The Peroxisome Proliferator–Activated Receptor γ Agonist Pioglitazone Preserves Bone Microarchitecture in Experimental Arthritis by Reducing the Interleukin-17–Dependent Osteoclastogenic Pathway




To investigate the effect of pioglitazone on inflammation-induced bone loss and changes in bone microarchitecture in rats with adjuvant-induced arthritis (AIA), focusing on the contribution of interleukin-17 (IL-17) and the balance of RANKL and osteoprotegerin (OPG).


Male Lewis rats sensitized with Freund's complete adjuvant were treated orally for 21 days with 30 mg/kg/day of pioglitazone or vehicle. Arthritis severity was evaluated by clinical and histologic examination. Bone mineral density (BMD) was assessed by dual x-ray absorptiometry. The therapeutic effect of pioglitazone on changes of the bone architecture was determined by micro–computed tomography (micro-CT). Levels of RANKL, OPG, and IL-17 were determined by serum immunoassay and by synovial tissue immunohistochemistry. Messenger RNA for IL-17 and retinoic acid receptor–related orphan nuclear receptor γt (RORγt) was evaluated by quantitative reverse transcription–polymerase chain reaction and IL-17 promoter activity by gene-reporter assay.


Micro-CT analysis revealed that pioglitazone treatment reduced arthritis severity and bone erosion scores and increased BMD in comparison to vehicle treatment. Cortical bone thickness was preserved, although the major beneficial effect of pioglitazone was on indices of the trabeculae, especially trabecular separation. Pioglitazone reduced the ratio of RANKL to OPG, in both the serum and the inflamed synovium. Circulating levels of IL-17 were significantly reduced by pioglitazone treatment, as were the percentages of IL-17–positive cells, mainly polymorphonuclear cells, in the inflamed synovium. Induction of IL-17 was strictly dependent on the binding of RORγt to IL-17 promoter, and lentiviral overexpression of peroxisome proliferator–activated receptor γ (PPARγ) reduced the expression of RORγt.


Pioglitazone decreased the level of inflammatory bone destruction and protected the bone microarchitecture in rats with AIA by controlling the circulating and local expression of IL-17, with a subsequent decrease in the RANKL-to-OPG ratio. Along with the inhibition of RORγt expression after PPARγ overexpression, these findings provide evidence of the major contribution of reduced IL-17/RANKL-dependent osteoclastogenesis.

Rheumatoid arthritis (RA) is a chronic autoimmune disease characterized by symmetric synovitis, inflammatory exudates in joint cavities, and profound cartilage and bone changes ([1]). One of the hallmarks of the disease is the destruction of periarticular bone, which leads to functional disability of the joint ([2]). Patients with RA also develop progressive systemic bone loss. Indeed, the formation of osteoclasts is enhanced in inflammatory conditions and is not balanced by increased activity of bone-forming osteoblasts. Such an imbalance results in joint destruction and increased fracture risk, both of which are major causes of functional disability in RA ([2]).

Th17 cells have been shown to be involved in the pathogenesis of autoimmune diseases, including RA, and are characterized by their ability to produce interleukin-17 (IL-17). The differentiation and survival of Th17 cells are regulated by IL-6, transforming growth factor β, IL-1β, IL-21, and IL-23 under the control of retinoic acid receptor–related orphan nuclear receptor γt (RORγt) ([3]). Murine models suggest that IL-17 may be a critical mediator of joint destruction, since its blockade reduces arthritis severity ([4]). The osteoclastogenic potential of human IL-17 is now well established through the modulation of the RANK/RANKL/osteoprotegerin (OPG) pathway ([5, 6]). The ratio of RANKL to OPG in the bone marrow is an important determinant of bone mass control in normal and disease states.

Peroxisome proliferator–activated receptors (PPARs) are members of the nuclear hormone receptor superfamily that behave as ligand-activated transcription factors in response to endogenous fatty acids and eicosanoids ([7]) or to isotype-selective synthetic compounds, such as fibrates or thiazolidinediones ([8]). Besides their ability to reduce the metabolic syndrome, several studies have shown that activation of one or more PPAR isotypes (PPARα, PPARβ/δ, or PPARγ) had variable power to control inflammation ([9, 10]). We and other investigators have demonstrated an antiinflammatory effect of PPARγ agonists on experimental polyarthritis ([11-14]), which was supported mainly by the findings of reduced activation of the proinflammatory cytokine network, including tumor necrosis factor α (TNFα) ([14, 15]), IL-6, or IL-1β ([11-14, 16]). Cell studies and animal studies further showed that most of this cytokine-suppressive capacity was linked to the ability of PPAR agonists to inhibit transcription factor pathways by mechanisms that included the formation of direct complexes with activator protein 1 (AP-1) and with NF-κB family members ([11, 17]).

