Adjuvant-induced arthritis (AIA) in the rat is an experimental model reproducing some immunological aspects of rheumatoid arthritis (RA) such as genetic linkage and T-cell dependence 1, as well as several pathological features including chronic inflammation, involvement of peripheral joints, polysynovitis and secondary destruction of cartilage and bone 2. Its relevance to the pathogenesis of RA is further supported by the demonstration that pro-inflammatory cytokines are highly expressed in the developing arthritic process 3 and that clinically relevant anti-cytokine therapy decreased arthritis severity when used alone 45 or in combination 67. In addition, the AIA model reproduces most of the bone changes found in RA 1, including inflammatory bone loss, which has been linked to an increased risk of fracture 8. Finally, the administration of osteoprotegerin, a decoy receptor, prevented cortical and trabecular bone loss in arthritic rats 9, suggesting that induction of osteoclast differentiation by receptor activator of nuclear factor κB ligand (RANKL) in inflammatory synovium could have a role 10. This model is therefore suitable for the study of the anti-arthritic and bone protective effects of drugs thought to regulate cytokine expression at the gene level, such as peroxisome proliferator-activated receptor (PPAR)-γ agonists.

PPARs are ligand-inducible nuclear trans-acting factors belonging to the steroid receptors family 11. Among the three characterized isotypes, PPAR-α is expressed essentially in tissues contributing actively to the catabolism of fatty acids (mainly in liver, and less markedly in brown fat, kidney, heart and skeletal muscle), where it regulates the expression of genes involved in fatty acid uptake and ω-oxidation or β-oxidation 12. PPAR-α is also expressed in endothelial and vascular smooth muscle cells, as well as in macrophages and foam cells, where it contributes to the control of inflammation 1314. PPAR-β/δ is expressed ubiquitously and takes part in the reverse transport of cholesterol and the oxidation of fatty acids 15. It has profound anti-obesity and anti-diabetic actions in animal models 16 and has also been linked to wound healing 17. PPAR-γ is highly expressed in adipose tissue, where it has a pivotal role in adipocyte differentiation and lipid storage 12. Its activation has been linked to insulin-sensitizing properties that have entered the clinics 1418 and to the suppression of the release of cytokines, resulting in anti-inflammatory effects 19.

The anti-arthritic potency of PPAR agonists has been demonstrated only rarely in patients with RA 20. In contrast, several studies have demonstrated the ability of PPAR agonists to decrease the severity of experimental polyarthritis 21 with a major effect on the expression of inflammatory genes 222324 or on oxidative stress 2224. However, some data were obtained with 15-deoxy-Δ^12,14-prostaglandin J₂ 2425, which is known to have anti-inflammatory properties independently of PPAR-γ activation 26. Moreover, synthetic agonists were sometimes administered in a non-classical way such as the oral use of 10% dimethylsulphoxide as a vehicle 24 or repeated intraperitoneal administration 25, which could possibly interfere with the arthritic process 27. Finally, daily doses of thiazolidinediones (TZDs) as high as 100 mg/kg/day were reported to be effective in experimental arthritis 2125 although these doses are far above those required to restore insulin sensitivity. As glitazones are used primarily as antidiabetic agents, we decided to study the effects of rosiglitazone and pioglitazone on the arthritic process, cartilage changes and secondary bone loss when given orally at doses including those shown previously to be effective as insulin sensitizers 28293031.

In the present study we show that rosiglitazone 10 mg/kg/day or pioglitazone 30 mg/kg/day were required to decrease inflammation-induced fever and arthritis severity. At these anti-inflammatory doses, TZDs decreased synovitis and the expression of several cytokines and growth factors (TNF-α, IL-1 and basic fibroblast growth factor (bFGF)) without affecting neovascularization. However, none of the TZDs decreased proteoglycan changes in arthritic cartilage while preventing bone erosions and inflammatory bone loss. Both molecules induced PPAR-γ-dependent responses in adipose tissue but the maximal anti-arthritic effect was accompanied by increased fatness in animals. These data demonstrate that the anti-inflammatory potency of TZDs is of poor relevance to their insulin-sensitizing properties and suggest that a strong activation of PPAR-γ may expose arthritic patients to drawbacks secondary to excessive adipocyte differentiation.

