Deletion of activin A in mesenchymal but not myeloid cells ameliorates disease severity in experimental arthritis.

OBJECTIVE: The aim of this study was to assess the extent and the mechanism by which activin A contributes to progressive joint destruction in experimental arthritis and which activin A-expressing cell type is important for disease progression.
METHODS: Levels of activin A in synovial tissues were evaluated by immunohistochemistry, cell-specific expression and secretion by PCR and ELISA, respectively. Osteoclast (OC) formation was assessed by tartrat-resistant acid phosphatase (TRAP) staining and activity by resorption assay. Quantitative assessment of joint inflammation and bone destruction was performed by histological and micro-CT analysis. Immunoblotting was applied for evaluation of signalling pathways.
RESULTS: In this study, we demonstrate that fibroblast-like synoviocytes (FLS) are the main producers of activin A in arthritic joints. Most significantly, we show for the first time that deficiency of activin A in arthritic FLS (ActβAd/d ColVI-Cre) but not in myeloid cells (ActβAd/d LysM-Cre) reduces OC development in vitro, indicating that activin A promotes osteoclastogenesis in a paracrine manner. Mechanistically, activin A enhanced OC formation and activity by promoting the interaction of activated Smad2 with NFATc1, the key transcription factor of osteoclastogenesis. Consistently, ActβAd/d LysM-Cre hTNFtg mice did not show reduced disease severity, whereas deficiency of activin A in ColVI-Cre-expressing cells such as FLS highly diminished joint destruction reflected by less inflammation and less bone destruction.
CONCLUSIONS: The results highly suggest that FLS-derived activin A plays a crucial paracrine role in inflammatory joint destruction and may be a promising target for treating inflammatory disorders associated with OC formation and bone destruction like rheumatoid arthritis. © Author(s) (or their employer(s)) 2022. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.

High levels of activin A in serum, synovial fluid and synovial tissue of patients with rheumatoid arthritis (RA) were observed.

Activin A promotes osteoclast development.

Fibroblast-like synoviocytes (FLS) seem to be the main producer of activin A in arthritic joints.

Deletion of activin A in FLS but not in myeloid cells highly diminishes joint pathology in experimental arthritis displayed by less inflammation and less bone destruction, confirming for the first time a significant role for activin A in arthritis development in vivo.

Blockade of activin A or corresponding signalling pathways may be a promising treatment option for diseases associated with inflammatory bone destruction such as RA.

Introduction

Rheumatoid arthritis (RA) is a common type of inflammatory arthritis characterised by chronic inflammation, culminating in destruction of the joint. In the pathological state of RA, the synovial lining layer becomes hyperplastic due to influx of macrophage-like synoviocytes and increased cell division of fibroblast-like synoviocytes (FLS).1–3 The continuous presence of cytokines, chemokines, growth factors and other molecular mediators leads to the activation of FLS resulting in an aggressive tumor-like transformation4 5 contributing to pannus formation, a highly destructive tissue located at the interface between synovium, cartilage and bone. FLS and macrophages predominate this tissue and mediate the process of joint destruction by expression of degradative enzymes including matrix metalloproteinases and collagenases.6

However, bone destruction is mediated by osteoclasts (OCs) located at the interface between the pannus and periarticular bone surface.7 Many factors such as tumour necrosis factor alpha (TNF-α), interleukin (IL)-1, IL-6, IL-17 and receptor activator of nuclear factor κB ligand (RANKL) produced directly by FLS or within the pannus tissue by infiltrated immune cells are able to enhance the differentiation of cells of the monocyte–macrophage lineage into OC, thereby disturbing the balance of bone remodelling and shifting the remodelling process towards bone resorption.7–13

Another factor shown to influence bone destruction in RA by directly promoting OC formation is activin A.14–19 Activin A belongs to the transforming growth factor beta (TGF-β)-like group of the TGF-β superfamily consisting of two disulfide-linked inhibin βA subunits.20–22 Interestingly, proinflammatory cytokines such as TNF-α, IL-1β and TGF-β have been found to increase levels of activin A in RA.23 24 In accordance with this, high levels of activin A in serum, synovial fluid and synovial tissue of patients with RA, compared with patients with osteoarthritis (OA), were observed.25–28 Besides FLS, activin A is produced also by macrophages29 and not only is induced by proinflammatory cytokines but in turn also stimulates the production of inflammatory mediators like TNF-α, IL-1β, IL-6, nitric oxide and prostaglandin E2,25 28 30 31 suggesting an important role of activin A in the pathology of RA. However, the extent to which activin A influences arthritis development and progression in vivo has not been elucidated so far, and even the most significant activin A-producing cell type involved in disease progression has not yet been identified. Therefore, the aim of this study was to investigate whether activin A contributes to joint inflammation and progressive bone loss in experimental arthritis and which activin A-expressing cell type within the inflamed joint is important for disease progression.

Methods

Detailed experimental procedures are described in the online supplemental material.

