The role of bone morphogenetic proteins in ankylosing spondylitis

Abstract

Ankylosing spondylitis (AS), the best-known form of spondyloarthritis (SpA), is a remodelling arthritis characterized by chronic inflammation and bone formation. Ankylosis of the axial skeleton and sacroiliac joints leads to an impairment of spinal mobility, progressive spinal fusion and an increased risk of spinal fractures. The nature of the relationship between inflammation and new bone formation in AS has been controversial and questions remain as to whether there is a direct relationship between inflammation and new bone formation. Like others, we have hypothesized that the molecular pathways underlying ankylosis recapitulate the process of endochondral bone formation and that bone morphogenetic proteins (BMPs) play a key role in this process in AS. Furthermore, we discuss the entheseal stress hypothesis, which proposes that inflammation and ankylosis are linked but largely independent processes, and consider observations from mouse models and other human diseases which also imply that biomechanical factors contribute to the pathogenesis of AS. As current therapeutics, such as tumour necrosis factor inhibitors do not impede disease progression and ankylosis in AS, it is the pathways discussed in this review that are the now the focus for the identification of future drug targets.

Introduction

Bone morphogenetic proteins (BMPs) are growth factors, morphogens and cytokines that were originally identified as proteins that can induce a complete cascade of endochondral bone formation [Urist, 1965; Wozney et al. 1988]. In this essential developmental process, progenitor cells first differentiate into chondrocytes, building a cartilage template in which the cells progressively differentiate towards hypertrophy, their terminal differentiation state. Subsequently, blood vessels and osteoblast precursors invade the chondrocyte matrix and replace the cartilaginous template by bone [Hall and Miyake, 2000]. Injection of purified BMPs into the muscle of rodents is sufficient to recapitulate this process in vivo and ectopically [Luyten et al. 1989; Sampath et al. 1990; Urist, 1965; Wozney et al. 1988]. BMPs belong to the transforming growth factor superfamily and have multiple roles beyond their critical function in skeletal development and growth. For instance, BMPs are essential in early development and determine the ventral–dorsal boundary [Hogan, 1996].

Ankylosing spondylitis (AS) is the best-known form of spondyloarthritis (SpA). The SpA disease concept groups distinct diagnostic entities that share clinical, genetic and pathophysiological features. This concept also includes psoriatic, reactive, undifferentiated and juvenile-onset arthritis, as well as arthritis associated with inflammatory bowel disease. Predominant axial disease affecting the spine and the sacroiliac joints and extra-articular manifestation such as acute anterior uveitis, enthesitis, inflammatory bowel disease and psoriasis, clearly set this group of diseases apart from other forms of chronic arthritis [Braun and Sieper, 2007]. AS onset is common between 18 and 30 years, affects more males than females, and is strongly associated with the HLA-B27 antigen. The disease usually presents with inflammatory back pain (worsening with rest and improving with activity), leading to significant pain, loss of function and disability [Machado et al. 2010]. From a pathological perspective, enthesitis, osteitis and synovitis exemplify the inflammatory processes in SpA. Structural damage to the skeleton is one of the factors that defines the clinical outcome of disease [Braun and Sieper, 2007]. Structural damage in AS and related SpAs is characterized by extensive new cartilage and bone formation leading to progressive ankylosis of the spine and the sacroiliac joints. Recent data clearly demonstrate that both persistent inflammation and progression towards ankylosis contribute to loss of function and disability [Machado et al. 2010]. The clinical picture of the individual patient is determined by both features: for instance, stiffness secondary to muscle tension linked to underlying inflammation may explain how spinal movement is restricted far more than can be accounted for solely by bony ankylosis.

Treatment of ankylosing spondylitis

Treatment options for patients with AS include the use of nonsteroidal anti-inflammatory drugs (NSAIDs). These drugs are capable of limiting or suppressing inflammation in a large number of patients but often do not result in sufficient symptom control. Clinical trial data suggests that sustained therapy with celecoxib, a specific inhibitor of the cyclo-oxygenase 2 enzyme, not only reduces signs and symptoms of disease but may also affect progression of ankylosis [Wanders et al. 2005].

