Many studies on the potential significance of HLA for GD in Asians were performed. Since 1978, several of them have postulated the impact of HLA-B*46 in the development of GD (31–33). However, with application of serological methods, in some studies, HLA*Bw46 significance was demonstrated in males only, with lack of correlation in females (34–36). Similar gender-limited relationship was observed between GD and HLA-DR9 and DQB1*03:03 (36). A metanalysis of 14 studies performed by Li et al. in 2013 demonstrated association between HLA-B*46 and GD in Asians (31). Although, most of the authors of the included studies applied serological methods, more recent reports confirmed this correlation (37–40). The overview of the published results on HLA significance in GD are demonstrated in Table 1 . It is clearly visible that the obtained results are not consistent and potential high risk alleles or protective ones are different even in similar populations (i.e. in the Chinese, the Japanese, the Korean etc.). Several antigens other than Bw46 were postulated by authors who used serological methods, including, for example, HLA-DRw8, -DQw4, -B5, Dw12 and –A11 antigens (44–46). At the end of the 20th century, polymerase chain reaction (PCR) sequence specific oligonucleotide probe (SSOP) method was introduced and many authors applied it at least for MHC class II analysis. Similarly to the serology-based reports, the studies with combined serological and SSOP methods revealed inconsistent results. Chan et al. reported that the risk of GD was higher in patients with HLA-A*2, Cw1, DRB1*16:02, DRB1*03:01, DRB1*14:05, DRB5*02, DQB1*05:02, while the presence of HLA-DRB1*15:01 and -DQB1*03:01 played a protective role (35). On the other hand, Japanese authors presented that the most important factor was the presence of HLA-DPB1*05:01 and/or HLA-A*2, with the risk being the highest in carriers of both of them (46). In one Taiwanese study, HLA-A*02:07 was found a GD risk factor (37), while other Taiwanese authors showed correlation between GD and HLA-B*46:01, DPB1*05:01, DQB1*03:02, DRB1*15:01 and DRB1*16:02, with the strongest relationship existing with HLA-DPB1*05:01 (38).

There are discrepancies in the obtained results, not only between methods applied by the authors or between subpopulations based on patients’ nationality, but also between studies performed with the same method and among the same subpopulation (40–57). Additionally, many studies included MHC class II alleles only, or even one group mostly HLA-DRB1 (48–52). Therefore, direct comparison of the results, or any attempt to include the studies into metanalysis, would be subjected to high risk of error. The most reliable results obtained with NGS method are scarce. Katahira et al. did not compare their GD patients to a control group but only to patients with Hashimoto thyroiditis. Therefore, they are not suitable for comparison with others. Ueda et al. presented results as a list of alleles associated with high risk of GD and a list of 3 haplotypes which showed significant protective effects against the development of GD in Japanese population (HLA-A*24:02-C*12:02-B*52:01-DRB1*15:02-DQB1*06:01-DPB1*09:01 and HLA-A*24:02-C*07:02-B*07:02-DRB1*01:01-DQB1*05:01-DPB1*04:02, and HLA-A*33:03-C*14:03-B*44:03-DRB1*13:02-DQB1*06:04-DPB1*04:01). Taking into account the difficulties in direct comparison of the results, we decided to analyze a number of studies which revealed a given allele as the high risk one or the protective one. In order to obtain the most reliable results we included all studies which used methods which allowed to obtain allelic specificity. We did not include any of the results obtained by serological methods. As it was stated above, a metanalysis of those studies previously confirmed a significance of HLA-B46 only.

Our current analysis demonstrated that the following alleles were most commonly reported as related to high risk of GD in Asians: B*46:01 (7 studies), DPB1*05:01 (7 studies), DRB1*16:02 (4 studies), DRB1*08:02/03 (4 studies), DRB1*14:03 (3 studies), DRB1*04:05 (3 studies), DQB1*05:02 (3 studies), DQB1*03:03 (3 studies, including two studies with significance for children only and one study with significance for males only) ( Table 2 ). On the other hand, the following alleles were reported as potentially protective: DRB1*07/DRB1*07:01 (4 studies), DRB1*01:01 (3 studies), DRB1*13:02 (3 studies), DRB1*12:02 (3 studies) ( Table 2 ).

