The RANK/RANKL pathway is critical in bone destruction in conditions such as rheumatoid arthritis (RA). Since RANK/RANKL-deficient mice show major lymph node (LN) abnormalities, undertook this study to investigate the expression of RANK/RANKL in paired samples of synovium and LNs from RA patients.
Using immunohistochemistry, RANK/RANKL expression by dendritic cell (DC) and T cell subsets was studied in this unique set of samples and in RA synoviocytes stimulated with interleukin-1β (IL-1β), tumor necrosis factor α (TNFα), and IL-17.
In RA synovium, RANKL+ cells were detected in the lining layer and the lymphocytic infiltrates, whereas RANK expression was restricted to the perivascular infiltrates. In LNs, RANK+ and RANKL+ cells were diffusely expressed in the T cell zone and in germinal centers. Double staining of paired RA synovium and LN sections showed that some immature CD1a+ DCs expressed RANK and RANKL, while some mature DC-LAMP+ DCs expressed only RANK. Some CD3+, CD4+, interferon-γ+, and IL-17+ cells expressed RANKL, while none expressed RANK. Treatment of synoviocytes with TNFα or IL-1β in combination with IL-17 was particularly potent at inducing RANKL expression.
This study shows the involvement of RANK/RANKL in DC–T cell interactions during an inflammatory process. RANK expression appears to be limited to the sites of immune reaction, both in the synovium and in LNs. Therapeutic control of these targets may have both positive and negative consequences for the immune system.
Rheumatoid arthritis (RA) is a chronic inflammatory disease in which inflammation of the synovium leads to joint destruction (1). Interactions between RANK, expressed by osteoclasts, and its ligand RANKL, expressed by osteoblasts, have major effects on bone. RANKL-deficient mice lack osteoclasts and exhibit osteopetrosis, whereas transgenic mice show severe osteoporosis (2, 3).
In addition to its well-studied contribution to bone metabolism, this RANKL/RANK system is involved in dendritic cell (DC)–T cell interactions (4). These are critical in lymph node (LN) formation, since RANKL- and RANK-deficient mice have a complete lack of LNs with the related immune defects (3, 5). However, this important phenomenon has not been given much attention in humans, chiefly because of limited access to LNs. Since RA synovium has several but not all of the characteristics of a lymphoid organ (6), in the present study we compared the RANKL/RANK expression pattern in RA synovium with that in LN samples. Although a study of paired samples would be easy to perform in the mouse, such a study in human RA would be much more difficult if not impossible to perform. However, when we examined the pathology department database, we were able to select 11 paired samples that had been obtained ∼20 years ago when control of disease was difficult. At the time of joint surgery, LN biopsies had been performed to eliminate lymphomas in patients with reactive LNs, which are often observed in active, uncontrolled RA. Using this unique set of samples, we used immunohistochemistry to investigate RANK and RANKL expression by immature CD1a+ and mature DC-LAMP+ DC subsets and by subsets of CD3+ and CD4+ T cells expressing the Th1 cytokines interleukin-17 (IL-17) and interferon-γ (IFNγ).
MATERIALS AND METHODS
Collection of paired RA synovium–LN samples
A computer database from the pathology department was used to select paired samples using the key words “rheumatoid synovium” and “lymph nodes.” From this database, 11 paired samples were obtained. Synovial samples had been obtained from patients with RA (defined according to the 1958 revised criteria of the American College of Rheumatology [formerly, the American Rheumatism Association] ) who had been undergoing joint surgery. The 11 paired LN samples had been obtained at or around the time of joint surgery in order to eliminate a possible lymphoma. Pathology was that of reactive LNs and excluded lymphoma. Samples were fixed in 4% phosphate buffered paraformaldehyde and embedded in paraffin. To detect antigen expression, antigen retrieval procedures were performed, including incubation in either citrate buffer (10 mM, pH 6) or EDTA buffer (1 mM, pH 8), followed by incubation in a microwave oven 3 times for 3 minutes each.
Detection of RANK and RANKL by immunohistochemistry
Paraffin sections were treated in xylene and rehydrated in a gradient of ethanol (once in 99% ethanol, once in 95% ethanol, and once in H2O). Endogenous peroxidase activity was blocked with 3% H2O2. The sections were then incubated with 5 μg/ml of a goat polyclonal anti-RANKL antibody (Santa Cruz Biotechnology, Santa Cruz, CA) or with 5 μg/ml of a goat polyclonal anti-RANK antibody (Santa Cruz Biotechnology). Specificity of the detection was studied in phytohemagglutinin/phorbol myristate acetate–activated peripheral blood mononuclear cells (PBMCs) in which RANK+ cells were defined as CD14+ monocytes. Conversely, RANKL was expressed in subsets of T cells (data not shown). After overnight incubation at 4°C and washing, the sections were incubated with biotinylated mouse anti-goat IgG antibodies for 30 minutes, followed by incubation with streptavidin–peroxidase complex (Dako, Glostrup, Denmark) for 15 minutes and by incubation with 3,3′-diaminobenzidine (Dako) for 20 minutes. The sections were then counterstained with Mayer's hematoxylin.
