HLA genotype-clinical phenotype correlations in multiple sclerosis and neuromyelitis optica spectrum disorders based on Japan MS/NMOSD Biobank data

HLA genotype-clinical phenotype correlations are not established for multiple sclerosis (MS) and neuromyelitis optica spectrum disorders (NMOSD). We studied HLA-DRB1/DPB1 genotype–phenotype correlations in 528 MS and 165 NMOSD cases using Japan MS/NMOSD Biobank materials. HLA-DRB1*04:05, DRB1*15:01 and DPB1*03:01 correlated with MS susceptibility and DRB1*01:01, DRB1*09:01, DRB1*13:02 and DPB1*04:01 were protective against MS. HLA-DRB1*15:01 was associated with increased optic neuritis and cerebellar involvement and worsened visual and pyramidal functional scale (FS) scores, resulting in higher progression index values. HLA-DRB1*04:05 was associated with younger onset age, high visual FS scores, and a high tendency to develop optic neuritis. HLA-DPB1*03:01 increased brainstem and cerebellar FS scores. By contrast, HLA-DRB1*01:01 decreased spinal cord involvement and sensory FS scores, HLA-DRB1*09:01 decreased annualized relapse rate, brainstem involvement and bowel and bladder FS scores, and HLA-DRB1*13:02 decreased spinal cord and brainstem involvement. In NMOSD, HLA-DRB1*08:02 and DPB1*05:01 were associated with susceptibility and DRB1*09:01 was protective. Multivariable analysis revealed old onset age, long disease duration, and many relapses as independent disability risks in both MS and NMOSD, and HLA-DRB1*15:01 as an independent risk only in MS. Therefore, both susceptibility and protective alleles can influence the clinical manifestations in MS, while such genotype–phenotype correlations are unclear in NMOSD.

www.nature.com/scientificreports/ information and biological samples, such as DNA, as a part of the Rare Disease Bank at the National Institutes of Biomedical Innovation, Health and Nutrition in Japan 16 . Various susceptibility and protective HLA class II alleles have been reported in MS and NMOSD, including in Japanese patients 11,13,15,17 . HLA genotype-clinical phenotype correlations have been reported, focusing on the major susceptibility alleles in Caucasian (HLA-DRB1*15:01) and Japanese MS patients (HLA-DRB1*15:01 and DRB1*04:05). However, phenotype correlations are unknown in NMOSD. In Caucasian patients with MS, HLA-DRB1*15:01 is associated with younger age at onset, higher white matter lesion volume, greater brain atrophy and impairment of cognitive function 10,18 , but its association with clinical course, disease severity, and prognosis is controversial [19][20][21][22][23] . In Japanese patients with MS, HLA-DRB1*15:01 is associated with a higher frequency of cerebrospinal fluid (CSF) IgG abnormality, including OCBs, while DRB1*04:05 is associated with younger age at onset, lower frequency of CSF IgG abnormality, milder disease course and fewer intracortical lesions 7,11,13,24 . In the present study, we aimed to identify relationships between various susceptibility/protective HLA alleles and clinical manifestations in MS and NMOSD using the newly established Japan MS/NMOSD Biobank data.

