TGF-β's are synthesized as large pro-proteins that have to be activated before receptor binding and signal via a heteromeric complex of transmembrane serine/threonine receptors, so-called type I and type II receptors, and accessory proteins.[8] The type I receptors, also called-activin receptor-like kinases (ALKs), are activated by the type II receptors and intracellular signaling is initiated by C-terminal phosphorylation of receptor-regulated (R)-Smad proteins by the type I receptors.[9] The canonical TGF-β type I receptor is ALK5 but also other ALKs have been shown to be able to propagate the TGF-β signal. In chondrocytes ALK1 has been identified as a TGF-β receptor.[10,11] Activation of ALK5 stimulates phosphorylation of Smad2 and Smad3 while ALK1 activates the Smad1/5/8 pathway.[10,11] Phosphorylated R-Smads form heterotrimers with Smad4 and this complex is transported to the nucleus. In the nucleus this complex, in association with enhancers and suppressors, regulates gene expression. Activation of the Smad2/3 and Smad1/5/8 pathway has differential effect on gene expression and regularly antagonize each other.

In many cell types it has been shown that the TGF-β concentration to which the cells are exposed determines the preferred-signaling route. In primary human fibroblasts low TGF-β concentrations (≤1 ng/mL) activate the Smad2/3 route while high concentrations (>5 ng/mL) mainly stimulate the Smad1/5/8 pathway.[12] A similar pattern was found in myofibroblast cell lines and endothelial cells.[13,14,15] This is most likely a result of the different affinities of TGF-β for different receptor complexes. However, how cellular signaling is determined by the active TGF-β concentration will depend on the absolute and relative number of different TGF-β receptors on the cell membrane of different cell types. A consequence of this is that exposure of cells to different TGF-β contractions will results in differential regulation of gene expression at low or high TGF-β concentrations.

Healthy human cartilage contains a large pool TGF-β that is normally not readily accessible for the chondrocytes.[16] The predominant form of TGF-β in articular cartilage appears to be TGF-β1 (60%-85% of total) and high quantities of total TGF-β1 have been detected in articular cartilage (68.5±20.6 ng/mL).[17] The first measurement of TGF-β activity in joint fluids of patients with joint diseases, using bioassays, has been published by Fava et al.[18]. They found high levels of active TGF-β in synovial fluid of rheumatoid arthritis patients (10 ng/mL), OA patients (4 ng/mL) and gout patients (8 ng/mL) but no significant TGF-β activity in synovial fluid of patients with avascular necrosis. Acid activation resulted in a 3 to 4-fold increase in TGF-β activity. Furthermore, the synovial fluid of normal temporomandibular joints does not contain detectable levels of active TGF-β but elevated levels are measured in patients with temporomandibular OA.[19] Moreover, there is evidence that TGF-β is proteolytically activated in synovial fluid and elevated levels of proteases are present in OA synovial fluid.[20] These results clearly indicate that active TGF-β levels will be very low or absent in normal articular joints and elevated in joint diseases such as OA.

It can be concluded that cells in normal joints are not exposed to high TGF-β levels but that this will change once OA develops. The elevated levels of TGF-β in OA joints will activate cells that are normally not exposed to high levels of active TGF-β and this will result in altered cellular differentiation and can contribute to pathogenesis if these high TGF-β levels are enduring.

Cells in joint tissues that are normally not exposed to active TGF-β are exposed to high levels in OA joints. Even in articular cartilage a situation of temporarily activation of TGF-β by loading, but fast inactivation of activated TGF-β by the cartilage matrix,[30] is altered to a state with a continuous exposure to elevated active TGF-β levels. Since high levels of TGF-β activate different receptor pathways than low levels, and continuous exposure will also alter chondrocyte activation, the role of TGF-β on cartilage will be different in an OA joint than in a normal joint. It can be expected that preferentially the Smad1/5/8 pathway will be activated (via ALK1), a pathway that is not activated in intact healthy cartilage by TGF-β (unpublished results). The changed situation of high TGF-β levels will drive chondrocytes preferentially in the direction of hypertrophy since it has been shown that the Smad1/5/8 pathway stimulates this in chondrocytes.[31,32,33,34] Due to the altered TGF-β concentrations to which the chondrocytes are exposed the function of TGF-β changes from a factor that blocks chondrocyte hypertrophy (Smad2/3 signaling) to a factor that facilitates chondrocyte hypertrophy (Smad1/5/8 pathway). As a consequence the protective effect of TGF-β on articular cartilage will be lost. We and others have found a switch from Smad2/3 to Smad1/5/8 signaling in age-related OA models which appears to confirm the altered role of TGF-β in OA cartilage compared to normal cartilage.[11,35]

