The Discovery of Autoantibodies in RA Patients and the Concept of Autoimmunity

The Mucosal Origin of Autoimmunity in RA

The RA-associated autoantibodies can be present in the serum years before the first signs and symptoms of synovitis [15], suggesting that the initial events resulting in the emergence of autoimmunity may occur outside the synovium. The strong association of pulmonary exposures (e.g., cigarette smoking, silica or textile dusts) with the development of ACPA and RA (discussed below) suggests that autoimmunity might be triggered in the lungs [16]. This hypothesis was further supported by studies in ACPA-positive subjects revealing (a) radiographic signs of parenchymal inflammation in the lung, (b) ectopic lymphoid structures in lung biopsies, and (c) presence of IgA ACPA and citrullinated proteins in bronchoalveolar lavage and sputum [17]. Cigarette smoking creates in the lung a permissive microenvironment for ACPA production: it activates peptidyl-arginine deiminase (PADI) inducing the emergence of citrullinated neoepitopes in a background of local inflammation [18]. Another scenario implicating inflammation in the oral mucosa or intestinal dysbiosis in triggering autoimmunity in RA requires further validation [19].

HLA Association and the Concept of Ag Presentation by the Shared Epitope

In 1969, it was first described that cross-reactivity in mixed lymphocytic reactions was decreased in most RA patients if the stimulating donor was another RA patient [20]. This observation suggested that RA patients share common HLAs. Serotyping experiments in the late 1970s identified that about 70% of RA patients share the same HLA-DR4 alleles [21, 22]. In the late 1980s, the shared epitope hypothesis was proposed following the discovery that the majority of RA patients share a 5-amino acid sequence motif in the DRβ chain of HLA-DR4 [23]. A recent study fine-mapped the strongest RA link to amino acids located within the antigen-binding groove of HLA-DRβ chain [24], suggesting that the shared epitope might contribute to disease pathogenesis via antigen presentation of pathogenic epitopes. Structural studies indicate that citrullinated peptides “fit better” within the groove of the shared epitope and are presented more efficiently to cognate T-cells [25, 26]. Thus, in shared epitope carriers, citrullination of self-peptides increases their immunogenicity and triggers ACPA production.

Discoveries from Genome-Wide Association Studies (GWAS)

Since the introduction of GWAS in the early 2000s, more than 100 genetic loci have been associated with RA [27]. Prominent among these predisposing loci are genes involved in antigen-mediated activation and co-stimulation of T-cells (e.g., HLA class II, PTPN22, CD28, CTLA4), post-translational modification of proteins (e.g., PADI), and cytokine signaling (e.g., TNF, IL6R, STAT4, TYK2, TNFAIP3, REL). Each of the non-HLA predisposing loci confers only modest increase in the risk of developing RA (odds ratios ≤2) [28]. Notably, the combination of the shared epitope with variants of PTPN22 and TRAF1-C5 increases >40-fold the risk of developing RA, suggesting an additive or synergistic effect when several risk-variants are present in the same individual [29, 30]. The long list of RA-associated genetic variations includes genetic loci predisposing also to other autoimmune and inflammatory diseases, suggesting pathogenetic similarities and potential common therapeutic targets among the spectrum of immune-mediated diseases [31]. Another conclusion from GWAS is that between ACPA-positive and ACPA-negative RA there are overlapping and distinct risk variants [32], with shared epitope and PTPN22 variants contributing primarily to ACPA-positive RA whereas PRL and NFIA are associated with ACPA-negative RA [17, 33].

Notably, 80% of the RA-associated genetic variations identified by GWAS are localized in non-coding regions [34]; thus it may predispose to RA without altering the amino acid sequence and the function of a gene product. Fine-mapping studies revealed that many of these non-coding variations colocalize with (a) expression quantitative trait loci (eQTL) and (b) DNA-binding sites for STATs [27, 35]. The latter indicates the involvement of Jak-STAT pathway in RA pathogenesis, which is now proven by the success of Jak inhibitors in the clinic [36]. In this context, the non-HLA genetic variations may contribute in RA pathogenesis via two potential mechanisms: (1) by altering the expression levels of genes (change the expression or DNA binding of transcription factors) and (2) by altering the amino acid sequence and the function of proteins [34].

