Tumor necrosis factor α and RANKL blockade cannot halt bony spur formation in experimental inflammatory arthritis

Abstract

Objective

To investigate the kinetics of bony spur formation and the relationship of bony spur formation to synovial inflammation and bone erosion in 2 rat arthritis models, and to address whether bony spur formation depends on the expression of tumor necrosis factor α (TNFα) or RANKL.

Methods

Analysis of the kinetics of synovial inflammation, bone erosion, osteoclast formation, and growth of bony spurs was performed in rat collagen-induced arthritis (CIA) and adjuvant-induced arthritis (AIA). In addition, inhibition experiments were performed to assess whether inhibition of TNFα and RANKL by pegylated soluble TNF receptor type I (pegTNFRI) and osteoprotegerin (OPG), respectively, affected bony spur formation.

Results

Bony spurs emerged from the periosteal surface close to joints, and initial proliferation of mesenchymal cells was noted as early as 3 days and 5 days after onset of CIA and AIA, respectively. Initiation of bony spur formation occurred shortly after the onset of inflammation and bone erosion. Neither pegTNFRI nor OPG could significantly halt the osteophytic responses in CIA and AIA.

Conclusion

These results suggest that bony spur formation is triggered by inflammation and initial structural damage in these rat models of inflammatory arthritis. Moreover, emergence of bony spurs depends on periosteal proliferation and is not affected by inhibition of either TNFα or RANKL. Bony spur formation can thus be considered a process that occurs independent of TNFα and RANKL and is triggered by destructive arthritis.

Arthritis is characterized by a massive influx of immune cells into the synovial membrane and its neighboring structures such as the tendons, ligaments, and the joint cavity (1). Chronic joint inflammation leads to profound changes in the joint architecture, which is the structural basis for a progressive impairment of function (2). In rheumatoid arthritis (RA), destruction of periarticular bone and the articular cartilage is the dominant feature of structural damage and is radiographically reflected by bone erosion and joint space narrowing (3). Conversely, inflammatory joint destruction is sometimes accompanied by the modeling of bony spurs, also termed osteophytes, which emerge at the joint margins in diseases such as psoriatic arthritis (PsA) and ankylosing spondylitis (AS) (4). The reason for the apparently divergent bone responses among the various inflammatory diseases has not been fully clarified, but appears to involve a differential regulation of local bone homeostasis in the course of joint inflammation.

Considerable insight into the pathologic mechanisms underlying catabolic patterns of joint damage has been obtained in the past few years. Particularly, osteoclasts have been recognized as the primary bone-resorbing cell in inflamed joints. Cytokines such as tumor necrosis factor α (TNFα) and RANKL, both of which are key enhancers of osteoclastogenesis, have been recognized to play a central role in animal models of arthritis and human RA (5–7). Up-regulation of RANKL has also been described in the inflamed joints of patients with osteoarthritis (OA) (8) and those with spondylarthritis (9). The catabolic effects of RANKL are blocked by osteoprotegerin (OPG), a soluble decoy receptor that prevents osteoclast formation, activation, and survival (10). The relative balance between RANKL and OPG is thought to be an important determinant of bone resorption (11), and reports of reduced OPG levels in the inflamed joints of RA patients provide further evidence that the RANKL/OPG axis regulates bone destruction in this condition (12).

Based on these insights, local formation of osteoclasts and their resorption of bone are considered to be the primary mechanism for the catabolic pattern of inflammatory joint damage that is typically seen in arthritis (13). In the context of joint inflammation, TNFα has been implicated in the up-regulation of genes related to matrix degradation in chondrocytes (14), and TNFα is present at elevated levels in serum and synovial fluid from RA patients (15, 16). The ability of TNFα inhibition to control inflammation in patients with arthritis is well established (17).

