To investigate the efficacy of single and combined blockade of tumor necrosis factor (TNF), interleukin-1 (IL-1), and RANKL pathways on synovial inflammation, bone erosion, and cartilage destruction in a TNF-driven arthritis model.
To investigate the efficacy of single and combined blockade of tumor necrosis factor (TNF), interleukin-1 (IL-1), and RANKL pathways on synovial inflammation, bone erosion, and cartilage destruction in a TNF-driven arthritis model.
Human TNF–transgenic (hTNFtg) mice were treated with anti-TNF (infliximab), IL-1 receptor antagonist (IL-1Ra; anakinra), or osteoprotegerin (OPG; an OPG-Fc fusion protein), either alone or in combinations of 2 agents or all 3 agents. Synovial inflammation, bone erosion, and cartilage damage were evaluated histologically.
Synovial inflammation was inhibited by anti-TNF (−51%), but not by IL-1Ra or OPG monotherapy. The combination of anti-TNF with either IL-1Ra (−91%) or OPG (−81%) was additive and almost completely blocked inflammation. Bone erosion was effectively blocked by anti-TNF (−79%) and OPG (−60%), but not by IL-1Ra monotherapy. The combination of anti-TNF with IL-1Ra, however, completely blocked bone erosion (−98%). Inhibition of bone erosion was accompanied by a reduction of osteoclast numbers in synovial tissue. Cartilage destruction was inhibited by anti-TNF (−43%) and was weakly, but not significantly, inhibited by IL-1Ra, but was not inhibited by OPG monotherapy. The combination of anti-TNF with IL-1Ra was the most effective double combination therapy in preventing cartilage destruction (−80%). In all analyses, the triple combination of anti-TNF, IL-1Ra, and OPG was not superior to the double combination of anti-TNF and IL-1Ra.
Articular changes caused by chronic overexpression of TNF are not completely blockable by monotherapies that target TNF, IL-1, or RANKL. However, combined approaches, especially the combined blockade of TNF and IL-1 and, to a lesser extent, TNF and RANKL, lead to almost complete remission of disease. Differences in abilities to block synovial inflammation, bone erosion, and cartilage destruction further strengthen the rationale for using combined blockade of more than one proinflammatory pathway.
Rheumatoid arthritis (RA) is a complex inflammatory disorder characterized by at least 3 major events: chronic synovial inflammation, cartilage destruction, and bone erosion. These 3 events are probably closely linked to each other; however, the underlying cellular and molecular processes may differ considerably. Given the complex nature of joint destruction in RA, it was highly surprising that selective blockade of a single proinflammatory cytokine, tumor necrosis factor (TNF), effectively inhibited the arthritic process (1, 2). These findings revised the assumption that proinflammatory cytokine networks are highly complex and therefore virtually impossible to target by a selective pharmacologic intervention.
As a consequence of these findings, the existence of “master players” within these proinflammatory cytokine networks, such as TNF, were postulated to essentially trigger joint inflammation and destruction (3). The critical number of such molecules that are necessary to promote the development of disease is not known. Studies in animals, such as the human TNF–transgenic (hTNFtg) mouse model, however, suggest that deregulated expression of a single proinflammatory cytokine is sufficient to trigger a complex, progressive, and destructive inflammatory process that closely resembles RA in humans (4). Thus, hTNFtg mice develop an arthritis that shows all 3 pathophysiologic processes that are found in human disease: chronic synovial inflammation, cartilage destruction, and bone erosion (4, 5).
It is still unclear why pharmacologic blockade of TNF does not fully stop the signs and symptoms of RA. Indeed, complete remission of signs and symptoms of RA is rarely achieved by any disease-modifying antirheumatic drug, including TNF blockers (1, 6–10). The most obvious explanation may be the complex nature of the disease, suggesting that several proinflammatory pathways act independently of TNF. Thus, the degree to which TNF fuels the RA disease process has not been completely clarified. Although, clinical studies have quantitatively assessed the efficacy of TNF blockers in RA (1, 6–9), it remains to be determined whether the lack of complete clinical response is due to a failure of TNF blockers to neutralize the entire biologic activity of TNF or is due to activation of TNF-independent pathways of inflammation.