Results

Increased levels of activin A in arthritis

Immunofluorescence staining revealed a higher expression of activin A in the synovial tissues of patients with RA compared with those of patients with OA (2.7-fold). Whereas in OA specimens activin A expression appeared mainly in cells of the synovial lining layer, activin A expression in RA was observed throughout the whole synovial tissue (figure 1A). More detailed analysis demonstrated an increase of activin A-expressing macrophages (2.7-fold), FLS (5.6-fold) and neutrophils (2.6-fold), whereby the latter one was not significant (figure 1C). Consistently, a TNFα-dependent chronic arthritis mouse model (hTNFtg) revealed dramatically increased levels of activin A in the hind paws (3.7-fold, figure 1B) and in sera (7.3-fold, figure 1D) compared with wild-type (WT) mice. Moreover, mRNA of the inhibin βA subunit was abundantly expressed in RA FLS as well as in FLS of arthritic mice (figure 1E). Together, these data indicate that under inflammatory conditions, activin A appears to be highly upregulated and that FLS may be the main producers of activin A within the inflamed synovial tissue.

Figure 1

High levels of activin A under inflammatory conditions. (A) representative images of fluorescence stainings of activin A in synovial tissue samples obtained from patients with OA and RA and corresponding quantification (n=6, unpaired t-test**). (B) Representative images of fluorescence stainings of activin A in hind paws of WT and arthritic hTNFtg mice and corresponding quantification (n=5–6, unpaired t-test**). (C) Representative fluorescence costainings of activin A with CD68, CD90 or MPO in synovial tissue samples obtained from patients with OA and RA and quantification of corresponding activin A-positive cells (n=6, unpaired t-test*,**). (D) Activin A concentrations in serum of WT and arthritic hTNFtg mice. Data represent means±SEM (n=7, unpaired t-test****). (E) PCR analysis of inhibin βA-subunit mRNA in FLS of patients with RA (n=4) and of hTNFtg mice (n=4). (F) Secretion of activin A by WT and (G) hTNFtg FLS on stimulation with IL-1α (20 ng/mL), IL-1β (20 ng/mL), TGF-β1 (20 ng/mL) and IL-17A (20 ng/mL) for 48 hours. (H) Comparison of activin A secretion by WT and hTNFtg FLS. (I) Secretion of activin A by WT and (J) hTNFtg BMDMs on stimulation with IL-1α (20 ng/mL), IL-1β (20 ng/mL), TGF-β1 (20 ng/mL) and IL-17A (20 ng/mL) for 48 hours. (K) Comparison of activin A secretion by WT and hTNFtg BMDMs. All data are means±SEM (n=3, paired t-test, comparison WT/hTNFtg unpaired t-test*). *P≤0.05, **P≤0.01, ****P≤0.0001. Ctrl, control; FLS, fibroblast-like synoviocytes; IL, interleukin; ns, not significant; OA, osteoarthritis; RA, rheumatoid arthritis; WT, wild type.

Since an inflammatory environment obviously causes an increase in activin A levels, WT and hTNFtg mouse FLS were stimulated with the proinflammatory cytokines IL-1α, IL-1β, TGF-β, and IL-17A. Indeed, all cytokines were able to significantly enhance the secretion of activin A compared with untreated controls (4.3-fold to 9.7-fold higher) with the exception of IL-17A, where the secretion of WT FLS was identical to the untreated control (figure 1F). Moreover, hTNFtg FLS showed an even stronger increase in activin A secretion on stimulation with proinflammatory cytokines compared with WT FLS (11.6-fold to 17.5-fold higher, figure 1G). Because of the increased number of activin A-producing macrophages in RA and hTNFtg synovial tissue, regulation of activin A by inflammatory cytokines was additionally analysed in bone marrow-derived macrophages (BMDM). In contrast to FLS, a distinct increase in activin A secretion by BMDMs was exclusively observed on stimulation with TGF-β1, and no significant differences in secretion were found between WT (increase by 4.1) and hTNFtg (increase by 6.5) (figure 1I–K). It should be stressed that BMDMs secreted considerably lower amounts of activin A than FLS (pg/mL vs ng/mL range; figure 1H, K), further supporting the hypothesis that FLS are the main producers of activin A within the inflamed synovium.

Activin a highly enhances OC formation and activity

To investigate the impact of activin A on osteoclastogenesis, primary BMDMs isolated from WT mice were differentiated into OC by stimulation with RANKL, activin A or both (figure 2A). Interestingly, concomitant treatment of BMDMs with RANKL and activin A led to enhanced OC differentiation (2.6-fold, figure 2B) associated with a higher number of nuclei per OC (3.6-fold, figure 2C) as well as an increased OC size (2.5-fold, figure 2D), leading to an increased total OC area (6.9-fold, figure 2E) compared with RANKL treatment alone. Of note, activin A alone was not able to induce OC differentiation (figure 2A). Consistently, resorption analysis revealed an increased formation of resorption pits (figure 2F, G) and total resorption area (figure 2H) on calcium phosphate plates on differentiation with both RANKL and activin A simultaneously compared with the RANKL-treated control (2.6-fold and 5.9-fold, respectively). Most interestingly, the additional increase in resorption area per pit by 2.3-fold compared with the RANKL-treated control clearly demonstrate that the observed higher resorption is not only due to higher OC numbers but also additionally based on increased OC activity (figure 2I). Finally, figure 2J exclude that the higher number of OC on stimulation with activin A was dependent on a higher number of BMDMs or available OC precursors provoked by proliferation.