Much less is known about bony spur formation, which is a prominent feature of inflammatory joint diseases such as AS and PsA and has been recognized in degenerative joint diseases such as OA or hemochromatosis arthropathy. Bony spurs represent spots of new bone formation that emerge from periosteal sites close to joints (where they are called osteophytes) or intervertebral spaces (where they are called spondylophytes or syndesmophytes depending on their pattern of growth) (18). Bony spur formation is considered a process of endochondral bone formation, requiring the differentiation of mesenchymal cells into hypertrophic chondrocytes and, finally, into osteoblasts, which are cells that produce extensive matrix for building up new bone. Although bony spurs can be considered a response-to-stress reaction of the joint, they may also contribute to the disease burden itself when they lead to fusion of the entire joint and loss of motion (19).

Bony spur formation appears to depend on molecules involved in bone formation, such as transforming growth factor β (TGFβ), bone morphogenetic proteins, and the Wnt protein family (20–22). Although the relative role of these pathways in bony spur formation and the effects of their mutual interaction are poorly defined, it is evident that these essential bone-forming molecular pathways are turned on when joints become inflamed or are subjected to mechanical stress. Direct therapeutic intervention in these pathways is not a treatment strategy that is currently applied, and may also have drawbacks because these pathways elicit important antiinflammatory functions and are required for physiologic bone formation and maintenance of bone mass.

More attention, however, is currently being placed on the interaction between inflammation, bone metabolism, and bony spur formation and their relevance to current treatment strategies. This is now the focus for several reasons. 1) Bony spurs are the basis for structural outcome parameters such as those measured in AS, and therefore therapeutic modification leading to improvement is of clinical interest (23, 24). 2) TNFα inhibition is of key importance in controlling chronic arthritis, but its effect on bony spur formation is poorly defined. Recent data in fact suggest that the blockade of TNFα may not have a major effect on inhibiting the formation of syndesmophytes in the vertebral column of patients with AS (25). 3) Bony spur formation is linked to enhanced bone metabolism, with increased bone formation and bone resorption to shape the newly created bone (20–22), and therefore interventions in bone metabolism, such as osteoclast inhibition, may affect the formation of bony spurs and also modify structural joint damage.

Most of the animal models of inflammatory arthritis display bony spur formation. Adjuvant-induced arthritis (AIA), collagen-induced arthritis (CIA), and also the K/BxN serum transfer model of arthritis are all characterized by formation of bony spurs along the joints. The kinetics of bony spur formation and the relationship of bony spur formation to inflammation and bone resorption, however, are poorly understood. Moreover, it is not known whether inhibition of inflammation and/or inhibition of bone resorption in these models could affect bony spur formation. The introduction of TNFα blockade was not effective in blocking bony spur formation in the male DBA/1 mouse arthritis model, which is an experimental model characterized by minimal inflammation and extensive bone growth, suggesting that these mechanisms may be uncoupled (26).

We were therefore interested in studying whether therapeutic interventions to block inflammation or bone resorption would be effective in modifying the formation of bony spurs in 2 standard models of inflammatory arthritis, AIA and CIA. To accomplish this, we first defined the kinetics of bony spur formation in AIA and CIA. In addition, we investigated whether inhibition of TNFα, as a strategy to decrease inflammation, or inhibition of RANKL, as a strategy to inhibit bone resorption, could affect the formation of bony spurs in these 2 forms of inflammatory arthritis.

MATERIALS AND METHODS

Animals and induction of arthritis.

Young adult Lewis rats (54 males and 54 females), weighing 80–100 grams, were purchased from Charles River (Wilmington, MA). Animals were acclimatized for 1 week under normal environmental conditions and fed a pelleted rodent chow (no. 8640; Harlan Teklad, Madison, WI) with tap water ad libitum. Initially, a total of 60 rats was assigned to a time-course experiment. AIA was induced in male rats (n = 30) by a single intradermal injection of 0.5 mg heat-killed mycobacteria H37Ra (Difco, Detroit, MI), suspended in paraffin oil, into the tail base. CIA was induced in female rats (n = 30) by multiple intradermal injections with a total of 1 mg porcine type II collagen (Chondrex, Redmond, WA), emulsified 1:1 in Freund's incomplete adjuvant (Difco), into the skin of the back. Rats subjected to these inductions for AIA or CIA were killed at disease onset (day 0) or on days 1, 2, 3, 4, 5, 10, 14, 20, or 27 after disease onset.