Link with Cigarette Smoking and the Stochastic Model of Genes-Environment Interaction

The concordance rate of RA between identical twins is only 12–15% [37], suggesting that strong environmental effects (e.g., by cigarette smoking, silica exposure, and microbiome) cooperate with the disease-predisposing genetic background for the full-blown development of RA [38]. Cigarette smoking is the best characterized environmental risk factor and its link with RA was first described in 1987 [39]. Smoking increases ≥20-fold the risk of developing RA in the presence of the shared epitope or variants of PTPN22, revealing the stochastic synergy between genes and environment in RA pathogenesis [40, 41]. Smoking induces citrullination of self-proteins in the lung and increases the risk of developing ACPA, especially in shared epitope carriers [42].

Discoveries Leading to the Concepts of Cytokine Network and Cytokine Hierarchy

Seminal observations by the group of Marc Feldnamm and Ravinder Maini established the pathogenetic role of TNF in RA [43]. At first it was shown that in RA, but not osteoarthritic synovial cell cultures, anti-TNF antibodies dramatically reduced the production of a range of proinflammatory cytokines, such as IL-1, GM-CSF, IL-6, and IL-8 [44, 45]. Subsequent proof-of-concept clinical trials demonstrated that ant-TNFs are clinically effective and rapidly reduce serum levels of IL-6 [46, 47]. These observations introduced the concepts of the cytokine network and cytokine hierarchy in RA pathogenesis: within the RA synovium although many proinflammatory cytokines are highly expressed and share similar signaling pathways there is no biological redundancy; instead there is a cytokine network with hierarchical organization [48]. Hence, there are cytokines (e.g., TNF or IL-6) operating in the context of RA upstream of others (e.g., IL-1) driving the cascade of cytokine production within the inflamed joints. Targeting these upstream cytokines cripples the downstream cascade and has higher therapeutic potential. This concept is further supported by the differential therapeutic effects in RA of the various strategies targeting cytokines expressed in RA synovium: targeting TNF and IL-6 is effective and targeting GM-CSF appears promising [49], whereas targeting IL-1 is moderately effective [50] and trials targeting IL-17 or IL-12/IL-23 have not provided convincing benefit so far [51, 52].

Discoveries Leading to the Concept of Cytokine-Driven Osteoclastogenesis

Three types of bone loss are typically observed in RA: (a) marginal erosions in the interface of synovium with adjacent bone, (b) periarticular demineralization of bones near the inflamed joints, and (c) systemic bone loss [53]. The foundational discoveries of osteoimmunology, the myeloid origin of osteoclasts (OC) and the cytokine-driven osteoclast differentiation/activation, have established the mechanistic link between inflammation and bone destruction [54]. In 1998 it was first described that a cytokine of the TNF-superfamily, now known as RANKL, is the master-regulator of osteoclast generation and function [55]. Subsequent studies in RA revealed abundance of osteoclast precursors and expression of RANKL in the areas of pannus invasion into bone, and have identified activated fibroblast-like synoviocytes (FLS), T-cells, B-cells, and osteoblasts as the sources of RANKL within RA synovium [56–59]. The effectiveness of denosumab, an antibody against RANKL, in inhibiting the progression of erosions provides in vivo evidence for the contribution of RANK/RANKL-axis in bone destruction during RA [60]. Treatments blocking TNF and IL-6 also prevent bone erosions in RA, indicating that cytokines other than RANKL are also involved in bone destruction. Expansion of OC-precursors’ pool, induction of RANKL, and synergy with RANKL pathway are proposed mechanisms for the cytokine-driven osteoclastogenesis in RA [53].

Discoveries Leading to the Concept of ACPA-Driven Osteoclastogenesis

The association of ACPAs with higher risk of erosions was known for years in established RA [53]. Recently, it was shown that the presence of ACPAs is also associated with lower bone density in healthy individuals [61]. A mechanistic link between ACPA and osteoclastogenesis has been suggested by the following observations: (1) administration of human ACPA in mice induces systemic bone loss in vivo in the absence of any apparent synovitis [62, 63], (2) ACPA or Fab fragments with ACPA-specificity promote human osteoclastogenesis in vitro, (3) fluorescently labelled ACPA bind to osteoclasts and osteoclast-precursors in murine joints [64], (4) protein citrullination and surface expression of citrullinated proteins is part of the normal process of osteoclast differentiation [62], (5) ACPA with specificity against citrullinated vimentin promote human osteoclastogenesis in vitro [62], and (6) pharmacologic inhibition of citrullination or neutralization of IL-8 diminish ACPA-driven osteoclastogenesis in vitro [63]. These observations suggest Fab-mediated direct binding of ACPA to citrullinated proteins on osteoclasts and osteoclast-precursors and IL-8-mediated pathways as mechanisms for ACPA-driven bone loss.