Subchondral bone is one of the tissues that shows alterations in OA joints. Accelerated remodeling, indicated by cyst formation and subchondral sclerosis are hallmarks of OA joints.[36,37,38] Several studies indicate that TGF-β plays a role in the alterations in the subchondral bone in OA. A correlation between total TGF-β and OA severity has been reported for human hip OA and increased mRNA of TGF-β1 and TGF-β3 subtypes has been shown in osteoblasts in subchondral bone from knee joints with OA.[39,40] Rats with experimental OA (anterior cruciate ligament transection [ACLT] model) treated with alendronate showed inhibition of subchondral bone resorption, vascular invasion and the local release of active TGF-β.[41] Activation of TGF-β has been shown in a mouse model of OA (ACLT model).[42] These high active TGF-β concentrations induced the formation of nestin-positive mesenchymal stem cell clusters in subchondral bone. Furthermore, elevated expression of TGF-β1 in osteoblast induced OA while TGF-β inhibition resulted in decreased development of OA in mice with experimental OA.[42] Treatment of mice and rats with post-traumatic OA (ACLT model) with halofuginone, a TGF-β inhibitor, attenuated subchondral bone deterioration.[43] Moreover, in a guinea pig model of spontaneous OA changes in subchondral bone were in parallel with active TGF-β expression. These data indicate that high active levels of TGF-β contribute to the pathogenetic changes observed in subchondral bone in OA joints.

In OA not only bone remodeling takes place but also the formation of new bone as a result of chondrogenesis and ostechondral ossification, osteophyte formation. Osteophytes are a typical characteristic of OA and are not formed in joints affected by rheumatoid arthritis. Osteophytes are bony outgrowths that originate from progenitor cells being located in the periosteum at the border line of the cartilage and bone.[44] Bolus injections of TGF-β1 or adenoviral overexpression of TGF-β1 in mouse knee joints induce the formation of osteophytes with structure and localisation similar to the osteophytes found in spontaneous murine OA.[45,46] Injection BMP's also results in osteophyte formation but these osteophytes appear to mainly originate from the growth plate chondrocytes.[47,48]

Inhibition of TGF-β activity, either blocking receptor binding by the use of the soluble type II receptor or blocking intracellular signaling using Smad7, significantly decreased osteophyte formation in experimental OA in mice.[49,50] Mice with drug-inducible chondrocyte-specific overexpression of BMP-2 did not show more osteophyte formation than control mice.[51] However, osteophyte formation was greatly boosted in BMP-2 overexpressing mice when OA (destabilization of medial meniscus model) was induced in these mice. When under these conditions TGF-β1 activity was inhibited also the formation of osteophytes was strongly reduced (unpublished). These results indicate, that at least in experimental models, TGF-β is the main driver of osteophyte formation but that this process can be enhanced by BMP activity. Elevated levels of active TGF-β, absent in normal joints, activate progenitor cells residing in joint tissues, such as periosteum, to undergo chondrogenesis and make new bone. This is confirmed by in vitro studies which confirm that TGF-β is the most potent cytokine to induce chondrogenesis in progenitor cells.[52,53]

Patients with OA complain of pain but also in many cases of joint stiffness.[54,55] Joint stiffness can be a result of fluid accumulation in the affected joint but also of fibrosis of the joints capsule. Fibrosis can be considered as a exaggerated repair process of damage tissues. TGF-β is known to be the most potent profibrotic factor. Either injection or adenoviral expression of active TGF-β result in synovial fibrosis in mouse knee joints.[47,56] Remarkably, fibrosis is more pronounced after adenoviral overexpression of TGF-β, prolonged exposure to moderate levels, while osteophyte formation is more pronounced after TGF-β injection, short exposure to high concentrations. This suggest that different cells types are activated by different regimes of TGF-β exposure, progenitor cells move forward into the direction of chondrogenesis by peak levels of TGF-β and fibroblasts are induced to proliferate and synthesize matrix molecules by prolonged elevated concentrations of TGF-β.