Results
Comparison of demographic features between MS and NMOSD. We enrolled 739 patients and collected 731 DNA, 528 plasma and 566 serum samples. Overall, 19,8, and 1 patients were excluded from the analyses because of a lack of clinical information, a lack of DNA samples, and double registration, respectively. Clinical information and genotypes of HLA-DRB1 and -DPB1 alleles were available from 528 MS and 183 NMO/ NMOSD patients based on the 2010 revised McDonald criteria 25 and NMO and NMOSD criteria advocated by Wingerchuk in 2006 and 26,27 , respectively. Of 183 NMO/NMOSD patients enrolled, 165 patients fulfilled the international consensus diagnostic criteria of NMOSD published in 2015 28 . In the context of the current clinical situation, we conducted the following analyses of NMOSD using data from these 165 patients. MS patients included 438 relapsing-remitting MS (RRMS), 70 secondary progressive MS (SPMS), and 19 primary progressive MS (PPMS) cases. Comparing clinical and laboratory information between MS and NMOSD (Table 1), MS patients were younger at onset and at registration compared with NMOSD patients (both p < 0.001). The female to male ratio in NMOSD was higher than in MS (p < 0.001). Disease duration was longer in MS than in NMOSD patients (p = 0.004). There was no significant difference in the total number of relapses between the two groups, but annualized relapse rate (ARR) was higher in NMOSD than in MS (p = 0.001). NMOSD patients experienced optic neuritis more frequently than MS patients, while MS patients had higher frequencies of cerebrum, cerebellum, and brainstem involvement than NMOSD patients. Myelitis, especially episodes of transverse myelitis and the presence of longitudinally extensive spinal cord lesions (LESCLs), was more frequent in NMOSD compared with MS (p = 0.031, p < 0.001 and p < 0.001, respectively). NMOSD patients had higher Kurtzke's Expanded Disability Status Scale (EDSS) scores 29 and progression index (PI) values than MS patients (both p < 0.001). Of the seven functional system (FS) scores 29 , visual, pyramidal, sensory, and bowel and bladder FS scores were significantly higher in NMOSD than in MS, while brainstem, cerebellar and cerebral FS scores were significantly higher in MS than in NMOSD. Patients with MS more frequently had CSF IgG abnormalities including OCBpositivity and/or high IgG index (> 0.658) 30 than those with NMOSD (p < 0.001). Anti-aquaporin 4 (AQP4) antibody was positive in 95.1% of NMOSD patients. In this study, four cases with anti-AQP4 antibody were diagnosed as MS. These diagnoses were ultimately made by each doctor-in-charge based on the patients' clinical and MRI characteristics and clinical courses, and they were treated with interferon-β (IFN-β). Excluding these cases from the MS group did not significantly change our findings. Half of the MS patients and one-quarter of the NMOSD patients showed a benign course. Co-existing autoimmune diseases and/or autoantibodies, especially Sjögren syndrome, anti-nuclear antibody (ANA), and anti-SS-A/SS-B antibodies, were more frequent in NMOSD patients compared with MS patients (all p < 0.001), while frequencies of co-existing thyroid disease and/or thyroid-related antibodies did not differ between the two groups. Almost all NMOSD patients in the cohort were treated with oral corticosteroids, and two thirds with immunosuppressant drugs. More than 85% of MS patients were treated with disease-modifying drugs (DMDs). Twenty-seven NMOSD patients were treated with IFN-β. Out of the 23 patients for whom data on IFN-β efficacy was obtained, IFN-β was not effective in 18 (78.3%). The other five patients who showed effective treatment responses to IFN-β were additionally treated with corticosteroids.

Comparison of the demographic and clinical features of MS patients between northern and southern Japan.
