considered such low amounts as baseline level in our study since control values for BAFF in unstimulated RASF were between 0 and 10 pg/ml. Since we found an upregulation of cytokine receptors TACI, BAFF-R, BCMA, ObR and IFN-γR after stimulation with TNF/IFN-γ, we next investigated whether ligands for these receptors have any impact on proliferation or cytokine production by RASF. We found IFN-γ to further enhance TNF-induced proliferation with a concomitant reduction in IL-6 and IL-8 production. With respect to RASF proliferation this contrasts results from Alvaro-Garcia et al., who showed reduced proliferation when TNF was combined with IFN-γ 40 . This might be due to different culture conditions since we used RPMI medium with only 2% FCS while the above mentioned study used DMEM with 10% FCS. We found only minimal basal proliferation using our culture conditions, with strong induction of proliferation upon addition of cytokines. Using less stringent culture conditions, where basal proliferation is already significant, IFN-γ might exert different effects. Additionally, Alvaro-Garcia et al. demonstrated that IFN-γ reduced collagenase expression in RASF, which supports our observations regarding IL-6 production, since collagenase and IL-6 are both induced by TNF and employ overlapping transcriptional regulation of mRNA synthesis 41,42 . In line with the inhibitory effects of IFN-γ on RASF, studies from Schlaak et al. and Ohta et al. showed reduced IL-6 production when IL-4 was combined with IFN-γ and diminished IL-8 production when IFN-γ and TNF were added together, respectively 43,44 . Our results demonstrated a dual role of IFN-γ on inflammation: It cooperates with TNF to up-regulate membrane-bound cytokines and receptors on RASF while concomitantly decreasing the release of cytokines IL-6 and IL-8. IFN-γ is usually produced in a TH1-biased environment during the development of the adaptive immune response and in this situation it might be detrimental to attract innate immune cells such as neutrophils that cannot support a T cell/ B cell response. In fact, it has been demonstrated that IFN-γ priming of macrophages before an LPS challenge reduced the production of mediators necessary for leukocyte recruitment 45 .
We further examined the impact of TWEAK, APRIL, BAFF and leptin on TNF and TNF/IFN-γ induced proliferation and IL-6/IL-8 production. TWEAK supported TNF but not TNF/IFN-γ induced proliferation and reduced IL-6 production only when induced by TNF but not by TNF/IFN-γ. While Yamana et al. also found TWEAK to reduce TNF-induced proliferation and IL-6 synthesis in RASF, TWEAK increased TNF-induced proliferation in our study 17 . In contrast to Yamana et al., we incubated RASF for 72 h instead of 48 h and included different TNF concentrations. In addition, like mentioned above with IFN-γ, we used only 2% FCS in culture medium for proliferation assays, which might change the effects of TWEAK from inhibitory to stimulatory. TWEAKs ability to promote proliferation is in line with in vivo results, where TWEAK has been shown to stimulate wound-healing 46,47 . APRIL modestly increased TNF-induced proliferation but not IL-6 or IL-8 production. In contrast, data from Nagatani et al. showed that APRIL fosters IL-6 mRNA expression in RASF, but this study investigated the effects of APRIL without concomitant stimulation with TNF 15 . In addition, the same study demonstrated significant APRIL mRNA production by RASF, which was not confirmed on protein level of soluble APRIL by our results or by Alsaleh et al. 37 . However, APRIL might be present in a membrane-bound form, which has been demonstrated on the surface of macrophages 48 . Overall, the effects of APRIL on RASF are weak and might only be relevant before the adaptive immune response takes over. BAFF exerted no influence on proliferation, IL-6 and IL-8 production of RASF. In HT1080 fibrosarcoma cells however, BAFF serves as an autocrine stimulator of basal NFκB activity by binding to intracellular BAFF-R 49 . Since we only employed surface staining of BAFF-R, there might still be a receptor pool located intracellularly. Addition of exogenous BAFF likely activates membrane-bound BAFF-R exclusively, which does not exclude a specialized role of intracellular BAFF in RASF, which needs to be defined in future studies.