For patients that do not respond sufficiently to NSAID therapy, biological treatments targeting the proinflammatory cytokine tumour necrosis factor (TNF) have become available. The soluble receptor etanercept, as well as the monoclonal antibodies infliximab, adalimumab and golimumab are highly successful in controlling signs and symptoms in patients with AS. In contrast, no effect on structural disease progression characterized by ankylosis has been demonstrated in patients with established disease over a 2-year period [van der Heijde et al. 2008a, 2008b, 2009]. Although rapid progression towards spinal ankylosis may only be clinically relevant for a subset of patients, many questions remain regarding the effect of TNF inhibition and structural disease progression in patients with early disease. Therefore, these observations suggest that the search for additional targets is warranted.

Research into new bone formation

We and others have hypothesized that the molecular pathways underlying tissue remodelling and ankylosis, as seen in AS patients, recapitulate the process of bone formation from development and growth [Lories et al. 2009]. Considering the key role of BMPs in endochondral bone formation, a role for this protein family in the progression of AS seems likely.

Several issues arise when considering the use of human tissues in the study of AS. First, patient tissues from the spine or sacroiliac joints are difficult to obtain and progression of disease in patients is a relatively slow process. Second, precise monitoring of disease progression is hampered by current imaging protocols, which are less than perfect. Fortunately, mouse models of arthritis provide a useful alternative to understand some of the molecular mechanisms underlying progressive joint or spine ankylosis and its relationship with inflammation. The spontaneous arthritis model in aging male DBA/1 mice (SpAD) has been extensively characterized [Corthay et al. 2000; Lories et al. 2004; Nordling et al. 1992]. It occurs with an incidence approaching 100% by the age of 26 weeks after grouped caging of different males from distinct litters at the age of 10 weeks. The arthritis is fairly mild, and only appears in the hind paws. Although earlier reports stressed the occurrence of ankle involvement, we found the model in our animal facility limited to arthritis of the toes, mostly nonsymmetrical and affecting the fourth and fifth digits. A short phase of inflammation is suggested by histomorphological analysis and is characterized by neutrophil and mononuclear cell influx in close proximity to joint-associated entheses and by subcutaneous oedema [Lories et al. 2004]. This presentation shows remarkable similarities with dactylitis, a typical clinical feature of psoriatic arthritis. The clinically apparent arthritis is revealed under the microscope as a surprisingly rapid cascade of endochondral bone formation that occurs at the enthesis. First, entheseal cells appear to proliferate and commit to chondrogenic differentiation. These cartilage-like cells further differentiate towards hypertrophic chondrocytes and are then progressively replaced by bone. As the process is occurring on both the proximal and distal side of the joint, the end result is ankylosis and loss of joint function. Although the model does not mimic all aspects of AS, such as HLA-B27 genetic associations, it is extremely useful to study molecular aspects of ankylosis and the relationship between ankylosis and inflammation. In this context, differences in disease localization, peripheral arthritis in the mouse versus axial lesions in humans, also emphasize that some caution needs to be in place when translating concepts from the model towards the patients. Ideally, datasets and concepts should be corroborated by ex vivo or in vivo patient studies.

Bone morphogenetic protein signalling and ankylosing spondylitis

BMPs can activate different signalling cascades in the cell [Gilboa et al. 2000]. Of these, the SMAD signalling pathway has been most extensively studied. Upon ligand receptor binding, the kinase activity of the type I BMP receptors phosphorylates SMAD1, 5 or 8. Subsequently the phosphorylated SMAD associates with common SMAD4 and translocates to the nucleus where it can associate with different transcription enhancers or repressors [Nohe et al. 2002]. We used immunofluorescence and immunohistochemistry to detect phosphorylated SMAD proteins in the spontaneous arthritis model in DBA/1 mice and found evidence for active BMP signalling in the early stages of the disease process, when entheseal cells are proliferating and committing towards chondrogenic differentiation [Lories et al. 2005, 2006]. These first steps result in local production of different BMPs by the differentiating entheseal cells. However, a role for other cell types, including inflammatory cells in producing BMPs cannot be excluded. Based on this observation we performed a series of experiments in which we used overexpression of the BMP antagonist noggin to block the signalling cascade in the mouse model [Lories et al. 2005].