Characterization of RANK- and RANKL-producing cells in RA synovium and LNs
To clarify the cell subsets expressing RANK and RANKL, double-labeling experiments were performed by first staining for RANK or RANKL, followed by staining for T cell markers with mouse monoclonal anti-CD3 (IgG1; Dako) and anti-CD4 (IgG2a; Novocastra, Newcastle-upon-Tyne, UK) antibodies, for Th1-producing cells with anti-IFNγ (IgG1; Genzyme, Cambridge, MA) and anti–IL-17 (IgG1; R&D Systems Europe, London, UK) antibodies, for monocyte marker with anti-CD14 antibodies (IgG2a/2b; Dako), and for immature and mature DC subsets with anti-CD1a (IgG2b; Becton Dickinson, Le Pont de Claix, France) and anti–DC-LAMP (IgG1; Immunotech, Marseilles, France) antibodies, respectively, as described (8). After staining for RANK or RANKL as described above, incubation with one of these monoclonal antibodies was performed, and then purified rabbit anti-mouse IgG (Dako) and mouse alkaline phosphatase–anti–alkaline phosphatase (Dako) were applied. Alkaline phosphatase was revealed using fast blue as chromogen (blue color; Vector, Peterborough, UK).
To isolate synoviocytes, synovium fragments were cut into small pieces and digested with 1 mg/ml collagenase and DNase in RPMI 1640 for 2 hours at 37°C. After centrifugation, cells were suspended in RPMI 1640 complete medium supplemented with 10% fetal calf serum, 2 mML-glutamine, and a 1% (volume/volume) mixture of antibiotics (100 units/ml penicillin and 10 μg/ml streptomycin) and cultured in 100-mm petri dishes. At confluence, cells were trypsinized and passaged. Cells were grown in 75-cm2 flasks in RPMI 1640 complete medium. They were plated at a density of 1 × 104/cm2 onto glass coverslips (12-mm diameter) or on 6-well plates and were stimulated with IL-1β (100 pg/ml), TNFα (100 pg/ml), or IL-17 (100 ng/ml) obtained from Sigma (St. Quentin Fallavier, France) or with combinations of these cytokines. These concentrations were determined to be optimal in previous studies (9). Cells were fixed in 4% paraformaldehyde for immunocytochemistry with anti-RANKL antibody 48 hours after addition of cytokine.
Expression of RANK and RANKL in paired RA synovium and LN sections. In RA synovium sections, RANKL was expressed in the lining layer but also in the sublining layer by fibroblast-like synoviocytes and cells in the lymphocytic infiltrates (Figure 1A), while RANK+ cells were detected exclusively in the lymphocytic perivascular infiltrates and not in the lining layer (Figure 1B). As observed in activated PBMCs, some RANK+ cells were identified as CD14+ cells (results not shown). In LN sections from RA patients, RANKL+ and RANK+ cells were diffusely expressed both in the T cell zone and in germinal centers (Figures 1C and D, respectively).
Coexpression of RANK/RANKL and DC/T cell markers in paired RA synovium and LN sections. To further clarify the phenotype of RANKL+ and RANK+ cells, double staining was performed with anti-RANKL or anti-RANK antibodies and DC markers (CD1a, DC-LAMP) or T cell markers (CD3, CD4, IL-17, IFNγ) both in RA synovium and in LN sections. Some immature CD1a+ and mature DC-LAMP+ DCs expressed RANK (Figures 2A and B, respectively). Double staining with anti-RANK and anti-CD3, anti-CD4, anti-IFNγ, or anti–IL-17 antibodies showed that none of the CD3+ (Figure 2C), CD4+ (Figure 2D), IL-17+ (Figure 2E), or IFNγ+ (Figure 2F) T cells expressed RANK. Double staining with anti-RANKL and anti-CD1a or anti–DC-LAMP antibodies showed that some RANKL+ cells expressed the immature DC marker CD1a (Figure 2G). However, none of the mature DC-LAMP+ DCs expressed RANKL (Figure 2H). Double staining with the T cell markers showed that some CD3+ (Figure 2I), CD4+ (Figure 2J), IL-17+ (Figure 2K), or IFNγ+ (Figure 2L) T cells expressed RANKL. These results indicated an association between RANK expression and T cell interactions, since such expression was specific to the perivascular infiltrates of the synovium. The same picture was observed both in RA synovium and in LN sections.