In addition, AIA (n = 24) or CIA (n = 24) was induced in another group of 48 rats, and these animals were randomly assigned to 1 of the following 3 treatment groups (n = 8/group): pegylated soluble TNF receptor type I (pegTNFRI, or pegsunercept; 4 mg/kg/day by daily subcutaneous [SC] bolus), OPG (consisting of the RANKL-binding portion of native OPG fused with the constant [Fc] domain of IgG; 3 mg/kg/day given every other day by SC bolus), or vehicle control. Treatments were started 4 days after the onset of clinical disease and continued for 10 days. Moreover, in a total number of 24 mice, we performed a preventive treatment with 3 different doses of OPG (0.1, 1, and 10 mg/kg; each n = 8/group), in comparison with a vehicle control group (n = 8/group). These treatments were started at the onset of clinical disease (day 0) and continued for 10 days.

This study was conducted in accordance with federal animal care guidelines. Approval for the study was provided by the Amgen Institutional Animal Care and Use Committee.

Assessment of paw swelling.

Swelling of the hind paws was assessed daily from disease onset to day 20 after disease onset. In AIA, paw swelling was measured by water plethysmography, as previously described (27). In CIA, paw swelling was quantified using calipers (Fowler Sylvac Ultra-Cal Mark III; Sylvac, Crissier, Switzerland) to measure the ankle diameter of the hind paws.

Conventional histology and detection of osteoclasts.

At necropsy, the right hind paws of rats with AIA or CIA were removed at the fur line just proximal to the tibiotarsal (hock) joint. The paws were then fixed in zinc formalin for 2 days, decalcified with a 1:4 mixture of 8N formic acid and 1N sodium formate, and then divided longitudinally along the median axis, processed into paraffin, and cut serially at 4 μm. One section was stained with hematoxylin and eosin (H&E) to allow for conventional histopathologic assessment. The other section was studied immunohistochemically to visualize osteoclasts, using an indirect immunoperoxidase procedure for the detection of cathepsin K, an osteoclast-specific protease.

Immunohistochemistry was performed on an automated tissue stainer (Model Mark 5; DPC, Flanders, NJ) according to a standard method (28). Briefly, sections were pretreated with 0.1% trypsin in 1% CaCl₂ (Sigma, St. Louis, MO) for 15 minutes, blocked with CAS Block (Zymed Laboratories, San Francisco, CA) for 10 minutes, and incubated with a proprietary rabbit polyclonal anti–cathepsin K antibody (1 μg/ml; Amgen, Thousand Oaks, CA) for 60 minutes. The primary antibody was localized using sequential 30-minute incubations with biotin-conjugated goat anti-rabbit polyclonal secondary antibody (Vector, Burlingame, CA) (used at 1:200), peroxidase-blocking solution (Dako, Carpinteria, CA) for 25 minutes, and avidin–biotin complex and peroxidase reagents (ABC Elite Kit; Vector). The reaction was visualized using diaminobenzidine (DAB+Substrate Chromagen System; Dako) for 3 minutes. The osteophytic proliferative response was assessed using an antibody against the proliferation antigen Ki-67 (Novocastra, Newcastle-upon-Tyne, UK) according to the protocol described above. Immunostaining for type X collagen was done by incubating sections with a specific antibody against type X collagen (29) (kindly provided by Klaus Von der Mark, Erlangen, Germany) overnight at 4°C. The sections were then incubated with alkaline phosphatase–streptavidin for 30 minutes at room temperature before detection with fast red Texas Red/naphthol solution (Sigma), resulting in red staining.

