HLA-B Alleles

The association of HLA-B27 with AS was described in 1973, and remains one of the strongest genetic associations with any common human disease. Nonetheless, only a minority (likely <5%) of B27-positive individuals develop AS. The discovery that allelic variation of HLA-DRB1*01 and *04 influenced the risk of rheumatoid arthritis stimulated research into variation in HLA-B27 itself. There are now known to be more than 90 subtypes of HLA-B*27 , which have arisen from the common ancestral subtype, HLA-B*2705. Unlike the situation in rheumatoid arthritis, for the most part in AS B*27 subtype variation plays little role in influencing disease risk. There is strong evidence to suggest that HLA-B*2706 (found in east Asian populations) and B*2709 (found in Sardinia) have reduced strength of association with AS. The common white European subtypes, B*2705 and B*2702, are equally strongly associated with AS. The primarily Asian subtype B*2704 is at least as strongly associated with AS as B*2705 in the same populations, with some studies suggesting that it may be more strongly associated. B*2707, also mainly found in Asians, seems equally strongly associated with AS as B*2705. Although AS cases have been reported carrying many other B27 subtypes, for most alleles the number of cases reported is too few to definitely comment on their relative strength of association with the disease.

There are currently four main theories as to how HLA-B27 is involved in AS etiopathogenesis. The arthritogenic peptide hypothesis proposes that HLA-B27 presents a pathogenic peptide that initiates disease. This hypothesis is consistent with the antigen presentation function of HLA-B27, and also is consistent with the gene-gene interaction (epistasis) seen with ERAP1 (discussed later). Despite extensive efforts, no definitive “arthritogenic peptide” has been identified. There are many potential explanations for this including that the peptide may only be present at particular phases in the disease pathogenesis or at particular sites, that it may only represent a small fraction of HLA class I presented peptides, or that more than one peptide may be involved. Because the arthritogenic peptide is proposed to be presented to CD8 T-lymphocytes, the finding that in the HLA-B27 transgenic rat model of SpA disease is independent of CD8 cells is inconsistent with this hypothesis. However, no animal model perfectly captures human AS, and the relevance of this finding to human AS is not entirely certain.

It has been shown that HLA-B27 heavy chains, either alone or as heavy chain homodimers, can form on the cell surface and then interact with antigen-presenting cells carrying receptors, such as killer-cell immunoglobulin-like receptors. These antigen-presenting cells can then initiate a pathogenic T-helper 17 (Th17) response. These homodimers are thought to occur when unstable HLA-B27:peptide complexes dissociate on the cell surface. ERAP1-deficient cells have more unstable HLA-peptide complexes on the cell surface, which one would expect to promote cell surface homodimer formation, but the AS-protective alleles of the ERAP1 variants are associated with decreased function, which is inconsistent with this hypothesis.

Endoplasmic reticulum (ER) stress, which occurs when misfolding leads to accumulation of HLA-B27 heavy chains in the ER, precipitates a stress response called the unfolded protein response. The unfolded protein response is a homeostatic mechanism that the cell initiates to clear the misfolded proteins and return the ER environment to normal. ER stress has been shown to be present in the HLA-B27 transgenic rat model of SpA, and has been shown to induce interleukin (IL)-23 production, providing a potent link between HLA-B27 and AS.

Finally, HLA-B27 may tag a nearby disease-causative gene, the association of HLA-B27 with AS being caused by linkage disequilibrium with this nearby “linked gene.” This theory was made much less likely by the findings of the Australian-Anglo-American (TASC) genomewide association study (GWAS) that confirmed the highest association with AS was with HLA-B27 and not a linked gene.

There are mixed reports on whether homozygosity for HLA-B27 influences clinical manifestations, and although some reports have suggested an increased risk of AS among HLA-B27 homozygotes, the sample sizes involved in these studies were not sufficient to produce definitive results either way. An association with HLA-B60 has also been described in HLA-B27–positive and HLA-B27–negative individuals, although the strength of association reported was not definitive.

The huge volume of genetic information produced by GWASs has allowed researchers to examine further questions of interest relating to heritability of disease and disease-genotype correlations. If one considers the known AS associations, there is not a higher burden of genetic associations in familial AS than sporadic AS, except for HLA-B27.

HLA-B27 –negative AS makes up only about 10% of AS cohorts, but does demonstrate that an essentially identical disease can be evident without the major genetic risk factor being present. HLA-B27 –negative AS is less likely to be familial, has a later disease onset, and is less likely to respond to anti–tumor necrosis factor (TNF) treatment, but controlling for disease duration has similar disease severity (measured by the Bath Ankylosing Spondylitis Functional Index), activity (measured by the Bath Ankylosing Spondylitis Disease Activity Index), and radiographic severity (measured by the modified Stoke Ankylosing Spondylitis Severity Score). HLA-B27 –negative AS has been shown to have similar, although not identical, genetic associations with HLA-B27 –positive AS, the main exception being the association with ERAP1 , which is restricted to HLA-B27 –positive AS.