Interestingly, the inhibitory potential of TNF blockade varies in the individual patient, depending on which pathophysiologic processes are addressed. TNF blockade has been shown to arrest bone erosion in a large number of patients whose clinical signs of inflammation show no response (6). The relative role of TNF in joint inflammation, bone erosion, and cartilage destruction may therefore differ and may depend on differences in the molecular mechanisms that underlie each of these 3 events. Two major target molecules of TNF, interleukin-1 (IL-1), and RANKL, may have at least a predilection for 1 of these 3 mechanisms. Blockade of IL-1 has been shown to have its strongest effects on the reduction of joint space narrowing, which is considered a surrogate marker for cartilage damage (11). Experimental blockade of RANKL has been shown to be highly effective in preventing local bone erosion, but not synovial inflammation, in animal models of arthritis (12–15).
In this study, we used hTNFtg mice as a model of erosive arthritis to gain new insights into the mechanisms of TNF-driven synovial inflammation, cartilage destruction, and bone erosion. We addressed the specific potential of TNF blockade to interfere with each of these pathologic features, and we compared it with the effects of blocking 2 of the central downstream mediators, IL-1 and RANKL. We also examined potential additive effects of TNF, IL-1, and RANKL blockade on these 3 features.
The heterozygous Tg197 TNF-transgenic mice (C57BL/6) used in the present study have been described previously (4). These mice develop a chronic inflammatory and destructive polyarthritis within 4–6 weeks after birth. A total of 64 mice were examined. Two independent experiments, one including 48 mice and the other including 16 mice, were performed. All animal procedures were approved by the local ethics committee.
Mice were divided into 8 groups of 8 mice each and were treated according to the following protocols: group 1 received phosphate buffered saline (PBS; negative control), group 2 received anti-TNF (infliximab, a chimeric monoclonal antibody; Centocor, Leiden, The Netherlands), group 3 received IL-1 receptor antagonist (IL-1Ra) (anakinra, a recombinant IL-1Ra; Amgen, Thousand Oaks, CA), group 4 received osteoprotegerin (OPG) (an OPG-Fc fusion protein; Amgen), group 5 received anti-TNF plus IL-1Ra, group 6 received anti-TNF plus OPG, group 7 received IL-1Ra plus OPG, and group 8 received anti-TNF plus IL-1Ra and OPG.
Anti-TNF (10 mg/kg) was administered 3 times/week by intraperitoneal injection. IL-1Ra was given by continuous infusion at a dosage of 5 mg/kg/hour, using a subcutaneously implanted minipump (Alzet; Durect, Cupertino, CA). OPG (10 mg/kg) was administered 3 times/week by intraperitoneal injection. Therapy was started at the onset of symptoms (week 5) and lasted for 4 weeks.
Clinical evaluation was performed weekly, starting at 4 weeks after birth. Arthritis was evaluated in a blinded manner as described previously (5). Briefly, joint swelling was examined in all 4 paws, and a clinical score of 0–3 was assigned (0 = no swelling, 1 = mild, 2 = moderate, and 3 = severe swelling of the toes and ankle). In addition, grip strength was examined in each paw, using a 3-mm–diameter wire, and was scored on a scale of 0 to −4 (0 = normal grip strength, −1 = mildly reduced, −2 = moderately reduced, −3 = severely reduced, and −4 = no grip strength). Groups were matched for arthritis severity at the beginning of treatment. At the end of the treatment period, animals were killed by cervical dislocation, blood was withdrawn by heart puncture, and all paws were harvested for histologic examination.
Front paws, back paws, and right knees were fixed overnight in 4.0% formalin and then decalcified in 14% EDTA (Sigma, St. Louis, MO) at 4°C (pH adjusted to 7.2 by addition of ammonium hydroxide [Sigma]) until the bones were pliable. Serial paraffin sections (2 μm) of all 4 paws and the right knee joint were stained with hematoxylin and eosin (H&E), toluidine blue, or tartrate-resistant acid phosphatase (TRAP) (all mice, respectively) and analyzed by immunohistochemical methods (3 representative mice per group). TRAP staining was performed as previously described (14, 16).