Since thiazolidinediones (TZDs) have become clinically available for the treatment of type 2 diabetes mellitus, several epidemiologic studies have suggested that these compounds increased the rate of bone loss, and therefore the risk of fracture, at least in women with the disease ([18]). The implication that activation of PPARγ may have adverse effects on the skeleton was further supported by animal studies showing that prolonged administration of rosiglitazone ([19]) or pioglitazone ([20]) to healthy rodents resulted in a significant decrease in bone mineral density (BMD), or that PPARγ haploinsufficiency increased bone mass throughout the life of the mice ([21]). More recently, PPARγ was also shown to regulate osteoclastogenesis, since its targeted deletion in osteoclasts using a TIE-2/Cre/loxP-flanked mouse model caused osteopetrosis ([22]), which further suggests that PPARγ activation may hinder bone homeostasis.

In contrast to the capacity of TZDs to decrease bone mass in patients with diabetes mellitus, a few animal studies support the notion that TZDs may paradoxically display some bone-sparing properties in severe inflammatory conditions. Thus, PPAR agonists with an established antiarthritic effect have also been shown to decrease osteoclast formation from bone marrow progenitors ([23]) or peripheral blood mononuclear cells ([17]). Our previous study also demonstrated that at a dosage that reduced arthritis severity, pioglitazone prevented inflammation-induced bone loss, as assessed by dual x-ray absorptiometry (DXA) ([13]).

In the present study, we used adjuvant-induced arthritis (AIA) in rats as a model of chronic inflammation to investigate the effect of pioglitazone on bone microarchitecture and circulating mediators of osteoclastogenesis. We found that an antiarthritic dose of pioglitazone reduced bone loss in the lumbar spine and femur, and it prevented trabecular weakening of the bone more efficiently than cortical thinning. Pioglitazone decreased circulating levels as well as joint expression of IL-17 in inflammatory cells while reducing the ratio of RANKL to OPG in serum and synovium. We also demonstrated that the expression of IL-17 in the Jurkat T cell line was strongly controlled by PPARγ activation, in a RORγt-dependent way, further supporting the notion of an inhibitory effect of pioglitazone on transcriptional activation of IL-17. These results are the first to demonstrate that the paradoxical “bone-protective” effect of pioglitazone during chronic arthritis occurred through the control of IL-17/RANKL–driven osteoclastogenesis.



Twenty-two inbred male Lewis rats (Charles River France) weighing 150–175 gm were acclimated for 1 week in the laboratory before experimentation. Animals were housed in groups of 3 or 4 in solid-bottomed plastic cages, with access to tap water and standard pelleted rodent chow (Scientific Animal Food and Engineering) ad libitum. Room temperature was set at 23°C (±1°C), and animals were subjected to a 12-hour cycle of light/dark. All experiments were performed in accordance with national animal care guidelines and were preapproved by the local ethics committee. Arthritis sensitization, BMD measurements, and blood sampling at necropsy were performed while the animals were under general anesthesia with the volatile anesthetic isoflurane (AErrane; Baxter France).

Induction of arthritis and treatment regimen

Arthritis was induced according to the protocol we have previously described ([13]). Rats with AIA were randomly assigned to treatment with vehicle (vehicle-treated arthritic controls) or treatment with 30 mg/kg/day of pioglitazone. Naive animals without AIA were treated with vehicle and served as additional controls (vehicle-treated controls). Pioglitazone was given orally, at a volume of 1 ml/100 gm of body weight, from the day of sensitization until necropsy (day 21). Treatment was prepared daily from commercially available pills (Actos; Takeda) as a suspension in 0.5% carboxymethylcellulose (CMC). Naive rats and arthritic controls received CMC only.

Assessment of arthritis

The severity of arthritis was assessed by the periodic monitoring of total body weight and arthritis scores, which were performed at the indicated times, as previously described ([13]).

Histologic analysis of ankle joints

Ankle joints were processed as described elsewhere ([13]) and stained with hematoxylin–eosin–saffron (HES) or toluidine blue. Bone erosion and synovium inflammation were graded as previously described ([24]).

Immunohistochemical analysis

Ankle sections were rehydrated in a graded ethanol series and then incubated for 10 minutes in 3% hydrogen peroxide. Sections were washed 3 times in water and incubated overnight at 70°C in citrate buffer, pH 6 (Zytomed Systems). After washing in water and phosphate buffered saline (PBS), slides were incubated for 20 minutes in blocking serum and then overnight at 4°C with IL-17 (Tebu-Bio), OPG (Santa Cruz Biotechnology), and RANKL (Santa Cruz Biotechnology) polyclonal primary antibodies. After washing in PBS, slides were covered for 30 minutes with secondary antibodies conjugated to a micropolymer containing horseradish peroxidase (ImmPRESS kit; Vector) and then washed in PBS. Staining was developed with 3,3′-diaminobenzidine (0.05% in hydrogen peroxide) (Novostain Super ABC kit; Novocastra). All slides were counterstained for 10 seconds with Harris' hematoxylin (Réactifs RAL). Identification of IL-17–positive cells on HES-stained serial sections was performed by 2 pathologists who were blinded with regard to the treatment.