Semiquantitative lesion scoring.

Synovial inflammation and bone erosion were assessed in H&E-stained sections using a semiquantitative scoring system as previously described (30). Inflammation was scored according to the following criteria: 0 = normal, 1 = presence of a few inflammatory cells in perisynovial tissue, 2 = mild inflammation, with a few small focal aggregates and modest buildup in perisynovial tissue, 3 = moderate inflammation, with many small aggregates and extensive buildup in perisynovial tissue, 4 = marked inflammation, with large aggregates and extensive buildup in perisynovial tissue. Bone erosion in AIA was scored as follows: 0 = normal, 1 = minimal, with a few erosion sites in tarsal bones, 2 = mild, with a modest number of erosion sites in tarsal bones, 3 = moderate, with many erosion sites in tarsal bones, 4 = marked, with partial destruction of the tibia and extensive destruction of tarsal bones, 5 = extensive, with fragmentation of tarsal bones and full-thickness cortical penetration of the tibia. Bone erosion in CIA was quantified as follows: 0 = normal, 1 = minimal, with 1–2 small, shallow erosion sites, 2 = mild, with 1–4 erosion sites of medium size and depth, 3 = moderate, with ≥5 erosion sites partially extending through the cortical bone, 4 = marked, with multiple foci partially or completely extending through the cortical bone, 5 = extensive, with cortical penetration at >25% of the bone length. Analysis included the tibiotarsal articulation and all intertarsal joints. Osteoclasts in AIA and CIA were quantified according to the following scores: 0 = normal (no osteoclasts), 1 = presence of a few osteoclasts (lining fewer than 5% of most affected bone surfaces), 2 = some osteoclasts (lining 5–25% of most affected bone surfaces), 3 = many osteoclasts (lining 30–50% of most affected bone surfaces), 4 = abundant osteoclasts (lining >50% of most affected bone surfaces).

Histomorphometric analysis of bony spurs.

In addition, the size of the entire bony spur of the navicular bone as well as its bony fraction was analyzed quantitatively by histomorphometry. Previous studies have validated the navicular bone as a sensitive indicator for the extent of arthritic changes in experimental arthritis in rats (28, 31). Bony spurs are most prominent at the navicular bone as well. In addition, the area of the periosteal surface of the navicular bone covered by the bony spur and the numbers of cathepsin K–labeled osteoclasts (multinucleated cells attached to bone) were analyzed quantitatively by histomorphometry. All parameters were analyzed using commercial image-analysis software (OsteoMeasure program 2.2; Osteometrics, Atlanta, GA) as previously described (32).

Radiographs.

The left hind paws were placed in position on Kodac X-OMAT TL high-resolution specimen-imaging film (Eastman Kodak, Rochester, NY) and radiographed with a Faxitron X-ray system (Model 43855A; Faxitron X-ray, Buffalo Grove, IL). Images were shot at 26 kV for 10 seconds.

Statistical analysis.

All results are expressed as the mean ± SEM. Groups were compared by nonparametric Kruskal-Wallis test using GraphPad Prism software (version 4; GraphPad Software, San Diego, CA). P values less than or equal to 0.05 were used to delineate significant differences between groups.

RESULTS

Emergence of bony spur formation following synovial inflammation and bone destruction.

To assess the kinetics of bony spur formation in arthritis, we first performed a sequential analysis of the development of bony spurs in the AIA model, as well as in CIA. In each rat model of arthritis, massive synovial infiltration in the hind paws occurred in conjunction with the onset of clinical symptoms of arthritis (day 0) (results for AIA are shown in Figures 1A–D, while those for CIA are available from the corresponding author upon request). Onset of periosteal proliferation was observed as early as 5 days after the onset of disease in the AIA model (Figure 1A) and as early as 3 days after the onset of disease in the CIA model (results not shown). Moreover, initial formation of small bone erosions could be detected before the bony spurs had emerged (Figure 1B), suggesting that spur formation in AIA and CIA may depend on an initial resorptive phase of arthritis.