Figure 2

Enhanced RANKL-mediated differentiation and activity of OCs by activin A. (A) Representative images of TRAP staining after 4 days of OC differentiation in the presence of 30 ng/mL macrophage colony-stimulating factor (M-CSF, control) together with activin A (30 ng/mL) or RANKL (50 ng/mL) or RANKL plus activin A (scale bar 100 µm). (B) Corresponding OC numbers, (C) number of nuclei per OC, (D) OC size and (E) total OC area per well (n=4). (F) Representative images of resorption pit formation of WT BMDMs after 6 days of OC differentiation using calcium phosphate as substrate on stimulation with RANKL or RANKL plus activin A (scale bar 500 µm). (G) Number of resorption pits, (H) total resorption area and (I) resorption area per pit after 6 days of OC differentiation (n=3). (J) Cell numbers of BMDMs with or without activin A stimulation after 1, 2 and 3 days (n=3). (K) Representative TRAP staining of human OCs after 15 days of differentiation in the presence of M-CSF (30 ng/mL) and RANKL (50 ng/mL) with and without 100 ng/mL activin A (scale bar 200 µm) and corresponding OC number and OC size. All data are means±SEM (t-test). *P≤ 0.05, **P≤0.01, ***P≤ 0.001. OC, osteoclast; RANKL, receptor activator of nuclear factor κB ligand.

Altogether, these data indicate that activin A strongly enhances the RANKL-mediated osteoclastic resorption due to increased numbers as well as increased activity of in vitro differentiated primary OCs. In order to bridge to humans, the effects of activin A on human OC differentiation were investigated. Consistent with the effects of activin on murine osteoclastogenesis, activin A also enhanced the RANKL-induced differentiation of human peripheral blood mononuclear cells (PBMCs) into OCs by about 5-fold, associated again with an increase in OC size by about 6-fold, indicating that activin A must be considered an important cross-species factor in OC differentiation (figure 2K).

Since both FLS as well as macrophages produce activin A, we next wanted to unravel whether autocrine or paracrine activin A is important for joint destruction in arthritis. To this end, we compared the impact of activin-deficiency in myeloid cells and FLS on the formation of OCs in vitro as well as on arthritis development and progression in vivo.

Lack of activin A in myeloid cells has no influence on OC differentiation

To confirm a myeloid lineage-specific as well as effective deletion of activin A in the ActβAd/d LysM-Cre mice, activin A secretion by FLS (figure 3A) and myeloid cells (figure 3B) was measured by ELISA. As expected, ActβAd/d LysM-Cre FLS showed no changes in activin A secretion compared with FLS from ActβAflox/flox (ActβAf/f) mice, whereas a significant, almost complete reduction of activin A secretion by BM cells, BMDMs, pre-OCs (pOCs) and OC of ActβAd/d LysM-Cre mice was observed, indicating that activin A is effectively deleted in cells of the myeloid lineage but not in mesenchymal cells such as FLS.

Figure 3

Deficiency of activin A in BMDMs had no impact on OC differentiation in vitro. (A) Secretion of activin A by FLS from ActβAf/f and ActβAd/d LysM-Cre mice after 48 hours. (B) Secretion of activin A by BM cells, BMDMs, pOCs and OCs from ActβAf/f and ActβAd/d LysM-Cre mice after 48 hours. BM cells were not stimulated; BMDMs were stimulated with M-CSF (30 ng/mL) for 3 days; pOCs were stimulated for 3 days with M-CSF (30 ng/mL) followed by stimulation with M-CSF and RANKL (50 ng/mL) for further 2 days. OCs were generated by stimulation of pOCs for a further 2 days with RANKL. All data are means±SEM (n=3–5, Mann-Whitney U test*,**). (C) Representative images of TRAP staining after 5 days of differentiation of BMDMs from ActβAf/f and ActβAd/d LysM-Cre mice (scale bar 100 µm). (D) Corresponding OC number and total OC area per well. All data are means±SEM (n=4, Mann-Whitney U test). (E) Representative images of TRAP staining after 5 days of differentiation of BMDMs from ActβAf/f hTNFtg and ActβAd/d LysM-Cre hTNFtg mice (scale bar 100 µm). (F) Corresponding OC number and total OC area per well. All data are means±SEM (n=4 and 3, respectively; Mann-Whitney U test). *P≤0.05, **P≤0.01. ActβAf/f, ActβAflox/flox; FLS, fibroblast-like synoviocytes; IL, interleukin; OC, osteoclast; RANKL, receptor activator of nuclear factor κB ligand.