It is worth indicating, that among the alleles found as related to high risk of GD on the basis of our present review, HLA-DRB1*16:02 is in linkage disequilibrium with DQB1*05:02 (58), so they cannot be considered independent risk predictors. Additionally, among alleles demonstrated as related to lower risk HLA-DRB1*01:01 is in linkage disequilibrium with DQB1*05:01, DRB1*12:02 is in linkage disequilibrium with DQA1*06:01 and DRB1*13:02 is in linkage disequilibrium with DQB1*06:04 (58). These correlations must be taken into account if independent associations are analyzed.

Similarly to GD, the results concerning HLA associations with GO development are inconsistent. The overview of the published results is presented in Table 3 . Inoue et al. used serological method in Japanese patients and demonstrated that only DQw3 seemed to be a GO risk factor (59). However, in a study published a year later, Inoue et al. found that the risk of GO was associated with the following sets of HLA antigens: HLA-DQw4 without presence of -A31, HLA-A11 without presence of -DPw2, and with a co-presence of HLA-B5 and -Dw12 (45). Other authors applied serological methods, in another group of patients from Japan, and obtained entirely different results (60). They reported that predisposition to severe GO can be related to a presence of HLA-DR14 and DQ1 antigens, while HLA-B35, B54, DR4, and DQ4 were postulated to play a protective role (60).

In Iranian cohort, no differences in allele distribution between patients with and without GO was postulated (61).On the contrary, in a Korean study, HLA-C*03:03 allele had higher frequency of occurrence in GO as compared to non-GO patients (14). In a Chinese cohort described by Huang et al., the HLA-B*38:02, DRB1*16:02, DQA1*01:02 and DQB1*05:02 were postulated as GO high risk alleles (62). No consistent results were found ( Table 3 ), therefore, no allele can be unequivocally selected as actually GO related. Taking into account the significance of the applied method and the size of groups, the results presented by Huang et al. in 2021 can be considered the most reliable, as they were performed by NGS method in 82 GO patients, 272 GD patients and 411 healthy controls ( Table 3 ) (62). Therefore the alleles HLA-B*38:02, DRB1*16:02, DQA1*01:02 and DQB1*05:02 can be considered associated with increased risk of GO in Asians, although such relationship was demonstrated in one study only and requires further confirmation. Moreover, HLA-DRB1*16:02 is in linkage disequilibrium with DQA1*01:02 and DQB1*05:02 (58), so their associations with GO cannot be considered independent. On the other hand, the significance of HLA-B*38:02 seems to be entirely independent. In this study, no potentially protective alleles were selected, however there was a concordance in two previous studies in regard to HLA-B*54/HLA-B*54:01 which was demonstrated to play a protective role (14, 60). Therefore HLA-B*54:01 can be considered as associated with lower GO risk in Asian population.

In Caucasians, GD was initially reported to be associated with allelic group DRB1*03, i.e. with B8 and Dw3antigens (63). Many serological studies performed between 1978 and 1991 confirmed the significance of B8and DRw3 (63–75). In different Caucasian populations, the frequency of DRB1*03 in GD patients was demonstrated to range between 40% and 55%, while in the general population it ranged between 15% and 30%. The relative risk (RR) of GD in DRB1*03 carriers was assessed as approximately 4 (76). Since 1993, HLA-DQA1*05 has been reported as associated with GD (77, 78) and a few years later, HLA-DQB1*02 was also demonstrated to be related to GD in Caucasians (6). For many years, the results of studies were concordant mainly with regard to the increased risk of GD in carriers of HLA-DRB1*03 and of the alleles DQA1*05:01, DQB1*02:01 which are in linkage disequilibrium with HLA-DRB1*03 in Caucasians (6, 74, 77). Both case-control and family studies, demonstrated that a haplotype DRB1*03- DQA1*05-DQB1*02 can be considered a predictor of the development of GD (78).