Inhibition of BMP signalling by noggin resulted in protection against arthritis and ankylosis, not only in a preventive setting in which gene transfer was performed before onset of disease but also in a therapeutic setup where gene transfer was performed at the onset of clinical symptoms. Further ex vivo analysis using biopsy material from SpA patients with extra-articular enthesitis, further corroborated our initial observation that BMP signalling is active in the early stages when progenitor cells are committing towards chondrogenic differentiation.

Inflammation and new bone formation

The nature of the relationship between inflammation and new bone formation in AS has been controversial. The recent observation that 2 years of treatment with TNF blocking agents does not affect radiographic progression in patients with established disease has stimulated further research [van der Heijde et al. 2008a, 2008b, 2009]. The question remains whether there is a direct relationship between inflammation and new bone formation. Imaging studies have demonstrated on the one hand that the majority of syndesmophytes, e.g. at the vertebral margins, occur in vertebrae in which prior MRI images did not reveal active inflammation. On the other hand, syndesmophytes are more likely to appear in vertebrae in which inflammation has disappeared but was detected at baseline [Maksymowych et al. 2009]. Our work in the DBA/1 model further suggested that inhibition of TNF using the human soluble receptor etanercept does not affect incidence or severity of ankylosis [Lories et al. 2007]. Furthermore, inhibition of osteoclasts using bisphosphonates also has no impact on ankylosis in the DBA/1 model [Lories et al. 2008]. These observations further support the view that the pathological bone formation in AS and its models is not due to abnormalities in the normal bone remodelling cycle. From this perspective, the remarkable paradox between concomitant bone loss, due to inflammation-associated osteoporosis and new bone formation, as is often seen in patients with AS, can be explained by different biological mechanisms steering on the one hand, the bone remodelling cycle and, on the other hand, new bone formation originating from the entheses [Carter and Lories, 2011].

Based on these observations, we have put forward the entheseal stress hypothesis [Lories et al. 2009]. Our view suggests that inflammation and ankylosis are linked but largely molecularly independent processes. Data from mouse models and concepts from human disease implies a biomechanical factor contributes to the pathogenesis of AS. We therefore propose that entheseal ‘stress’, resulting in cell death or damage, forms a trigger for both inflammation and new bone formation. The inflammatory process is steered by cytokines such as TNF and may be supported or amplified by the presence of additional factors including infection, bowel inflammation and genetic susceptibility. On the other hand, ankylosis or its severity may also be a genetic trait as similar forms of new bone formation can be seen in patients with diffuse idiopathic skeletal hyperostosis (DISH) without clinically relevant inflammation. As outlined above in the context of the spontaneous arthritis model, it is not clear to which extent structural changes to the skeleton in DISH patients mimic processes in patients with AS. Further research in this area is certainly warranted but is hindered by the difficult access to patient tissues.

Links between inflammation and bone morphogenetic proteins?

Activation of mitogen activated protein kinases (MAPK) is an alternative intracellular signalling cascade induced by BMP ligands [Gallea et al. 2001; Guicheux et al. 2003]. As MAPK activation, and in particular p38 activation, has been associated with both BMP and proinflammatory cytokine signalling, we recently studied the effects of p38 inhibition on chondrogenesis and osteogenesis induced by BMPs [Braem et al. 2011]. Our in vitro data strongly support a role for p38 signalling in the endochondral cascade as blocking of p38 activity inhibited differentiation of progenitor cells. However, in vivo experiments paradoxically demonstrated accelerated new bone formation in the group of mice treated with a p38 inhibitor. We hypothesize that this paradox can be explained by the relatively short half-life of the compound in vivo resulting in compensatory mechanisms leading to increased MAPK activation. This observation could teach us an important lesson for the development of therapies that aim to control structural progression of the disease by suggesting that continuous suppression of bone-forming processes is a critical issue.