Figure 1. Relationship of bony spur formation to inflammation and bone erosions in rat adjuvant-induced arthritis (AIA). A, Sequential assessment of hind paw inflammation was performed by measurement of paw diameter in rats with AIA from day 0 to day 27 after disease onset. B and C, Semiquantitative sequential assessment of bone erosion (B) and osteoclast counts (C) was performed in the same mice. D, Quantitative measurement of bony spur size was carried out by histomorphometry. Bars show the mean and SEM. ∗ = P < 0.05 versus day 0. Results for rats with collagen-induced arthritis are available from the corresponding author upon request.

Increased osteoclast counts also preceded the appearance of bony spurs (Figure 1C), suggesting that an initial resorptive stimulus paves the way for growth of bony spurs. In the late stages of disease, when growth of bony spurs was most pronounced (Figure 1D), osteoclast counts gradually decreased, suggesting a switch of bone metabolism to more bone formation and less resorption. Interestingly, the pattern was almost identical between AIA (Figure 1) and CIA (results not shown).

Microarchitecture of growth of bony spurs in AIA and CIA.

Initial lesions in AIA and CIA were characterized by a proliferation of mesenchymal cells at periosteal sites in the vicinity of the joint space (Figure 2). Bony spurs showed a rapid and consistent growth in both models over time, peaking in their size at the final observation time point, 27 days after the onset of arthritis. The size of the lesions was much more pronounced in AIA as compared with CIA, but microarchitectural changes were very similar between the 2 models. The surface of the lesions contained dense accumulations of mesenchymal cells that showed high proliferative activity, with almost all of the cells in both models positively staining for the proliferation marker Ki-67 (results available from the corresponding author upon request). Mesenchymal cells are known to form a densely packed multilayer comprising the outer surface of the bony spur, which determines the growth of the bony spur. Underneath this dense and proliferating mesenchymal lining layer of the spur, we observed hypertrophic chondrocytes producing an extensive amount of matrix and expressing type X collagen (results available from the corresponding author upon request). Finally, the inner regions of the spur, which were closest to the former periosteal surface, appeared to be remodeled into bone.

Figure 2. Sequence of bony spur formation in rat antigen-induced arthritis (AIA) (top) and collagen-induced arthritis (CIA) (bottom). Photomicrographs of hematoxylin and eosin–stained sections of hind paws of rats with AIA or CIA show the periosteum of the navicular bone (original magnification × 5 in top; × 10 in bottom). Sections were obtained 5, 10, 20, and 27 days after disease onset. Arrows indicate the proliferation front. Note that bony spurs in AIA are much larger than in CIA.

The periosteal surface of the navicular bone is an area that reproducibly shows marked growth of bony spurs. Assessment of the periosteal surface revealed that this process affected a major part of the navicular surface early on and affected virtually the entire navicular bone at later stages (results not shown). In fact, the periosteum appears to be essential to allow the bony growth in these 2 inflammatory arthritis models. Disruption of the integrity of the periosteum and cortical bone completely prevented the proliferative response. This was evident at sites where osteoclasts had penetrated cortical bone (results available from the corresponding author upon request). At these sites, no proliferative response was found, whereas the neighboring sites with intact periosteal bone interphase showed a massive osteophytic proliferation. This suggests that bony spurs require an intact periosteum covering cortical bone for their formation.

Simultaneous occurrence of bone deposition and osteoclast influx into bony spurs.