We could show that inhibition of TGF-β activity in experimental OA inhibited synovial fibrosis.[50] Furthermore, exposure of human fibroblasts to TGF-β or overexpression of TGF-β in murine knee joints resulted in up regulation of expression of the same genes that were found to be up regulated in synovium of knee OA patients.[57] It appears that these genes are central in synovial fibrosis and are stimulated by TGF-β. Moreover, elevated expression of procollagen-lysine, 2-oxoglutarate 5-dioxygenase 2, an enzyme increasing the number of collagen hydroxypyridinone cross links, was both detected in experimental murine OA as well as mouse knee joints with adenoviral overexpression of TGF-β.[58]

We have shown that elevated TGF-β concentrations preferentially activate the Smad1/5/8 pathway in fibroblasts.[12] In our in vitro experiments stimulation of fibrotic genes in fibroblast was mainly mediated by ALK5/Smad2/3 signaling. The role of ALK1/Smad1/5/8 signaling in fibrosis is not clear yet. It has been reported that ALK1 activity decreases fibrosis in a number of experimental set ups.[59,60] However, on the other hand BMP9, signaling via ALK1 mainly, has been reported to be a profibrotic factor.[61] The role of ALK1, being either pro or antifibrotic, is still an enigma but can be dependent on additional factors that modulate the function of the Smad proteins.[62]

OA is long considered as a non-inflammatory wear and tear disease mainly involving cartilage degradation but is now regarded as a whole joint disease with inflammation as in integral component in many patients. TGF-β can be a stimulator of synovial inflammation and hyperplasia. It has been shown that intra-articular injection of rat knee joints with TGF-β resulted in swelling and erythema within one day.[63] The TGF-β-induced infiltrate consisted after 24 hr mostly of mononuclear phagocytes with some lymphocytes and neutrophils, while after 4 hr infiltration of mainly polymorphonuclear leukocytes was observed.[64] Antagonism of TGF-β with a neutralizing antibody in a model of erosive polyarthritis blocked cell infiltration and tissue pathology.[65] These data indicate that elated levels of TGF-β in OA joints lead to TGF-β-related inflammation that can contribute to joint damage in OA.

On the other hand, although TGF-β can attract inflammatory cells to the joint TGF-β also has a strong anti-inflammatory action. As mentioned above, mice lacking TGF-β1 develop massive inflammation resembling autoimmune disorders.[4,5,66] In lamina propria cells of the gut TGF-β suppressed nuclear factor (NF)-α-induced NF-κB p65 accumulation in the nucleus, NF-κB binding DNA activity, and NF-κB-dependent gene activation.[67] Moreover, in a mouse model of zymosan-induced arthritis injection of TGF-β had no effect on inflammation but stimulated proteoglycan content and protected against proteoglycan loss.[68] These varying result show that TGF-β attracts inflammatory cells to the joint and stimulate synovial hyperplasia and fibrosis but that it is not clear what is the end effect of TGF-β on the progression of tissue destruction. This might be related to the (Osteo) arthritis model and the timing of the TGF-β or anti-TGF-β treatment.

The role of TGF-β in the normal or the OA joint is strikingly different. In a normal joint TGF-β plays an important, protective, role in maintaining the differentiated chondrocyte phenotype while in an OA joint TGF-β will have a deleterious action due to its continuous presence at high levels. This changing action of TGF-β makes therapy targeting TGF-β a challenge. In general OA is a focal process, being confined to one or a defined number of joints and even in a specific joint the OA process can be at a different stage at different locations within this joint. Systemic inhibition of TGF-β has the danger of negatively affecting normal joints while the addition of extra active TGF-β can result in the pathogenic effects of TGF-β as described above. Only local inhibition in joints with severe OA appears to be an option although one has to keep in mind that TGF-β also has a role in cartilage repair as a stimulus of chondrogenesis of precursor cells. With this in mind, human trials with TGF-β or anti-TGF-β have to very well controlled, also measuring the effect of the treatment on active TGF-β levels, since expected findings can without doubt happen to the duality and context dependent action of this important cytokine.