Objectives: Chronic inflammation is a major risk factor for systemic bone loss leading to osteoporotic fracture and substantial morbidity and mortality. Inflammatory cytokines, particularly tumour necrosis factor (TNF) and interleukin-1 (IL1), are thought to play a key role in the pathogenesis of inflammation-induced bone loss, but their exact roles are yet to be determined.
Methods: To determine whether TNF directly triggers bone loss or requires IL1, human TNFα mice (hTNFtg) were crossed with mice lacking IL1α and IL1β (IL1−/−hTNFtg). Systemic bone architecture was evaluated using CT scanning, static and dynamic bone histomorphometry and serum markers of bone metabolism.
Results: hTNFtg mice developed severe bone loss accompanied by a severe distortion of bone microarchitecture. Bone trabeculae were thinner and decreased in numbers, resulting in increased trabecular separation. Histomorphometric analyses revealed strongly increased bone resorption in hTNFtg mice compared with wild-type mice. In contrast, IL1−/−hTNFtg mice were fully protected from systemic bone loss despite still developing inflammation in their joints. Lack of IL1 completely reversed increased osteoclast formation and bone resorption in hTNFtg mice and the increased levels of RANKL in these mice. Structural parameters and osteoclast and osteoblast numbers were indistinguishable from wild-type mice.
Conclusions: These data indicate that IL1 is essential for TNF-mediated bone loss. Despite TNF-mediated inflammatory arthritis, systemic bone is fully protected by the absence of IL1, which suggests that IL1 is an essential mediator of inflammatory osteopenia.
Statistics from Altmetric.com
Rheumatoid arthritis (RA) is an autoimmune disease which gives rise to a symmetrical and destructive polyarthritis.1 Apart from local damage caused by degradation of articular cartilage and invasion of the inflammatory synovial tissue into the periarticular bone, a substantial proportion of patients with RA are affected by systemic bone loss leading to premature osteoporosis and increased risk of fracture.2
Systemic bone loss in RA is the result of a multifactorial process involving age, menopausal state, functional disability and corticosteroid use. However, several points of evidence suggest that inflammation per se is an independent risk factor for bone loss: (1) cross-sectional and longitudinal studies have shown that high disease activity, which reflects a high degree of inflammation, is independently associated with bone loss in RA3 4; (2) population-based studies have revealed that even slight elevations of C-reactive protein increase the risk of fracture in the general population5 6; and (3) virtually all forms of chronic inflammatory diseases other than RA, such as inflammatory bowel disease, systemic lupus erythematosus and ankylosing spondylitis, are characterised by progressive bone loss.7 8 9
Histomorphometric and serum biomarker analyses suggest that increased osteoclast activation occurs in patients with RA, indicating that the inflammatory process might directly affect systemic bone resorption.10 11 Key triggers for systemic bone resorption are cytokines. The introduction of cytokine blockers as a treatment for RA has clearly established a pivotal role for cytokines as drivers of arthritis and structural damage.12 TNF is considered as a key molecule involved in the pathogenesis of RA. A large number of clinical trials and daily clinical practice has proved that TNF blockade is a bone protective treatment. Anti-TNF treatment can arrest systemic bone loss and downregulates systemic bone resorption markers in patients with RA. The impact of TNF blockade on bone formation is less clear when assessed using biomarkers such as serum osteocalcin.13 Moreover, overexpression of TNF in rodents is sufficient to trigger a destructive form of arthritis as well as systemic bone loss.14
However, it is unclear whether TNF acts directly on bone or orchestrates other pro-inflammatory mediators that facilitate systemic bone loss. Interleukin-1 (IL1), originally discovered as an important osteoclastogenic cytokine in 1986,15 is a key candidate molecule since TNF can induce IL1 production and IL1 has been shown to mediate local destructive effects of TNF in arthritis models.16 We hypothesised that IL1 might be important in triggering systemic bone loss downstream of TNF. We generated mice overexpressing TNF but lacking IL1 and investigated systemic bone loss in these animals.