We next studied whether the formation of new bone within the bony spur is linked to the emergence of osteoclasts, which are generally required for the remodeling of mineralized tissue. Assessment of bone growth in bony spurs in AIA and CIA showed a kinetic pattern that was similar to the size of the expansion of the entire lesion (Figures 3A–C). The bony part of the spur, however, was consistently smaller than the entire lesion, which indicated a consistent growth of the lesions as well as consistent remodeling of fibrous and cartilage-like tissue into bone. Small deposits of bone were found in lesions even in the early phase of disease (day 5). Emergence of bone within the osteophytic lesion triggered the appearance of osteoclasts in the lesions. The marked increase in bone size within bony spurs from day 5 to day 14 was then accompanied by a dramatic accumulation of osteoclasts in these lesions, with levels peaking on day 20 in AIA and day 14 in CIA, before gradually decreasing thereafter.

Figure 3. Bone formation and osteoclast accumulation in bony spurs in rat AIA and CIA. A and B, Newly formed bone within bony spurs (A) and osteoclasts within bony spurs (B) were assessed quantitatively in rat AIA and CIA by histomorphometry. Bars show the mean and SEM. C, Photomicrographs of bony spurs show results of hematoxylin and eosin staining of hind paws in AIA and CIA (first and third panels, respectively) and staining with an antibody against cathepsin K in AIA and CIA (second and fourth panels, respectively). Osteoclasts are indicated in brown within the bony spur (broken arrows; the margin of the lesion is indicated with solid arrows) (original magnification × 10). S = synovium; C = cartilage-like tissue; B = bone; P = periosteum (see Figure 2 for other definitions).

Lack of effect of TNFα and RANKL inhibition on growth of bony spurs.

To address whether growth of bony spurs depends on inflammation or osteoclast generation, we used a potent antiinflammatory approach (TNFα inhibition) as well as an effective method to block osteoclasts (RANKL inhibition) in the AIA and CIA models. Inhibitory treatments in both AIA and CIA were started in the early phase of arthritis (day 3), when initial bone erosions had started to form. Blockade of TNFα using pegTNFRI significantly reduced inflammation but did not affect the formation of bony spurs (Figures 4A and B), suggesting that inhibition of inflammation does not inhibit the periosteal bone response. This finding was identical in both AIA and CIA.

Figure 4. Effects of tumor necrosis factor α (TNFα) and RANKL blockade on bony spur formation in rat adjuvant-induced arthritis (AIA) and collagen-induced arthritis (CIA). Hind paws of rats were treated with vehicle (blue line), pegylated soluble tumor necrosis factor receptor type I (PEG sTNFRI) (green line), or osteoprotegerin (OPG) (orange line). A, Clinical assessment of joint swelling in AIA and CIA. Arrow indicates initiation of treatment in each group. B, Quantitative assessment of bony spur formation in AIA and CIA by histomorphometry. Bars show the mean and SEM.

OPG treatment of rats with AIA or rats with CIA resulted in marked (>95%) reductions in osteoclast numbers in the hind paws (results not shown), as described in a previous report related to this same study (30). This level of reduction of osteoclasts was not associated with changes in the size of the bony spurs, as evident on radiographs, in either AIA or CIA (Figures 5A and B). Moreover, preventive treatment with OPG was not effective in blocking bony spur formation.

Figure 5. Radiographic evidence of bony spurs after TNFα and RANKL blockade. The hind paws of rats with AIA (A) and rats with CIA (B) were treated with vehicle, OPG, or pegylated TNFRI, and radiographs were obtained at the end of the study. Representative results are shown. See Figure 4 for definitions.

Bony overgrowth was found after treatment with all 3 doses of OPG, ranging from 0.1 mg/kg to 10 mg/kg, and was as pronounced as that in vehicle-treated mice (Figures 6A–D). Even at higher doses, when OPG completely blocked the formation of osteoclasts and bone resorption, fully formed bony spurs were observed, suggesting that the presence of osteoclasts is not essential for the formation of bony spurs.