BMD measurements

BMD was determined in vivo by DXA, using a QDR-4500A densitometer (Hologic) equipped with a small-animal module. Rats were anesthetized as described above and scanned as described elsewhere ([13]). Data were expressed as changes in BMD over the duration of the study, with each rat being used as its own control.

Micro–computed tomographic (micro-CT) analysis

Rats were killed, and 1 femur from each rat was carefully dissected and fixed in 10% phosphate buffered formalin, pH 7.4. Ten days later, the femurs were dehydrated successively in 70%, 95%, and 100% ethanol, cut through the middle of the diaphysis, and measured without additional preparation using a SkyScan 1072 micro-CT apparatus at 21× magnification, as previously described ([25]).

Sandwich bead immunoassay for mediators of osteoclastogenesis

Serum levels of IL-17, soluble RANKL (sRANKL), and OPG were determined on day 21 with the use of a Milliplex kit (Millipore). Assays were run in triplicate according to the manufacturer's protocol. Data were analyzed using a Bio-Plex system (Bio-Rad). A 5-parameter regression formula was used to calculate the concentrations in the samples from the standard curves (from 5 pg/ml to 20,000 pg/ml). The limits of quantification (manufacturer's data) were 1.6 pg/ml for IL-17, 1.0 pg/ml for RANKL, and 2.3 pg/ml for OPG.

Lentiviral vector production and viral transduction

A ViraPower lentiviral expression system (Invitrogen) was used to generate replication-incompetent human immunodeficiency virus 1 (HIV-1)–based lentivirus to express PPARγ and β-galactosidase proteins. Generation of lentiviral vectors was accomplished using 3 packaging plasmids, pLP1, pLP2, and pLP/VSVG (Invitrogen), which supply the helper functions as well as the structural and replication proteins required to produce the lentivirus.

Vector production

We used 293 T cells to produce lentivirus according to the manufacturer's protocol (Invitrogen). The viral supernatants were filtered and concentrated 250-fold by ultracentrifugation at 29,000 revolutions per minute for 90 minutes at 4°C in a Sorvall Wα Ultra 90 ultracentrifuge (Thermo Fisher Scientific). The pellets were resuspended overnight in PBS and then stored at −80°C. The titer of the lentiviral vector was determined with an HIV p24 enzyme-linked immunosorbent assay kit (Cell Biolabs).

Viral transduction

Cell transduction was performed in growth medium supplemented with 6 μg/ml of Polybrene (catalog no. H9268; Sigma-Aldrich). The virus-containing medium was removed the following day and replaced with fresh growth medium. On day 3, medium was replaced by growth medium containing 4 μg/ml of blasticidin (Gibco Invitrogen) to select stably transduced cells. Medium was replaced every 3 or 4 days with growth medium containing antibiotic until resistant cells could be identified.

IL-17 expression and regulation

IL-17 promoter assay

Jurkat cells were electroporated with plasmids encoding for either wild-type or RORγt binding site–deleted murine IL-17 promoter and pCMV-Renilla luciferase (Promega) using a Nucleofector kit (Amaxa/Lonza) according to the manufacturer's protocol. Briefly, 3 × 106 cells were gently mixed with 2 μg of plasmid and with 5 ng of pCMV-Renilla luciferase before electroporation using a Nucleofector kit, program X-001. Immediately after transfection, cells were split equally into 3 wells containing 10% fetal calf serum–RPMI 1640 growth medium, left to recover for 24 hours, and IL-17 promoter activity was analyzed.

Reporter gene assay

IL-17 promoter activity was measured using a dual-luciferase reporter assay kit (Promega) according to the manufacturer's protocol. Results were expressed as the mean ratio of firefly to Renilla luciferase activity.

Jurkat T cell assay for cells with the Th17-like phenotype

Jurkat T cells were stimulated with 20 nM phorbol 12-myristate 13-acetate (PMA) and 1 μM ionomycin. The concentrations of PMA and ionomycin were determined by a preliminary dose-ranging study. The duration of stimulation was 6 hours for assessment of IL-17 promoter activity and 48 hours for quantification of messenger RNA (mRNA). Cells were challenged when indicated with 30 μM pioglitazone (a generous gift from Takeda Japan).

Assays for IL-17 and RORγt expression

To quantify the expression of mRNA for the genes of interest, a quantitative reverse transcription–polymerase chain reaction (qRT-PCR) analysis was performed using LightCycler (Roche) technology. PCR was performed with a SYBR Green Master Mix system (Qiagen). The gene-specific primer pairs used were 5′-TGTCCAAACGCCGAGGCCAAT-3′ (forward) and 5′-AAGCGAGCTCCAGAAGGCCCT-3′ (reverse) for IL-17 and 5′-CGGGCCTACAATGCTGACA-3′ (forward) and 5′-GCCACCGTATTTGCCTTCAA-3′ (reverse) for RORγt. Results were expressed as the ratio of the mRNA level of each gene to that of the housekeeping gene RP29.