Heterozygous human tumour necrosis factor transgenic (hTNFtg) mice (strain Tg197) and IL1−/− mice (both genetic background C57BL/6) were interbred to receive IL1−/−hTNFtg mice.16 The resulting four genotypes (wild-type, IL1−/−, hTNFtg and IL1−/−hTNFtg mice) of the F2 generation were comparatively analysed and, for all of the experiments, only females and littermates were used. Thirty-two animals were included in clinical, histological and cellular analyses and animals were killed at 12 weeks of age by cervical dislocation. Blood was withdrawn by cardiac puncture.
The following primers were used for genotyping of mice: hTNF transgene: 5′-TACCCCCTCCTTCAGACACC-3′ and 5′-GCCCTTCATAATATCCCCCA-3′; IL1α/β primers were as follows: IL1αF, 5′-CTTGGCCATACTGCAAAGGTCATG-3′, IL1αR, 5′-CAGGTCATTTAACCAAGTGGTGCTG-3′; IL1α/β KO, 5′-GAGGTGCTGTTTCTGGTCTTCACC-3′; IL1βF, 5′-GCGAATGTGTCACTATCTGCCACC-3′; IL1βR, 5′-CCATGCCACAGTCCCTCCAC-3′;
Quantitative imaging studies
Micro-CT images were acquired on a vivaCT40 (Scanco Medical, Bassersdorf, Switzerland). The scanner generates a cone beam at 5 μm spot size and operates at 50 keV. A region of 402 slices was imaged at 10 μm isotropic resolution starting from the proximal end of the tibia. Cortical and trabecular bone mineral density (BMD) and cortical geometry in mouse tibiae were determined by peripheral quantitative CT (pQCT) on an XCT Research SA+ (Stratec Medizintechnik, Pforzheim, Germany) fitted with a 0.5 mm collimator as described by Gasser et al.17 The following instrument settings were chosen for the measurements: voxel size 0.1 mm ×0.1 mm ×0.5 mm; scan speed, scout view 10 mm/s, final scan 2 mm/s; contour mode 1, peel mode 2, cortical threshold 400 mg/cm3. A slice located 2.5 mm distal from the intercondylar tubercle in the proximal tibia metaphysis was analysed. Parameters of BMD end points measured by pQCT were total cross-sectional BMD (in mg/cm3), volumetric cortical BMD (in mg/cm3) and mean cortical thickness (in mm).
Tibiae were fixed in 4% formalin for 24 h followed by 100% ethanol and undecalcified embedded in methymetacrylate (Technovit; Heraeus Kulzer, Wehrheim, Germany). After polymerisation, 3–4 μm sections were prepared with a Jung microtome (Jung, Heidelberg, Germany), deplastinated in methoxymethylmetacrylate (Merck, Darmstadt, Germany) and stained using von Kossa’s stain, Goldner trichrome and toluidine blue. Histomorphometry was performed using a microscope (Zeiss Axioskop 2, Zeiss, Germany) equipped with a video camera and digital analysis system (Osteomeasure, Osteometrics, Decatur, Atlanta, Georgia, USA).18 The following parameters were measured according to international standards: fraction of bone volume of the total sample volume (bone volume/tissue volume (BV/TV)), trabecular number (Tb.N), trabecular thickness (Tb.Th), trabecular separation (Tb.Sp), number of osteoclasts per bone perimeter (No.OC/B.pm), osteoclast surface per bone surface (OC.S./BS), number of osteoblasts per bone perimeter (No.OB/B.pm) and osteoid surface per bone surface (OS/BS).19
Dynamic labelling of bone
All mice at 12 weeks of age were given injections of calcein green (Sigma Aldrich Diagnostics) at a dose of 30 mg/kg body weight 7 days apart and killed 2 days after the final injection. Left tibial bones were prepared as described above and embedded in methoxymethylmetacrylate. Measurements were performed on the entire marrow region within the cortical shell using an image analysis system (Osteomeasure) and the mineral apposition rate (μm/day) was calculated.
Serum markers of bone metabolism
Serum levels of deoxypyridinoline crosslinks (Quidel, San Diego, California, USA), RANKL and OPG (both R&D Systems, Minneapolis, Minnesota, USA) were measured by enzyme immunoassay according to the manufacturers’ instructions.
Data are presented as mean ± SEM. For group comparisons, one-way factorial analysis of variance was used with the Bonferroni-Dunn test or the Mann-Whitney U test. A p value <0.05 was considered significant.