However, HLA-DRB1*03 is well known to be associated with an increased risk of many autoimmune diseases, not exclusively with GD. Increased risks of Hashimoto’s thyroiditis, myasthenia gravis, Addison’s disease, diabetes mellitus type 1, systemic lupus erythematosus (SLE) were reported (79).

Many other alleles were postulated, but hardly any consistency was found between the results. Vita et al. found that frequencies of HLA-C*07, -C*17 and -DRB1*04 are significantly higher in patients with GD than in controls (5), while Heward et al. suggested that HLA-DQB1*03:01/04 and -DQB1*02 can play a role in GD occurrence (6).

Recently, the significance of B*08:01, B*39:06, B*37:01, C*07:01, C*14:02, C*03:02, C*17:01, DRB1*03:01, DRB1*11:01, DRB1*13:03, DRB1*01:03, DRB1*14:01, DQB1*03:01, DQB1*02:01 was demonstrated by our research team with application of NGS method (13). Unfortunately, it was the first study which applied NGS in Caucasian patients, therefore, no comparison with other NGS-based studies is possible. The overview of studies focused on associations between HLA and GD in Caucasian population is presented in Table 4 .

Similarly to Asian population, there were discrepancies in the applied methods, group size and analyzed MHC classes (13, 63–91). Most of the studies included only one MCH class, mainly class 2, while some studies included even only one gene, predominantly HLA-DRB1 ( Table 4 ). Therefore, due to difficulties in direct comparison of the results, we analyzed numbers of studies which revealed a given allele as a high risk one or a protective one. In order to obtain the most reliable results we included studies which used methods which allowed to obtain allelic specificity, or studies which reported allele group concordant with results in which a specific allele was provided (e.g. DRB1*03 and DRB1*03:01/04). We did not include any of the results obtained by serological methods.

Our present analysis demonstrated that the following alleles were most commonly reported as related to high risk of GD in Caucasians: DQA1*05:01 (7 studies), DRB1*03, including DRB1*03:01/04 (6 studies), C*07/C*07:01 (3 studies), DQB1*02/DQB1*02:01 (3 studies) ( Table 5 ). On the other hand, the following alleles were reported as potentially protective: DRB1*07/DRB1*07:01 (5 studies), DQA1*02:01 (3 studies), C*03/C*03:04 (2 studies), DQB1*02/DQB1*02:02 (2 studies) ( Table 5 ).

The significance of HLA-DRB1*03:01 as GD high risk allele is the only one which is common for both Asian and Caucasian populations. The alleles HLA-DRB1*03:01, DQA1*05:01 and DQB1*02:01 are in strong linkage disequilibrium in Caucasian population (58). Therefore, their significance cannot be considered entirely independent, however, the single presence of any of them can be correlated with the risk of GD. Moreover, some studies demonstrated a common association of HLA-B*08:01, which was also reported as GD high risk allele, with DRB1*03:01 and DQB1*02:01 (6, 13, 29, 30, 77). In our recent study, the combination of these three alleles occurred in 22% of patients with GD, while it was found only in 5.87% of the healthy control group (13). Interestingly, in patients in whom HLA-B*08:01 was present, it occurred with alleles other than the ones described above only in 1.9% of GD patients (13). HLA-B*08:01 is in linkage disequilibrium with HLA-C*07:01 – another high risk allele, whose significance in GD was also demonstrated (92, 93).

Therefore, it can be concluded that HLA-B*08:01, C*07:01, DRB1*03:01, DQA1*05:01, DQB1*02:01 and can be considered the most important predictors of GD development in Caucasians.

On the other hand, HLA-DRB1*07:01 was the allele most commonly demonstrated to be a protective one. Interestingly, this observation is the only one consistent with results of studies in Asians in regard to the protective alleles. It should be indicated that HLA-DRB1*07:01 is in linkage disequilibrium with two other potentially protective alleles – HLA-DQB1*02:02 and DQA1*02:01 (58). HLA-C*03:04, which is in no linkage disequilibrium with other discussed alleles, seems to be the only entirely independent protective allele (34, 92). Therefore, the alleles HLA-C*03:04, DRB1*07:01, DQB1*02:02 and DQA1*02:01 can be considered a group of GD protective alleles.