To analyse whether activin A deficiency in myeloid cells affects osteoclastogenesis in vitro, BMDMs from ActβAf/f and ActβAd/d LysM-Cre mice as well as from ActβAf/f hTNFtg and ActβAd/d LysM-Cre hTNFtg mice were stimulated with RANKL and subsequently stained for TRAP after 5 days of differentiation (figure 3C, E). Of note, the deficiency of activin A in myeloid cells did not show significant differences in OC number and area neither from non-arthritic (figure 3D) nor from arthritic mice (figure 3F) which leads to the assumption that autocrine activin A is not important for OC differentiation. Because activin levels were rather low, cells were additionally stimulated with TGF-ß1, previously shown to enhance activin secretion in BMDMs (figure 1I, J). However, additional stimulation with TGF-ß1 had no effect on OC formation (online supplemental figure 1). Moreover, to exclude whether residual activin A, due to incomplete deletion in LysM-Cre BMDMs, may interfere with the effects of deletion, cells were additionally treated with activin A antibodies. Indeed, blocking activin A did not influence OC development of ActβAd/d LysM-Cre BMDMs, whereas blocking of myostatin, another member of the TGF-ß superfamily and known to stimulate OC formation, led to decreased OC differentiation (online supplemental figure 4).

Deficiency of activin A in LysM-Cre-expressing cells does not significantly affect disease severity in hTNFtg mice

Taking the involvement of activin A in bone remodelling and inflammation into account, we first asked whether the loss of activin A in cells of the myeloid lineage would influence the development of inflammatory bone destruction in arthritis. Because the complete knockout of activin A is lethal,32 ActβAf/f mice were used to generate conditional knockouts by breeding these mice with the LysM-Cre mouse line.33 34 The resulting mice with a cell-specific deletion of activin A (ActβAd/d LysM-Cre) were subsequently cross-bred with hTNFtg mice to generate deleted arthritic mice. hTNFtg mice are overexpressing the hTNFα transgene and thereby develop a chronic destructive arthritis that shares many characteristics with human RA.35

As expected, ActβAf/f hTNFtg and ActβAd/d LysM-Cre hTNFtg mice showed an increase in paw swelling (figure 4A) and loss of grip strength (figure 4B) during disease development. However, no significant differences in the clinical symptoms between the two hTNFtg genotypes could be observed during the course of the disease. ActβAf/f and ActβAd/d LysM-Cre showed no signs of arthritis.

Figure 4

Deficiency of activin A in LysM-Cre-expressing cells does not substantially affect disease severity in hTNFtg mice. (A) Paw swelling and (B) grip strength measured in ActβAf/f (n=6), ActβAd/d LysM-Cre (n=6), ActβAf/f hTNFtg (n=6) and ActβAd/d LysM-Cre hTNFtg (n=6) mice over 5–12 weeks. All data are means±SEM (two-way analysis of variance, Bonferroni’s multiple comparison test). (C) Representative images of µCT analysis from the front and back (n=4, each genotype), (D) toluidine blue-stained sections (scale bar 200 µm) and (E) TRAP-stained sections (scale bar 200 µm) from the hind paws of 12-week-old ActβAf/f (n=3), ActβAd/d LysM-Cre (n=3), ActβAf/f hTNFtg (n=6) and ActβAd/d LysM-Cre hTNFtg (n=7) mice. Quantitative histomorphometric assessment of (F) synovial pannus formation, (G) bone erosion and (H) number of OCs in tarsal joints. All data are means±SEM (Mann-Whitney U test). µCT, micro-CT; ActβAf/f, ActβAflox/flox; OC, osteoclast.

Moreover, micro-CT (µCT) and histomorphometry showed no obvious differences in joint destruction as well as OC numbers in 12-week-old arthritic mice with deletion of activin A compared with the hTNFtg mice (figure 4C–E). Quantitative morphometric evaluation confirmed the lack of significant differences in inflammation, bone erosion and OC numbers (figure 4F–H and online supplemental figure 6). However, a tendency towards less inflammation (31.4 %) in the hind paws of activin A-deleted compared with non-deleted hTNFtg mice could be observed. Moreover, evaluation of trabecular and cortical bone parameters in non-arthritic ActβAf/f, LysM-Cre and ActβAd/d LysM-Cre mice revealed that activin A also does not influence physiological bone remodelling (online supplemental figure 2). Of note, activin A serum levels were not significantly reduced, although a tendency towards lower levels (about 27%) could be observed in ActßAd/d LysM-Cre hTNFtg mice (online supplemental figure 3A).