Preserved bone mass in IL1−/−hTNFtg mice
We have recently shown that deletion of IL1 has no significant influence on the arthritic process caused by TNF overexpression, but local bone erosions were significantly diminished in these mice.16 We therefore hypothesised that IL1 might also be crucial for TNF-induced systemic bone loss.
We first investigated bone volume in wild-type mice and IL1−/− mice, both of which have no inflammation. Three-dimensional microcomputed tomography (μCT) showed no significant differences in bone mass between wild-type and IL1−/− mice (mean bone volume per total volume (BV/TV) ± SEM 7.3±0.2% and 8.9±1.8%, respectively), although there was a trend to higher bone volume in IL1−/− mice, suggesting that IL1 might play some role on normal bone homeostasis. We next assessed systemic bone loss in arthritic hTNFtg mice. Trabecular BMD was significantly diminished compared with wild-type controls (BV/TV 2.7±1.1%; mean reduction –63%; p<0.05 vs wild-type). In contrast, BMD was completely maintained in hTNFtg mice lacking IL1 (mean BV/TV 8.0±2.6%; p = NS vs wild-type; fig 1). Representative micro-CT images are shown in fig 2.
Preserved cortical bone in IL1−/−hTNFtg mice
To determine the effect of chronic arthritis on cortical bone, we also assessed cortical BMD and cortical thickness by peripheral quantitative computed tomography (pQCT). There was a non-significant trend for decreased cortical BMD in hTNFtg mice compared with wild-type mice (mean BMD for hTNFtg 628.1±17.6 mg/cm3 vs wild-type 662.6±31.4 mg/cm3). Cortical BMD was, however, similar among wild-type, IL1−/− (696.3±26.1 mg/cm3) and IL1−/−hTNFtg mice (695.9±17.4 mg/cm3). Most interestingly, cortical thickness was significantly diminished in hTNFtg compared with wild-type mice (mean cortical thickness 0.33±0.04 mm vs 0.44±0.01 mm, respectively, p<0.05) but not in IL1−/− (mean 0.50±0.04, p = NS vs wild-type) and IL1−/−hTNFtg mice (mean 0.43±0.05, p = NS vs wild-type), suggesting that TNF-induced thinning of cortical bone is mediated by IL1.
IL1 deficiency rescues inflammation-induced changes of bone microarchitecture
We then assessed the trabecular microarchitecture of systemic bone by μCT. As shown in fig 3, IL1−/− mice again showed a very similar bone structure to that of wild-type controls. In contrast, analysis of hTNFtg mice revealed severe alterations of trabecular architecture: trabeculae were thinner (mean reduction in thickness −24%, p<0.05 vs wild-type), fewer in number (wild-type 4.3±0.08/mm vs hTNFtg 3.4±0.4, p<0.05) and more separated from each other (wild-type 0.23±0.004 μm vs hTNFtg 0.30±0.03 μm, p<0.05), resulting in reduced connectivity density of bone. Moreover, the microarchitecture of the bone changed to a more rod-like than a plate-like trabecular bone structure in hTNFtg mice, as evidenced by an increased structural model index (SMI). Conversely, we found no significant change in bone structural parameters in hTNFtg mice lacking IL1. The number (mean 4.5±0.33/μm, p = NS vs wild-type), thickness (mean 0.04±0.004 μm, p = NS vs wild-type) and separation of trabeculae (mean 0.22±0.02 μm, p = NS vs wild-type) were comparable to wild-type mice. Likewise, connectivity density and SMI were not significantly different from wild-type and IL1−/− mice.