Statistical analysis

Data are expressed as the mean ± SEM. Arthritis scores and histologic grades were analyzed with the Kruskal-Wallis test, using StatView version 5.0 software (SAS Institute). All other data were compared by analysis of variance followed by the Bonferroni post hoc test. P values less than 0.05 were considered significant.


Effect of pioglitazone on arthritis severity

As shown in Figure 1a, arthritis became obvious by day 11 and was maximal around day 20 after sensitization. At the end of the experiment, the arthritis score averaged 12 in vehicle-treated animals, reflecting severe arthritis. This was confirmed by histologic analysis of the ankle joints (Figure 1b). Administration of pioglitazone (30 mg/kg/day) significantly reduced the arthritis score (by almost 50% on day 21) as compared to that of vehicle-treated rats with AIA. Noteworthy, the major bone erosion seen in the vehicle-treated rats with AIA (mean ± SEM 3.00 ± 0.02) as compared to the vehicle-treated controls (0.00 ± 0.02) was significantly reduced in the pioglitazone-treated rats with AIA (1.60 ± 0.11; P < 0.05 for each comparison). In addition, synovial inflammation seen in the vehicle-treated rats with AIA (mean ± SEM 5.50 ± 0.08) as compared to the vehicle-treated control rats (1.00 ± 0.32) was also reduced in the pioglitazone-treated rats with AIA (3.75 ± 0.88), but the difference was not significant.

Figure 1.

Effect of pioglitazone (PIO) on the severity of adjuvant-induced arthritis (AIA) in male Lewis rats. Rats were sensitized subcutaneously at the base of the tail with a single injection of 1 mg of Mycobacterium tuberculosis. One group of rats with AIA (n = 14) was treated once daily with 30 mg/kg of pioglitazone orally. A second group of rats with AIA (n = 13) was treated with vehicle alone (0.5% carboxymethylcellulose). A group of naive rats without AIA (n = 11) was treated with vehicle alone (controls). Shown are a, arthritis scores, b, representative ankle joint sections demonstrating synovial inflammation, c, bone mineral density values, and d, representative ankle joint sections demonstrating osteoclast density at the synovium–bone margin. Histologic sections were stained with hematoxylin–eosin–saffron; original magnification × 400. Values in a and c are the mean ± SEM. ∗ = P < 0.05 versus vehicle-treated controls; # = P < 0.05 versus vehicle-treated arthritic rats.

Bone mineral density (BMD) increased over the duration of the study in the 3 regions of interest (lumbar spine, right femur, and left femur) in the vehicle-treated controls (Figure 1c). The mean gain in BMD in arthritic vehicle-treated rats was 61% lower in the lumbar spine and 54% lower in the femur as compared to the values in vehicle-treated controls. Beyond its ability to decrease arthritis severity, pioglitazone efficiently reduced bone erosion and significantly increased the BMD in the femurs of arthritic rats almost to the levels in vehicle-treated controls. Furthermore, histologic analysis of the ankle bones showed a marked decrease in osteoclast density (∼60%) at the synovium–bone margin in arthritic animals treated with pioglitazone (Figure 1d).

Effect of pioglitazone on arthritis-induced bone changes

Three-dimensional (3-D) reconstruction of the distal femur showed that vehicle-treated arthritic animals displayed a thinning of the cortices and a massive disorganization of the trabecular network, with a large area having no trabeculae (Figure 2a). In arthritic animals treated with pioglitazone, the cortices were preserved, and the trabecular network was globally conserved, except for some holes corresponding to areas of loss of connectivity. These 3-D images were highly consistent with the bone architecture seen on the corresponding histology sections (Figure 2b).

Figure 2.

Effect of pioglitazone (PIO) on arthritis-induced changes of bone morphometry in male Lewis rats. Rats were sensitized subcutaneously at the base of the tail with a single injection of 1 mg of Mycobacterium tuberculosis. One group of rats with adjuvant-induced arthritis (AIA; n = 8) was treated once daily with 30 mg/kg of pioglitazone orally. A second group of rats with AIA (n = 7) was treated with vehicle alone (0.5% carboxymethylcellulose). A group of naive rats without AIA (n = 7) was treated with vehicle alone (controls). Shown are a, representative 3-dimensional images of the distal femur obtained by micro–computed tomography on day 21 of the study, and b, representative histologic sections of the uncalcified femoral metaphysis, which was embedded in methylmethacrylate and then stained with modified Goldner's trichrome. Original magnification × 200.