For immunohistochemistry, deparaffinized, ethanol-dehydrated tissue sections were placed in a 700W microwave oven, boiled for 2 minutes in 10 mM sodium citrate buffer (pH 6.0), allowed to cool to room temperature, and then rinsed in detergent solution (0.5% Tween in PBS) for 10 minutes. Tissue sections were blocked for 20 minutes in PBS containing 20% rabbit serum, followed by incubation for 1 hour at room temperature with the following antibodies: rat monoclonal antifibroblast antibody (1:40 dilution; Biogenesis, Poole, UK), rat monoclonal antimacrophage (F4/80) antibody (1:80 dilution; Serotec, Raleigh, NC), rat monoclonal anti-CD3 antibody (1:100 dilution; Novocastra, Newcastle, UK), and mouse monoclonal anti–matrix metalloproteinase 3 (anti–MMP-3), anti–MMP-9, and anti–MMP-13 antibodies (1:100 dilution each; NeoMarkers, Fremont, CA). After rinsing, endogenous peroxidase was blocked for 10 minutes with 0.3% hydrogen peroxide in Tris buffered saline (10 mM Tris HCl, 140 mM NaCl, pH 7.4). This was followed by 30 minutes' incubation with a biotinylated species-specific anti-IgG secondary antibody (Vector, Burlingame, CA). Sections were then incubated for another 30 minutes with the appropriate Vectastain ABC reagent (Vector), using 3,3′-diaminobenzidine (Sigma) for the color reaction, which resulted in brown staining of antigen-expressing cells.
Synovial inflammation, bone erosions, osteoclast numbers, and cartilage destruction were quantified with the use of a Zeiss Axioskop 2 microscope (Zeiss, Marburg, Germany) equipped with a digital camera and image analysis system (model KS300; Zeiss), as described previously (17). The area of inflammation was quantified in H&E-stained sections (5 per mouse). Total scores were calculated as the sum of the areas of inflammation in all digital, carpal, and tarsal joints as well as the right knee joint in each mouse. Erosions were quantified in same H&E-stained sections. The number of osteoclasts was counted in TRAP-stained serial sections. The areas of total and negatively stained cartilage were measured in toluidine blue–stained sections. The proportion of damaged cartilage was calculated by dividing the area of negatively stained cartilage by the area of total cartilage. Total cell numbers and numbers of positively stained cells were counted in at least 3 different sites of each tarsal, carpal, and knee joint pannus in each mouse. MMP expression in chondrocytes from at least 3 different sites of the tarsal, carpal, and knee joint cartilage of each mouse was also assessed. A total of 24 mice from the 8 treatment groups (n = 3 per group) and 3 baseline controls animals (killed before initiation of treatment at week 6) were evaluated in these immunohistochemical analyses.
Serum levels of human TNF, murine TNF receptor type I (murine TNFRI), IL-1, and sRANKL were measured by ELISAs, according to the manufacturer's recommendations (all from R&D Systems, Minneapolis, MN). The lower limits of detection were 0.18 pg/ml for human TNF, 5 pg/ml for murine TNFRI, 3 pg/ml for IL-1, and 5 pg/ml for sRANKL. Analyses were performed in a total of 24 mice from the 8 treatment groups (n = 3 per group) and 3 baseline control animals (killed before initiation of treatment at week 6).
Data are reported as the mean ± SEM. Group mean values for the histologic data were compared by analysis of variance with the Bonferroni correction. The nonparametric Wilcoxon signed rank test was used for comparison of the results of clinical assessments.
To investigate the effects of single and combined blockade of TNF, IL-1, and RANKL on clinical signs of arthritis, joint swelling and grip strength scores were obtained (Figures 1A and B). During the course of disease, joint swelling scores increased continuously in untreated hTNFtg mice, with a mean ± SEM swelling score of 1.25 ± 0.07 at week 10. Joint swelling in mice treated with IL-1Ra, OPG, or the combination of IL-1Ra/OPG showed a similar course as that in the control mice. Anti-TNF treatment, however, significantly reduced (P < 0.01) the development of joint swelling (1.0 ± 0.13). This effect was even more significant (P < 0.001) after combination therapy with anti-TNF/IL-1Ra and anti-TNF/OPG, as well as the triple combination therapy (0.69 ± 0.12, 0.82 ± 0.17, and 0.63 ± 0.14, respectively).
Similar results were obtained for grip strength (Figure 1B). Grip strength continuously decreased in untreated hTNFtg mice, with a mean ± SEM grip strength of −2.38 ± 0.13 at week 10. This decrease was slowed only to a small degree in mice treated with IL-1Ra, OPG, or the combination of IL-1Ra/OPG. Anti-TNF treatment, in contrast, was effective in slowing loss of grip strength (−1.35 ± 0.28; P < 0.05), and this effect became much stronger after combination therapy with anti-TNF/IL-1Ra and anti-TNF/OPG, as well as the triple combination therapy (−0.75 ± 0.1, −0.68 ± 0.07, and −0.61 ± 0.13, respectively; P < 0.01 for each comparison versus controls).