Some differences in the results between Caucasian and Asian populations seem unexpected. Among the alleles, which are in linkage disequilibrium with HLA-DQA1*05:01, only DRB1*03:01 was proved to be GD-related in both Asian and Caucasian populations (13, 30, 35, 54, 91). However, these linkage disequilibrium-based correlations are not exclusive for the Caucasian population, but are common in all analyzed populations, including Asians (58). Therefore, it remains unexplained why these correlations found in Caucasians are absent in Asians, in whom completely different alleles were reported as high risk of GD development. Furthermore, on the basis of the present review, a phenomenon of opposite roles of HLA-DQB1*02:01 can be observed. This allele was demonstrated as high risk in Caucasians (6, 13), but it was also reported as protective in Asian pediatric studies (48, 49). This population-dependent opposite correlation indicates the potential significance of other factors influencing GD risk in either population. Further studies with application of high resolution methods are required to solve this puzzle.

Similarly to the current state of art in Asians, the results concerning HLA associations with GO development in Caucasians are inconsistent ( Table 6 ). From the first study in 1980, DR3 antigen was being demonstrated in serological studies to be not only associated with GD high risk, but also with GO development (66). However, several studies, using different methods, both serological and genetic, did not confirmed such association (36, 72, 97) and the results were highly divergent (80, 94–99). Even a quite recent study performed by Yin et al. did not confirm the existence of HLA-related susceptibility to GO in the group of patients with GD and postulated the importance of environmental or epigenetic factors only (99). However, the authors of that study, similarly to some other earlier ones, focused only on the frequency of HLA-DR3, without assessing the frequencies of other alleles, and applied low resolution method. Recently, the significance of HLA-DRB1*03:01 was demonstrated by our research team with application of NGS method (15). We performed genotyping of both MCH classes and found that the frequency of several alleles is significantly higher in GO group as compared to either patients with GD without GO or healthy controls. Except for the previously mentioned HLA-DRB1*03:01, this group of alleles includes HLA-B*08:01, B*39:06, B*37:01, C*07:01, C*14:02, C*03:02, C*17:01, DRB1*11:01, DRB1*13:03, DRB1*01:03, DRB1*14:01, DQB1*03:01, DQB1*02:01. On the other hand, alleles HLA-C*04:01, C*03:04, C*07:02 and DRB1*15:02 were significantly less frequent in GO patients as compared to GD without GO or controls (15). Unfortunately, it was the first study which applied NGS in Caucasian patients, and therefore – similarly to the studies on GD described above – no comparison with other NGS-based studies is possible. The overview of studies focused on associations between HLA and GO in Caucasian is presented in Table 6 .

In our study, based on NGS method, HLA-DQB1*02:01 was also found to be a high risk allele (15). Such phenomenon could be expected as this allele is in linkage disequilibrium with HLA-DRB1*03:01 (58). Interestingly, another allele demonstrated to be GO-related in our study – i.e. HLA-DRB1*14:01, was previously observed to be GO-related in Japanese patients (60) ( Table ​1 ). This similarity is the only one between our results in Caucasian population and the already published data for Asian population in regard to GO. Nevertheless, such a lack of coherence between Asians and Caucasians could be expected, because HLA-related susceptibility for many autoimmune diseases differs between these two populations. In our NGS-based study, the highest risk of GO was associated with the presence of HLA-C*08:02 (OR 6.9) and -B*37:01 (OR 4.5) as well as DRB1*14:01 (OR 6.2) (15). There is no linkage disequilibrium between these three alleles (58, 92, 93), so the presence of any of them constitutes independent high risk factor.

Interestingly, in the same study, we demonstrated a protective effect of HLA-C*04:01 allele, which was previously reported as related to increased risk of subacute thyroiditis (SAT) (100, 101). This coincidence can be considered an explanation of the fact that SAT and GO actually occur together extremally rarely in clinical practice. It is worth noting that HLA alleles had already been suggested to be related with the course of SAT and GD in patients with simultaneous presence of both these diseases (102).