Bone morphometric parameters

In arthritic rats, the 2-D section images of the primary spongiosa, the secondary spongiosa, and the central diaphysis showed a significant reduction in cortical bone thickness in all areas (−37%, −42%, and −33%, respectively) as compared to vehicle-treated controls (Table 1). Treatment with pioglitazone preserved cortical bone thickness in 2 of the 3 zones studied (−23% in the secondary spongiosa and −14% in the diaphysis) and slightly improved bone thinning in the other zone (−32% in the primary spongiosa) as compared to vehicle-treated controls.

Table 1. Effect of pioglitazone on arthritis-induced changes in bone morphometric parameters in rats*
 Vehicle-treated control ratsVehicle-treated rats with AIAPioglitazone-treated rats with AIA
  1. The mean cortical thickness in the primary and secondary spongiosa and central diaphysis, as well as the bone volume fraction (BV/TV), the trabecular thickness (TbTh), the trabecular number (TbN), and the trabecular separation (TbSp) were calculated as described elsewhere ([25]). Values are the mean ± SEM.
  2. aP < 0.05 versus vehicle-treated control rats.
  3. bP < 0.05 versus vehicle-treated rats with adjuvant-induced arthritis (AIA).
Cortical bone thickness, μm   
Primary spongiosa365.8 ± 40.4231.9 ± 21.7a249.0 ± 27.3
Secondary spongiosa376.3 ± 21.5217.1 ± 6.9a290.6 ± 19.1b
Central diaphysis401.0 ± 17.7267.6 ± 8.5a344.7 ± 10.6b
BV/TV, %30.2 ± 3.52.43 ± 0.9a14.9 ± 1.9b
TbTh, μm57.0 ± 2.032.3 ± 0.1a39.6 ± 0.3b
TbN,/mm5.3 ± 0.30.73 ± 0.2a3.7 ± 0.3b
TbSp, μm135.8 ± 35.82,120 ± 767.8a249.5 ± 47.6b

The bone morphometric parameters derived from 3-D microtomography measurements confirmed that arthritis induced an overall loss of bone (Table 1). Thus, the bone volume ratio decreased significantly in vehicle-treated arthritic rats, as did the trabecular number and trabecular thickness, in comparison to the values in the vehicle-treated controls. Consistent with the reductions in the trabecular number and trabecular thickness, vehicle-treated arthritic animals displayed an increased trabecular separation as compared to that in the vehicle-treated naive controls. In rats administered pioglitazone at 30 mg/kg/day, the bone volume ratio recovered significantly, as did the trabecular number and trabecular thickness indices, in comparison to the values in the vehicle-treated rats with AIA. Pioglitazone-treated animals also had a remarkable decrease in trabecular separation as compared to the vehicle-treated arthritic controls.

Effect of pioglitazone on the RANKL/OPG balance

Circulating levels of sRANKL were increased in arthritic vehicle-treated animals as compared to those in the vehicle-treated controls (mean ± SEM 53.5 ± 4.9 pg/ml versus 24.4 ± 1.4 pg/ml), whereas OPG levels were not significantly reduced (mean ± SEM 264.2 ± 34.3 pg/ml versus 287.8 ± 20.8 pg/ml) (Figure 3a). In pioglitazone-treated arthritic rats, a significant reduction in sRANKL levels (33.0 ± 3.1 pg/ml) was observed, without a significant difference in the circulating levels of OPG (247.9 ± 26.9 pg/ml). As a consequence, the mean ratio of sRANKL to OPG increased from 0.09 to 0.22 in rats with established arthritis, and was reduced to 0.15 in rats treated with pioglitazone. Immunohistochemical analysis of the ankle joint synovium confirmed a major overproduction of RANKL in vehicle-treated arthritic animals as compared to vehicle-treated controls, whereas a modest increase in the density of OPG-positive cells was also observed (Figure 3b). Sections from pioglitazone-treated arthritic animals revealed a marked decrease in the number of RANKL-positive cells in the synovium, but no major reduction of the density of OPG-positive cells as compared to the arthritic vehicle-treated animals (Figure 3b).

Figure 3.

Effect of pioglitazone (PIO) on RANKL and osteoprotegerin (OPG) in male Lewis rats. Rats were sensitized subcutaneously at the base of the tail with a single injection of 1 mg of Mycobacterium tuberculosis. One group of rats with adjuvant-induced arthritis (AIA; n = 8) was treated once daily with 30 mg/kg of pioglitazone orally. A second group of rats with AIA (n = 7) was treated with vehicle alone (0.5% carboxymethylcellulose). A group of naive rats without AIA (n = 7) was treated with vehicle alone (controls). Shown are a, circulating levels of soluble RANKL (sRANKL) (left) and OPG (middle), as well as the ratio of sRANKL to OPG (right), and b, expression of OPG and RANKL in representative sections of the ankle joint synovium. Histologic sections were stained as described in Materials and Methods; original magnification × 400. Values in a are the mean ± SEM. ∗ = P < 0.05 versus vehicle-treated controls; # = P < 0.05 versus vehicle-treated arthritic rats.