In addition, a potential role for Fc/Fc Receptor-mediated mechanisms in ACPA-driven bone loss is indicated by the following observations: (1) the sialylation status of Fc fraction of IgGs modulates the binding of IgG-immune complexes to Fc Receptors and alters their osteoclastogenic potential [65], (2) the sialylation status of ACPA modulates in vitro and in vivo their osteoclastogenic potential [66], and (3) RA patients with lower levels of ACPA sialylation displayed lower bone density [66]. All together, these observations suggest that a combination of Fab-mediated and Fc-dependent mechanisms triggers ACPA-driven osteoclastogenesis in ACPA-positive individuals, even in the absence of overt synovitis [18].

Discoveries Leading to the Concept of Inflammation-Driven Inhibition of Osteoblasts

In RA patients, repair of existing erosions is infrequent [67], and histopathologic analysis in animal models revealed paucity of mature osteoblasts (OB) in areas of bone erosion [68, 69]. These observations suggest that synovial inflammation not only induces osteoclastogenesis but also impairs OB differentiation and function. The concept of inflammation-driven inhibition of OB is supported mechanistically by studies investigating the in vivo impact of inflammation on the regulation of Wnt pathway, which is the key driver of OB differentiation [70]. In animal models, synovial inflammation upregulates Wnt pathway-antagonists (e.g., DKK1, sclerostin, sFRP1, and sFRP2) with parallel downregulation of Wnt agonists (e.g., Wnt10b) [68, 69, 71]. In RA patients and animal models, the high levels of DKK1 are decreased by treatment with anti-TNF, indicating TNF as a key inducer of DKK1 [71, 72]. Along the same lines, pharmacologic inhibition of DKK1 or sclerostin in animal models prevents bone erosions [71, 73, 74]. These observations suggest that the inflammatory milieu of RA synovitis inhibits OB differentiation/function by deregulating the balance between antagonists and agonists of Wnt pathway.

The Concept of Fibroblast-Like Synoviocytes (FLS) as “Imprinted Aggressors”

The histopathologic hallmark of RA is pannus, which is a hyperplastic synovial lining mass that produces cytokines and proteases, erodes subchondral bone, and degrades the adjacent cartilage. Bone erosion is driven by OC and cartilage degradation is directly mediated by invading FLS [75]. Ex vivo models in immunodeficient mice show that transplanted FLS derived from RA patients retain for months their invasive, migratory, and cartilage-destructive capacity [76, 77]. In vitro studies show that passaged RA FLS retain higher spontaneous and triggered cytokine production, compared to FLS derived from osteoarthritis [78, 79]. These observations reveal that RA FLS possess or acquire an “arthritogenic memory”, which is remarkably stable and autonomous from exogenous stimulation. Subsequent studies in RA FLS identified a unique DNA methylation signature, which is dependent on disease duration, persists ex vivo, and impacts the expression of arthritogenic genes [80–82]. Additional epigenetic modifications, such as inflammation-induced changes in histone acetylation and microRNAs’ expression, have been associated with the persisting functional abnormalities observed in RA FLS [79, 83–85]. Taken together, these observations suggest that RA FLS are epigenetically imprinted, and the observed epigenetic marks (e.g., DNA-methylation and histone-acetylation) are either inherited or occur de novo as result of inflammation and environmental influences [86].

In this context, RA FLS not only respond passively to the surrounding microenvironment ( passive aggressors ) but also possess an autonomous arthritogenic phenotype due to epigenetic imprinting ( imprinted aggressors ) [86]. The consequence of this model is that RA FLS potentially retain their autonomous pathogenetic functions even under potent immunosuppression, hence the imprinted aggressors may drive disease flares or the residual synovial inflammation observed in inadequate responders to synthetic or biologic DMARDs [87]. In this group of resistant patients, therapeutic targeting of FLS, on top of the standard-of-care immunosuppression, emerges as an attractive option [88]. FLS-targeting approaches, including epigenetic modifiers and antibodies against FLS-expressing molecules, are currently in early stages of clinical development [89].