To determine whether the treatment effects on clinical signs of arthritis were reflected by synovial abnormalities, joint sections were quantitatively assessed for the extent of inflammatory infiltrates (Figure 2). Untreated hTNFtg mice showed extensive synovial inflammation, covering a mean ± SEM area of 5.5 ± 0.3 mm2 of each joint section. Blockade of TNF led to a significant reduction (−51%; P < 0.01) of synovial inflammation (2.7 ± 1.0 mm2), which indicates the efficacy of TNF blockade, but also shows that even high doses of TNF blockers do not completely overcome TNF-driven synovial inflammation. In contrast to TNF-blockade, monotherapy with IL-1Ra and OPG did not show significant effects on synovial inflammation (3.9 ± 0.4 mm2 [−29%] and 4.3 ± 0.2 mm2 [−22%], respectively).
Among the combined therapeutic approaches, the combination of TNF blockade with IL-1Ra was the most effective and almost completely prevented synovial inflammation in hTNFtg mice (0.48 ± 0.1 mm2 [−91%]). The triple combination was not superior to the anti-TNF/IL-1Ra combination (0.45 ± 0.1 mm2 [−92%]). Interestingly, the combination of TNF blockade with OPG also showed additive effects (1.0 ± 0.4 mm2 [−81%]; P < 0.001), reducing synovial inflammation significantly more efficiently than either agent alone. The combination of IL-1Ra with OPG yielded significant antiinflammatory effects compared with control (3.1 ± 0.4 mm2 [−44%]); however, it was not superior to any of the monotherapies.
Results from a second independent experiment (n = 16 mice) were similar. The data from the second experiment differed by <5% from the data shown above. Representative H&E-stained sections of paws obtained from each of the study groups are shown in Figure 3.
To address whether the cellular composition of inflamed tissue is influenced by inhibition of TNF, IL-1, and RANKL, the expression of cell-specific markers for fibroblasts, macrophages, and T cells was quantitated by immunohistochemistry (Table 1). No significant differences were seen, which suggests that these therapies did not influence the cellular composition of inflamed tissue. In contrast, levels of MMPs 3, 9, and 13, which served as surrogates for the effector molecules of cytokine-driven synovial inflammation, decreased significantly after single and combined blockade of TNF (Figure 1B). MMP-9, a key MMP in osteoclasts, was also decreased after single and combined treatment with OPG, and MMP-3 expression was also down-regulated by IL-1Ra. These data indicate that inhibition of TNF, IL-1, and RANKL not only reduces the amount of inflamed tissue, but it also affects the molecular composition, and thus the invasive properties, of inflamed tissue, which facilitates destruction of bone and cartilage (5).
|Treatment||Cellular composition of inflamed synovial tissue||MMP expression of inflamed synovial tissue|
|Fibroblasts (antifibroblast)||Macrophages (F4/80)||T cells (anti-CD3)||MMP-3||MMP-9||MMP-13|
|Baseline (6-week-old mice)||42.3 ± 1.3||39.9 ± 2.2||3.9 ± 1.1||19.3 ± 2.4||24.5 ± 2.2||20.4 ± 3.8|
|Control||44.7 ± 3.7||41.2 ± 3.7||6.0 ± 1.7||33.3 ± 4.8||46.2 ± 1.7||33.8 ± 3.7|
|Anti-TNF||48.5 ± 2.6||39.2 ± 3.5||3.2 ± 1.0||20.3 ± 1.4†||25.6 ± 5.7†||25.0 ± 5.0†|
|IL-1Ra||46.5 ± 4.7||39.4 ± 3.5||5.0 ± 1.5||24.3 ± 5.3†||36.3 ± 2.8||32.6 ± 4.3|
|OPG||47.5 ± 2.7||36.6 ± 0.8||4.3 ± 0.3||27.6 ± 1.8||29.0 ± 3.0†||34.0 ± 6.8|
|Anti-TNF + IL-1Ra||<1‡||<1‡||<1‡||<1‡||<1‡||<1‡|
|Anti-TNF + OPG||41.7 ± 3.8||43.4 ± 2.1||4.0 ± 0.5||12.3 ± 3.7†||18.1 ± 3.0†||19.6 ± 2.7†|
|IL-1Ra + OPG||42.2 ± 2.7||38.6 ± 2.2||3.0 ± 0.8||19.6 ± 1.7†||29.0 ± 3.0†||32.0 ± 6.5|
|Anti-TNF + IL-1Ra + OPG||<1‡||<1‡||<1‡||<1‡||<1‡||<1‡|
Having gained evidence of differences in the capacity of TNF, IL-1, and RANKL blockade to reduce joint inflammation, we were interested in the effects of these therapies on bone erosion (Figure 4A). Untreated hTNFtg mice showed a large amount of destruction of articular bone (mean ± SEM area of resorbed bone 0.47 ± 0.06 mm2). TNF blockade was highly effective in slowing this process (0.1 ± 0.05 mm2 [−79%]; P < 0.001), suggesting that TNF blockade inhibited bone erosion even more effectively than it inhibited synovial inflammation. Although IL-1Ra showed a trend toward reduced bone erosion (0.28 ± 0.04 mm2 [−40%]), this was not significant. OPG treatment yielded significant protection of articular bone (0.19 ± 0.03 mm2 [−60%]; P < 0.001).