Effect of pioglitazone on IL-17 expression

As shown in Figure 4a, circulating levels of IL-17 were significantly increased in vehicle-treated arthritic animals as compared to vehicle-treated controls (mean ± SEM 117.2 ± 5.7 pg/ml versus 20.8 ± 7.8 pg/ml). A significant reduction in the IL-17 level (52.7 ± 11.7 pg/ml) was observed in pioglitazone-treated arthritic rats on day 21.

Figure 4.

Effect of pioglitazone (PIO) on interleukin-17 (IL-17) expression in male Lewis rats with adjuvant-induced arthritis (AIA). One group of rats with adjuvant-induced arthritis (AIA; n = 8) was treated once daily with 30 mg/kg of pioglitazone orally. A second group of rats with AIA (n = 7) was treated with vehicle alone (0.5% carboxymethylcellulose). A group of naive rats without AIA (n = 7) was treated with vehicle alone (controls). Shown are a, circulating levels of IL-17, b, percentages of IL-17–positive cells in ankle joint sections, as determined by immunohistochemical analysis, c, immunohistochemical analysis of ankle joint synovial tissue in arthritic rats, and d, cellular profile of the inflamed synovium in arthritic rats. In c, C1 = IL-17 in synovium, C2 = hematoxylin–eosin–saffron (HES)–stained section of synovium, C3 = HES-stained section of bone marrow, and C4 = IL-17 in bone marrow. Original magnification × 1,000. Values in a and b are the mean ± SEM. ∗ = P < 0.05 versus vehicle-treated controls; # = P < 0.05 versus vehicle-treated arthritic rats. PMN = polymorphonuclear cells.

Immunohistochemical analysis of ankle joint sections confirmed that IL-17 was weakly expressed in the synovium of vehicle-treated controls but that 70% of the cells in the inflamed synovium of vehicle-treated arthritic animals stained positively for IL-17 (Figure 4b). HES staining of serial sections examined at high magnification revealed that IL-17–positive infiltrating cells were mainly polymorphonuclear cells (PMNs) and, to a lesser extent, lymphocytes (Figures 4c and d). Bone histologic analysis showed that all IL-17–positive cells in the bone marrow were from the granulocyte lineage and that 100% of osteoclasts stained positively for IL-17.

IL-17–positive cells were mainly located at the periphery of the synovial tissue, along the area of bone invasion. Pioglitazone treatment decreased the percentage of IL-17–positive cells by 37% (Figures 4b and d), affecting essentially the IL-17–positive PMN population.

Contribution of RORγt to the inhibitory effect of pioglitazone on IL-17.

To confirm the specificity of the inhibitory effect of pioglitazone, we further investigated the regulatory mechanisms of IL-17 expression in an in vitro model of T cells. After 48 hours of stimulation with PMA/ionomycin, we found that Jurkat cells transiently expressed mRNA for IL-17, which remained undetectable in the resting cells (Figure 5a). In our T cell line model, IL-17 induction by PMA/ionomycin was strictly dependent on RORγt binding to the IL-17 promoter, since transactivation was suppressed when we used a mutated IL-17 promoter construct in which the RORγt binding site had been deleted (Figure 5b). In addition, incubation of Jurkat cells with pioglitazone reduced the stimulating effect of PMA/ionomycin on RORγt expression by 43% (Figure 5c). Finally, in PPARγ-overexpressing T cells, no induction of RORγt was noted in response to PMA/ionomycin challenge (Figure 5c).

Figure 5.

Effect of pioglitazone (PIO) on interleukin-17 (IL-17) expression in the Jurkat T cell line. a, Induction of IL-17 mRNA expression by Jurkat T cells after stimulation with phorbol 12-myristate 13-acetate (PMA)/ionomycin (Iono), as determined by real-time quantitative reverse transcription–polymerase chain reaction (RT-PCR). Data were expressed as the ratio of mRNA for the IL-17 gene to that of the RP29 housekeeping gene. b, Dependence of PMA/ionomycin-induced transactivation of the IL-17 promoter on retinoic acid receptor–related orphan nuclear receptor γt (RORγt) binding to the promoter. Jurkat cells were nucleofected with either wild-type (prom-IL-17-Luc) or RORγt binding site–deleted (prom-ΔRORγt-IL-17-Luc) murine IL-17 promoter. After 24 hours, cells were left untreated or were treated with 30 μM pioglitazone and then were left unchallenged or were challenged for 6 hours with PMA/ionomycin. The ratio of firefly to Renilla luciferase (Luc) was determined by luminometry. c, Pioglitazone reduction of RORγt expression via activation of peroxisome proliferator–activated receptor γ (PPARγ). Jurkat cells were lentivirally transduced with β-galactosidase or PPARγ and then selected with blasticidin. Transduced cells were then left untreated or were treated with 30 μM pioglitazone and then were left unchallenged or were challenged with PMA/ionomycin. After 48 hours, RORγt expression was analyzed by real-time quantitative RT-PCR. Data were expressed as the ratio of mRNA for the RORγt gene to that of the RP29 housekeeping gene. Values are the mean ± SEM of 3 experiments. ∗ = P < 0.05 versus unstimulated cells in a, and for the indicated comparisons in b and c; # = P < 0.05 for the indicated comparisons. ND = not detected.