The Three Phases of RA Pathogenesis

The clinical syndrome of rheumatoid arthritis results from a sequence of events, which unfold in three phases (Fig. 1.1). In the preclinical phase , individuals carrying disease-predisposing genetic variants and epigenetic modifications (individuals at risk) develop over time an asymptomatic immune response characterized by autoantibodies (e.g., ACPA, RF, anti-CarP) and systemic production of cytokines/chemokines [90]. This phase is triggered by environmental factors (“first hit”) and occurs outside the joints (e.g., in the lungs). The duration of preclinical phase is variable and could be many years before the onset of the first symptoms and signs of arthritis. The transition from preclinical to early RA is characterized by serological changes (discussed below), systemic bone loss, arthralgias, and synovitis [30]. Gradually, the disease enters the third phase of established RA characterized by destructive synovitis and extra-articular comorbidities [91].

It appears that different pathogenetic pathways drive each phase. Ag stimulation of T-cells, activation of B-cells, and antibody production drive the preclinical phase [90]. Evolution of the adaptive immune response to modified self-proteins (epitope spreading and isotype class switch) together with synovial cell activation drives the phase of early RA [30]. Finally, established RA involves the interaction of resident synoviocytes with recruited immune cells, synovial cytokine networks, and local production of antibodies with deposition of immune complexes that trigger complement activation [30, 91].

Homing to Synovial Joints: The Unique Anatomy of the Joint and the “Second Hit Hypothesis”

The autoantigens in RA (e.g., citrullinated vimentin, enolase, fibronectin, fibrinogen) are not synovial tissue-specific, and autoimmunity emerges outside the joints (e.g., in the lung or oral cavity) [18]. In addition, ACPA alone although directly trigger bone loss are not sufficient to initiate arthritis in humans and mice [18]. In this context, the tropism of RA for synovial joints cannot be fully explained by the development of autoantibodies and by the classical concept of organ/tissue-specific autoimmunity. An alternative hypothesis suggests that the unique anatomy of synovial joints renders them vulnerable to a “second hit” [87].

The synovial lining consists of macrophage-like synoviocytes (MLS) and FLS, which are loosely attached to each other without tight-junctions or basement membrane [92]. The juxtaposition of FLS/MLS favors the development of cytokine networks with feed-forward autocrine and paracrine loops [93]. The lack of physical barriers allows free trafficking of cells and macromolecules (such as antibodies and immune complexes) from the vasculature to the joint. The articular cartilage is a “sticky surface” favoring the non-specific binding of antibodies and immune complexes and, due to lack of complement inhibitors, allows the unopposed activation of complement system [94]. In addition, the extracellular matrix of synovial joints contains long-lived immobile proteins with very low turn-over rate, favoring the accumulation over time of neoepitopes via post-translational modifications [14]. Finally, the presence of cortical bone pores potentially allows the spread of inflammation from the subchondral bone and juxta-articular bone-marrow to the adjacent synovium [95]. All these unique anatomical features render the joint (a) vulnerable to a “second hit” and (b) a “fertile soil” that fuels local inflammation.

The nature of the “second hit” is obscure, but joint inflammation due to local microtrauma or transient infections has been suspected. Then it follows synovial endothelium activation, increased vascular permeability, synovial recruitment of immune cells, and deposition of antibodies and immune complexes [30, 96]. Notably, around the time of transition from preclinical to early RA, there are changes in the status of ACPA (increase in serum titers, isotype-class switch, epitope spreading, modulation of the glycosylation status) [97, 98]. These observations suggest that changes in the effector functions and the antigen specificity of ACPA might contribute to the homing of the disease to synovial joints.

The Spectrum of Spondyloarthritides

In 1970s, the unifying concept of spondyloarthritis (SpA) was introduced to describe a spectrum of clinical phenotypes [99], which share some key clinicopathological, genetic, and etiopathogenetic features. Based on the clinical presentation and the underlying etiology, SpA is classified into six endotypes: (1) ankylosis spondylitis (AS) [100], (2) psoriatic arthritis (PsA) [101], (3) enteropathic arthritis [102], (4) reactive arthritis (ReA) [103], (5) undifferentiated SpA (USpA) [104], and (6) juvenile SpA (JSpA) [105]. The three hallmarks of all endotypes are (a) strong association with HLA-B27 [106], (b) enthesitis (inflammation of the insertion sites of tendons and ligaments into bone) [107], and (c) abnormal new bone formation (enthesophytes or syndesmophytes) [108].

