Classical and Paradoxical Effects of TNF-α on Bone Homeostasis

Biology of osteoclasts

Osteoclasts are hematopoietic cells derived from the monocyte lineage. They undergo a series of differentiation steps to become mature bone-resorbing cells. Intracellular signaling cascade of transcriptional regulation of osteoclast-specific genes is activated by macrophage-colony-stimulating factor (M-CSF) and receptor activator of nuclear factor kappa-B ligand (RANKL) (16) (Figure 1). Upon differentiation, OC precursor cells express strongly RANK, the receptor for RANKL.

Biology of osteoblasts

Osteoblasts are derived from pluripotent stem cells called “mesenchymal stem cells (MSCs)” capable of differentiating into adipocytes, chondrocytes, fibroblasts, skeletal muscle cells, and tendon cells (17). A number of factors can stimulate the differentiation toward an osteoblastic direction. The first step includes the proliferation of osteoprogenitors leading to the production of proteins such as histones, fibronectin, and type I collagen. Subsequently, osteogenic matrix is formed through the expression of genes responsible for the production of alkaline phosphatase (ALP), bone sialoprotein, and type I collagen. Finally, extracellular matrix is mineralized by activation of genes involved in the release of osteocalcin (OCN), osteopontin, and collagenase. This complex cascade of gene expression and cellular differentiation is orchestrated by several transcription factors including the runt-related transcription factor-2 (Runx2) and surface expression of RANKL (Figure 2A).

Biology of osteocytes

Osteocytes are derived from MSCs through OB differentiation. They are located within the bone matrix and represent 90–95% of all bone cells in adults (18). In addition, they have the longest life span among bone cells, up to decades, within their mineralized environment (19). OYs are involved in the dynamics of bone structure through expression of different markers that are released during OB-to-OY ontogeny. In addition, OYs affect the function of other organs such as phosphate transport in kidney (20). Furthermore, OYs are now known to be the principal sensors for mechanical loading of bone. They translate the canalicular flow resulting from bone loading into OCs and OBs recruiting signals (21). Due to the high expression of RANKL by OYs, they have a more significant contribution to osteoclastogenesis as compared to OBs and bone MSCs (22, 23).

Effects of TNF-α on osteoblasts

Previous in vitro studies have indicated an inhibitory effect of TNF-α on OB differentiation (Table 1). This blockage of osteogenic differentiation resulted from the inhibition of insulin-like growth factor-1 (IGF-1), osterix (Osx also known as SP7), and Runx2 (24–26). The inhibition of Runx2 seemed to be mediated by the TNF-induced up-regulation of Smurf1, a negative regulator of OB differentiation that causes the degradation of Runx2 (27). Moreover, TNF-α inhibited MSC differentiation into OBs via the ubiquitin protein ligase Wwp1 in TNF transgenic mice expressing human TNF-α (28). Another human study demonstrated an inhibitory effect of TNF-α on Runx2 and collagen expression. However, an increase in alkaline phosphatase activity and matrix mineralization was noted (29).

In full contrast with these findings, other murine and human MSC studies have reported that TNF-α can also induce osteogenic differentiation (Table 2). In three in vitro rodent models, low concentrations of TNF-α increased osteogenic differentiation via an up-regulation of Runx2, Osx, OCN, and ALP levels (38, 39). Moreover, four human experimental models indicated a similar osteogenic activity for TNF-α. In these models, osteogenic differentiation was promoted via induction of BMP-2, Osx, Runx2, and OCN (40–43). In addition, human models revealed that the dual role of TNF-α on osteogenic differentiation was directly dependent on the concentration of TNF-α, the cell type, and the exposure time (Table 2). Furthermore, a recent study showed that TNF-α stimulated the expression of Wnt5a, which was directly associated with an increase in tissue non-specific alkaline phosphatase (TNAP) levels and mineralization. This suggests an autocrine stimulation of OB activity by Wnt5a in response to TNF in hMSCs (44). This paradoxical effect of TNF-α in inhibiting or activating osteoblastogenesis, lies in the differentiation stage of the responding cells. MSCs can differentiate into several lineages. At this early stage, TNF-α binds its receptors and favors osteogenic differentiation through activation of multiple signaling pathways, notably NF-kB (40) (Figure 2B) (Table 2). On the other hand, TNF-α inhibits osteoblastogenesis in the presence of pre-OB cells, which are already on the differentiation process. This inhibition occurs at different levels, where TNF-α induces DKK-1 expression that inhibits Wnt pathway (45), or activates Smurf that inhibits BMP-2 pathway (31) (Figure 2A) (Table 1).

Effects of TNF-α on osteocytes

Limited results are available on the effect of TNF-α on OYs, which are known to be affected by the surrounding environment. TNF-α and interleukin-1 (IL-1), which increase with estrogen deficiency, can induce OYs apoptosis (48). In a fluid shear stress study, in which the mechanical bone loading was mimicked by applying pulsating fluid flow, the TNF-α-induced apoptosis observed in OYs was inhibited by mechanical loading. However, this effect was not observed in OBs and periosteal fibroblasts. Since apoptotic OYs attract OCs, these results suggest a key role for OYs apoptosis in osteoclastic bone resorption during bone remodeling that is in part modulated by TNF-α (49). Additionally, TNF-α was shown to inhibit the increase in nitric oxide (NO) production and intracellular calcium while strongly reducing F-actin content. This resulted in a reduction of OYs stiffness and provided a possible mechanism to explain the contribution of inflammation to loss of bone mass (50).