The combination of TNF blockade and IL-1Ra was the most effective combination treatment and led to complete blockade of bone erosion (0.01 ± 0.005 mm2 [−98%]). Since blockade by TNF alone, however, produced a strong reduction in bone erosion, the additive effect of this combination was not significant. The anti-TNF/OPG combination yielded results that were similar to those with TNF blockade alone (0.07 ± 0.03 mm2 [−85%]), and the results of IL-1Ra/OPG combination therapy were similar to those with OPG monotherapy. Thus, for both of these combinations, there was no significant benefit beyond the respective monotherapies. The triple combination completely prevented bone erosion (0.008 ± 0.004 mm2 [−99%]; P < 0.001) and yielded results similar to those with the anti-TNF/IL-1Ra combination. In a second experiment (n = 16 mice), the results remained robust, showing a difference of <5% from the values shown above.
Based on the fact that osteoclasts are essentially involved in local bone erosion, we quantitatively assessed the numbers of osteoclasts within the inflamed synovial tissue (Figure 4B). As reported previously (14), untreated hTNFtg mice showed numerous osteoclasts at sites of bone erosion (mean ± SEM 94 ± 10 osteoclasts/section). Blockade of TNF significantly reduced osteoclast numbers at sites of erosions (41 ± 14 [−56%]). Corresponding to the effects on bone erosion, IL-1Ra did not significantly influence the number of synovial osteoclasts (74 ± 10 osteoclasts/section [−21%]). In contrast, OPG reduced the number of synovial osteoclasts to a similar extent as TNF blockade (53 ± 11 [−43%]).
Among the combination treatments, IL-1Ra/OPG combination therapy (40 ± 10 osteoclasts/section [−57%]) was as effective as OPG alone. Although anti-TNF/OPG combination therapy (27 ± 8 osteoclasts/section [−71%]) showed a trend toward lower numbers of osteoclasts, it was not superior to anti-TNF or OPG monotherapy. Again, the combination of anti-TNF/IL-1Ra was the most effective in reducing the number of osteoclasts (11 ± 2 osteoclasts/section [−88%]), showing an additive effect on blockade of osteoclast formation in vivo. Similar results were also found with the triple combination therapy (15 ± 5 osteoclasts/section [−83%]).
Thus, the capacity of the various treatments to inhibit synovial osteoclast formation was well matched to their antiresorptive potential. Representative TRAP-stained sections of paws obtained from each of the study groups are shown in Figure 5.
To determine whether TNF, IL-1, and RANKL blockade affected cartilage degradation to the same extent as bone resorption, we quantitatively assessed proteoglycan loss by toluidine blue staining of articular cartilage (Figure 6). Untreated hTNFtg mice showed a significant loss of proteoglycans (mean ± SEM 21 ± 3.0% of total cartilage area). Blockade of TNF significantly reduced proteoglycan loss (12 ± 2.9% of total cartilage area [−43%]); however, its effect on cartilage was markedly weaker compared with its effect on bone resorption. IL-1Ra treatment also led to a marked reduction in proteoglycan loss (14 ± 1.8% of total cartilage area [−35%]), although again, this effect did not reach statistical significance. OPG, in contrast, did not influence cartilage damage (21 ± 2.2% of total cartilage area).
The combination of TNF blockade with IL-1Ra showed an additive effect and protected articular cartilage significantly better than each agent as monotherapy (4 ± 0.8% of total cartilage area [−80%]). The combination of anti-TNF/OPG (8 ± 0.9% of total cartilage area [−63%]) was not statistically significantly different from TNF blockade alone. The combination of IL-1Ra/OPG (11 ± 1.7% of total cartilage area [−48%]) was not significantly different from IL-1Ra monotherapy. The triple combination strongly inhibited proteoglycan loss (3.7 ± 0.7% of total cartilage area [−82%]) to a similar extent as the anti-TNF/IL-1Ra combination.