Since we found synergistic and antagonistic effects of IFN-γ when combined with TNF on RASF, we were interested how RASF treated with these cytokines regulate B cell survival in co-cultures. We used murine B cells isolated from spleens as they do respond to human BAFF and CD40L, but only minimally to human TNF and not to human IFN-γ minimizing own effects of these cytokines [50][51][52][53] . Murine B cells cultured alone showed very little survival after 48 h, but survival was enhanced when conditioned medium of RASF was added, suggesting a pro-survival effect of a RASF-derived soluble factor. When murine B cells were cultured with RASF directly survival rates quadrupled compared to conditioned medium. This confirms results from Alasaleh et al. that showed that cell-cell contact is a prerequisite for class switch recombination in B cell RASF co-cultures and they identified BAFF as a major factor 37 . We also demonstrated that addition of BAFF alone or to co-cultures increased B cell survival and this was inhibited by the addition of the monoclonal anti-BAFF antibody belimumab. Although belimumab was considered to only block soluble BAFF, subsequent studies found that this antibody also inhibits membrane-bound BAFF 54 , and there are hints that inhibition of membrane BAFF might also be relevant in a clinical setting 55 . An effect on membrane bound BAFF might explain its inhibitory effect on B cells cultivated with unstimulated RASF, as they produce no soluble BAFF but do display membrane-bound BAFF under these conditions. We identified IFN-γ as the sole inducer of soluble BAFF in SF and therefore tested whether RASF stimulated with this cytokine increase B cell survival. Indeed, IFN-γ-stimulated RASF enhanced murine B cell survival and this effect was inhibited by belimumab, suggesting an influence of BAFF. Since membrane-bound BAFF was not modified by IFN-γ, soluble BAFF might be important as it is induced by IFN-γ. While TNF-stimulated RASF did not increase B cell survival, the combination of TNF and IFN-γ reduced survival compared to treatment with each cytokine alone. This was puzzling, since this combination was the strongest inducer of membrane-bound (2020) 10:780 | https://doi.org/10.1038/s41598-020-57772-7 www.nature.com/scientificreports www.nature.com/scientificreports/ BAFF which is supposed to be indispensable for B cell survival 12 . However, RASF not only support B cell survival by BAFF but they also express other proteins involved in regulating B cell function. Another regulator of B cell survival, vascular cellular adhesion molecule 1 (VCAM-1), which is the cellular receptor for integrin α4β1 on B cells, is expressed in RASF 56 . Interestingly, we found TNF and to a lesser extent TNF/IFN-γ to up-regulate cell-bound VCAM-1, but only the combination strongly induced soluble VCAM-1. This suggests that IFN-γ when combined with TNF induces the shedding of VCAM-1 possibly by regulating a specific enzyme e.g. ADAM17 23 . Soluble VCAM-1 might then compete with membrane-bound VCAM-1 on RASF inhibiting cell-cell contact and thereby diminishing the pro-survival effect of RASF. Another cytokine that regulates B cell function is TGF-β and it is produced by RASF 57 . We detected the membrane-bound form on RASF, which was up-regulated by TNF and even more so by the combination of TNF and IFN-γ. TGF-β regulates many aspects of B cell function from Ig production to apoptosis and an up-regulation on RASF might reduce B cell survival 26 .
In addition to RASF, we also analyzed SF from patients with osteoarthritis (OA) as a non-chronically inflamed control. In all assays, we didn't find any difference between OASF and RASF (data not shown). However, in vivo this might be completely different, since high levels of pro-inflammatory cytokines (e.g. TNF, IFN-γ) are only found in RA but not OA synovial tissue 58 . Hence, the in vitro observed up-regulation of receptors and cytokines of the TNF family in response to TNF/IFN-γ likely occurs in RASF exclusively.

conclusion
In this study we demonstrated synergistic effects of IFN-γ and TNF on the expression of TNF family member receptors, associated cytokine receptors and RASF proliferation. In contrast, cytokine production was negatively regulated by the combination of TNF and IFN-γ: While BAFF production was inhibited by TNF, IL-6 and IL-8 production was reduced by IFN-γ. Similarly, survival was increased by TNF or IFN-γ-stimulated RASF but reduced by the combination of both cytokines. These findings along with the shedding of VCAM-1 and the up-regulation of TGF-β by the combination of IFN-γ and TNF confirm a partial anti-inflammatory effect of IFN-γ in a TNF-driven pro-inflammatory environment in RA. In fact, it has already been shown that IFN-γR knockout mice show exacerbated arthritis 11 . Interestingly, in lupus erythematodes (SLE), IFN-γ is a prerequisite for disease development and this might be due to the cell type mainly responsible for disease initiation 59 . While lupus is mainly driven by the activity of autoreactive B cells, RA is dependent on T cell activation and innate mechanisms 60,61 . This is also evident in the therapeutic approach of both diseases, where RA profits from TNF inhibition, while SLE is combatted by B cell-depleting agents, like belimumab, whereas anti-TNF treatment exacerbates or even induces SLE. IFN-γ also induced soluble VCAM-1 when combined with TNF, which might, similar to soluble ICAM-1 alter leucocyte recruitment dependent on cell type and inhibit B cell attachment by disrupting VCAM1/integrin α4β1 interactions 62,63 . Additionally, increased membrane-bound TGF-β under the same culture conditions might not only alter B cell but also T cell function und support T-reg generation leading to an overall anti-inflammatory phenotype 64 .

Materials and Methods
patients. 28 patients with long-standing RA fulfilling the American College of Rheumatology revised criteria for RA 65 were included in this study. The RA group comprised of 21 females and 7 males with a mean age of 63.6 years ± 7.6 years and 67.3 ± 9.8 years, respectively. C-reactive protein was 21.9 mg/dl ± 33.4 mg/dl for females and 67.7 mg/dl ± 61.4 mg/dl for males. 20 out of 28 patients received non-steroidal anti-inflammatory drugs, 20 out of 28 glucocorticoids, 5 out of 28 methotrexate, 1 out of 28 sulfasalazine, 3 out of 28 biologicals and 1out of 28 JAK inhibitor. All patients underwent elective knee joint replacement surgery, and they were informed about the purpose of the study and gave written consent. The study was approved by the Ethics Committees of the University of Düsseldorf (approval number 2018-87-KFogU) and Regensburg (approval number 15-1 01-021). We confirm that all experiments were performed in accordance with relevant guidelines and regulations.