Effects of TNF-α on osteoclasts

The role of TNF-α as a stimulator of osteoclastogenesis is well established (51–55). The expression of a number of transcription factors, including NF-κB, is critical for osteoclastogenesis (56, 57) (Figure 1). Early bone marrow progenitors embark on a path toward pre-OCs or monocytes–macrophages under the influence of M-CSF while TNF-α, IL-1, and RANKL promote a progression toward the functional OCs phenotype. Thus, both RANKL and TNF-α are not only required for, but also synergize to induce osteoclastogenesis. The requirement for both TNF-α and RANKL allows for a more precise control of OCs numbers through a dual level of regulation. First, they affect the selective signaling of the precursor cells via the two different receptors, and subsequently their synergy further increases the activation of NF-κB and AP-1 signaling. Furthermore, the effect of the pro-inflammatory cytokines TNF-α, IL-17, IL-6, IL-1β, and IL-23 on osteoclastogenesis in vitro demonstrated that they have specific characteristic osteoclastogenic properties. However, they exhibited osteoclastogenesis-related activity only in the presence of permissive levels of RANKL. TNF-α appears to have osteoclastogenic properties in the early stage of osteoclastogenesis when bone marrow-derived macrophages (BMMs) are at the stage of osteoclast precursor cells (58).

Rheumatoid arthritis

Rheumatoid arthritis is a chronic disease characterized by joint inflammation leading to destruction of cartilage and bone (71). In RA, focal bone loss is due to an excessive bone resorption by OCs (72). Moreover, bone formation by OBs is greatly impaired at the erosion site in RA (73). TNF-α plays an important role in the bone destruction observed in RA patients. TNF inhibitors delay appearance of bone erosion in RA patients with no progression of bone destruction in responders and possibly to some extend, in non-responders (9, 74–76). In a murine CIA model, neutralization of TNF-α reduced joint pathology, especially when the treatment was started early after the onset of disease (77). Moreover, simultaneous inhibition of TNF-α and IL-17A further improved protection against joint damage even at the late stages of CIA (78). Mechanistic models of human TNF transgenic mice confirmed that ongoing TNF-driven inflammation suppresses new bone formation by Wnt signaling through up-regulation of DKK-1, and then the release of this inhibition allows new bone formation (45). Furthermore, several murine autoimmune arthritis models provided evidence that TNF deficiency reduces disease incidence (79–81).

The pro-inflammatory influence of TNF-α in RA acts at several levels. It induces other pro-inflammatory cytokines such as IL-1 and IL-17 and regulates osteoclastogenesis via several mechanisms. TNF-α accelerates cartilage destruction through inducing the production of matrix metalloproteinases (MMPs) and aggrecanases (ADAMTSs), which are enzymes that degrade components of the cartilage extracellular matrix. As an example, in human RA patients and in the human TNF-α transgenic mouse model, TNF-α was shown to regulate MMP3 and 13 by the small ubiquitin-like modifier-1 (SUMO-1) through NF-κB, which led to bone destruction (82). In addition, TNF-α stimulates OCs maturation both directly and indirectly by inducing RANKL on synovial fibroblasts, the synovial lining cells, and T-cells in the synovium (52, 83–85). This increased OCs maturation, and in turn, osteoclastic bone resorption. Moreover, TNF-α possibly promotes OCs activity via enhancing oxidative stress. The production of reactive oxygen species (ROS) by OCs helps accelerate the destruction of calcified tissue (86). This was confirmed in RA patients using an anti-TNF-α antibody therapy. In these patients, reduced levels of oxidative stress markers were associated with the control of joint and bone damage (87). Thus, TNF-α in RA has a destructive function as net effect.

Ankylosing spondylitis

Ankylosing spondylitis is the second most common form of chronic inflammatory arthritis. The unique hallmark of the disease is pathologic new bone formation (88). AS and related spondyloarthritis are characterized by a paradoxical and simultaneous bone destruction and formation occurring in proximal anatomical sites. The strongest evidence for a key role of TNF-α in AS pathophysiology comes from in vivo inhibition of TNF in AS patients. TNF blockers reduce signs and symptoms of AS, which translates into better physical function and quality of life. However, little or no effect on structural remodeling is achieved (89, 90).

In AS, relevant human samples are not readily available even so TNF expression has been shown at the site of sacroiliitis. Thus, most observations are only made in rodent models. Recently, mice that specifically overexpress the transmembrane but not the soluble form of TNF-α, so called tmTNF transgenic mice, developed clinically spinal abnormalities similar to those in AS (91). These mice developed pronounced local inflammatory infiltration of the axial and peripheral joints (88). In sharp contrast with the other TNF overexpression models, these mice do not develop systemic inflammation, yet display pronounced axial and peripheral new bone formation leading to ankylosis (92). This model strongly argues for a role of tmTNF in the pathogenesis of AS but the exact molecular mechanisms still need to be elucidated. AS patients are particularly prone to develop syndesmophytes, a bony growth originating inside a ligament, and enthesitis, an inflammation at the sites where tendons or ligaments insert to bone. Therefore, in these settings, OCs may not be in direct contact with tendon-derived OBs and TNF-α may exhibit an osteogenic effect there. This is different from the effect found in whole bone where massive bone loss is seen. Thus in AS, even with TNF-α inhibition, ectopic bone formation is still observed, meaning that additional factors are governing this effect.

Osteoporosis

Osteoporosis is characterized by a decrease in bone mass and density leading to bone weakness and prone to fracture. In RA patients, several studies have shown an increased incidence of osteoporosis. This could be due to the highly systemic inflammatory environment observed in these patients and the use of glucocorticoid treatment, which can affect bone remodeling (93, 94). Despite ectopic bone formation, AS patients show massive systemic osteoporosis (95). Among other cytokines, TNF-α seems to be an important player implicating in osteoporosis via TNFR1 (96). In RA patients treated with TNF-α inhibitors, control of bone loss was observed in parallel to that seen on arthritis levels (76, 97).