HLA-B27 Typing for Clinical Practice

Accurate HLA-typing is technically challenging and difficult to establish as a high throughput method. This has reduced enthusiasm for the use of HLA-B27 in population screening for risk of AS. Recently, an major histocompatibility complex (MHC) tag single nucleotide polymorphism (SNP) rs4349859 was shown to be able to identify HLA-B27 in those of European descent with a sensitivity of 98% and a specificity of 99%, within the likely boundaries of accuracy of direct HLA-B27 genotyping itself. Another SNP, rs13202464, was then reported that showed high sensitivity and specificity in east Asian populations. The use of these SNPs has significant advantages over the current methods for HLA-B27 typing in cost and complexity. Further research incorporating other ethnic groups may lead to additional ethnicity-neutral HLA-B27 tag SNPs. This discovery has implications for potential screening of high-risk cohorts either in the primary care or population-based settings. It may be able to be integrated into referral strategies, by taking advantage of point-of-care testing, which is currently being developed.

Antigen-Presentation Genes

ERAP1 is a member of the MHC class I presentation pathway and trims peptides before presentation on MHC class I molecules, such as HLA-B27 ( Fig. 1 ). ERAP1 has been robustly associated with AS in multiple studies and populations including Europeans, Hungarians, Portuguese, Taiwanese, Han Chinese, and Koreans. Recently, it has been demonstrated that the association of ERAP1 with AS is restricted to HLA-B27 –positive disease. ERAP1 is also associated with psoriasis, and the association in psoriasis is restricted to HLA-Cw6 carriers. The AS-protective SNPs in ERAP1 result in reduced peptide trimming function of the ERAP1 enzyme. It is not yet clear whether the protective variants of ERAP1 lead to just quantitative reductions in trimmed peptide availability, or if they also lead to qualitative changes in the peptides.

The antigen presentation pathways with ERAP1 (and ERAP2) trimming the peptide before loading onto the MHC class I molecule.

The ERAP1 enzyme has also been described to have two other functions. First, it has been described to act as a sheddase to cleave cytokine receptors, such as IL-6, TNF, and IL-1β, from the cell surface. Studies of ERAP1-deficient mice have demonstrated that the levels of TNF receptor and IL-6 receptor are no different to control animals and in patients with AS there is no difference in serum cytokines based on ERAP1 genotypes. Second, ERAP1 has also been described to be secreted from macrophages in response to interferon-γ and lipopolysaccharide and assists in phagocytosis. Deficiency in phagocytosis could impair responses to commensal or invasive microbes and HLA-B27’s restricted repertoire may interact to exacerbate this, or push it over a disease-causing threshold.

ERAP2 encodes an aminopeptidase, which is encoded at chromosome 5p15 immediately adjacent to ERAP1 , and has also been shown to be associated with AS, although whether this is independent of the ERAP1 association is not clear. An ERAP2 association has been described with Crohn disease. ERAP2 is an aminopeptidase similar to ERAP1, which also trims peptides in the ER before their MHC class I presentation on the cell surface. It has a different peptide preference from ERAP1 and has been shown to form heterodimers with ERAP1.

T-Helper 17 Pathway Genes

The association of multiple genes in the pathogenic T-helper 17 (Th17) cell pathway, including IL23R , STAT3 , and IL12B , suggests this is an important pathway in AS. The preliminary report of the effective use of anti–IL-17 therapy is also a pragmatic demonstration that clinically this is a pathway that deserves further attention.

IL-23 is made up of two subunits, IL-23p19 and IL-12p40. IL-12p40 is encoded by IL12B . IL-23 signals through its receptor IL-23R, present on a wide range of cells, but importantly on gamma-delta T cells and Th17 cells. This receptor, once activated, signals through STAT3 by promoting its phosphorylation. This STAT3 phosphorylation then promotes IL-17 production by Th17 cells.

In cells of the innate immune system and γσ T cells, pattern recognition receptors, such as dectin-1, signal through CARD9 after β-glucan stimulation ( Fig. 2 ). The SKG mouse model develops an SpA phenotype when stimulated with β-glucan, characterized by axial and peripheral spondyloarthritis, IBD, and unilateral iritis. After this pathway is activated this promotes the production of prostaglandin E ₂ . Prostaglandin E ₂ is a proinflammatory mediator that can signal through the arachadonic acid pathway and promote inflammation. Prostaglandin E ₂ can also upregulate IL-23 and IL-17 by signally through prostaglandin E receptor 4, subtype EP4 (PTGER4). This receptor has been associated with AS at genomewide levels of significance. Nonsteroidal anti-inflammatory drugs inhibit cyclooxygenase enzymes and consequently the production of prostaglandins. The finding that nonsteroidal anti-inflammatory drugs retard radiographic progression in AS and that nonsteroidal anti-inflammatory drugs are used to reduce heterotopic ossification is a clinical demonstration of the importance of this pathway to bone formation and homeostasis.

