Introduction

Spondyloarthritis (SpA) is a multifaceted chronic joint disease that includes different diagnostic entities that share clinical, genetic, radiographic, and pathophysiological characteristics . The recent proposal to redefine SpA by distinguishing two important disease subforms (axial and peripheral SpA – axSpA and pSpA) linking the main clinical manifestation is gradually replacing the long list of distinct but oft overlapping disorders that made up the original SpA concept . New insights into the disease processes and signs based on magnetic resonance imaging (MRI) and improved classification criteria have boosted this new approach to disease with axSpA encompassing ankylosing spondylitis as well as non-radiographic axSpA that are typically affecting the spine and sacroiliac joints, and pSpA referring to arthritis and enthesitis that are predominant in the peripheral joints and that were most commonly classified before as reactive arthritis, inflammatory bowel disease-associated SpA, many forms of psoriatic arthritis, and undifferentiated SpA.

Among chronic inflammatory joint diseases, SpA stands out by the widespread involvement of the axial skeleton and the characteristic radiological evolution of disease . Erosive damage to joints or vertebra is relatively limited but new bone formation with bony outgrowths is dominant, for example, syndesmophytes in the spine. This process of new bone formation, sometimes described as “osteoproliferation,” can lead to progressive joint or spine ankylosis. In axSpA, progressive ankylosis strongly contributes to signs and symptoms, loss of function, and to disability in particular in patients with long-standing disease . In pSpA, new bone formation can easily be recognized in many patients but its overall impact on short- and long-term outcome of the disease, including signs and symptoms as well as loss of function and disability, has not been clearly defined. Whereas inflammation appears to be the dominant feature that impacts the quality of life of patients in the early phases of disease, over time, structural damage becomes the dominant factor in determining patient status and outcome .

The introduction of targeted therapies such as antibodies or soluble receptors directed against pro-inflammatory cytokine tumor necrosis factor-alpha (TNFα) has dramatically improved the overall outlook for patients with SpA . Nevertheless, prevention of structural damage is still a challenge as many questions with regard to this particular type of skeletal tissue damage remain. In this overview, we aim to summarize and provide novel insights into the genetic, cellular, and molecular nature of progressive ankylosis and its role in active disease. In general, two issues complicate enormously the study of ankylosis in SpA. First, human tissues from affected sites in the axial skeleton are extremely difficult to obtain due to ethical considerations. Second, ankylosis is a relatively slow process often evolving over decades. These kinetics as well as the high interindividual variation provide additional hurdles for human and translational studies. In addition, as outlined below, animal models are not the perfect answer to this challenge. Taking this perspective into account, we discuss genetic factors as potential players in determining the extent and severity of ankylosis, summarize current insights into the molecular and cellular players involved trying to develop a tissue-level understanding, revisit old and new data on the relationship between inflammation and new bone formation, to finally phrase novel conceptual questions that may help define the research agenda in this field.

Genetic factors and their impact on new bone formation and ankylosis

The strong association of HLA-B27 with AS was established more than 40 years ago . However, the pathogenic role of B27 is not known. Apart from being a susceptibility factor does HLA-B27 affect the severity of AS? This question has been addressed in several studies with variable results. HLA-B27 has not stood out as a strong genetic factor affecting radiographic progression in AS. It should however be pointed out that the vast majority of AS patients carry the HLA-B27 gene, while this is not the case in non-radiographic axSpA. This could be because HLA-B27 is a risk factor for sacroiliitis progression. However, only C-reactive protein (CRP) was a significant predictor of progression in a study on 95 patients who progressed to AS at a rate of 10% every 2 years . In other large studies looking at spinal progression of AS HLA-B27, positivity did not predict progression . Interestingly, association with other major histocompatibility complex (MHC) loci has been identified in some studies on radiographic severity in AS .

Genome-wide association studies (GWAS) have identified a number of novel genetic associations with AS, but none of the early Caucasian studies identified bone-specific markers that would be obviously pathogenic in syndesmophyte formation. Two genes, ANTXR2 and PTGER4 , which were strongly associated with AS in the initial GWAS studies, could be linked to osteoproliferation. ANTXR2 codes for anthrax toxin receptor 2, which is the receptor for the anthrax toxin, and aids in the toxin’s entry into cells. ANTXR2 is also called capillary morphogenesis 2 protein (CMG2) and is also responsible for maintaining the integrity of the basement membranes and aiding developing capillaries. ANTXR2-mediated internalization of the toxin requires the low-density lipoprotein receptor 6 (LRP6) . LRP5 and 6 are co-receptors in the Wnt–beta-catenin signaling pathway, a growth factor cascade with a role in skeletal development, homeostasis, and disease . Excess Wnt signaling can stimulate bone formation and LRP5 point mutations lead to the osteoporosis and pseudoglioma (OPPG) syndrome in humans . A direct link between ANTXR2 and osteoproliferation is yet to be established.