We previously reported a difference in the clinical characteristics of MS patients between northern and southern Japan 1,13 ; therefore, MS patients were stratified by their resident area into two groups (northern and southern Japan) at a latitude of 37° North. Of 528 MS patients, 125 and 371 patients lived in northern and southern Japan, respectively (residence information was missing for 32 patients). As shown in Supplementary Table S1, northern MS patients had lower numbers of relapses (p < 0.001), annualized relapse rate (p = 0.001), and FS score of cerebellar function (p = 0.050), but they experienced cerebrum relapses more frequently than southern MS patients (p = 0.005). Although disease disability and severity did not differ between northern and southern MS patients, patients with a benign disease course (EDSS ≤ 3 at 10 or more years of disease duration) were more frequently observed in southern Japan than in northern Japan (p = 0.027). Patients with a high IgG index were more frequent in northern Japan than in southern Japan (p = 0.039), but the OCB positivity rate was similar between groups. Patients who experienced transverse myelitis and autoantibody positivity, especially for ANA and anti-SS-A and/or -SS-B antibodies, were more frequent in southern Japan than in northern Japan (p = 0.007, 0.001, 0.001 and 0.026, respectively). Although a small number of PPMS patients were registered only in southern Japan, the above-mentioned differences were similar if these 19 PPMS patients were excluded (data not shown). We previously reported that the effect of HLA-DRB1*04:05 on MS patients differed between northern and southern Japan 13 ; therefore, we additionally compared MS patients from northern and southern Japan with and without HLA-DRB1*04:05 (Supplementary Tables S2, S3). CSF IgG abnormality was less frequent in HLA-DRB1*04:05-positive than -negative patients from northern or southern Japan (p = 0.005 and p < 0.001, respectively) similar to that for all of Japan. In southern Japan, MS patients with HLA-DRB1*04:05 were younger at disease onset (p = 0.009) with higher visual FS scores (p = 0.003) and tended to have optic nerve involvement (optic neuritis) more frequently than patients without DRB1*04:05 (p = 0.090). In northern Japan, HLA-DRB1*04:05positive patients had higher pyramidal FS scores (p = 0.029) and experienced fingolimod treatment more frequently than HLA-DRB1*04:05-negative patients (p = 0.006). Moreover, when excluding PPMS, as in a previous study 13 , PI and MS severity scores (MSSS) 31 were lower in southern MS patients with HLA-DRB1*04:05 than in those without the allele (p = 0.22 and 0.050, respectively) (Supplementary Table S4). When we compared the clinical characteristics of MS patients with HLA-DRB1*04:05 (excluding PPMS) between northern and southern Japan, northern patients had a lower number of relapses (p < 0.001) and annualized relapse rate (p = 0.006) but higher PI (p = 0.018) and MSSS (p = 0.008). Patients who experienced transverse myelitis and who had ANA were more frequent in southern Japan than in northern Japan (p = 0.023 and 0.012, respectively). Southern MS patients with HLA-DRB1*04:05 were less frequently treated with DMD, especially with fingolimod, compared with northern patients (p = 0.018 and 0.009, respectively) (Supplementary Table S5).
HLA-DRB1*15:01-positive MS patients had CSF IgG abnormalities more commonly than DRB1*15:01-negative patients (p = 0.036) ( Table 7) and they experienced attacks in the optic nerve (optic neuritis) and cerebellum more frequently (p = 0.044 and 0.031, respectively), and also tended to have a higher FS score of visual function (p = 0.060) compared with those without the allele. FS scores of cerebellar function were similar between these two groups. FS scores of pyramidal function and PI were significantly higher in DRB1*15:01-positive patients than in DRB1*15:01-negative patients (p = 0.027 and 0.031, respectively), although EDSS scores and MSSS were not significantly different between them. MS patients with DRB1*15:01 had anti-SS-A and/or -SS-B antibodies more frequently than those without (p = 0.007). HLA-DPB1*03:01 carriers had higher FS scores of brainstem and cerebellar function than non-carriers (p = 0.035 and 0.004, respectively) (Supplementary Table S6).
With respect to resistance alleles, MS patients with the HLA-DRB1*01:01 allele had a lower female to male ratio than those without (p = 0.006) (Supplementary www.nature.com/scientificreports/ www.nature.com/scientificreports/   Table S11). Moreover, HLA-DRB1*08:02 carriers were more prone to having autoimmune diseases and/or autoantibodies, especially anti-SS-A and/or -SS-B antibodies, compared with non-carriers (p = 0.033 and 0.035, respectively). NMOSD patients with HLA-DPB1*05:01 had longer disease durations and a larger number of relapses compared with those without (p = 0.016 and 0.001, respectively) (Supplementary Table S12), while the ARR was comparable between them. Moreover, they tended to have more frequent involvement of the cerebrum and cerebellum compared with those without (p = 0.090 and 0.077, respectively). PI was lower in patients with the HLA-DPB1*05:01 allele compared with those without (p = 0.034), although EDSS scores were not statistically different. They had Sjögren syndrome and anti-SS-A and/or -SS-B antibodies more often than DPB1*05:01-negative patients (p = 0.044 and 0.028, respectively). With respect to protective alleles, HLA-DRB1*09:01-positive NMOSD patients had co-existing autoimmune diseases, especially Sjögren syndrome, more frequently than DRB1*09:01-netagive patients (p = 0.030 and 0.034, respectively) (Supplementary Table S13).