Increased bone resorption in hTNFtg mice is dependent on IL1
To further define the systemic bone turnover in TNF-mediated arthritis, histomorphometric analysis of the proximal tibia was performed. In concordance with μCT analysis, hTNFtg but not IL-1−/−hTNFtg mice showed a significant loss of trabecular bone compared with wild-type and IL-1−/− mice (fig 4). In conjunction with the loss of trabecular bone, hTNFtg mice showed an accumulation of bone-resorbing osteoclasts compared with wild-type mice (mean No.OC/B.pm hTNFtg 7.3±1.4 vs wild-type 4.0±0.4, p<0.05). This was not accompanied by an increase in the activity and number of osteoblasts. Osteoid surface (OS/BS) and number of osteoblasts (OB/B.Pm) were not significantly increased in hTNFtg mice compared with wild-type mice (mean OS/BS hTNFtg vs wild-type 11.3±4.4% and 7.7±1.8%, respectively, p = NS; mean OB/B.Pm 10.6±2.8 vs 5.6±1.1, p = NS). Also, we could not detect significant changes in the mineral apposition rate (MAR) in wild-type and hTNFtg mice as evidenced by dynamic histomorphometry. This indicates increased osteoclast activity in the trabecular bone of hTNFtg mice. When IL1−/−hTNFtg mice were analysed, no sign of an increase in osteoclast numbers was found, which were significantly lower than in hTNFtg mice (p<0.05 for No.OC/B.pm and OC/BS) and comparable to wild-type mice. This suggests that IL1 is responsible for TNF-driven induction of systemic osteoclast-mediated bone resorption.
Restoration of bone turnover markers in hTNFtg mice lacking IL1
We next determined serum markers of bone metabolism to further define the consequences of TNF overexpression on systemic bone turnover. hTNFtg mice had significantly higher serum levels of deoxypyridinoline crosslinks (DPD) than wild-type mice (mean level 123±9 ng/ml vs 63±11 ng/ml for wild-type mice, p<0.05), whereas IL1−/− and IL1−/−hTNFtg mice had serum DPD levels comparable to wild-type mice (mean level 43±11 ng/ml and 67±10 ng/ml for IL1−/− and IL1−/−hTNFtg mice, respectively; both p = NS vs wild-type mice). We also determined serum levels of receptor activator of nuclear factor κB ligand (RANKL), an essential osteoclast differentiation factor. Similar to serum DPD levels, hTNFtg (mean level 29.6±3.5 pg/ml, p<0.05 vs wild-type, IL1−/− and IL1−/−hTNFtg mice) but not IL1−/− and IL1−/−hTNFtg mice (mean level 17.4±3.0 pg/ml and 16.1±1.7 pg/ml, respectively) showed elevated serum RANKL concentrations compared with wild-type mice (mean 23.3±2.7 pg/ml), suggesting that TNF-mediated elevation of systemic RANKL levels is accomplished through IL1. Serum levels of osteoprotegerin (OPG), a natural antagonist of RANKL, showed a somewhat different result. Both hTNFtg (mean 1028±61 pg/ml) and IL1−/−hTNFtg mice (mean 1050±64 pg/ml) had elevated OPG levels compared with wild-type (mean 832±40 pg/ml) and IL1−/− mice (mean 851±41 pg/ml), which indicates a general induction of OPG by chronic inflammation which is not affected by IL1. These findings are depicted in fig 5.
The following findings in this study show that IL1 is a major cytokine in TNF-induced systemic bone loss: (1) loss of trabecular bone mass and cortical thinning in hTNFtg mice is completely prevented by the deletion of IL1; (2) trabecular microarchitecture is completely normal in IL1−/−hTNFtg mice and indistinguishable from wild-type mice; (3) upregulation of osteoclastogenesis in hTNFtg mice is prevented in the absence of IL1; and (4) increased serum markers of bone resorption and serum RANKL levels are normalised in the absence of IL1.
Systemic bone loss causes substantial morbidity and mortality in patients with inflammatory diseases, particularly RA.20 21 The pathophysiological steps leading to the development of decreased BMD are not fully understood but appear to require the inflammatory process itself. Proinflammatory cytokines are known to interfere with bone homeostasis and modulate key pathways of bone resorption and bone growth.1 A deterioration in the quantity and quality of bone in patients with RA has been observed for many years. Studies performed in patients with RA have shown a high prevalence of osteoporosis.2 4 Risk factors for systemic bone loss are heterogeneous in chronic arthritis, but inflammatory disease activity is considered a major driver for systemic bone loss beside glucocorticoid therapy.3 An increased risk of fracture in patients with RA is especially supported by a recently published longitudinal study by van Staa et al, which showed that there is an independent role of disease activity and duration on the risk of osteoporotic fracture in many skeletal sites including the hip.22 Moreover, epidemiological studies investigating the level of C-reactive protein have shown it to predict osteoporotic fracture independent of established risk factors.5 6 These data suggest that inflammation directly precipitates systemic bone loss.