Pioglitazone is an insulin-sensitizing thiazolidinedione, acting as a selective agonist for the PPARγ subtype. Aside from cardiovascular safety considerations and the possible risk of bladder cancer ([26]), alterations in bone metabolism, such as increased skeletal fragility, have become a major question about the therapeutic use of TZDs in type 2 diabetes mellitus, since it remains to be clarified whether it is dependent on the drugs ([27]) rather than on the association between disease complications ([28]). On the one hand, PPARγ activation can promote the diversion of pluripotent marrow progenitors from an osteoblastic to an adipocytic lineage ([29]). On the other hand, PPARγ agonists may modulate osteoclastogenesis, but again, reported data have been the subject of controversy, with rosiglitazone being able to increase osteoclast numbers and bone resorption in aged animals ([30]) and 15-deoxy-Δ12,14-prostaglandin J2 being able to reduce osteoclast formation and activity in human and mouse models of osteoclastogenesis ([31]).

Despite the well-established antiarthritic properties of PPARγ ligands ([11]), no previous study has addressed the effect of PPARγ ligands on bone structure in chronic inflammatory conditions. The experimental model we used reproduces several pathologic features of RA, such as chronic synovitis with secondary joint destruction and systemic osteoporosis ([32]). The results of the DXA analysis are consistent with our previous report of an overall bone loss in the course of AIA ([13]) and confirm that both the axial skeleton and the peripheral joints were involved. In arthritic animals, a thinning of the cortices and a weakening of the trabecular architecture were obvious by day 21, and bone erosions were observed mainly at the joint margins. The severe fragmentation of trabecular bone ([33]) was characterized by decreased trabecular number, trabecular thickness, and bone volume ratio and increased trabecular separation. These findings are indicative of increased osteoclast resorption, which is consistent with the kinetics of osteoclastogenesis reported in this model ([34]) and suggests decreased bone strength in arthritic animals ([35]). Interestingly, arthritic animals treated with pioglitazone showed improvements in all bone morphometric parameters, although with a more marked effect on trabecular indices, especially trabecular separation, rather than on bone thickness.

Thus, data from the current study suggest that pioglitazone prevented bone loss by decreasing bone resorption under inflammatory conditions, although the drug was previously shown to disrupt bone homeostasis in healthy animals when given at a similar dosage ([20]). Despite some variability originating from differences in both the sex of the animals studied ([36]) and treatment duration ([20]), we suggest that such discrepancies are likely attributable to the variable level of inflammation. Thus, polyarthritis is characterized by deterioration of mechanical bone properties, resulting in increased fragility ([37]), which can be prevented by the antiarthritic potency of TZDs, although PPARγ activation is able to impair bone mechanical strength under basal conditions ([20]). In clinical situations, this proposed mechanism would make sense with regard to the increased risk of fracture reported in type 2 diabetes mellitus patients who have a low level of chronic inflammation, while suggesting that TZDs might be beneficial in the prevention of inflammation-induced bone loss in RA patients, particularly during disease flares.

Osteoclastogenesis and bone turnover are regulated mainly by the balance between RANKL/RANK and their soluble decoy receptor OPG ([38]). An increase in the ratio of sRANKL to OPG in the synovium was reported following an increase in the severity of joint inflammation and bone destruction in animals with collagen-induced arthritis ([39]). We demonstrated that PPARγ activation lowered the ratio of circulating sRANKL to OPG (−59%) toward a normal level while reducing the severity of the arthritis. A similar trend was observed in inflamed joint tissue, since a marked decrease in the number of cells staining positive for RANKL was observed, whereas OPG expression remained poorly affected. As OPG therapy was shown to be ineffective with regard to clinical and histopathologic measures of inflammation ([40]), our data support the idea that pioglitazone reduced arthritis-induced bone loss through direct modulation of osteoclastogenesis. This proposed mechanism is further substantiated by the previous demonstration that rosiglitazone, another TZD, prevented inflammatory periodontal bone loss by inhibiting osteoclastogenesis and decreasing RANKL expression ([41]). Soluble RANKL was shown to be a reliable circulating biomarker and a mediator of local and systemic bone loss in experimental arthritis ([42]), highlighting the idea that the reduction in the ratio of circulating sRANKL to OPG induced by pioglitazone reflected its ability to reduce osteoclastogenesis.