PTGER4 (prostaglandin E receptor 4, EP4) is one of four receptors for prostaglandin E2 (PGE2). PGE2 can induce mineralized bone nodule formation through the EP2 and EP4 receptors, which appears to be independent of bone morphogenic protein (BMP) pathways . In addition, EP4 activation can augment BMP-mediated bone formation . The association of PTGER4 with AS becomes even more interesting considering the recent reports of a possible disease-modifying potential of non-steroidal anti-inflammatory drugs (NSAIDs) (see below). Thus, PGE2 activation of EP4 could be an important mediator of osteoproliferation in AS.

A GWAS study on Han Chinese AS patients suggested novel associations with HAPLN1 , EDIL3 , and ANO6 . HAPLN1 codes for hyaluronan and proteoglycan link protein 1 (HAPLN1). HAPLN1 stabilizes aggrecan and hyaluronan aggregates in the cartilage extracellular matrix . Mice deficient in Hapln1 develop musculoskeletal abnormalities with suggested abnormalities in endochondral bone formation . Interestingly, HAPLN1 was found to be associated with spinal disc degeneration as well as osteophyte formation . However, this association is deemed to have its effect through cartilage abnormalities rather than through a direct effect on osteophyte formation.

EDIL3 codes for EGF-like repeats and discoidin I-like domains 3, an integrin receptor involved in angiogenesis and prevention of the neutrophil adhesion cascade . EDIL3 deficiency in a mouse model of periodontitis and associated bone loss led to increased IL-17 production, which is of interest in AS pathogenesis . EDIL3 is also considered important for cartilage development and organogenesis by modulating the Wnt and BMP pathways . The links with both bone loss and bone formation identify EDIL3 as a receptor of specific interest in SpA.

ANO6 codes for Anoctamin 6, a multi-pass transmembrane protein mediating the calcium-dependent exposure of phosphatidylserine on cell surface. ANO6 is expressed in differentiating and mature osteoblasts . ANO6 deletion leads to skeletal abnormalities in mice, in particular decreased mineral deposition . Again, a direct link between these novel genes and osteoproliferation in AS has yet to be established but the pathways outlined above could be potential areas to explore.

Genes associated with AS and antigen-processing pathway were specifically investigated in the context of AS disease progression . An association between the proteasome gene LMP2 and ERAP1 and progression of spinal disease was documented. LMP2 has been linked previously to uveitis and extra-spinal disease and ERAP1 has been shown to affect HLA-B27 expression in AS patients . These findings have to be replicated in larger studies to further establish them as risk factors for severe ankylosis in SpA patients.

In a recent study on radiographic severity of AS, more than 2000 patients were pooled from multiple centers . After analyzing association with 498 single-nucleotide polymorphisms (SNPs) in selected genes associated with bone homeostasis, 15 top hits were assessed in the replication cohort. Two SNPs were considered significantly associated with radiographic severity in the study. The first SNP was in the gene RANK (Receptor Activator of Nuclear Factor kB). RANK is an essential receptor in osteoclast development. This raises the intriguing hypothesis that new bone formation and radiographic progression could be linked to local or systemic bone loss (see also below). The second association was with the gene PTGS1 (Prostaglandin Endoperoxide Synthase 1) or Cyclogenase 1. This again brings in the possibility of prostaglandins playing a key role in osteoproliferation in AS. However, considering the borderline significance of the associated genes and the issues around multiple testing, the findings of this study have to be validated independently Fig. 1 .

Ankylosing spondylitis susceptibility genes with links to new bone formation.