These data were confirmed by a second experiment (n = 16 mice), which showed almost identical results, differing by <5% from the values shown above. Representative toluidine blue–stained sections of paws obtained from each of the study groups are shown in Figure 7.
Inhibition of proteoglycan loss was associated with gradually decreased expression of MMPs 3, 9, and 13 in articular chondrocytes (Table 2). MMP expression was significantly inhibited by anti-TNF and IL-1Ra, but not OPG, and was more effectively inhibited by combination therapy.
|Baseline (6-week-old mice)||3.8 ± 1.3||3.0 ± 1.1||5.3 ± 1.9|
|Control||23.0 ± 3.6||46.0 ± 7.0||47.5 ± 7.5|
|Anti-TNF||8.6 ± 2.9†||18.3 ± 3.7†||28.1 ± 5.8†|
|IL-1Ra||9.0 ± 2.0†||19.0 ± 3.2†||32.0 ± 1.5†|
|OPG||22.3 ± 2.6||38.3 ± 7.7||37.3 ± 4.7|
|Anti-TNF + IL-1Ra||2.6 ± 0.3†||3.6 ± 1.2†||2.6 ± 1.1†|
|Anti-TNF + OPG||8.0 ± 2.1†||14.6 ± 2.9†||15.0 ± 3.5†|
|IL-1Ra + OPG||14.5 ± 1.5†||20.6 ± 2.6†||16.0 ± 8.9†|
|Anti-TNF + IL-1Ra + OPG||1.3 ± 0.8†||2.7 ± 0.9†||1.1 ± 0.5†|
To directly compare the effects of single and combined TNF, IL-1, and RANKL blockade on synovial inflammation, bone erosion, and cartilage damage, we plotted the blocking potentials on these 3 processes against each other (Figure 8). Comparing the effects on synovial inflammation and bone erosion (Figure 8A), a shift toward more pronounced effects on bone erosion was evident, suggesting that these cytokine-based therapies more effectively inhibited bone erosion than synovial inflammation. This was especially true for anti-TNF and OPG monotherapy. However, combination therapy, such as anti-TNF/IL-1Ra, completely blocked both pathologic features and thus appeared to be especially efficient. Synovial inflammation plotted against cartilage damage (Figure 8B) showed similar effects of most treatments, except that OPG weakly affected synovial inflammation but not cartilage damage. Comparison of bone erosion and cartilage damage (Figure 8C) showed that most therapeutic approaches except IL-1Ra monotherapy had a propensity toward more effective inhibition of bone erosion than cartilage destruction. This was especially true for therapies involving OPG, as well as for monotherapy with anti-TNF.
To assess whether single and combined TNF, IL-1, and RANKL blockade induced a change in the levels of expression of the human TNF transgene, the amount of circulating human TNF was measured in sera from all animals (Table 3). No significant difference in serum levels of human TNF was found between any of the treatment groups, indicating that blockade of TNF, IL-1, and RANKL had no influence on the expression of human TNF. In addition, serum levels of murine sTNFRI in the various treatment groups were not significantly altered. Levels of IL-1, however, were significantly decreased in the groups that received anti-TNF, suggesting that blockade of TNF activity inhibited downstream expression of IL-1. Similar results were obtained for sRANKL levels, which were also decreased after anti-TNF treatment as well as IL-1Ra treatment.
|Treatment||Human TNF||Murine sTNFRI||IL-1||sRANKL|
|Baseline (6-week-old mice)||30.0 ± 3.3||411 ± 22||10.9 ± 2.1||101 ± 34|
|Control||29.3 ± 1.4||460 ± 66||15.3 ± 1.8||530 ± 81|
|Anti-TNF||33.3 ± 2.7||516 ± 24||<3†||139 ± 36†|
|IL-1Ra||29.3 ± 2.0||520 ± 45||13.0 ± 3.2||269 ± 80†|
|OPG||38.0 ± 4.5||401 ± 20||9.6 ± 3.7||517 ± 99|
|Anti-TNF + IL-1Ra||39.6 ± 2.8||356 ± 53||<3†||91 ± 39†|
|Anti-TNF + OPG||35.1 ± 1.1||453 ± 43||<3†||129 ± 28†|
|IL-1Ra + OPG||34.0 ± 1.0||416 ± 16||10.3 ± 3.2||185 ± 91†|
|Anti-TNF + IL-1Ra + OPG||33.0 ± 3.2||574 ± 89||<3†||99 ± 1†|
In the present study, we analyzed the efficacy of TNF, IL-1, and RANKL blockade, either as monotherapy or as combined therapy, in inhibiting synovial inflammation, bone erosion, and cartilage destruction in TNF-driven arthritis. Comparison of the effects of 8 different therapeutic regimens yielded 3 major conclusions: complete remission of disease is hardly achieved by any of the monotherapies (18), not even by TNF inhibition; combined therapies are highly superior to monotherapy in inhibiting inflammation, bone erosion, and cartilage damage; and these therapies vary in their ability to affect these 3 pathologic processes. The effects of the monotherapies and the double-combination therapies on the 3 pathologic processes are summarized in Figure 9.