The endotypes of SpA target in various combinations five anatomical domains: (1) axial skeleton (e.g., sacroiliitis, axial enthesitis, spondylitis), (2) peripheral musculoskeletal structures (e.g., peripheral synovitis, enthesitis, dactylitis), (3) skin (e.g., psoriasis), (4) gut (e.g., inflammatory bowel disease or subclinical intestinal inflammation), and (5) eye (e.g., uveitis) [109]. According to the localization of the dominant musculoskeletal manifestations, SpA is subclassified into axial SpA (axSpA), if the involvement of the axial skeleton predominates [110], and peripheral SpA (perSpA) if the main manifestations are peripheral synovitis, enthesitis, or dactylitis [111]. Recently, the concept of non-radiographic SpA (nrSpA) was introduced to allow early diagnosis and treatment based on signs of early inflammation in sacroiliac joints captured by MRI, before the occurrence of radiographically visible structural damage [112]. This chapter focuses on the pathogenetic themes which are common among the SpA endotypes.

The Complex Genetic Landscape of SpA: Strong Polygenic Heritability

The role of heredity in the pathogenesis of SpA was first suspected when Bechterew’s disease (now known as AS) was observed to be segregated in families [113]. In the 1970s, three independent studies revealed that 88–96% of AS patients were carriers of HLA-B27 [114–116]. Subsequent studies in identical twins described a concordance rate of 63% for AS and estimated heritability in SpA at 90% [117, 118]. Although HLA-B27 is the strongest genetic risk factor, it explains only 20% of the total genetic predisposition [119]. Using next-generation sequencing technologies, more than 100 polymorphisms in >40 genes were associated with the various SpA endotypes [120–125]. Despite this progress, more than 70% of the genetic contribution remains unknown [119].

Fine mapping of the SpA-associated loci revealed that only a minority are missense variants that change the function of the encoded protein [119]. The best studied example of a missense variant is rs11209026, a single nucleotide polymorphism (SNP) of IL23R, which encodes a substitution of arginine with glutamine at position 381 that results in a loss-of-function variation of IL-23 receptor and confers substantial disease protection [120, 126]. Notably, the majority of risk variations are colocalized with epigenetic marks of enhancers and predispose to SpA by regulating the expression of genes [119]. One example is rs6600247, a SNP that reduces the recruitment of IRF4 and decreases the expression of RUNX3 in CD8 positive T-cells [127].

Presentation of “Arthritogenic” Antigens by HLA Class I

In further support of the “antigen presentation theory”, among the SpA-predisposing variants there are polymorphisms in genes involved in peptide processing and loading to HLA class I, primarily polymorphisms in the gene encoding endoplasmic reticulum aminopeptidase 1 (ERAP1) (more details on these genes in Table 1.3) [132]. ERAP1 is an enzyme operating as a “peptide ruler”, trimming peptides to the optimal size for HLA class I binding [133]. Notably, ERAP1 is in epistasis with HLA-B27 and confers risk for AS only in the presence of HLA-B27 [129], suggesting a pathogenetic partnering of the “peptide ruler” with the “peptide presenter”. The locus of ERAP1 is highly polymorphic and the SpA-predisposing variants increase the enzymatic activity or the expression of ERAP1, modulating the repertoire of peptides that bind to HLA-B27 [132]. In this context, inhibition of ERAP1 emerges as an attractive future therapeutic strategy for SpA.

HLA-B27 Beyond Antigen-Presentation: Misfolding and Homodimerization

Peptides with convincing arthritogenic potential have not been identified yet, and there is very limited evidence so far about the presence of HLA-B27-restricted arthritogenic CD8+ cells in patients with SpA [134]. In addition, HLA-B27 transgenic animal models do not require CD8+ cells or transporters of antigen processing (TAP) for disease development [135–137]. All together, these observations have challenged the concept of presentation of arthritogenic antigens by HLA-B27 to CD8+ T-cells. Two alternative concepts (Table 1.4) have emerged recently suggesting that HLA-B27 contributes to SpA pathogenesis by intracellular misfolding or by cell-surface homodimerization [106]. According to the misfolding hypothesis , intracellular accumulation of misfolded HLA-B27 triggers endoplasmic reticulum stress, autophagy, and IL-23 production [138–143]. Notably, a recent study indicates that intracellular accumulation of HLA-B27 and autophagy-triggered IL-23 production occur in the gut of AS patients [144]. According to the homodimerization hypothesis , beta-2-microglobulin free heavy chains (FHC) of HLA-B27 homodimerize on the cell-surface and are recognized by killer-immunoglobulin-like receptors (KIR) expressed on NK-cells and CD4+ T-cells, inducing their activation and IL-17 production [145–149]. In support of this scenario, CD4+ T-cells expressing KIR-3DL2 and markers of Th17 lineage commitment were identified in patients with AS [150]. The antigen-presentation theory and the misfolding and dimerization hypotheses should not be considered mutually exclusive, instead all three mechanisms may contribute in different degrees depending on the context.