Genetic factors and their impact on new bone formation and ankylosis

The strong association of HLA-B27 with AS was established more than 40 years ago . However, the pathogenic role of B27 is not known. Apart from being a susceptibility factor does HLA-B27 affect the severity of AS? This question has been addressed in several studies with variable results. HLA-B27 has not stood out as a strong genetic factor affecting radiographic progression in AS. It should however be pointed out that the vast majority of AS patients carry the HLA-B27 gene, while this is not the case in non-radiographic axSpA. This could be because HLA-B27 is a risk factor for sacroiliitis progression. However, only C-reactive protein (CRP) was a significant predictor of progression in a study on 95 patients who progressed to AS at a rate of 10% every 2 years . In other large studies looking at spinal progression of AS HLA-B27, positivity did not predict progression . Interestingly, association with other major histocompatibility complex (MHC) loci has been identified in some studies on radiographic severity in AS .

Genome-wide association studies (GWAS) have identified a number of novel genetic associations with AS, but none of the early Caucasian studies identified bone-specific markers that would be obviously pathogenic in syndesmophyte formation. Two genes, ANTXR2 and PTGER4 , which were strongly associated with AS in the initial GWAS studies, could be linked to osteoproliferation. ANTXR2 codes for anthrax toxin receptor 2, which is the receptor for the anthrax toxin, and aids in the toxin’s entry into cells. ANTXR2 is also called capillary morphogenesis 2 protein (CMG2) and is also responsible for maintaining the integrity of the basement membranes and aiding developing capillaries. ANTXR2-mediated internalization of the toxin requires the low-density lipoprotein receptor 6 (LRP6) . LRP5 and 6 are co-receptors in the Wnt–beta-catenin signaling pathway, a growth factor cascade with a role in skeletal development, homeostasis, and disease . Excess Wnt signaling can stimulate bone formation and LRP5 point mutations lead to the osteoporosis and pseudoglioma (OPPG) syndrome in humans . A direct link between ANTXR2 and osteoproliferation is yet to be established.

PTGER4 (prostaglandin E receptor 4, EP4) is one of four receptors for prostaglandin E2 (PGE2). PGE2 can induce mineralized bone nodule formation through the EP2 and EP4 receptors, which appears to be independent of bone morphogenic protein (BMP) pathways . In addition, EP4 activation can augment BMP-mediated bone formation . The association of PTGER4 with AS becomes even more interesting considering the recent reports of a possible disease-modifying potential of non-steroidal anti-inflammatory drugs (NSAIDs) (see below). Thus, PGE2 activation of EP4 could be an important mediator of osteoproliferation in AS.

A GWAS study on Han Chinese AS patients suggested novel associations with HAPLN1 , EDIL3 , and ANO6 . HAPLN1 codes for hyaluronan and proteoglycan link protein 1 (HAPLN1). HAPLN1 stabilizes aggrecan and hyaluronan aggregates in the cartilage extracellular matrix . Mice deficient in Hapln1 develop musculoskeletal abnormalities with suggested abnormalities in endochondral bone formation . Interestingly, HAPLN1 was found to be associated with spinal disc degeneration as well as osteophyte formation . However, this association is deemed to have its effect through cartilage abnormalities rather than through a direct effect on osteophyte formation.

EDIL3 codes for EGF-like repeats and discoidin I-like domains 3, an integrin receptor involved in angiogenesis and prevention of the neutrophil adhesion cascade . EDIL3 deficiency in a mouse model of periodontitis and associated bone loss led to increased IL-17 production, which is of interest in AS pathogenesis . EDIL3 is also considered important for cartilage development and organogenesis by modulating the Wnt and BMP pathways . The links with both bone loss and bone formation identify EDIL3 as a receptor of specific interest in SpA.

ANO6 codes for Anoctamin 6, a multi-pass transmembrane protein mediating the calcium-dependent exposure of phosphatidylserine on cell surface. ANO6 is expressed in differentiating and mature osteoblasts . ANO6 deletion leads to skeletal abnormalities in mice, in particular decreased mineral deposition . Again, a direct link between these novel genes and osteoproliferation in AS has yet to be established but the pathways outlined above could be potential areas to explore.

Genes associated with AS and antigen-processing pathway were specifically investigated in the context of AS disease progression . An association between the proteasome gene LMP2 and ERAP1 and progression of spinal disease was documented. LMP2 has been linked previously to uveitis and extra-spinal disease and ERAP1 has been shown to affect HLA-B27 expression in AS patients . These findings have to be replicated in larger studies to further establish them as risk factors for severe ankylosis in SpA patients.