Lack of activin A in FLS effectively decreases OC differentiation

In order to analyse the deletion efficiency and specificity of activin A in FLS, secretion of activin A by FLS and myeloid cells was measured via ELISA. As anticipated, ActβAd/d ColVI-Cre FLS showed a significant reduction in activin A secretion by about 85% compared with FLS from ActβAf/f mice (figure 5A). In contrast, activin A secretion by myeloid lineage cells from ActβAd/d ColVI-Cre mice were not significantly reduced compared with those from ActβAf/f mice (figure 5B).

Figure 5

Deficiency of activin A in FLS significantly reduces OC formation. (A) Secretion of activin A by FLS from ActβAf/f and ActβAd/d ColVI-Cre mice after 48 hours. (B) Secretion of activin A by BM cells, BMDMs, pOCs and OCs from ActβAf/f and ActβAd/d ColVI-Cre mice after 48 hours. BM cells were not stimulated; BMDMs were stimulated with M-CSF (30 ng/mL) for 3 days; pOCs were stimulated for 3 days with M-CSF (30 ng/mL) followed by stimulation with M-CSF and RANKL (50 ng/mL) for a further 2 days. OCs were generated by stimulation of pOCs for a further 2 days with RANKL. All data are means±SEM (n=3–5, Mann-Whitney U test**). (C) Representative images of TRAP staining after 5 days of differentiation of BMDMs from ActβAf/f and ActβAd/d ColVI-Cre mice (scale bar 100 µm). (D) Corresponding OC number and total OC area per well. All data are means±SEM (n=4, Mann-Whitney U test). (E) Representative images of TRAP staining after 5 days of differentiation of BMDMs from ActβAf/f hTNFtg and ActβAd/d LysM-Cre hTNFtg mice (scale bar 100 µm). (F) Corresponding OC number and total OC area per well. All data are means±SEM (n=4, Mann-Whitney U test). (G) Representative fluorescence images of WT-EGFP BMDMs cocultured with FLS from ActβAf/f hTNFtg and ActβAd/d ColVI-Cre hTNFtg mice for 5 days to induce OC differentiation in presence of 1 µM PGE2 (scale bar 100 µm). (H) Corresponding OC number and total OC area per well. All data are means±SEM (ActβAf/f hTNFtg n=4, ActβAd/d ColVI-Cre hTNFtg n=5, unpaired t-test**). **P≤0.01. FLS, fibroblast-like synoviocytes; OC, osteoclast; RANKL, receptor activator of nuclear factor κB ligand.

As expected, no differences in OC formation between BMDMs from ActβAf/f and ActβAd/d ColVI-Cre mice,36 neither of non-arthritc nor of arthritic mice, was observed (figure 5C–F). In contrast, cocultures of BMDMs with FLS from ActβAd/d ColVI-Cre hTNFtg FLS led to the formation of smaller and less OCs compared with the cocultures with ActβAf/f hTNFtg FLS, which generated a higher number of OCs that were also larger in size (figure 5G). Quantitative analysis revealed that the number of OCs and the total OC area were significantly decreased in cocultures with ActβAd/d ColVI-Cre hTNFtg FLS by 25.6% and 36.7 %, respectively, compared with the ActβAf/f hTNFtg FLS, suggesting that activin A is an important paracrine factor for OC formation (figure 5H).

Of importance, FLS do not only express activin A but also myostatin, another member of the TGF-ß superfamily and also a promoting factor for OC differentiation.37 To clarify the relative contribution of activin and myostatin on OC development, cocultures of arthritic FLS with BMDM in the presence of blocking antibodies against activin A or myostatin were performed. Cocultures with FLS from ActβAf/f hTNFtg mice treated with antibodies against activin A showed a reduction of OC development by about 48%, whereas treatment with antibodies against myostatin showed a tendency towards reduced OC formation (about 20%), pointing to a more important role of paracrine activin A. As expected, the already decreased osteoclastogenesis observed in cocultures with ActβAd/d ColVI-Cre hTNFtg FLS could not be further reduced by anti-activin AB but interestingly, also no further reduction could be seen on treatment with anti-myostatin AB, confirming that activin A is the determining factor in FLS-mediated OC differentiation (online supplemental figure 5).

Taken together, these data indicate that less bone destruction in ActβAd/d ColVI-Cre hTNFtg mice is caused by reduced OC formation due to less activin A secretion by FLS.

Deficiency of activin A in ColVI-Cre-expressing cells highly ameliorates inflammation and bone destruction in hTNFtg mice

Next, we investigated whether a deletion of activin A in mesenchymal cells via ColVI-Cre has an effect on arthritis severity in vivo. To this end, ActβAd/d ColVI-Cre mice were cross-bred with hTNFtg mice to generate conditional deleted arthritic mice. Again, no significant differences in the clinical scores between ActβAf/f hTNFtg and ActβAd/d ColVI-Cre hTNFtg mice was observed. As expected, ActβAf/f and ActβAd/d ColVI-Cre showed no signs of arthritis (figure 6A, B). Evaluation of serum confirmed significantly reduced activin A levels by about 68% in mice with SF-specific activin A deletion compared with non-deleted arthritic mice (online supplemental figure 3A).