Antibodies and cytokines. All antibodies used in this study are listed in Table 1, cytokines are listed in Table 2. Antibodies used in this study (Table 1): www.nature.com/scientificreports www.nature.com/scientificreports/ Antibodies were used in a 1:11 dilution as recommended by the supplier. Cytokines used in this study ( Table 2): Synovial fibroblast and tissue preparation. Samples from RA synovial tissue were collected immediately after opening the knee joint capsule and tissue was prepared for cell isolation thereafter 66 . In brief, excised synovial tissue was cut into small fragments, minced and treated with liberase (Roche Diagnostics, Mannheim, Germany) at 37 °C overnight. The cell suspension was filtered (70 µm) and centrifuged at 300 g for 10 min. After that, the pellet was treated with erythrocyte lysis buffer (20.7 g NH 4 Cl, 1.97 g NH 4 HCO 3 , 0.09 g EDTA ad 1 l H 2 O) for 5 min, recentrifuged for 10 min and then resuspended in RPMI-1640 (Sigma Aldrich, St. Louis, USA) with 10% FCS. After overnight incubation, RPMI medium was replaced with fresh medium to wash off dead cells and debris.
Stimulation of RA synovial fibroblasts. 6000 cells were seeded onto 96 well microtiter plates, grown for three days and were then stimulated with either TNF (0.1-10 ng/ml), IFN-γ (0.1-10 ng/ml) or a combination (IFN-γ 10 ng/ml + TNF 10 ng/ml) for 72 h (in RPMI medium containing 2% FCS to minimize proliferation; for all assays). After that, cells or supernatants were used for flow cytometry (cells) and ELISA (supernatants). IL-6, IL-8, sVCAM-1, sCD40L, sTACI, sBCMA, BAFF and APRIL ELISA. Cell culture supernatants were used for ELISAs 72 h after addition of respective stimulants. Tests were conducted as described by the supplier (BD, OptEIA, Heidelberg, Germany (IL-6, IL-8) and R&D/Biotechne, Wiesbaden, Germany (sVCAM-1, sCD40L, sTACI, sBCMA, APRIL, BAFF)). Inter-and intraassay coefficient of variation was below 10%. proliferation of RASf. Proliferation was assessed by the cell titer blue viability assay (promega). We analyzed proliferation in RPMI medium containing only 2% FCS. This was important since FCS contains up to 28 µM lysophosphatidic acid (LPA). LPA has been reported to induce proliferation, calcium mobilization and IL-8 production in a wide variety of cell types 67,68 and therefore, we minimized this influence by only using 2% FCS. flow cytometry. Data were acquired on a MACSQuant analyzer 10 (Miltenyi, Bergisch Gladbach, Germany). After detachment of RASF with accutase (Life-Technologies/Thermo, #A11105-01) at 37 °C for 20 min, RASF were incubated with primary labelled antibodies for 10 min at 4 °C in PBS/1% BSA and analyzed thereafter. Cells were single-stained with respective antibodies. isolation of murine B cells. Murine B cells were isolated from spleens using the B Cell Isolation Kit mouse (miltenyi #130-104-443) according to the suppliers' instructions.
Animals. Male DBA/1 J mice (6-8 weeks old) were purchased from Janvier (France). They had unrestricted access to chow and water. Five animals lived in one cage, and they were acclimated to the environment for one week before commencement of experiments. The University of Düsseldorf approved all experiments according to institutional and governmental regulations for animal use (project: O57/15). All procedures were performed in accordance with the German Animal Welfare Act and the European Directive 2010/63/EU on the protection of animals used for scientific purposes.
RASf co-culture with murine B cells. Co-culture experiments were performed in 48 well plates (Cellstar, Greiner bio-one, Kremsmünster, Austria). In brief, 20.000 RASF were seeded in 750 µl RPMI-1640 (Sigma-Aldrich) and grown for 72 h. Then, 250.000 isolated murine B cells were added for 48 h and cells were stimulated with cytokines as indicated. After that, B cell survival and apoptosis were assessed by Annexin V/ propidium iodide staining.
Apoptosis assay. B cell survival and apoptosis were assessed using the FITC Annexin V assay kit (BD biosciences, Heidelberg, Germany) according to manufacturers' instructions.
Statistical analysis. Statistical analysis was performed with SigmaPlot 13 (Systat Software Inc., San Jose, USA) and SPSS 25 (IBM, Armonk, USA). The statistic tests used are given in the figure legends. When data are presented as box plots, the boxes represent the 25th to 75th percentiles, the lines within the boxes represent the median, and the lines outside the boxes represent the 10th and 90th percentiles. When data are presented as line plots, the line represents the mean and error bars are depicted as standard error of mean. The level of significance was p < 0.05.