The association analysis between HLA-A*26 and Behçet’s disease

The strongest genetic risk factor of Behçet’s disease (BD) is HLA-B*51. Our group previously reported that HLA-A*26 is independently associated with the risk of the onset of BD apart from HLA-B*51. Here, we re-evaluated the association between HLA-A*26 and BD in the Japanese population. We also performed a comprehensive literature search and meta-analyzed the extracted published data concerning the relationship between HLA-A*26 and BD to estimate the odds ratio (OR) of HLA-A*26 to BD. In this study, we genotyped 611 Japanese BD patients and 2,955 unrelated ethnically matched healthy controls. Genotyping results showed that the phenotype frequency of HLA-A*26 was higher in BD patients than in controls (OR = 2.12, 95% CI: 1.75–2.56). Furthermore, within the HLA-B*51-negative populations, the phenotype frequency of HLA-A*26 was significantly higher in BD patients than in controls (OR = 3.10, 95% CI: 2.43–3.95). Results obtained from meta-analysis combined with our data showed that the modified OR of HLA-A*26 became 1.80 (95% CI:1.58–2.06), whereas within the HLA-B*51-negative population, the modified OR became 4.02 (95% CI: 2.29–7.05). A subgroup analysis arranged by the geographical regions showed HLA-A*26 is in fact associated with the onset of BD in Northeast Asia (OR = 2.11, 95% CI: 1.75–2.56), but not in the Middle East or in Europe.


HLA-A*26 genotyping.
Allele and phenotype frequencies of HLA-A*26 are shown in Table 1. Both allele and phenotype frequencies of HLA-A*26 were significantly higher in the patient group as compared to the healthy controls. Allele frequency: 18.43% in BD vs. 10.93% in controls (OR = 1.84, 95% CI: 1.56-2.18). Phenotype frequency: 35.55% in BD vs. 20.68% in controls (OR = 2.12, 95% CI: 1.75-2.56). In addition to HLA-A*26 allele, -A*11, -A*31, -A*33 were also statistically significant, and the frequencies in the patient group and controls indicated that the genetic association of -A*31 as risk and -A*11, -A*33 as protective type.

HLA-A*26 frequency within the HLA-B*51-negative population.
In the current genotyping study, 314 BD patients (51%) and 2,433 controls (82%) did not carry the HLA-B*51 antigen. The allele and phenotype frequencies of HLA-A*26 within the HLA-B*51-negative populations are shown in Table 2. Both allele and phenotype frequencies of HLA-A*26 were significantly higher in the BD group as compared to the controls. Allele frequency: 23.57% in BD vs. 11.16% in controls (OR = 2.48, 95% CI: 2.02-3.04). Phenotype frequency: 45.22% in BD vs. 21.05% in controls (OR = 3.10, 95% CI: 2.43-3.95). Besides HLA-A*26, -A*33 were statistically significant, and the frequency in the patient group and controls indicated HLA-A*33 as protective type.
Retrieved HLA-A*26 and -*B locus haplotype analysis in the Japanese population. 2 Fig. S1). Assessment of risk of bias for included studies was done through Newcastle-Ottawa Scale, and all included studies were of high quality with scores ranging from 7 to 9 (Table 4). Publication bias was assessed using a funnel plot ( Supplementary Fig. S2). A synthesized analysis of data from a number of publications showed that the OR of HLA-A*26 among BD patients was 1.62 (95% CI: 1.09-2.39). Combined with our genotyping data, the modified OR became 1.80 (95% CI: 1.58-2.06). (Fig. 1). Additionally, within the HLA-B*51-negative populations, the entire OR combined with our genotyping data became 4.02 (95% CI: 2.29-7.05) (Fig. 2).

Discussion
Despite its worldwide presence, BD has a much higher prevalence in countries along the ancient Silk Route that extends from the Mediterranean basin to the far Eastern Asia. Turkey has the highest prevalence of BD in the world, with 20-420 cases per 100,000 reported [14][15][16][17] . Compared to that, the prevalence rates range from 7.3-30.5 cases per 100,000 in Korea, China, Iran, Saudi Arabia, and Japan 18 . Strong genetic association between HLA-B*51 and BD has been identified in numerous ethnicities and counties 3,18 . We have previously reported that HLA-A*26 was significantly and independently associated with the risk of BD, apart from HLA-B*51 in the Japanese population 12   www.nature.com/scientificreports www.nature.com/scientificreports/ and negative associations between HLA-A*26 and BD [4][5][6][7][8]11,12,[19][20][21][22][23][24][25][26][27] . In our current study, we aimed to summarize how HLA-A*26 is involved in the risk of BD within various geographical regions and ethnicities. In our current genotyping study, within a total of 611 BD patients, almost half of the patients did not carry HLA-B*51 alleles, and within these HLA-B*51-negative populations, almost half of the patients carried HLA-A*26. This indicated that almost a quarter of the total BD patients were independently influenced by HLA-A*26 in the Japanese population.