Effect of the proinflammatory cytokines IL-1β, TNFα, and IL-17 on the expression of RANKL by RA synoviocytes. Following the study of RANK and RANKL expression by DC or T cell subsets in samples of inflamed tissue, we switched to an in vitro model to reproduce some of the interactions observed in vivo. In particular, we looked at the contribution of the proinflammatory cytokines produced by monocytes (IL-1β, TNFα) and by T cells (IL-17) to RANKL expression by RA synoviocytes (9). Immunostaining of stimulated synoviocytes for RANKL showed that staining intensity of RANKL was enhanced in IL-1β–stimulated synoviocytes compared with that in TNFα- or IL-17–stimulated synoviocytes, and that combinations of these cytokines were more effective (Figure 3). These results indicate that the increased expression of RANKL, particularly in the lining layer and the perivascular infiltrates, reflects the local effect of inflammatory cytokines.
The involvement of RANK–RANKL interactions has been clearly identified in bone destruction. Although deficient mice show major LN and immune defects (3, 5), these observations have not been given much attention. RA is the perfect context in which to study these phenomena, given the combination of local inflammation and bone destruction (10). In addition, LN hypertrophy is a common feature of RA disease activity.
The active RA synovium and LNs have many similarities as well as differences (6). Planning a study to compare the two sites would be impossible today. In addition, better therapy for RA with the treatments used today may modify the spontaneous picture observed in active disease. Accordingly, we used a set of paired synovium and LN samples to investigate the anatomic localization of RANK+ and RANKL+ cells in untreated (or at best poorly treated) RA patients. This choice has obvious limitations that we had to consider, including limited access to patient information, limited quality of fixation, and reduced amount of material left for new sections.
Using immunohistochemistry with paraffin sections, we focused on RANK and RANKL expression by DC and T cell subsets. In RA synovium, RANKL+ cells were detected in the lining layer and the lymphocytic infiltrates, whereas RANK expression was restricted to the perivascular infiltrates, in accordance with previous reports (11, 12). In LN sections, in which this had not been studied before, RANK+ and RANKL+ cells were diffusely expressed both in the T cell zone and in germinal centers. Thus, RANK expression appears to be associated with an active immune reaction.
Accordingly, we next focused on DC and T cell interactions, using markers previously tested in this context (8). Double staining showed that some immature CD1a+ DCs expressed RANK and RANKL, while some mature DC-LAMP+ DCs expressed only RANK. Although RANK and RANKL expression are not specific, it appears interesting that immature DCs can be distinguished by their expression of both RANK and RANKL from mature DCs, which express only RANK.
When we examined T cells, double staining showed that some CD3+, CD4+, IFNγ+, and IL-17+ cells expressed RANKL, while none of them expressed RANK. Accordingly, these subsets of T cells can interact directly with RANK+ cells such as DCs, as first observed between osteoclasts and osteoblasts. In addition, such interaction can be mediated through the production of Th1 cytokines (13). This can take place both in the synovium, leading to increased inflammation, and in juxtaarticular bone, leading to bone destruction, as well as in LNs.
To reproduce some of these interactions observed in vivo, we investigated the contribution of the proinflammatory cytokines produced by monocytes (IL-1β, TNFα) and by T cells (IL-17) to RANKL expression by RA synoviocytes. IL-1β was the most potent of the 3 cytokines when tested alone. Treatment with TNFα or IL-1β in combination with IL-17 was particularly potent at inducing RANKL expression, indicating the enhancing contribution of IL-17–producing cells and the role of synoviocytes in osteoclastogenesis (14). Indeed, blocking of IL-17 with an antibody and a soluble receptor in cultures of RA juxtaarticular bone explants led to a reduction in bone destruction, indicating the contribution of IL-17 to osteoclast activation (15). This effect was increased when inhibitors of IL-1, TNFα, and IL-17 were combined, indicating that these factors were more potent when interacting than when acting separately.
The down-regulation of RANKL expression through the actions of antiinflammatory cytokines or cytokine inhibitors (antibodies or soluble receptors) could be a strategy for reducing the involvement of RANK/RANKL, as observed in the effect of such treatment on joint destruction. There are drawbacks to such an effect, however, as can be seen with the side effects of anti-TNF treatment, mostly involving opportunistic infections. Such adverse effects must be considered when evaluating inhibition of the RANK/RANKL pathway for treatment of patients with RA.
In conclusion, in chronic inflammatory diseases such as RA, RANK expression appears to be restricted to mature DCs. In turn, these DCs can interact with RANKL+ T cells, some of which concomitantly express the Th1 cytokines involved in cell-mediated immunity and bone destruction. Inhibition of proinflammatory cytokines has been associated with positive effects on inflammation and destruction. Controlling the RANK/RANKL pathway may allow an improved potent result by acting on cell interactions present at synovium and bone sites. Conversely, induction of defects in LN function may have adverse consequences.