The pathophysiological pathways causing inflammation-induced bone loss have not been clearly established. As seen in histomorphometric studies of iliac crest biopsies from patients with RA, bone loss is attributed to increased bone resorption associated with increased bone turnover.11 23 This is supported by studies investigating biochemical markers of bone metabolism showing an increase in collagen type I breakdown products as a surrogate parameter for osteoclast activity in patients with RA.10 This is in line with observations from animal models of chronic arthritis which show increased bone resorption. This molecular link between inflammation and bone resorption is not fully established, but cytokines such as TNF, IL1 and IL6 play an essential role. These mediators link inflammation to bone loss by inducing the expression of factors which are essential for osteoclast formation such as M-CSF and RANKL, but they also directly stimulate these bone-resorbing cells.24 25 For instance, overexpression of TNF is sufficient to trigger systemic bone loss.18
TNF is considered as a major enhancer of osteoclastogenesis by driving the expression of RANKL and by directly increasing the differentiation of osteoclast precursors and enhancing the resorptive activity.26 However, it has been less clear whether TNF might act indirectly by stimulation of other cytokines. As TNF is a potent inducer of IL1, it might use IL1 to act on bone. In fact, elegant studies by Wei et al have shown that TNF induces the expression of IL1 and IL1 receptor in mesenchymal cells, which supports their osteoclastogenic effects on mononuclear cells.27 Thus, IL1 induces RANKL expression in mesenchymal cells and additionally acts directly on osteoclasts by enhancing the expression of RANK.27 16 Blockade of IL1 by its soluble receptor antagonist (IL1ra) or by using mice deficient in the type I IL1 receptor strongly reduced the potential of TNF to induce osteoclast formation, suggesting that IL1 represents a major link between TNF and osteoclast formation in vitro and in vivo.27 The central role of IL1 in inflammatory osteoclastogenesis is also supported by the destructive nature of arthritis models which depend on IL1, such as collagen-induced arthritis or serum transfer-induced arthritis.28 Thus, for instance, deficiency of the type I IL1 receptor does not only achieve excellent protection from inflammatory signs of arthritis in the serum transfer model but also protection from local bone destruction. Even in arthritis, where inflammatory signs of disease are fully TNF-dependent (TNFtg mice) and which do not require IL1, this cytokine is pivotal for cartilage degradation and local bone erosion.16 These previous data, as well as the observation that overexpression of IL1 causes osteopenia,29 suggest that IL1 also plays a central role in TNF-mediated systemic inflammatory bone loss. Our data show that this is indeed the case, as mice deficient in IL1 were completely protected from inflammatory bone loss induced by TNF.
In summary, this study provides evidence that IL1 is crucially involved in inflammation-induced systemic bone loss, implicating a certain hierarchy for cytokines with IL1 being a central one (fig 6). Even in cases of high levels of TNF and inflammatory arthritis, the absence of IL1 fully protects against systemic bone loss. This suggests that effective strategies to inhibit IL1 may prove valuable in protecting bone in inflammatory disease.
The authors thank Dr Yoichiro Iwakura (Center for Experimental Medicine, Institute of Medical Science, University of Tokyo, Tokyo, Japan) for kindly providing the IL-1−/− mice; Dr George Kollias (Alexander Fleming Biomedical Sciences Research Center, Vari, Greece) for providing the hTNFtg mice and Birgit Türk, Ivana Mikulic, Martin Steffen, Cornelia Stoll and Barabara Roy for their excellent technical assistance.
Funding This work was supported by an Austrian Science Fund START prize (to GS), Interdisziplinäres Zentrum für Klinische Forschung Erlangen Project C5 (to GS and JZ), Deutsche Forschungsgemeinschaft Grant FOR 661 (to GS) and SFB 423 project A18 (to GS and JZ) and Austrian Science Fund Grant P18223 (to KR).
Competing interests None.
Ethics approval The local ethics committee for animal studies approved all animal procedures.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.