Figure 6

Deficiency of activin A in ColVI-Cre-expressing cells ameliorates disease severity in hTNFtg mice. (A) Paw swelling and (B) grip strength measured in ActβAf/f (n=6), ActβAd/d ColVI-Cre (n=6), ActβAf/f hTNFtg (n=6) and ActβAd/d ColVI-Cre hTNFtg (n=6) mice over 5–12 weeks. All data are means±SEM (two-way analysis of variance, Bonferroni’s multiple comparison test). (C) Representative images of µCT analysis from the front and back (n=4, each genotype), (D) Toluidine blue-stained sections (scale bar 200 µm) and (E) TRAP-stained sections (scale bar 200 µm) from the hind paws of 12-week-old ActβAf/f (n=3), ActβAd/d ColVI-Cre (n=3), ActβAf/f hTNFtg (n=6) and ActβAd/d ColVI-Cre hTNFtg (n=6–8) mice. Quantitative histomorphometric assessment of (F) synovial pannus formation, (G) bone erosion and (H) number of OCs in tarsal joints. All data are means±SEM (Mann-Whitney U test). *P≤0.05, **P≤0.01. µCT, micro-CT; OC, osteoclast.

Most importantly, µCT and histomorphometry at the age of 12 weeks (figure 6C–E), as well as its quantitative assessment demonstrated a strong reduction of inflammation displayed by less formation of pannus tissue (46.3 %) and a distinct reduction of bone erosion (33.8 %) associated with a significant reduction in the number of OCs (46.2 %) in ActβAd/d ColVI-Cre hTNFtg mice compared with ActβAf/f hTNFtg mice (figure 6F–H and online supplemental figure 6). In consideration of the fact that activin A also stimulates the production of central players in inflammation such as IL-1 and IL-6,28 30 corresponding serum levels in activin A-deleted and non-deleted arthritic mice were assessed. Indeed, deletion of activin A caused a strong reduction in IL-1 and IL-6 serum levels (51.7% and 47.5%, respectively), confirming an anti-inflammatory effect of activin A via the downregulation of proinflammatory cytokines (online supplemental figure 3B). Assuming that a deletion of activin A in ColVI-Cre-expressing cells within the joint will mainly led to a deletion in FLS, the results strongly indicate that FLS-derived activin A plays an important role in hTNFα-driven RA-like pathogenesis in mice.

Activin A enhances expression of key differentiation genes via Smad2-dependent nuclear translocation of NFATc1

To shed light on the mechanisms by which activin A exerts its effect on OC formation, we focused on potential signalling pathways and the expression of the key markers for OC differentiation.

Western Blot analysis of P-JNK, P-ERK and P-p38 demonstrated that RANKL, but not activin A, induces MAP kinase activation (figure 7A), phosphorylation of NF-κB and degradation of IκBα (figure 7B), indicating that the stimulatory effect of activin A on osteoclastogenesis did not originate from increased MAP kinase and nuclear factor kappa B (NF-κB) signalling. In consideration of the fact that TGF-β-like proteins activate the Smad2/3 signalling pathway, the ability of activin A to activate Smad2 in primary BMDMs was verified. Activin A induced Smad2 phosphorylation after 30 min, while RANKL had no effect on Smad2 activation (figure 7C). Of note, Smad2 phosphorylation was still present on stimulation with activin A for 3 and 4 days (figure 7D). Due to the enhanced osteoclastogenesis associated with activated Smad2 signalling by activin A, it was speculated that activated Smad2 may interact with the key transcription factor in OC differentiation NFATc1, which then could enhance the expression of a set of genes that are essential for osteoclastogenesis. To demonstrate that Smad2 and NFATc1 interact on stimulation with RANKL and activin A, coimmunoprecipitation experiments were performed. As shown in figure 7E, NFATc1 was clearly bound to P-Smad2 on costimulation of BMDMs with RANKL and activin A. Subsequent analyses of the key differentiation makers integrin αν, integrin β3, DC-STAMP, NFATc1 and cathepsin K showed that RANKL induced the expression of all these differentiation markers, whereas activin A alone did not induce any of these marker genes. However, activin A was clearly able to enhance the RANKL-induced expression of all differentiation markers during osteoclastogenesis (figure 7F).