In the present study, neither rosiglitazone nor pioglitazone delayed the onset or reduced the incidence of arthritis, suggesting that the immunological spread of the disease was not impaired by the dose regimen used. A previous report showed that PPAR-γ agonists remained anti-arthritic when given after the early sensitization phase of adjuvant arthritis 21, although a decreased immunological response contributed partly to the efficacy of THR0921, a novel PPAR-γ agonist, in mice developing collagen-induced arthritis 39. The dose-ranging study of TZDs was performed by biotelemetry, which provides a unique opportunity to measure body temperature and locomotive activity continuously in freely moving conscious rodents over the study duration 33. To some degree, telemetry reproduces the clinical evaluation of RA because locomotive activity contributes to algofunctional indices, whereas fever is not unusual during acute flares 40. The arthritic animals displayed a biphasic febrile response that was previously shown to reflect inflammation linked to sensitization and immune arthritis, respectively, and to follow the systemic release of pro-inflammatory cytokines 32. In our experimental model, rosiglitazone and pioglitazone decreased both fever peaks at the highest doses tested but they remained ineffective towards loss of mobility, which is a pain-dependent event 32. Although TZDs are multistep inhibitors of the inducible arachidonic acid cascade 41, these data do not support a major contribution of prostaglandin E₂, since cyclo-oxygenase inhibitors were shown to improve the hypomobility of arthritic animals 33.

These anti-inflammatory properties of TZDs may be due to a decreased production of pro-inflammatory cytokines such as TNF-α, IL-1β and IL-6, which have a key role in fever 42 and were shown previously to decrease in the bloodstream of arthritic mice treated with rosiglitazone 23. In support of this proposal, we demonstrated that the highest doses of TZDs decreased the expression of IL-1β and TNF-α in inflamed synovium, which is a primary source for systemic inflammatory cytokines 3. The major finding of this dose-ranging study was therefore to indicate that the doses of TZDs required to decrease inflammation were equal to or greater than those sufficient to restore insulin sensitivity, as reported separately for troglitazone 25.

The study of the pathophysiological findings of polyarthritis further demonstrated that rosiglitazone at 10 mg/kg/day and pioglitazone at 30 mg/kg/day decreased most aspects of arthritis severity. These data confirmed that agonists of PPAR-γ are potential therapeutic agents for arthritis 21232539, although they displayed a variable ability to activate this receptor subtype 3943. From that point of view, rosiglitazone was shown to be at least 10-fold more active than pioglitazone in transactivation assays with murine or human PPAR-γ chimeric receptors 43. In addition, pharmacokinetics studies demonstrated that a single oral dose of 10 mg/kg/day in rats provided a maximal plasma concentration that was twice as high with rosiglitazone 44 as with pioglitazone 45. In addition, the plasma levels of pioglitazone were shown previously to be lower in male rats than in female rats for a given dose 45. As a consequence, one might expect that the threefold higher dose of pioglitazone would provide circulating levels in the same range as rosiglitazone in our arthritic male rats. This threefold dose ratio between pioglitazone and rosiglitazone was also consistent with their therapeutic use in patients with type 2 diabetes, although it cannot be considered a reliable indicator of their transactivating properties on PPAR-γ. Of particular note is our result that animals receiving 30 mg/kg/day of pioglitazone had a significantly higher gain in body weight before arthritis onset. To our knowledge, this is the first description of such a side effect in rodents developing arthritis, although it was highly consistent with the weight gain observed in diabetic patients treated with TZDs 46. Furthermore, the DEXA analysis revealed a higher than normal percentage of fat mass in these animals, which reproduced the increased fatness seen in the clinics, even if a favourable redistribution of fat from visceral to subcutaneous depots was also reported 47. This drawback was probably attributable to the ability of PPAR-γ to induce adipocyte differentiation 48, because we demonstrated that both TZDs stimulated the expression of adiponectin, a PPAR-γ-sensitive gene 49, in the adipose tissue of arthritic rats. However, the increased fatness was observed only in the group displaying the less severe arthritis. This suggests that a lower caloric cost of the disease could also contribute to the changes in body weight observed in animals treated with 30 mg/kg/day of pioglitazone 50.

