HLA-B*51 the primary risk in Behçet disease

Behçet disease (BD) is a multisystem autoinflammatory condition characterized by mucosal ulceration and predominantly neutropilic inflammation of immune privileged sites including the eye, brain, and synovial joint (1). In PNAS, Ombrello et al. analyze a combination of directly ascertained and imputated single nucleotide polymorphisms (SNPs) in the HLA-B region in Turkish patients with BD (2). HLA-B*51 was the largest single risk factor for BD, with smaller, independent effects of HLA-B*15 and HLA-B*27. HLA-B*15 was also significantly associated in HLA-B*51^−ve patients. Analysis of 32,869 SNPs across the MHC region identified HLA and MHC class I polypeptide-related sequence A (MICA) as the strongest association. Conditioning on HLA-B*51 and rs79556279, the strongest associated SNPs, 4.5 kb upstream of HLA-B loci showed no other significant SNPs. Moreover HLA-B*51 and rs79556279 are in strong linkage disequilibrium. Two other regions, one tetrameric to HLA-C and the second in a region that includes HLA-A, were associated with BD, but the former was lost on conditioning for rs79556279. Associations with HLA-A, particularly HLA-A26, have been previously reported, particularly in Japanese patients with BD.

Finally, using logistical regression and conditioning for HLA-B*51, HLA-A*03 was shown to be significantly protective for BD, and HLA-A*49 was protective in HLA-B*51^+ve individuals. To further understand the difference to between susceptible and protective alleles, amino acid motifs were assessed. Amino acids at positions 67, 97, 116, and 152 significantly and independently influenced the risk of developing BD.

Genetic Basis of Behcet Disease

BD has clear geographical prevalence, with the incidence being between latitudes 30° and 45° North in Europe and Asia, and it is very rare in Africa, South America, and Australia. This geographical distribution led to BD being classified as the Silk Road disease by Ohno, as these ancient trade routes ran from the Mediterranean to Northern China. Studies by Ohno et al. in patient groups along the Silk Road identified HLA-B*5 on chromosome 6 as associated with BD, and further analysis showed HLA-B*51, a split allele of B*5, to be the linked gene and was confirmed in every other patient group tested. Distribution of BD is strongly related to expression levels of HLA-B*51, with prevalence being highest in Europe and the Middle and Far East, where HLA-B*51 is found in >15% of individuals, and the prevalence being lowest in Africa,

The work by Ombrello et al. greatly strengthens the case for a primary causative role for HLA-B*51 in BD.

Oceania, and South America, where expression is low or nil (3, 4). Early studies using microsatellite markers in Japanese, Greek, and Italian patients patients with BD showed that HLA-B was the strongest association in each group. Three markers, which lie between MICA and HLA-B genes, were also associated with BD in each group. However, when HLA-B was stratified for the confounding effects of these markers, a significant association with B*51 remained the primary association (5). Significantly, a meta-analysis of 78 independent studies confirm HLA-B*51 as the major risk factor in BD (6).

This was supported by candidate gene studies of genes in the MHC region on chromosome 6 that identified in particular the MICA*009 allele to be associated with BD. MICA is expressed on stressed cells such as tumor or virally infected cells and delivers an activating signal to cytotoxic cells, i.e., NK cells, CD8^+ve cells, and γδ T cells, via natural killer glycoprotein 2D (NKG2D), to induce killing. However, the suggested association was shown to be due to linkage disequilibrium with HLA-B*51 (7). This was supported by analysis of MICA*009 in Middle Eastern patients where the allele was in linkage with both HLA-B*51 and HLA-B*52, which only differ by two amino acids, yet only the B*51/MICA*009 combination was found in patients with BD. By comparison, B*52/MICA*009 was found in 25% of controls (8). Recently genome-wide association studies (GWASs) in different patient populations showed specific genes linked to BD in different ethnic groups; however, all the studies identified the HLA-B loci as the strongest association (9–11). However, in a recent study, deep sequencing of the HLA region identified an SNP, rs116799036, close to MICA that gave the strongest association and was independent of HLA-B*51 (12). By comparison with Ombrello et al., rs116799036 had a much weaker association with BD than HLA-B*51, which remained significant after the data were conditioned for this SNP. The difference between the Ombrello et al. and Hughes et al. papers may be due to the number of cases tested, the use of geographical controls, and the reference panel used. Taken together, the data strongly support a principal causative role for HLA-B*51 in BD.