The significance of amino acid variants of HLA molecules were also postulated in the context of risk of GD development. In a study by Shin et al., both Leu35 (OR = 23.38, P = 0.0002) and Glu55 (OR = 23.38, P = 0.0002) of HLA-DPB1 were strongly associated with GD (14). These authors concluded that amino-acid signatures of the HLA-DP β chain, might contribute to the molecular pathogenesis of GD (14). However, in Japanese population, amino acid variants in HLA-DRB1 allotypes were examined, and critical significance of position 9 amino acid variants for GD development was found (57). Glu-9 variant, which was reported as related to GD high risk, is encoded by HLA-DRB1*03:01, DRB1*04:05 and -DRB1*14:03 alleles, which confer susceptibility to GD in several studies. On the other hand, Cys-9 variant, which was reported as protective against GD, is encoded by HLA-DRB1*07:01, DRB1*01:01 and DRB1*15:02 alleles, which were demonstrated to be protective against GD (57). It is speculated that changes in the charge and polarity of amino acids at these positions modify the three-dimensional structure of the DR peptide-binding pocket, which leads to susceptibility to or protection against GD (57, 119). Negatively charged Glu-9 was demonstrated to confer susceptibility to GD, while uncharged Cys-9 conferred protection against GD (57).

It was postulated that HLA-DR3 (HLA-DRB1*03) Arg-74 is the critical amino acid for the development of GD (88, 120). The group HLA-DR3 includes more than 30 alleles. A HLA-DRβ (beta chain of HLA-DR3) pocket variant with arginine at position 74 (HLA-DRβ-Arg74) is associated with an increased risk of GD, while glutamine (HLA-DRβ-Gln74) variant was protective (88, 119, 120). Three-dimensional computer modeling of the HLA-DR pocket allowed proving that electrostatic potential of HLA-DRβ-Arg74 generates a more positively charged P4 HLA-DR pocket than HLA-DRβ-Gln74. Therefore, the susceptibility to GD may be related to electrostatic potential of HLA-DR pocket and positive charge can be a high risk factor. The significance of this phenomenon can be confirmed by the finding that, among DRB1 molecules, DRB1*03 contains a positively charged Arg at β74, and DRB1*07 contains a non-charged Gln at β74 position (88). This positive charge difference is believed to facilitate auto-antigen presentation and T-cell activation because of its increased binding affinity of pathogenic self-peptides to the HLA-DR pocket (121). Candidate pathogenic peptides in GD include peptides derived from TSHR and Tg. This interaction between peptides of thyroidal origin and the HLA-DRβ-Arg74 pocket was considered as potential therapeutic target (121). Such an antigen-specific immunotherapy may constitute a crucial progress in the GD therapy, as current treatment modalities do not reverse the autoimmune process. It is well known that TSHR is the major autoantigen in GD. A TSHR peptide, designated TSHR.132, was found to be a dominant TSHR peptide in GD (122–124). TSHR.132 was demonstrated to bind with high affinity to recombinant HLA-DRβ-Arg74 and to cells that express HLA-DRβ-Arg74 (121). Cepharanthine is a small molecule compound which was demonstrated to inhibit T-cell activation by TSHR.132 ex vivo in splenocytes isolated from humanized mice induced with EAGD (experimental autoimmune Graves’ disease) (123, 124). Cepharantine directly interacts with Arg74 and blocks the HLA-DR3 pocket variant associated with AITD including GD. Due to the significant role of HLA-DR3 in many autoimmune diseases, cepharantine may be a potentially efficient causative therapy for GD and other autoimmune disorders such as Addison’s disease, Hashimoto’s thyroiditis, myasthenia gravis or SLE (121, 123, 124).

Additional various HLA polymorphisms were also described as associated with GD. In two independent Caucasian populations, an intronic variant (rs3094228) in HLA complex P5 (HCP5) was demonstrated to be associated with GD susceptibility and age of onset, which indicates a potential role of long non-coding ribonucleic acids, including HCP5, in GD pathogenesis (125, 126). HCP5 gene polymorphism (rs3094228) was associated with an earlier age of GD onset and patients with higher number of the HCP5 risk alleles tend to have a significantly earlier onset of GD (126).