Systemic overexpression of TNF is sufficient to trigger a molecular process that leads to chronic synovitis, bone erosion, and cartilage destruction. The clinical and histopathologic changes induced by TNF in hTNFtg mice therefore closely resemble the changes found in human RA. Although they are considered to be highly complex, the pathophysiologic events in RA in humans are crucially influenced by TNF, as has been demonstrated by the clinical efficacy of TNF blockers. Assessments of surrogate markers for all 3 events, synovial inflammation (swelling and tenderness seen clinically), bone erosion (erosions seen on radiographs), and cartilage destruction (joint space narrowing seen on radiographs), have unraveled 2 interesting features of TNF blockade in human RA. First, attempts to block TNF alone barely achieve complete remission of disease (1, 6–9). Second, the efficacy of TNF blockade varies among the 3 pathologic features, as illustrated by the fact that in many patients, the effects of TNF on bone are independent from a clinical response in the signs and symptoms of disease (6). Why are these effects so different? Are inflammation, bone erosion, and cartilage destruction governed by different pathophysiologic mechanisms? Is TNF not a “master switch” in these events?
In the present studies, we used the hTNFtg mouse model to ensure that all pathophysiologic processes were based on the overexpression of the single cytokine TNF. This enabled us to selectively study how TNF-driven synovial inflammation, bone erosion, and cartilage destruction were influenced by blockade of TNF itself, by IL-1 and/or RANKL as 2 essential downstream mediators of TNF, or by a combination of these approaches. To yield optimal effects, we used very high doses of anti-TNF antibody, IL-1Ra, and OPG. Despite these high doses and early initiation of treatment (at the onset of clinical arthritis), anti-TNF treatment did not completely overcome synovial inflammation. The same results were observed with cartilage damage, which was inhibited, but not completely blocked, by anti-TNF treatment. These observations are in contrast with the fact that selective TNF blockade almost completely blocked bone erosion. Moreover, the findings are astonishing, since TNF is the only primary pathogenic factor in this animal model.
Thus, it appears that complete inhibition of synovial inflammation and cartilage damage is hardly achievable by just blocking TNF alone. The reason for this is not entirely clear; however, 2 mechanisms should be considered.
First, the effects of TNF may be reduced, but not completely abrogated, by anti-TNF antibody. This is supported by a) the presence of histologic signs of inflammation and cartilage damage and b) increased levels of molecules downstream of TNF, such as MMPs 3, 9, and 13 in synovial tissue and articular cartilage as well as IL-1 in serum, in hTNFtg mice treated with anti-TNF. It therefore appears that for synovial inflammation and cartilage destruction, even minimal local amounts of TNF are sufficient.
Second, a 2-phase model can be assumed whereby in the first phase, transgene-driven production of TNF before the appearance of clinical signs of arthritis might lead to synovial cell activation, increased cytokine production, and increased cytokine receptor levels. In the second phase, even low local levels of TNF (i.e., despite anti-TNF antibody treatment) might still trigger membrane-bound TNF receptors on activated cells to induce the production of high levels of IL-1. This idea is strengthened by the presence of inflammatory infiltrates in the synovium of hTNFtg mice at baseline, which had a cellular composition similar to that found in later stages of disease. In addition, MMP expression by synovial cells and chondrocytes and systemic levels of human TNF, mouse sTNFRI, and IL-1 were elevated at baseline, suggesting that the first phase of human TNF overexpression already entailed profound molecular and cellular changes in the synovial tissue before the arthritis became clinically apparent. Molecules induced by TNF, such as IL-1, could then perpetuate arthritis even more efficiently and require only low levels of TNF to fuel the disease. For example, IL-1, but not TNF, can stably up-regulate factors important for MMP synthesis, such as the Ets family of transcription factors (19).