Figure 6. Preventive RANKL blockade does not affect bony spur formation in AIA. Rats with AIA were treated with vehicle or different doses of OPG starting at the onset of disease. Hind paws were scored for A, inflammation (synovitis), B, bone erosion, C, osteoclasts, and D, histomorphometrically assessed area of bony spur formation. Bars show the mean and SEM. ∗ = P < 0.05 versus vehicle. See Figure 4 for definitions.

DISCUSSION

Bony spurs are a frequently observed pathologic feature of both degenerative and inflammatory joint diseases and usually grow at the edges of the synovial joint or at insertion sites of tendons (entheses). Current concepts suggest that bony spur formation may represent a response-to-stress mechanism of the joints (4, 19). Both mechanical and inflammatory triggers can lead to formation of bony spurs, and mechanical triggers seem to be particularly important in the formation of bony spurs along the entheses (18, 33). Bone turnover in bony spurs is enhanced and reflects the type of spur typically seen in subchondral bone, where the rate of bone turnover exceeds the levels observed in epiphyseal and metaphyseal cancellous bone compartments (34). It is as yet unclear why bony spur formation is particularly prominent in certain forms of human joint disease such as AS and PsA, and why it is virtually absent in other forms of joint disease such as RA. Novel concepts, however, indicate molecular differences related to the expression of proteins involved in osteoblast differentiation as an underlying principle. The relationship of inflammation, bone erosion, and bony spur formation has not been completely elucidated.

For understanding the process of bony spur formation, it is of particular interest to characterize whether an initial erosive phase is necessary to promote the growth of these bony spurs. Herein we show that bony spur formation clearly follows inflammation, initial osteoclast formation, and bone erosion in 2 typical models of inflammatory arthritis, suggesting that bony spur formation can indeed be regarded as a response of bone to joint inflammation. The induction of bony spurs is preceded by an erosive stimulus, which shares similarities with fracture healing. Both processes are initiated in response to bone damage and involve the production of new bone through endochondral ossification (35). In fracture healing, new bone formation occurs on periosteal surfaces that are adjacent to, but not directly within, the damaged bone areas, and this formation typically results in bridging of the fracture site (36). We observed that bony spurs formed only above intact periosteum, which seems to be analogous to the periosteal response observed in fracture healing. The periosteal surface is considered to be one of the most active sites of bone modeling (37), and bony spur formation is a stepwise process originating from the periosteum.

Another similarity between bony spur formation and fracture repair is that the bone formation responses in these models were not inhibited by OPG (38). It is therefore reasonable to suggest that osteoclasts are not essential for these bone formation responses to occur, even though osteoclasts are prominent histologic features in both models. Osteoclasts are clearly important for the remodeling of fracture calluses, since osteoclast inhibition typically results in a larger fracture callus (39, 40). The lack of effect of osteoclast inhibition on bony spur formation that was observed in the present study suggests that these lesions do not remodel, which could explain their persistent nature, in contrast to the transient nature of fracture calluses.

Animal models of arthritis vary in their potential to form bony spurs, as do human inflammatory joint diseases. Whereas TNFα-transgenic mice do not form bony spurs and the histologic pattern is purely erosive unless bony spur formation is stimulated by pharmacologic interventions, other arthritides in rodents, such as that in the male DBA/1 mouse model, are dominated by osteoproliferation and formation of bony spurs is the hallmark of the disease, with little inflammation (41, 42). In fact, Lories and colleagues have demonstrated that osteoproliferation in male DBA/1 mice cannot be blocked by inhibition of TNFα, indicating that, after the initial inflammatory trigger, bone formation might proceed uncoupled from inflammation (42). This would indeed support clinical data obtained in AS, showing that the formation of syndesmophytes (bone spurs along the vertebral column) are not affected by TNFα-blocking therapy.