Author
Year Country    www.nature.com/scientificreports www.nature.com/scientificreports/ Moreover, the OR of HLA-A*26 to BD was 2.12, and was as high as 3.10 within the HLA-B*51 negative subsets. As shown in Table 3 Of note, none of these -*B alleles showed significant independent risk association with BD in our current study. This may suggest the existence of other latent genetic risk alleles between HLA-A*26 and these -*B loci, for instance, possible involvement of retrieving HLA-*C, and -*E alleles 28 . Further studies will be necessary to clarify the possible involvement of these alleles to the susceptibility of BD.
As reported by Hughes, et al., no association was found between HLA-A*26 and BD in their larger number case-control study within the Turkish population. They concluded the lack of association owes to the low frequency of HLA-A*26 in the Turkish population 29 30 . HLA-A*26:01 has Arg at position 97 and that corresponds to the present risk residue mentioned above. HLA-B*51 has a Bw4 epitope in the α1-binding pocket which interact with the killer immunoglobulin-like receptors (KIR) 3DL1 and 3DS1 to regulate the activities of natural killer (NK) cells and a subset of cytotoxic T lymphocytes (CTLs) 31 . Their group also suggested a potential role of activating KIR3DS1 alleles in BD patients with ocular manifestations independent of HLA-B*51 32 . Though HLA-A*26 has a strong linkage with ocular manifestations in the Northeast Asian population, as far as KIR interaction is concerned, HLA-A*26 is not a direct ligand of KIR molecules. In another words, HLA-A*26 does not have an epitope which could be directly recognized by KIR to regulate the activation of NK cells or CTLs. That may suggest a different pathophysiology from HLA-B*51 related KIR interaction underlies the development of BD in the case of HLA-A*26 triggered KIR interaction pathways. Further studies are required to ascertain the hypothesis suggested above.
The worldwide distribution of HLA-A*26 is unique, as it is especially frequent in Northeast Asia, Oman, Georgia, and in the Israeli Jewish population (Fig. 4) 13 . In Northeast Asia, HLA-A*26 is more commonly found in the Western Pacific Rim, i.e. Taiwan, Ryukyu (Okinawa islands), and the Japan Islands. In our subgroup analysis arranged by geographical areas, positive association between HLA-A*26 and BD was found in Northeast Asia, but not in the Middle East or in Europe. This may owe to the higher distribution of this allele in the Northeast Asian region, especially in the Western Pacific Rim, we might be able to find the association between HLA-A*26 and BD in these areas more apparently.
In spite of the relatively high prevalence of HLA-A*26 in the Jewish population in Israel, no positive association between HLA-A*26 and BD was reported 25 . We believe this is due to the high heterogeneity of the Israeli Jewish population: A large proportion of them is of Ashkenazi Jewish origin, of which 21.7% were HLA-A*26 positive 33 . However, as reported in the previous papers, most of the Jewish BD patients were of non-Ashkenazi origin 25,34 . The lack of BD/HLA-A*26 association in this particular region might owe to the low frequency of HLA-A*26 in the non-Ashkenazi Jewish population. In addition, results of studies concerning the HLA genotyping of BD patients in Georgia and Oman were not found during our research.
In their GWAS results, Abi-Rached, et al. reported that between all three Neanderthals found in the Vindija Cave, northern Croatia, had the HLA-A*02, C*07:02, and C*16 alleles. Moreover, the pooling of these three Neanderthals sequence infers their possession of HLA-B*07, -B*51, and either HLA-A*26 or its close relative A*66 35 . It was suggested that the presence of HLA-B*51 in Eurasians, together with B*07, C*07:02, C*16:02, might be the result of admixture with the Neanderthals, which occurred after out-of-Africa migration until 40-30,000 years ago 35,36 . It is www.nature.com/scientificreports www.nature.com/scientificreports/ believed that the adaptive introgression of the Neandertal alleles has significantly involved in the construction of the modern humans' immune systems and contributed to mediate the host defense immune mechanism against lethal infectious agents which the ancestors of modern humans newly encountered in the frontiers of the Eurasian continent during the period of the great human expansion. It is of interest that both HLA-B*51 37 and HLA-A*26 alleles, which were identified as risk alleles of BD, might have been introgressed from the archaic human species, Neanderthals. Further studies will be needed for the better understanding of these mysterious relationships between HLA-B*51, HLA-A*26, Neanderthal alleles and BD. It is also worth noting that modern human's ancestors lived in an environment where infectious diseases were mostly endemic, and under the influence of endemic environmental agents, infective microbial organisms led to genetic selection in order to produce more effective pro-inflammatory response to encourage the resistance to specific infections. However, these effective and high-potency immune systems could lead to immune-mediated inflammatory disease as an undesirable adverse effect 38 . Yersinia Pestis, the cause of Plague, is reported to have evolved near China, 20-15,000 years ago 39 . Yersinia Pestis has been a lethal infectious agent to human beings, and several studies have suggested HLA-B*51:01 had a protective role in the host response against the Yersinia Pestis infection 40 . It is assumed that the bottleneck effect following the high mortality rate of plague epidemics might have led the expansion of HLA-B*51:01 associated increased pro-inflammatory phenotypes and reservation of this complex genetically determined trait 40 . In other words, HLA-B*51 associated BD is suspected to be a secondary and undesirable side effect of the immunological advantages rendered by HLA-B*51 in activating NK cells and CTLs in response to these lethal infections 37 . Currently, we do not have enough knowledge on how HLA-A*26 contributed to protect modern human ancestors from life-threatening infections in the human immune history, we believe comprehensive investigations and better understanding of HLA-A*26 will lead us to a better understanding of BD pathogenesis.
In conclusion, we have performed the genotyping of Japanese BD patients and confirmed that HLA-A*26 was the susceptibility allele for BD in the Japanese population. Especially in the HLA-B*51-negative BD populations, HLA-A*26 was significantly associated with the onset of BD. A combination of our genotyping data with other data extracted from publications showed the association of BD and HLA-A*26 was geographically significant in Northeast Asia, but not in the Middle East or in Europe.