Additional lessons about new bone formation may also come from observations of the rare monogenetic disorder, fibrodysplasia ossificans progressiva (FOP). This is a very severe recessive disease whereby any injury or trauma to the muscle can result in a local process of endochondral bone formation, which in the long term leads to the formation of an exoskeleton [Shore and Kaplan, 2008]. The events are painful and result in severe disability and a severely shortened life expectancy. Genetic studies have revealed that FOP is caused by mutations in a BMP receptor, the Activin A receptor type 1 (ACVR1 or ALK2) [Shore et al. 2006; Tsuchida et al. 1993]. These mutations are likely to make the receptor constitutively active. Yu and colleagues developed a mouse model in which the Acvr1 gene was mutated to obtain a constitutively active form of the protein [Yu et al. 2008]. To avoid general effects, the mutated receptor was rendered inactive by the introduction of a stop cassette. Removal of the stop cassette in the genome using a Cre-recombinase strategy by adenovirus transfer into the muscle resulted in the formation of new bone in the infected muscle. In contrast, systemic removal of the stop cassette using a tamoxifen-based approach did not result in a similar phenotype unless the muscle was locally challenged with a nonspecific adenovirus. These data strongly suggest that even in the presence of an overactive BMP signalling cascade, local triggers such as inflammation or cell stress and damage are necessary for the cascade to develop. Taken together these recent observations in FOP models not only support a critical role for abnormal BMP signalling in pathology but also link cell stress and damage to the aberrant response.

A complicated molecular network steering bone formation in AS

Bone formation during development, growth and disease is taking place in a complex network with different cell types and molecular signalling cascades playing an active role. The original identification of BMPs as morphogens capable of triggering a full cascade of endochondral bone formation [Urist, 1965; Wozney et al. 1988] highlights their likely key role in the initiation of these processes, not only in health but also in disease. The observation that proinflammatory cytokines such as interleukin-1 and TNF-α can upregulate BMPs in different mesenchymal cell types further supports this view [Fukui et al. 2006; Lories et al. 2003]. Recent studies have also associated Wnt signalling and its antagonists with new bone formation in AS. Inhibition of DKK1, an extracellular Wnt coreceptor antagonist, shifts the phenotype of the human TNF transgenic mouse model from a destructive towards a remodelling arthritis [Diarra et al. 2007]. Moreover, low levels of Wnt antagonists such as DKK1 but also sclerostin are found in patients with AS as compared with other joint diseases or controls [Appel et al. 2009; Heiland et al. 2011]. The interactions between BMP and Wnt signalling remain difficult to understand. Wnts stimulate direct bone formation with differentiation of progenitor cells into osteoblasts but have more complex effects on chondrogenesis most likely depending on local concentrations and gradients. Cytokines such as TNF may play an important role in establishing such gradients as both BMPs and DKK1 appear upregulated by TNF. A simplified conceptual overview is given in Figure 1 illustrating some of the known interactions with focus on the role of BMPs during new bone formation.

Figure 1.

New bone formation in ankylosing spondylitis may result from both endochondral and direct bone formation. In the former process, skeletal progenitor cells first differentiate into chondrocytes building a cartilaginous template that is subsequently replaced by bone, a process triggered by cell death of the hypertrophic chondrocytes. In direct bone formation, progenitor cells directly differentiate into osteoblasts. BMPs appear essential in the early stages of endochondral bone formation. Wnts may stimulate or inhibit this process depending on specific ligands and concentrations. Wnt are essential in the direct osteogenic differentiation process. Wnt antagonists such as DKK1 and sclerostin may have important regulatory roles. In addition, proinflammatory cytokines such as TNF stimulate both BMPs and Wnt antagonists indicating again how local concentrations and gradients are likely to determine the overall outcome.

Conclusion

Recent progress in models of AS and related disorders supports the notion that BMP signalling has an important role in the structural progression of this disease. In addition, the entheseal stress hypothesis suggests that activation of BMP cascades may be a very early and even initiating event in AS and related SpAs. This may in the long term have an impact on the development of novel drugs and strategies.