Epithelial Barrier Disruption: Dysbiosis or Infection

Clinical observations showing that 10–20% of patients with IBD and 30% of patients with psoriasis develop SpA indicate a potential mechanistic link between epithelial barrier disruption and SpA [100]. In support of this concept, GWAS studies reveal overlapping genetic predisposition among IBD, psoriasis and SpA [128]. The most obvious pathogenetic consequence of a disrupted epithelial barrier (gut or skin) is the entrance into the host of entire microorganisms or their fragments (e.g., antigenic epitopes and molecular patterns) [19]. Along these lines, increased intestinal permeability has been observed in patients with AS [151]. According to this pathogenetic scenario, a “leaky” gut or skin licenses the colonizing microbiome to trigger an arthritogenic immune response or facilitates an arthritogenic infection [19, 152]. The essential role of microbiome in SpA pathogenesis is suggested by observations in animal models of SpA, where the disease does not develop under germ-free conditions [153, 154]. Recent studies reveal changes in the intestinal microbial composition (dysbiosis) in patients with SpA and animal models indicate a potential role of HLA-B27 in this process (Table 1.4) [155–158].

The Gut-Joint Axis: Intestinal Production of IL-23 and Expansion of Arthritogenic Immune Cells

The concept of a gut-joint axis was coined to describe the mechanistic link between the development of SpA and events occurring in the gut, such as epithelial barrier disruption, dysbiosis, intestinal inflammation, and arthritogenic priming of the immune system [19]. In support of this concept, subclinical gut inflammation has been identified in all endotypes of SpA and was strongly correlated with the degree of spinal inflammation [159, 160]. The subclinical gut inflammation is characterized by mononuclear cell infiltration, lymphoid follicle formation, IL-23 production, and local expansion of innate immune cells, such as innate lymphoid cells of group 3 (ILC3), γδT-cells, and mucosal-associated invariant T (MAIT) cells [159, 161–163]. Notably, expanded ILC3s expressing the homing integrin α4β7 were identified not only in the gut but also in peripheral blood, synovial fluid and bone marrow of AS patients [164–166]. The expansion and differentiation of these intestinal subpopulations of innate cells depends on local production of IL-7 and IL-23 [167, 168]. Activated paneth cells and intestinal myeloid cells have been identified as the cellular sources of intestinal production of IL-23 in AS patients [163]. Intestinal activation of autophagy, potentially triggered by intracellular accumulation of misfolded HLA-B27, has been suggested as a molecular pathway inducing the intestinal production of IL-23 in SpA [144].

The above observations suggest that subclinical intestinal inflammation is not an epiphenomenon, but rather an etiologic event that contributes to SpA pathogenesis and indicate the following pathogenetic model: interaction of environmental factors (e.g., microbiome), with genetic factors (e.g., HLA-B27) result in intestinal dysbiosis, disruption of epithelial barrier, activation of autophagy pathway and intestinal production of IL-7 and IL-23 [159]. In genetically predisposed individuals (e.g., carriers of IL7R polymorphisms or of variants that increase IL-23R signaling), the cytokines IL-7 and IL-23 trigger intestinal expansion and activation of innate (e.g., ILC3, γδT-cells, and MAIT cells) and adaptive (e.g., Th17) immune cells and induce the expression of homing integrins (e.g, α4β7) and effector cytokines (e.g., IL-17, IL-22, and TNF) [19]. Then, these cells migrate to the axial and peripheral musculoskeletal structures and drive enthesial, synovial, and spinal inflammation [159]. According to this model, the gut (especially the gut-associated lymphoid tissue; GALT) operates as the anatomic location that primes immune cells for a subsequent arthritogenic response [167]. If this pathogenetic scenario is true, inhibition of IL-7 and targeting the effector cells producing IL-17, IL-22, and TNF emerge as promising future therapeutic strategies for SpA.