Taken together, these data may explain the partial therapeutic resistance to TNF blockers in RA. In fact, the observation that TNF blockade is a far more efficient blocker of bone erosion than synovial inflammation is consistent with clinical results showing radiologic responses to TNF blockade even in patients who have small clinical responses. Indeed, RANKL levels and osteoclast numbers were most effectively influenced by anti-TNF treatment in our experiments; moreover, these parameters were not increased in mice at the baseline assessment.
TNF is a potent inducer of IL-1 (20), which itself is considered to be a crucial mediator in the pathogenesis of destructive arthritis (21, 22). Hypothesizing that TNF-mediated up-regulation of IL-1 is responsible for at least part of the effects of TNF, it was surprising that even high doses of IL-1Ra administered by continuous infusion were not sufficient to significantly block synovial inflammation, bone erosion, or cartilage damage in hTNFtg mice. This weak response to IL-1Ra may be the result of TNF-mediated effects bypassing IL-1, or it may be due to the possibility that complete blockade of TNF-activated IL-1 is difficult to achieve in this model, even when a continuous infusion of IL-1Ra is used. In fact, the observation that the combination of anti-TNF/IL-1Ra completely blocked synovial inflammation, bone erosion, and cartilage damage points to the latter effect. It favors the concept that IL-1 is mediating a major part, if not most, of the TNF-induced pathophysiologic processes in the synovial membrane and its neighboring structures. However, with regard to the blockade of IL-1 by IL-1Ra, these data indicate a higher therapeutic efficacy of IL-1Ra if concentrations of TNF are low. Thus, at least in conditions where IL-1 expression is driven by TNF, a combined approach of reducing TNF and blocking IL-1 is much more efficient than a single therapeutic approach.
Data from clinical studies support these concepts. First, monotherapy with IL-1Ra, although of proven clinical benefit, does not eliminate the clinical signs and symptoms of arthritis in a majority of patients (23, 24). Second, the effects of IL-1Ra on surrogate markers of bone erosion and cartilage damage appear to be superior to its effects on inflammation (11). This observation is consistent with the findings of our quantitative histologic assessment of inflammation, bone erosion, and cartilage damage in hTNFtg mice, which showed a predilection for IL-1Ra to inhibit the latter 2 pathophysiologic processes. Third, preliminary results of combined blockade of TNF and IL-1 suggest that such approaches may have additive effects, at least in experimental models (25, 26).
Blockade of RANKL-mediated osteoclast differentiation and function by OPG treatment (12–14) or by genetic deletion of RANKL (15) has proved to be powerful tool for blocking arthritic bone erosion without affecting joint inflammation. Data from our previous studies have shown that OPG was effective in hTNFtg mice (13) and suggested that RANKL, as key downstream mediator of TNF, is crucially involved in the formation of local bone erosion in the course of TNF-driven arthritis. TNF itself is not capable of triggering osteoclastogenesis on its own (27, 28), but it induces profound osteoclastogenesis in the presence of only minimal amounts of RANKL; the similar efficacy of anti-TNF and OPG in reducing the number of osteoclasts and bone erosions is consistent with these observations. Synovial inflammation and cartilage destruction, however, were not affected by OPG treatment, which suggests a selective role in arthritic bone erosion.
The effects of OPG, and thus blockade of RANKL–RANK interactions, may be more complex under conditions of TNF blockade: Whereas OPG was ineffective as monotherapy, the combination of OPG and anti-TNF was highly effective in reducing synovial inflammation. Thus, RANKL–RANK interactions regulated by TNF may also exert a certain proinflammatory role in the synovium. Blocking these interactions by OPG alone, however, seems difficult to accomplish and requires an additional reduction of TNF, whereas blocking bone erosion is at least partly achievable by OPG monotherapy. The cellular basis of this process remains to be elucidated, but may involve the interaction of RANKL-expressing T cells and fibroblasts with RANK-positive cells (16, 29), such as cells of the monocyte/macrophage lineage or dendritic cells (30). Mature osteoclasts, however, are unlikely to have a major impact on synovial inflammation, since TNF-mediated synovial inflammation is unaffected by the lack of osteoclasts (17). The main target of OPG, however, remains bone erosion. This is underlined by its lack of effect on cartilage degradation, although RANKL and its receptor are expressed by chondrocytes (31).
In summary, we have shown that IL-1 and RANKL are key downstream mediators of TNF-driven arthritis. Blockade of TNF combined with inhibition of either molecules, preferentially IL-1 and to a lesser extent RANKL, may prove to be a suitable strategy for achieving complete remission of symptoms in a majority of patients with RA.
We thank Birgit Türk for excellent technical assistance.