The reduction of synovitis in animals treated with 10 mg/kg/day of rosiglitazone or 30 mg/kg/day of pioglitazone extended the general meaning that inflamed synovium was a target for the anti-arthritic potency of TZDs 2122232539. The decrease in synovial hyperplasia seemed unlikely to have resulted from an increased apoptosis of synovial fibroblasts 25, because rosiglitazone had been shown to decrease apoptotic cells in connected tissues around the joints of arthritic mice 22. As mentioned previously, we demonstrated that both TZDs reduced the expression of IL-1β and TNF-α in arthritic synovium, a finding consistent with a recent report of their decrease by the PPAR-γ agonist THR0921 in joints of mice with established collagen-induced arthritis 39. This could be due to the ability of PPAR-γ agonists to inhibit the NF-κB pathway in arthritic tissues 22. However, differences between arthritis models might explain why we failed to observe any significant decrease in the chemokine monocyte chemoattractant expression in TZDs-treated rats, in contrast with mice treated with THR0921 39. We showed further that TZDs decreased the expression of bFGF, which is a powerful mitogen for both synovial fibroblasts and endothelial cells whose neutralization was reported to attenuate arthritis severity 51. Its reduced expression could therefore contribute to the anti-arthritic potency of these molecules. Finally, we demonstrated that neovascularization and expression of VEGF were reduced by neither pioglitazone nor rosiglitazone, despite a marked decrease in synovitis. Angiogenesis has a pivotal role in RA by increasing the exchange of cells, cytokines and growth factors from the bloodstream to the joint cavity, thereby promoting tissue infiltration and pannus growth 52. Pro-inflammatory cytokines 53 and bFGF 54 are strong inducers of VEGF release, and their inhibition by TZDs may lower neovascularization in the same way as anti-cytokine therapies in rheumatoid patients 53 or arthritic rats 55. In contrast, TZDs were reported to stimulate VEGF expression in several cell types 56 and to increase its plasma level in type 2 diabetic patients 57, therefore suggesting that they could counterbalance the impact of their anti-inflammatory effect on the formation of new vessels.

The decrease in ³⁵S-sulphate incorporation into proteoglycans 58 and secondary loss of proteoglycan content in cartilaginous tissue 59 are hallmarks of the chronic phase of adjuvant arthritis. In addition, biochemical changes in patellar cartilage were shown to be representative of the entire arthritic joint 60 and to be very sensitive to the inhibitory effect of pro-inflammatory cytokines 61. Thus, our data were highly homogeneous because the decreased proteoglycan synthesis and aggrecan expression in cartilage resulted in a significant loss of proteoglycans and decreased cartilage staining in arthritic controls. However, the anti-inflammatory effect of TZDs failed to provide significant protection to cartilage in both the knee and ankle joints. Our results differ from previous reports of decreased histological lesions in arthritic joints of rodents treated with glitazones 2325, although this difference could be supported, at least in part, by the scoring system used. Indeed, cartilage and bone changes were always assessed concomitantly 2325, whereas we checked separately for cartilage 35 and bone 36 changes in our arthritic rats. Had we considered both changes together, we would have demonstrated a possibly protective effect of TZDs in arthritic joints but it would have been supported by an improvement in bone lesions rather than in cartilage lesions. In addition, we demonstrated recently that selective agonists of the three PPAR isotypes decreased the anabolic response of chondrocytes to TGF-β 62. Because rat chondrocytes embedded in alginate beads are close to cartilage samples 63, one cannot exclude the possibility that the lack of restoration of proteoglycan metabolism by TZDs could be also supported by a diminished response of arthritic cartilage to TGF-β.