Type 3 Inflammation: Organ-Specific Role for IL-23 and IL-17

The signature cytokine of type 3 inflammation is IL-17A [166]. The cellular sources of IL-17A are innate (e.g., ILCs, γδT-cells, MAIT cells, and neutrophils) and adaptive (e.g., CD4 + Th17 and CD8 + IL-17+ cells) immune cells, which express the transcription factor RORγt [166, 169]. The IL-23/IL-23R/Jak2-Tyk2/STAT3-axis plays a critical role in the establishment of type 3 inflammation and IL-17 production [166]. In addition, IL-23-independet pathways have been identified to trigger and maintain IL-17 production, especially in ILCs, γδT-cells, and neutrophils [170–173]. Prostaglandin E2 (PGE2) and its receptor EP4 are among the non-IL-23 regulators of IL-17 production [174–176], but whether PGE2/EP4-axis operates independently of IL-23 is not clear.

Human genetics provide the strongest support for the role of type 3 inflammation in SpA pathogenesis [118]. Among the SpA-associated variants, there are polymorphisms in at least 10 genes, which are involved in the regulation of IL-17 production (Table 1.3). Several studies in animal models, reveal the implication of IL-23 and IL-17 in SpA pathogenesis. For example, systemic over-expression of IL-23 in mice was sufficient to induce IL-17-dependent enthesial inflammation and IL-22-dependent enthesial ossification [177]. Cellular and molecular profiling in SpA patients revealed various types of IL-17-producing cells and high levels of IL-23 and/or IL-17A in the gut, PBMC, serum, synovial tissue, synovial fluid, entheses, and facet joints [163, 164, 168, 178–181].

Recent evidence from clinical trials has challenged our initial linear perceptions about the IL-23/IL-17 pathway and its role in the pathogenesis of SpA [182], especially in driving axial skeleton inflammation and IBD [183–186]. Direct targeting of IL-17 pathway (anti-IL17A or anti-IL-17RA) was proven impressively effective in psoriasis and gave favorable responses in peripheral and axial SpA [187–192]. In contrast, trials in IBD either failed or were discontinued due to disease worsening [185, 186]. On the other hand, targeting IL-23 pathway (anti-p19 or anti-p40) was proven very effective in psoriasis and effective in IBD as well as peripheral SpA [193–196], but failed in axial SpA [183, 184]. These observations indicate that skin and peripheral-musculoskeletal manifestations of SpA are IL-17- and IL-23 dependent, axial inflammation is IL-17-dependent and IL-23-indpendent, whereas clinically overt intestinal inflammation is IL-23-dependent and IL-17-independent.

The differential effects of anti-IL-23 and anti-IL-17 in axial skeleton and intestinal inflammation reveal a tissue/organ-specific role for IL-23 and IL-17 [166, 182]. In this context, treatment decisions (targeting IL-17 versus IL-23 pathways) in the spectrum of SpA endotypes should be based on the anatomic domains that are affected in each individual patient [182]. In axial SpA, anti-IL-17 should be considered for treatment, but not anti-IL-23. In peripheral SpA (synovitis, enthesitis, and dactylitis) both anti-IL-17 or anti-IL-23 (anti-p19 and anti-p40) are acceptable therapeutic options, and due to their impact in the skin could be preferable in PsA with extensive skin involvement. Along the same lines, the status of intestinal involvement should be evaluated in SpA patients before considering inhibitors of IL-17. Since there is evidence for homeostatic functions of IL-17 in the gut (e.g., by maintaining the epithelial barrier integrity) [172], blocking IL-17 pathway is not recommended in SpA patients with IBD.

The Pathogenetic Role of TNF

Since anti-TNF therapies have become a standard of care for all endotypes of SpA [197, 198], the essential role of TNF in driving inflammation in the axial and peripheral skeleton, skin, intestine, and eye is well established. The pathogenetic pathways triggered by TNF have been reviewed recently [43]. Notably, the two modalities of blocking TNF (monoclonal antibodies and TNF receptor) display differences in effectiveness, depending on the anatomic location of inflammation [199]. Monoclonal antibodies are more effective from etanercept in intestinal inflammation and uveitis [200, 201], whereas there is equivalent effectiveness in psoriasis and inflammation of axial/peripheral skeleton.