Factors associated with disability. Finally, we evaluated which clinical parameters, laboratory findings and HLA subtypes were associated with disability. In RRMS, we focused on HLA-DRB1*04:05 and DRB1*15:01 as genetic factors. Multivariable models showed that old age at onset, long disease duration, many relapses, and being a carrier of the HLA-DRB1*15:01 allele were independent risk factors for high EDSS scores (Table 8). For NMOSD, old age at onset, long disease duration and many relapses were risk factors for high EDSS scores, but episodes of bilateral optic neuritis, experience of LESCL and harbouring the HLA-DPB1*05:01 allele were not (Supplementary Table S14).

Discussion
We used the Japan MS/NMOSD Biobank and the largest cohort of Japanese MS and NMOSD patients studied to date to show that HLA-DRB1*04:05, DRB1*15:01 and DPB1*03:01 are susceptibility alleles, while HLA-DRB1*01:01, DRB1*09:01, DRB1*13:02 and DPB1*04:01 are protective alleles for MS, which is consistent with previous Japanese studies on MS 11,13 . For NMOSD, HLA-DRB1*08:02 and DPB1*05:01 are susceptibility alleles and DRB1*09:01 is a protective allele. This is a novel finding for HLA-DRB1*08:02 and confirmatory for DPB1*05:01 and DRB1*09:01 15 . Importantly, we discovered new significant HLA genotype-clinical phenotype correlations in MS and NMOSD. In our MS cohort, the susceptibility alleles, DRB1*15:01, and DPB1*03:01, were associated overall with worsening functional disability scores. The finding that HLA-DRB1*15:01 increased the involvement of the optic nerve (optic neuritis) and cerebellum and worsened visual and pyramidal FS scores, resulting in greater PI, is a new finding in Asians and is compatible with several reports in Caucasians showing that DRB1*15:01 is associated with greater MSSS and more severe cognitive impairment 18,20 . The relationship of HLA-DRB1*04:05 to clinical manifestations has not been reported in Caucasians, primarily because this allele is rare in most European descendants 32,33 . MSSS was lower in southern MS patients with HLA-DRB1*04:05 than in those without the Table 5. Summary of HLA genotype-clinical phenotype correlations in NMOSD. If the effects were derived from only 5 or less than 5 positive cases, such findings were regarded as too preliminary, and therefore they are not described in the table. Ab antibody, AID autoimmune disease, ANA anti-nuclear antibody, NMOSD neuromyelitis optica spectrum disorders, OCBs oligoclonal IgG bands, PI Progression Index.  www.nature.com/scientificreports/ allele when excluding PPMS in this study, which is in accord with a previous finding that HLA-DRB1*04:05 was associated with a milder disease course as evaluated by EDSS scores and PI in Japanese MS patients, particularly from southern Japan 11,13 . Unexpectedly, we also found an association of this allele with increased tendency of involvement of the optic nerve (optic neuritis) and worsening of visual FS scores. Although these results appear contradictory, the fact that motor disability involving pyramidal and cerebellar functions is more contributory to EDSS scores than other systems, such as visual function, may explain this discrepancy. A comparison of the clinical characteristics of MS patients with HLA-DRB1*04:05 between northern and southern Japan showed that southern patients had a milder disease course and atypical presentation including episodes of transverse myelitis and having ANA, which may result in the lower frequency of DMD treatment. Given that they share HLA-DRB1*04:05 as a common genetic background, environmental factors including residential latitude may contribute in part to these differences in the clinical manifestations. In Caucasians, HLA-DRB1*15:01 has repeatedly been shown to be associated with younger age at onset in MS patients 10,18,34 , while in Japanese, HLA-DRB1*04:05 is associated with a younger age at onset 11,13 , which was confirmed in the present study. Thus, the contribution of HLA alleles to a younger age at onset could vary according to race. The association of HLA-DPB1*03:01 with an increase in