The interaction of microbial β-glucan and dectin-1, which signals through CARD9 to promote pathogenic proinflammatory cytokines.

Potential Skeletal Structure and Mineralization Genes

ANTXR2 , which encodes protein capillary morphogenesis protein 2, is an AS-associated gene with potential impacts on bone and the skeleton. Defects in this gene cause the human syndromes infantile systemic hyalinosis and juvenile hyaline fibromatosis. How variants of this gene are involved in AS is unclear.

A recent GWAS in east Asians found GWAS-significant associations in two loci harboring bone and cartilage related genes. The first associated locus harbors the genes HAPLN1 and EDIL3 ( P = 9 × 10 ⁻¹⁰ ; odds ratio [OR] = 1.2). HAPLN1 encodes hyaluronan and proteoglycan link protein 1, potentially relevant to AS etiology through bone effects. EDIL3 encodes EGF-like repeats and discoidin I-like domains 3, which promotes endothelial cell adhesion. The second association lies in an intron of ANO6 ( P = 2 × 10 ⁻⁸ ; OR = 1.3), which encodes a transmembrane protein involved in phosphatidylserine regulation on the cell surface. Phosphatidylserine exposure is involved in macrophage phagocytosis of apoptotic cells, potentially mediating immune responses. In addition, phosphatidylserine is involved in osteoclastogenesis. It will be valuable to see if these loci replicate in other east Asian or European cohorts.

TNF-Associated Genes

Multiple genes in the TNF pathway have been associated with AS, consistent with this pathway playing a major role in AS etiopathogenesis. Association has been described at the 17q21 locus, which was attributed to TBKBP1 , a member of the TNF signaling pathway. There are, however, two other plausible candidate genes at this locus include NPEPPS , an aminopeptidase similar to ERAP1 and ERAP2 , and TBX21 , a Th1 transcription factor. Further follow-up studies are required to clarify the association at this locus.

Association has also been reported at chromosome 12p13 at a locus containing two TNF-receptors, TNFRSF1A and LTBR . LTBR encodes the lymphotoxin beta receptor; lymphotoxin is a member of the TNF family. Further TNF genes that have been associated with AS include TBKBP1 , which is a component of the TNF signaling pathway, and TRADD , another TNF receptor protein has also been associated. An animal model of extreme supraphysiologic TNF overexpression causes an SpA phenotype; how this correlates to human SpA is not yet clear. Several factors implicate this biologic pathway in AS including raised TNF-α in patients with AS and the effectiveness of therapies that block TNF, such as anti-TNF biologics and thalidomide.

Other Genetic Associations

Two intergenic regions at 2p15 and 21q22 have now been robustly associated with AS at genomewide levels of significance. Proteasome assembly chaperone 1 ( PSGM1 ) gene is found near the 21q22 locus; the proteasome is part of the MHC class I presentation pathway. It is therefore plausible that the association operates through this gene. Against this, the association is not in close proximity to the gene. At chromosome 2p15 there is no nearby candidate gene. RNA-sequencing studies identified long noncoding RNA transcripts at both loci, and it may that the associations operate through cis – or trans -gene regulation, potentially through noncoding RNA (ncRNA).

RUNX3 encodes Runt-related transcription factor 3, which has been shown to be expressed in thymocytes on signaling by IL-7. These IL-7–stimulated CD4 and CD8 double-positive cells then differentiate into CD8 positive lymphocytes. Further evidence to support this finding is the moderate level of association found in the IL-7 receptor ( P = 8 × 10 ⁻⁵ ) in the TASC AS GWAS. Consistent with this it will be informative to examine other components of this pathway, such as the cytokine IL-7 itself, for association in future experiments.

IL1R2 encodes the IL-1 receptor 2, the biologic action of which is to inhibit IL-1 by acting as a decoy receptor. This protein exists in two forms, a long membrane-bound form and a shorter soluble form, which is produced by alternate splicing. The longer membrane-bound form is the functionally active inhibitory molecule. IL-1β is stimulated by conserved microbial sequences, such as pathogen-associated molecular patterns or damage-associated molecular patterns. Inhibition of an appropriate response to microbial colonization or infection may be the mechanism by which this association acts.

KIF21B has been associated with AS and with other autoimmune disorders including multiple sclerosis, Crohn disease, and ulcerative colitis. It is expressed in a variety of tissues, but best characterized in dendrites. It is involved in trafficking of components within the cell. It is also expressed in B cells, T cells, and natural killer cells. Preliminary evidence suggests KIF21B and a nearby open reading frame C1orf106 at the 1q32 locus are involved in ER stress and the NF-κβ pathway but further functional work is required.