Interestingly, standard models of inflammatory arthritis, such as AIA, CIA, or the serum transfer model of arthritis, also display formation of bony spurs, and thus these models do not exactly mimic the disease processes of RA. In fact, bony spur formation has not been rigorously studied in these models, because it has not been the focus of attention for therapeutic interventions targeting joint inflammation and structural damage such as bone erosion and cartilage degradation. Scharstuhl and colleagues performed an elegant analysis of bony spurs in murine CIA, and their results showed that TGFβ is an important mediator of growth of these lesions, suggesting that the release of growth factors from mesenchymal cells is indeed a key prerequisite for bony spur formation in CIA (21). Moreover, the inhibitory effect of nonsteroidal antiinflammatory drugs on new bone formation, by blocking the synthesis of prostaglandin E₂, has long been known (43).

Although it has been proven clinically effective, TNFα inhibition in the present study was not able to block bony spur formation in either AIA or CIA. This reinforces the current concept that bone formation is crucial for the development of bony spurs and that TNFα inhibition is unable to halt the process. TNFα itself is a potent down-regulator of bone formation, and its removal might be expected to increase bone formation (44, 45). However, TNFα blockade also did not increase bony spur formation. The clinical consequences of these observations are obvious, in that TNFα inhibition is not expected to change bony spur formation in diseases such as AS, PsA, and also, potentially, OA. Thus, TNFα inhibition does not actively prevent bony spur formation, which is potentially beneficial, since bony spurs allow a certain stabilization of affected joints. Furthermore, TNFα does not promote ankylosis and immobilization of joints. This finding has potential implications for the treatment of AS, but also of RA, which is sometimes associated with secondary OA and formation of bony spurs and osteosclerosis.

Another important finding is the lack of effect of osteoclast inhibition on bony spur formation. Because growth of bony spurs requires endochondral bone formation, which is characterized by the production of a cartilaginous scaffold containing hypertrophic chondrocytes followed by remodeling into bone, one could assume that osteoclasts are required for this process. Our data, as well as recent data in a nonerosive model of arthritis in which treatment with bisphosphonates was used (46), do not support this concept. Thus, RANKL inhibition by OPG did not influence growth of bony spurs in either model, which strongly suggests that osteoclasts are not necessary for the process of bony spur formation to occur. This notion is consistent with data from a study in nonhuman primates, which showed that osteoclast inhibition by estrogen had no significant impact on periarticular bony spurs (28). Furthermore, an observational study indicated that the antiresorptive effects of bisphosphonates were associated with neutral effects on OA symptoms and bony spur formation in postmenopausal women (47).

RANKL inhibition is considered a promising strategy to treat osteoporosis, bone metastasis, and arthritic bone erosions (48). Our observation that RANKL inhibition did not affect bony spur formation is of clinical interest, since bony spurs are frequently found in elderly patients and are a symptom of OA as well. These data suggest that RANKL inhibition would have a neutral role in terms of its effects on such lesions; instead, RANKL inhibition would allow the stabilization of an affected joint, because new bone formation would occur while bony ankylosis would not be provoked.

Another clinical implication of our findings relates to the confounding influence of bony spurs on the interpretation of bone densitometry evaluations by dual x-ray absorptiometry (49). Because the inhibition of RANKL and TNFα did not influence bony spurs, it is reasonable to suggest that those therapies also might not further complicate this scenario.

In summary, these data show that bony spur formation is a response-to-injury mechanism of the joint, which is turned on rapidly during initial joint damage. This mechanism occurred independently from TNFα, a major inflammatory stimulus, and RANKL, the triggering factor for osteoclast activation and bone loss. In fact, these observations reinforce current molecular and clinical concepts, which suggest that bony spur formation is not influenced by TNFα inhibition and follows distinct molecular processes that control bone formation.

AUTHOR CONTRIBUTIONS

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. Schett 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. Schett, Stolina, Zack, Kostenuik, Feige.

Acquisition of data. Schett, Stolina, Dwyer, Uderhardt, Krönke, Kostenuik, Feige.

Analysis and interpretation of data. Schett, Stolina, Dwyer, Zack, Kostenuik.