Methods
BD patients and controls. 611 Japanese BD patients and 2,955 unrelated ethnically matched healthy controls were enrolled in this study. The diagnosis of BD was established according to standard criteria 41 proposed by the Japan Behçet's disease Research Committee. All procedures, data collection, and handling were performed according to the principles of the Good Clinical Practice and Declaration of Helsinki. This study was approved by the Research Ethics Committee of the Medical Faculty, Yokohama City University. The study details were explained to all participants before obtaining the informed consent for genetic screening. Blood samples were collected after study participants agreed and signed informed written consent. Banked and de-identified samples were used for this study. www.nature.com/scientificreports www.nature.com/scientificreports/ HLA genotyping. We genotyped HLA-A and HLA-B alleles for 611 cases and 737 controls with Luminex reverse sequence-specific oligonucleotides and bead kits (One Lambda). For the remaining 2,218 controls, we performed an imputation analysis of HLA-A and HLA-B with our GWAS data using SNP2HLA 42 and a reference panel of 530 pan-Asian samples 43 . The χ 2 test was used to analyze categorical variables.
Literature Search and Meta-analysis. Meta-analysis was performed through the method proposed by the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) 44 , using the Review Manager software (version 5.3) for statistical analysis. This protocol has been registered in the international prospective register of systematic reviews (PROSPERO) as number CRD42017073887. Relevant studies were identified using the PubMed/Medline, Embase, Web of Science, CENTRAL database and through manual literature search in December 2018; no language restriction was used for published studies. Studies fulfilling the following inclusion criteria were included in the meta-analysis: (1) case-control studies; (2) studies reporting an association between HLA and BD; (3) genetic association studies; and (4) independent studies without repeat reports on the same populations or subpopulations. In this study, we performed the quantitative synthesis of the extracted data which contain the results of phenotype frequency, hence extracted data which contain only allele frequency results without phenotype frequency results were included in qualitative synthesis, but not into the quantitative synthesis. Two of the authors (JN and GI) individually assessed the bias of included-studies using the Newcastle-Ottawa Scale 45 . Pooled ORs and the corresponding 95% CIs were synthesized with the random-effects model. Heterogeneity was assessed using the I 2 statistic.