A major finding of the present work was that rosiglitazone and pioglitazone prevented arthritis-induced bone loss. PPAR-γ agonists were shown previously to decrease bone erosion 2324 and to improve the radiographic score 24 in arthritic rodents. The contribution of synovitis to bone loss, ranging from focal erosions to periarticular or generalized osteopenia, has become a very active topic in RA 24. Further confirming histological findings, the DEXA analysis showed that the protective effect of TZDs was detectable on the whole body and in selected regions of interest (data not shown), suggesting that they were active on cortical and trabecular bone. Bone loss is a classical feature of adjuvant arthritis, in which an increase in the number of osteoclasts and in the formation of osteoclast precursors from the monocyte/macrophage compartment in the synovium was reported within few days after disease onset 64. Focal bone erosions and generalized bone loss are thought to be secondary to an overload of pro-inflammatory cytokines 6, whose osteoclastogenic effect has recently been ascribed to RANKL 10. Thus, administration of osteoprotegerin, which prevents the binding of RANKL to RANK, decreased arthritis-induced bone loss 65, suggesting that TZDs may prevent bone loss, at least in part, by decreasing the production of inflammatory cytokines in inflamed synovium. In additon, PPAR-γ agonists were shown to decrease the phosphorylation of the inhibitor IκB in arthritic joints 22, and inhibition of the NF-κB pathway prevented inflammatory bone destruction in vivo by blocking osteoclastogenesis 66. Moreover, PPAR-γ agonists were shown to inhibit RANKL-dependent differentiation of osteoclasts in bone marrow cells 3967 or peripheral blood mononuclear cells 2168. A decrease in bone resorption was also supported by the decreased urinary excretion of deoxypyridinoline 38 in rats treated with TZDs. The bone-protective potency of TZDs is very provocative because these molecules were reported to cause bone loss in rodents 6970 or type 2 diabetic patients 7172 by switching the differentiation of bone precursors towards an adipogenic phenotype 6970. Our data therefore suggest that the differentiating effect of PPAR-γ agonists on adipocytes 48 may have a variable impact on bone depending on the inflammatory status of the body.

The present work confirms that an oral intake of the antidiabetics rosiglitazone or pioglitazone can reduce the severity of arthritis but indicates that the doses required for an anti-arthritic effect exceed those sufficient to restore insulin sensitization. This effect is due, at least in part, to their ability to decrease the expression of pro-inflammatory cytokines in inflamed synovium without affecting neovascularization. Thiazolidinediones prevent inflammatory bone loss but do not improve changes in proteoglycans in arthritic cartilage. However, activation of PPAR-γ in adipose tissue is concomitant with increased adiposity in arthritic animals, suggesting that a strong activation of PPAR-γ may expose arthritic patients to the drawbacks of excessive adipocyte differentiation.

ACO = Acyl-CoenzymeA oxidase; AIA = adjuvant-induced arthritis; ANOVA = analysis of variance; bFGF = basic fibroblast growth factor; BMC = bone mineral content; DEXA = dual-energy X-ray absorptiometry; IL = interleukin; MCP-1 = monocyte chemotactic protein-1; NF = nuclear factor; PLSD = protected least-squares difference; PIO = Pioglitazone; PPAR = peroxisome proliferator-activated receptor; RA = rheumatoid arthritis; RANKL = receptor activator of nuclear factor κB ligand; ROSI = rosiglitazone; RT-PCR = polymerase chain reaction with reverse transcription; TNF = tumour necrosis factor; TZD = thiazolidinedione; VEGF = vascular endothelial growth factor.

The authors declare that they have no competing interests.

MK performed all in vivo and molecular studies and drafted the manuscript. DM performed molecular studies and drafted the manuscript. AB and SS performed molecular studies and statistical analysis. MM performed the DEXA analysis. GW performed the DEXA analysis and contributed to the study design. PN supervised the study design and the manuscript. JYJ conceived the study and participated in its design and final presentation. All authors read and approved the final manuscript.