Figure 7

Activin A enhances RANKL-mediated OC formation and expression of OC differentiation markers by activation of Smad2. (A) Representative immunoblots of p-JNK, P-ERK and P-p38 of WT BMDMs stimulated with RANKL (50 ng/mL), activin A (30 ng/mL) or RANKL plus activin A in the presence of M-CSF (30 ng/mL) for 10 and 30 min. Unstimulated BMDMs served as control and glycerinaldehyde-3-phosphate dehydrogenase (GAPDH) as loading control (n=3). (B) Representative immunoblots of IκBα and NF-κB on stimulation of WT BMDMs with RANKL (50 ng/mL), activin A (30 ng/mL) or RANKL plus activin A for 10, 30 and 60 min always in the presence of M-CSF (30 ng/mL). Unstimulated BMDMs were used as control and GAPDH as loading control (n=3). (C) Representative immunoblots of P-Smad2 and Smad2 after stimulation of BMDMs with RANKL (50 ng/mL), activin A (30 ng/mL) or RANKL plus activin A for 10, 30 and 60 min (n=3) and (D) after 3 and 4 days (n=3). (E) Representative immunoblots of coimmunoprecipitation analyses. P-Smad2, coupled to protein G-labelled Dynabeads, was incubated equal amounts of protein lysates of BMDMs stimulated with RANKL, or RANKL plus activin A for 2 days was transferred to the antibody-bead complex. After incubation, pulldown was performed followed by western blot analysis against NFATc1 of the coimmunoprecipitated protein and equal amounts of the total protein lysates (input). Lysis buffer was used as negative control (neg ctrl) (n=3). (F) representative immunoblots of the OC differentiation markers integrin αν, integrin β3, DC-STAMP, NFATc1 and cathepsin K on stimulation of WT BMDMs with activin A (30 ng/mL), RANKL (50 ng/mL) or RANKL plus activin A for 3 and 4 days always in the presence of M-CSF (30 ng/mL). Unstimulated BMDMs served as control. GAPDH was used as loading control (n=3). (G) Representative images of TRAP staining after 4 days of differentiation. (H) Quantification of OC number and area on treatment of BMDMs with an ALK4/5/7 inhibitor together with RANKL (50 ng/mL) or RANKL plus activin A (30 ng/mL) for 4 days. For recovery experiments, ALK4/5/7 inhibitor was removed and BMDMs were stimulated with RANKL or RANKL plus activin A for a further 2 days. The ALK4/5/7 inhibitor was diluted in DMSO; therefore, DMSO-treated BMDMs served as control. All data are means±SEM (n=5–6, Mann-Whitney U test). All experiments were performed in the presence of M-CSF (30 ng/mL). *P≤0.05, **P≤0.01. OC, osteoclast; RANKL, receptor activator of nuclear factor κB ligand.

In order to verify whether the stimulatory effect of activin A on OC formation exclusively depends on receptor-mediated Smad2 signalling, a specific ALK4/5/7 inhibitor (SB431542) was used during OC differentiation. TRAP staining showed that treatment with ALK4/5/7 inhibitor strongly reduced the activin A-enhanced OC formation (figure 7G). Quantification revealed a strong inhibition of osteoclastogenesis either on stimulation with RANKL plus activin A compared with untreated differentiation controls, confirming a receptor-mediated and Smad-dependent impact of activin A on OC formation. Finally, recovery experiments showed that the inhibitor was not toxic and OC differentiation increased again after withdrawal of the inhibitor (figure 7H). However, OC differentiation was already reduced on stimulation with RANKL alone, assuming additional autocrine factors. Beside activin A, OC precursors also express myostatin, which similar to activin promotes OC differentiation by activating the Smad pathway. Since both myostatin and activin A signal through a combination of Acvr2b and/or Acvr2a and ALK4/5 or ALK4/7 activating the Smad pathway, the ALK inhibitor will block both the activin and the myostatin-induced Smad2 activation, which very likely cause the inhibition of basal RANKL-induced OC development.

Together, these results led to the conclusion that the stimulating effect of activin A on osteoclastogenesis is mediated by the interaction of P-Smad2 and the key transcription factor of osteoclastogenesis NFATc1, thereby enhancing the expression of osteoclastic genes and subsequently OC development.

Discussion

RA is common type of inflammatory arthritis characterised by chronic inflammation, culminating in joint damage due to cartilage and bone destruction. Identification of molecules and associated signalling pathways involved in these destructive processes has become increasingly important. A family of proteins that affects inflammatory processes and bone metabolism is the TGF-β superfamily, under which activin A has found to be markedly elevated in serum, synovial fluid and synovial tissue of patients with RA25–28 and to stimulate the production of inflammatory mediators.29 30 Moreover, activin A is expressed in cells that are involved in bone metabolism such as BM cells, macrophages, OCs and OBs,26 29 38–42 as well as in FLS that are abundantly present in the inflamed synovium of patients with RA,1 26 28 43 all suggesting a role of activin A in inflammatory bone remodelling. In accordance with this, we observed highly elevated levels of activin A in synovial tissues obtained from patients with RA as well as from arthritic mice, suggesting that the inflammatory environment leads to an upregulation of activin A expression. Interestingly, in patients with RA, activin A-expressing cells were mostly found in the synovial sublining layer, indicating that within the synovial tissue, FLS are the main producer of activin A.

In more detail, stimulation of FLS and BMDMs with proinflammatory cytokines showed that activin A secretion by FLS was increased on stimulation with proinflammatory cytokines, while activin A secretion by BMDMs was not affected. In addition, activin A secretion by FLS was considerably higher than by BMDMs and even still higher in arthritic FLS compared with WT FLS.