The Enthesial Link Between Mechanical Stress, Inflammation, and Abnormal Bone Formation

One of the clinicopathological hallmarks of all SpA endotypes is the tropism for the axial and peripheral entheses, manifested as inflammation (enthesitis) and ossification (enthesophytes) [107]. One potential explanation for this tropism is tissue-specific autoimmunity, but so far autoreactive T-cells with specificity against enthesial antigens have not been identified. An alternative pathogenetic scenario has emerged recently with an emphasis on the unique microanatomy of enthesis. According to this hypothesis, mechanical stress, in genetically predisposed individuals, triggers enthesial inflammation and mesenchymal tissue reaction resulting in abnormal bone formation (Fig. 1.2) [107, 202].

Entheses are essential structures, which enable stable anchoring of tendons and ligaments into bones and provide smooth transduction of mechanical forces from muscle to bone (in the case of tendons) and stability (in the case of ligaments) [107]. To fulfill these high mechanical demands, entheses have a unique microanatomy and structure. The bone at enthesial sites is thin and porous, with blood vessels allowing the communication of the entering tendon or ligament with the neighboring bone marrow [107]. In addition, the transition zone (terminal part of tendons and ligaments before entering the bone) of enthesis is comprised of fibrocartilage and resident mesenchymal stem cells (MSC). Repeated mechanical overload may result in enthesitis even in otherwise healthy individuals, but the unique features of SpA-associated enthesitis are remarkable degree of chronicity and ossification [107, 203].

According to the mechanical stress hypothesis (also known as the mechano-inflammation model; Fig. 1.2), repeated microtrauma and additional triggers (such as “leaky” epithelial barriers and immune cells primed in the gut) induce local inflammation in the perienthesial bone marrow (osteitis), which spreads in the neighboring enthesis (enthesitis) [107]. The key molecular mediators for these early local events are PGE2 and cytokines (e.g., IL-23, IL-17, IL-22, and TNF). It is speculated that PGE2 is locally produced by resident mesenchymal cells, which express cyclooxygenase 2 (COX2). PGE2 facilitates the development of osteitis and enthesitis by triggering local vasodilation. The anatomic and cellular sources of IL-23 are not well defined, but distant production in the gut (as described above) or local production by myeloid cells have been suggested. Resident enthesial IL-23R+ cell have been identified in mice and in humans [177, 204], but confirmatory studies are required to validate these observations. The combination of PGE2 and IL-23 fosters the local production of IL-17, IL-22 and TNF by resident and recruited innate (ILC3 and γδT cells) and adaptive immune cells (Th17 cells). IL-17 and TNF act as amplifiers of enthesitis and osteitis by inducing recruitment and activation of neutrophils and macrophages [107].

In SpA patients, the inflammatory phase (osteitis and enthesitis) is followed by a remarkable two-step mesenchymal tissue response. At first, there is a chondroblast-mediated formation of a cartilaginous scaffold, and then osteoblast-mediated ossification occurs [107]. The key effector cells during this tissue response are resident mesenchymal cells, which are differentiated into chondroblasts and osteoblasts [203]. It appears that the initial inflammatory phase fosters the tissue response, since IL-17 and IL-22 activate mesenchymal cells [177, 205–208], and PGE2 is a robust activator of osteoblast differentiation [209–211]. Whether effective targeting of inflammation halts or slows new bone formation is still controversial [212–214]. Studies with non-steroidal anti-inflammatory drugs (NSAIDs), which block PGE2 production, have shown effective control on pain and signs of inflammation, but conflicting results on new bone formation [215–218]. A series of studies with TNF inhibitors have shown that, despite the effective control in symptoms, signs and biomarkers of inflammation, there is no inhibition on new bone formation over 2 years [219–221]. Since new bone formation is a slow process, more recent studies revealed that an inhibitory impact of anti-TNF becomes apparent radiographically after longer-term follow-up [222–225]. Clinical trials with biologics blocking IL-17 pathway show promising results in arresting new bone formation [226, 227], but additional long-term confirmatory studies are required. The conclusion of all these studies is that early onset and long-term effective inhibition of inflammation has the potential to slow new bone formation.

The key anabolic mediators promoting chondroblast/osteoblast activation in SpA are the Hedgehog proteins, Wnt agonists, bone morphogenetic proteins (BMPs), parathyroid hormone-related peptide (PTHrP), and IL-22 (Fig. 1.2) [107]. In animal models, pharmacologic inhibition of Hedgehog, Wnt, BMP, or IL-22 pathways prevented new bone formation [177, 228–230]. In light of these observations, a combination of anti-inflammatory drugs with inhibitors of the bone anabolic pathways appears a promising strategy for efficient disease modification in SpA with more robust prevention of new bone formation.

Section 1

Section 2