However, IL-17 was recently shown to be an upstream regulatory cytokine with a major role in osteoclastic bone resorption. Indeed, IL-17 induced the production of TNFα and IL-1, up-regulated the production of RANKL, and stimulated the activity of matrix metalloproteinases and matrix catabolism ([43]). High levels of IL-17 were found in synovial fluid from RA patients, and the neutralization of culture supernatants from RA synovial tissue samples with an antibody against IL-17 suppressed their ability to promote osteoclast formation ([44]). We report here a 5.6-fold increase in serum levels of IL-17 in vehicle-treated arthritic rats, and we demonstrated that pioglitazone reduced this increase by 67%. Interestingly, the potency of pioglitazone to reduce the imbalance between sRANKL and OPG was in the same range, supporting the idea that its correcting effect was IL-17–dependent. We also demonstrated that IL-17 was highly expressed in the inflamed joint and that IL-17–positive cells consisted mainly of neutrophils, followed by lymphocytes. Our data are highly consistent with a recent report that IL-17A expression was localized to several immune cell subtypes in the inflamed synovium of RA patients ([45]). Infiltration of synovial tissue by neutrophils is a classic finding in inflamed joints in RA ([46]) and experimental arthritis ([47]). Pioglitazone treatment efficiently reduced the local production of IL-17, as shown by the marked decrease in cells staining positive for IL-17 in ankle joint synovium from arthritic animals.

The findings of several recent studies support the notion of a regulatory role of PPAR activation on IL-17 expression and/or Th17 activation ([48]). The nuclear orphan receptor RORγt is a key regulator of the differentiation program of proinflammatory IL-17–positive T helper cells ([49]) that acts synergistically with RORα to promote Th17 cell differentiation ([50]). Consistent with this is the recent report of a high-affinity synthetic ROR ligand that was shown to inhibit Th17 cell differentiation and function while reducing the severity of experimental allergic encephalopathy in mice ([51]). Activation of PPARγ is another way to regulate RORγt expression and function. Indeed, the PPARγ agonist pioglitazone was reported to repress the expression of RORγt in human and mouse T cells and to selectively suppress Th17 cell differentiation, but not Th1, Th2, or Treg cell differentiation ([52]). Our data are consistent with a similar RORγt-dependent reduction in IL-17 expression induced by pioglitazone during arthritis.

Thus, we demonstrated first that activation of the IL-17 promoter in transfected Jurkat cells required integrity of the RORγt binding capacity. Indeed, the deletion of the RORγt binding site abrogated IL-17 promoter activity under the experimental conditions used in our study, which is consistent with the ability of a novel isoform of RORγt lacking the hinge-encoding exons 5–8 to repress IL-17 gene transcription ([53]). Second, we demonstrated that RORγt expression was increased 3-fold in Th17 phenotype–like Jurkat cells and that this stimulatory effect was completely suppressed by lentiviral overexpression of PPARγ while being reduced by pioglitazone treatment in wild-type transduced cells. These data support the idea that PPARγ activation reduced arthritis severity, at least in part, by controlling IL-17 expression in a RORγt-dependent way. As RORγt activity is known to be induced by cytokines of the IL-6 family in a STAT-3–dependent manner ([54]), one cannot exclude the possibility that the reduction of circulating levels of IL-6 by treatment with the PPARγ agonist ([12]) may have also contributed to the reduced IL-17 expression in arthritic animals. However, our in vitro data in a T cell model of IL-17–induced expression demonstrate the transcriptional inhibition of IL-17 by pioglitazone that could result in correction of the RANKL/OPG imbalance, leading to a reduction in bone resorption. As mentioned above, this inhibitory effect was not restricted to T cells and, therefore, to Th17 cell differentiation, but involved all IL-17–producing cells, including neutrophils.

In conclusion, this study is the first to demonstrate a protective effect of pioglitazone treatment on arthritis-induced changes of the bone microarchitecture. This effect was concomitant with a reduction of arthritis severity and bone loss in selected regions of interest relevant for trabecular and cortical components. Based on the major suppressive effect of pioglitazone on IL-17 expression in blood and inflamed joint tissue, on local and systemic RANKL expression, and on the induced expression of IL-17 in Jurkat cells, we suggest that its bone-sparing effect during chronic arthritis is supported mainly by the modulation of IL-17/RANKL–dependent osteoclastogenesis. Even when considering the limitations of data extrapolation from animal models to clinical situations, our data support the idea that TZDs may “paradoxically” display bone-sparing properties in RA patients despite their adverse skeletal effects in diabetic patients.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Koufany had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Koufany, Netter, Jouzeau, Moulin.

Acquisition of data. Koufany, Chappard, Bastien, Weryha, Jouzeau, Moulin.

Analysis and interpretation of data. Koufany, Chappard, Netter, Bastien, Weryha, Jouzeau, Moulin.


We would like to thank Lydie Venteo (Labelhistologie, Reims, France) for technical assistance with the immunohistochemical analyses and François Plénat (Centre Hospitalier Régional et Universitaire, Vandoeuvre-lès-Nancy, France) for expertise with the histologic analysis.