Intriguingly, IL-17A strongly enhanced activin A secretion in hTNFtg but not in WT FLS. In this regard, former studies could show that IL-17-RA and IL-17RC are overexpressed in peripheral whole blood obtained from patients with RA, and these receptors are also highly expressed in the synovium of patients with RA, as well as by RA-FLS,44 45 probably explaining why arthritic FLS were more susceptible to IL-17A stimulation than WT FLS.

Altogether, these data strongly support the hypothesis that FLS-derived activin A may be an important regulator of inflammatory arthritis.

Since inflammation-induced bone destruction is one main feature of RA, we evaluated the contribution of activin A to OC differentiation in vitro and bone erosion in vivo. Indeed, activin A strongly increased RANKL-induced OC formation and resorptive activity in vitro, which is in line with others.14–19 Moreover, enhancement of osteoclastogenesis was mediated by the induction of Smad2 phosphorylation, which is consistent with studies from Murase et al and Kajita et al.15 19 However, this study showed for the first time that stimulation of primary BMDMs with activin A and RANKL led to an interaction of phospho-Smad2 with NFATc1, the key transcription factor of OC differentiation. Taking additionally the activin A-induced increase of OC differentiation and activity markers into account, we hypothesise that interaction of NFATc1 with phospho-Smad2 led to an increased translocation of NFATc1 into the nucleus, thereby enhancing the expression of OC-specific genes. Since this interaction has also been observed in myostatin-enhanced osteoclastogenesis, it seems to be a common mechanism by which members of the TGF-β family can regulate OC formation.37 Therefore, the interaction domains of phospho-Smad2 and NFATc1 might by interesting targets for the treatment of diseases with OC-mediated bone loss such as RA or osteoporosis. Finally, the fact that activin A also stimulates human OC differentiation further confirms the importance of activin A as an osteoclastogenic factor.

Based on the observation that activin A is upregulated within an inflammatory environment, we wondered whether the loss of activin A protects against bone damage during inflammatory destructive arthritis. Since FLS and macrophages appeared to be the main activin A-expressing cells within the inflamed synovium, we assessed the effect of activin A deficiency in ColVI-Cre-expressing and LysM-Cre-expressing cells on disease severity in the hTNFtg mouse model of arthritis.

The deficiency of activin A in myeloid cells (LysM-Cre) does not significantly affect inflammation, bone erosion and OC numbers in hind paws of hTNFtg mice, although a tendency towards a reduced inflammation was observed. The missing impact of this cell-specific deletion on OC formation and bone erosion is probably due to the very small amounts of activin A produced by myeloid cells compared with FLS (pg vs ng range). This assumption is further supported by the fact that the lack of autocrine activin A in BM cells, BMDMs, pOCs and OCs from ActβAd/d LysM-Cre or ActβAd/d LysM-Cre hTNFtg mice does not affect OC development.

In contrast, although the clinical scores were not affected by the loss of activin A, histomorphometric analyses clearly showed that deficiency of activin A in ColVI-Cre-expressing cells such as FLS strongly reduces OC formation (about 47%) in the hTNFtg model and thereby attenuating bone erosion (about 40%). These results highly suggest that paracrine, but not autocrine activin A, is important for inflammation-mediated bone erosion by directly acting on OC precursors and thereby regulating OC differentiation and activity. Results from coculture experiments demonstrating that the cell-specific deletion of activin A in FLS leads to less OC differentiation further substantiate the major role of paracrine activin A. However, additional in vivo studies would be helpful to further clarify the relative contribution of FLS-derived myostatin and activin A to OC differentiation in RA.

Moreover, serum levels of proinflammatory cytokines were highly reduced (about 50%), and pannus formation, which also reflects the degree of joint inflammation, was strongly reduced about 48% as well on lack of activin A, both confirming an anti-inflammatory effect of activin A deletion. In this context, activin A was shown to induce cell proliferation of RA-FLS,26 which potentially may contribute to the formation of pannus tissue within the inflammatory environment. Moreover, it is well known that FLS within the pannus tissue are activated by their proinflammatory environment, resulting in an aggressive phenotype that is maintained even in the absence of inflammatory stimuli.3 4 46 In this context, activated FLS produce inflammatory cytokines, chemokines, growth factors and degrading enzymes, thereby regulating joint inflammation and destruction. The fact that FLS appeared the be the main producers of activin A in the joint and that the lack of FLS-derived activin A exceedingly ameliorates disease severity confirm the prominent role of FLS in arthritis.

In conclusion, we could show for the first time that activin A significantly regulates disease severity in a mouse model of chronic arthritis. Deletion of paracrine but not autocrine activin A reduces OC formation and bone destruction and most interestingly reduces also joint inflammation in our arthritis model. Thus, inhibition of activin A may be a promising treatment option for arthritis and other diseases associated with inflammatory bone loss.