In a recent study on radiographic severity of AS, more than 2000 patients were pooled from multiple centers . After analyzing association with 498 single-nucleotide polymorphisms (SNPs) in selected genes associated with bone homeostasis, 15 top hits were assessed in the replication cohort. Two SNPs were considered significantly associated with radiographic severity in the study. The first SNP was in the gene RANK (Receptor Activator of Nuclear Factor kB). RANK is an essential receptor in osteoclast development. This raises the intriguing hypothesis that new bone formation and radiographic progression could be linked to local or systemic bone loss (see also below). The second association was with the gene PTGS1 (Prostaglandin Endoperoxide Synthase 1) or Cyclogenase 1. This again brings in the possibility of prostaglandins playing a key role in osteoproliferation in AS. However, considering the borderline significance of the associated genes and the issues around multiple testing, the findings of this study have to be validated independently Fig. 1 .

Ankylosing spondylitis susceptibility genes with links to new bone formation.

Nature of new bone formation in SpA – still enigmatic

Our understanding of disease-associated new bone formation or pathological joint remodeling remains surprisingly limited. This is in sharp contrast with our views on bone erosion. Basic, translational, and clinical research have mapped this process at the molecular, cellular, and tissue level with RANKL, with the osteoclasts and the synovitis as respective culprits. However, achieving a similar level of insight into new bone formation has been an enormous challenge. Different factors contribute to this problem. Human tissues directly involved in the disease process are not readily available as biopsies from the spine or sacroiliac joint are difficult to obtain. Most pathology studies have been performed in the previous century and were based on autopsy cases, including a limited number of patients and not being able to capture the kinetics of the disease process. Modern imaging methods such as MRI have been used with great success to document inflammation but specific imaging of processes going beyond anatomical changes in new bone formation have been difficult to document. With biomechanical factors increasingly proposed as players in the onset and progression of the disease process, commonly used mouse and rat models are far from having optimal translational value. Spondylitis and ankylosis are rarely seen in mouse models and many molecular mechanisms involved in the pathological bone formation process have been suggested mainly based on peripheral arthritis models.

The enthesis, the anatomical zone in which tendons and ligaments insert into the underlying bone, has been proposed as the primary disease localization in SpA . The enthesis concept has been extensively elaborated and fits well with many other concepts that are applied to SpA. Nevertheless, imaging and pathology studies also clearly demonstrated that not only enthesitis but also synovitis and osteitis can be documented and contribute to signs and symptoms . Based on the enthesitis concept, it is tempting to speculate that new bone formation specifically originates from this site and is closely linked to the biomechanical forces that are transduced through this multilayered stress-dissipating structure. The shape and localization of bony protrusions in the spine but also in peripheral joints and extra-articular sites suggest a close link to the enthesis. However, there is no solid evidence that entheseal cells are the cells that proliferate and differentiate into bone and cartilage. Adjacent and intimately connected tissues, in particular the periosteum and to a lesser extent the synovium, may be the source of the progenitor cells that undergo pathological differentiation. In this perspective, it should be noted that mesenchymal stem or progenitor cells with multilineage potential have been identified from virtually any joint tissue but that periosteal cells in particular have a strong chondrogenic and osteogenic differentiation potential . In addition, the existence of small channels between the enthesis, synovium, and the underlying bone marrow indicates that migration of bone marrow progenitor cells may be an unexpected contributor to onset and progression of ankylosis .

At the tissue level, three different differentiation and bone-forming processes have been proposed . Animal models strongly suggest that endochondral bone formation plays a key role . In this tissue formation cascade, essential during development and growth, new bone is formed through the intermediate formation of a cartilage template, in which the chondrocytes terminally differentiate, attract osteoblast precursors cells, and are progressively replaced by bone . In addition to its physiological roles in skeletal development and in the growth plate, endochondral bone formation is typically found when a large piece of bone needs to be formed. A good example is fracture healing where endochondral bone formation is dominant. Direct or membranous bone formation is largely based on juxtaposition of bone by osteoblasts . Human pathology studies provide some support for an important role in ankylosis but the specific contribution is unclear. During fracture healing its contribution is limited and growth through juxtaposition appears to be a relatively slow process. Nevertheless, some detailed imaging and pathology studies suggest that direct bone formation may be more important in human patients than originally proposed in animal models . Third, cartilage metaplasia with calcification of the extracellular matrix surrounding chondrocytes has been documented . The molecular mechanism and its relative contribution to ankylosis remain unknown.