Functional Role of HLA-B*51 in BD

The pathogenic role of HLA-B*51 in BD has yet to be determined. One possibility may be the presentation of antigen to CD8^+ve cells; however, this population of cells has not been implicated in BD. Moreover, no specific pathogen or autoantigen has been found for BD. A second possibility is that the inhibitory effect of HLA-B*51 binding to killer Ig-like receptor DL1 (KIR3DL1) on NK cells is important. This interaction is the normal default response for NK cells, where as long as cells express normal levels of HLA-B on their surface, they are protected from cytotoxicity. A reduction in HLA class I can be induced by tumor or viruses, leading to recognition and killing. Alternatively, increased MICA expression via an activating signal on binding to NKG2D can overcome the inhibitory effect of HLA class I. The combination of KIR and HLA alleles has been implicated in many diseases. In particular, expression of HLA-B*57 and KIR3DS1 has been linked to slow progression to AIDS in patients who are HIV positive. Similar combinations have been identified in response to hepatitis C infection and in some autoimmune diseases (13) Birdshot chorioretinopathy (BCR) is strongly associated with HLA-A*2902, which is 1 of 21 serologically defined variants of the HLA-A gene, with >95% of BCR patients expressing this allele. KIR-HLA pairs implicated for weak inhibition [i.e., KIR3DL1 + HLA-Bw4 (T80)] in combination with activating KIR (KIR2DS2, KIR2DS3, and KIR2DS4) were found to be associated with risk in BCR HLA A*29-positive patients. By comparison, association of strong inhibitory pairs, such as KIR3DL1+HLA-Bw4 (I80), in combination with weak activators, i.e., KIR3DS1+HLA-Bw4 (I80), was observed in HLA-A*29-negative controls. These results suggest that a profound effect of activating KIR2DS2/S3/S4 in the absence of strong inhibition may enhance the activation of natural killer cells contributing to pathogenesis of BCR (14). We have previously shown that NK cells from healthy control individuals kill MICA-transfected Chinese Hamster Ovary cells in different hierarchal patterns that are maintained over time. Such tuning suggests that individuals will only kill cells expressing MICA alleles at a certain concentration on the target cell (14). These data suggest several possibilities that should be addressed. First, the binding of KIR3DL1 to HLA-B*51 in BD patients may be stronger due to of KIR3DL1 subtypes with a higher affinity to HLA-B*51. Secondly, HLA-B*51 binds peptides with lower affinity than HLA-B*52 and therefore will present a more promiscuous pool of peptides, leading to increased surface expression of, and response to B*51. Third, most patients in GWASs have not been typed for the B*51 subtypes, and it is possible that combinations of B*51 and KIR3DL1 subtypes may influence killing responses in these assays. Finally, in patients with BD and healthy controls, NK cells would be subject to control by NKG2D which is also polymorphic (15).

The amino acid residues identified by Ombrello et al. are all located in the antigen binding site and therefore could be involved in either of these processes. Interestingly, amino acid 67 is one of the two residues that differ between HLA-B*51 and HLA-B*52, the alleles differently associated with BD. In support of an importance of HLA peptide binding, Kirino et al. recently reported that endoplasmic reticulum-associated animopeptides-1 (ERAP-1) was associated with BD. ERAP-1 is a ubiquitous metallopeptidase that removes NH₂-terminal residues from peptide precursors, optimizing their size to fit into class I HLA molecules associated with BD (16). ERAP-1-deficient mice had significantly less expression of HLA class I molecules on the cell surface and were less able to present self- and foreign antigens. Heterozygous ERAP-1^+/− mice showed an intermediate phenotype that suggest polymorphisms that affect ERAP-1 function could affect antigen presentation (17). ERAP-1 has recently been shown to be associated with anklyosing spondylitis, a condition strongly linked to another HLA class I molecule: HLA-B*27 (18). The relevance of these findings to BD includes the possibility that different peptides binding in the groove could either activate a different population of CD8^+ve cells or stabilize HLA molecules on the surface, potentially increasing the inhibitory signal to NK cells.

The work by Ombrello et al. greatly strengthens the case for a primary causative role for HLA-B*51 in BD. Although the mechanism is still not clear, several possibilities should be examined. Moreover, the novel findings of protective class I alleles and the involved amino acid residues provide a substantially more detailed understanding of the association and identifies new pathways for investigation to determine the functional role of these molecules.