Molecular aspects of new bone formation in SpA

Pathology analysis in human samples and mouse models provided the basis and rationale for studying a number of developmental molecular signaling cascades as factors that could play a role in new bone formation. Typical examples include BMP, Wnt, and Hedgehog signaling. BMPs were demonstrated in an endochondral bone formation process that originates from the enthesis and periosteum in a spontaneous mouse model of ankylosis . Overexpression of BMP antagonist noggin was effective in both preventive and therapeutic strategies in this model, thereby providing the first evidence that ankylosis can be selectively targeted . Wnt signaling has very complex effects on bone formation. Studies initiated to test the effect of antibodies against Wnt receptor antagonist Dickkopf-1 (DKK1) on inflammation-associated osteoporosis in the human TNF transgenic mouse model led to the serendipitous finding that such an approach radically changed the type of joint remodeling occurring in arthritic joints . In the absence of the blocking antibody, hTNF transgenic mice develop extensive joint destruction with typical bone erosions. Blocking of DKK1 resulted in increased active Wnt signaling and the formation of osteophytes or enthesophytes. This radical change in a genetic mouse model with transition from a rheumatoid arthritis (RA)-like joint damage pattern towards an SpA-like pattern identified DKK1 and Wnt signaling as critical masterswitches, at least in mouse models. Of interest, DKK1 expression is induced by TNF . Hedghog signaling is associated with a critical step in the endochondral bone formation process . A specific inhibitor of this cascade was demonstrated to be effective in inhibiting new bone formation in the post-inflammatory phase of a serum-transfer model . By specific targeting of this cascade and the hypertrophic chondrocytes, the authors argued that such a targeted approach would have benefits in terms of safety over targeting pivotal pathways such as BMP and Wnt signaling that clearly play homeostatic roles in other organ systems .

Different efforts have been undertaken to provide additional translational and clinical evidence in SpA patients to corroborate these in vivo mouse model observations. Unfortunately, patient studies, in particular when searching for biomarkers, have yielded conflicting results. Active BMP signaling was identified in a small number of entheseal biopsies and appeared to corroborate human data . Different BMPs have been measured in the serum of distinct patient groups with conflicting results. Increased levels of both BMP2 and BMP7 were found not only in patients with AS but also in patients with RA as compared to healthy controls. In AS patients, BMP2 levels correlated with disease activity measured by the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) questionnaire, whereas BMP7 levels correlated with the Bath Ankylosing Spondylitis Radiology Index (BASRI) . In another study, such relationships were not found for BMP4 and BMP6 . By contrast, Chen et al. confirmed higher levels of BMP2, −4, and −7 in AS patients with spinal fusions as compared to those without them . In addition, BMP7 levels appear to increase after anti-TNF therapy in SpA patients .

DKK1 levels were originally reported as very low in AS patients compared to RA patients and controls . More recently, it appeared that levels of active DKK1 are low, whereas the total amount of DKK1 in AS patients could be increased . Other studies did not show clear differences in DKK1 levels but a recent study suggested increased levels of Wnt3a in AS patients . Wnt antagonists such as DKK1 but also sclerostin have been proposed as biomarkers for radiographic progression in SpA but the study design has limitations. Taking into account the relative size and dynamics of the formation of single syndesmophytes and the biological role of the markers studied, it is not clear how serum concentrations of Wnt antagonists will relate to the specific anabolic process of interest and not to general skeletal remodeling and the effect of inflammation.

Value and limitations of animal models

Animal models, in particular in mice and rats, have been instrumental in providing insights into the molecular mechanisms that may underlie new bone formation in SpA. As highlighted above, the difficulties associated with access to human tissues have made the research community largely dependent on these models. However, clinicians and scientists should be aware of the limitations found with these models. Mouse and rats have an intrinsic repair or remodeling potential in the joints that is not commonly found in humans. In most models of transient joint inflammation, the resulting damage will trigger a repair response that usually does not respect the original anatomy of the joint but rather appears to stabilize the affected articulation. This may be a somewhat effective means to reduce pain associated with joint destruction. Therefore, models with non-sustained inflammation will most frequently show joint remodeling.

Another clear limitation is that most features of ankylosis, in particular in the spontaneously occurring ankylosing enthesitis, serum, or antibody transfer models, are studied in the peripheral joints and not in the spine or sacroiliac joints. In other models, such as proteoglycan-induced arthritis in mice and in the HLA-B27 transgenic rats, spine remodeling has been documented but molecular and cellular mechanisms involved remain largely unstudied.