brainstem and cerebellar FS scores is a novel finding, which further supports a contribution of the susceptibility alleles to aggravation of disease severity in Japanese MS patients. In contrast, the resistance alleles, HLA-DRB1*01:01, DRB1*09:01 and DRB1*13:02, were associated with decreased functional disability in the present cohort. HLA-DRB1*01:01 decreased involvement of the spinal cord and sensory FS scores, DRB1*09:01 decreased ARR, involvement of the brainstem and bowel and bladder FS scores, and DRB1:13:02 decreased involvement of the spinal cord and brainstem. HLA-DRB1*01 was reported to decrease PI and the frequency of malignant subtype and to extend the time to reach an EDSS score of 6 in Caucasian patients 35 , which is in line with our findings. Another resistance allele, HLA-DPB1*04:01, was reported to be associated with a reduced MSSS under corticosteroid or azathioprine in Brazilian patients 36 . This was not obvious in the present study, possibly because of the low frequency of HLA-DPB1*04:01 in this cohort. Collectively, these resistance alleles are indicated to decrease disease susceptibility and to decrease involvement of relevant CNS sites in different races.
We examined CSF IgG in our cohort and found that the resistance alleles, HLA-DRB1*13:02 and DPB1*04:01, decreased OCBs while one of the susceptibility alleles, DRB1*15:01, increased, and another, DRB1*04:05, decreased CSF IgG abnormalities. Potentiation of CSF IgG abnormalities by DRB1*15:01 was reported in both Caucasians 9,37,38 and Japanese patients with MS 11,13,39 , while mitigation of CSF IgG abnormalities by DRB1*04:05 has been repeatedly reported in Japanese patients with MS 7,11,13,39 . Collectively, even for susceptibility alleles, their influence on CSF IgG abnormalities, including OCBs, appears to be heterogeneous. The association of two resistance alleles, HLA-DRB1*13:02 and DPB1*04:01, with lower OCB frequency is a novel finding, which should be confirmed by a large scale study. HLA-DRB1*13:02 is a protective allele for systemic and organ-specific autoimmune diseases, such as rheumatoid arthritis, systemic lupus erythematosus, psoriasis, and autoimmune hepatitis 40 . Therefore, a person with HLA-DRB1*13:02 may have resistance to producing autoreactive immunoglobulins. Only a few studies of Caucasians reported such associations of distinct HLA alleles with CSF IgG abnormalities, which may be partly accounted for by the extremely high overall prevalence of OCBs (> 90%) 9,41 that obscures the effects of HLA alleles. Thus, investigation of a Japanese MS cohort, in which the positivity rate of OCBs is not very high (around 60%) 6,7 , for any association of HLA alleles with CSF IgG abnormalities is warranted.
Multivariable analysis also revealed HLA-DRB1*15:01 to be an independent risk factor for high EDSS scores in addition to old age at onset, long disease duration, and many relapses. Older age at onset, male sex, and high relapse rate especially in the early disease course, are well known poor prognostic factors in Caucasians [42][43][44][45] , but the natural history of MS in Asia is less well studied. We previously assessed the association between age at onset and disease severity using MSSS 13 . Old age at onset is thought to be an independent risk factor for increased disease severity and disability in both Caucasians and Japanese. In contrast to previous studies of Caucasian MS patients [42][43][44] , this and our previous studies did not show an effect of sex on disease severity or disability in Japanese MS patients 13 . Therefore, male sex can be a poor prognostic factor for Caucasians but not Japanese, although this should be confirmed in a large study, such as a nation-wide survey. The effect of the total number www.nature.com/scientificreports/ of relapses on prognosis is controversial 46 , but relapse frequency up to year 5 was reported to be predictive in some studies 44,47 . The median disease duration in RRMS patients was 8 years in our study, which may reflect relatively early disease activity. In NMOSD, no association between any HLA allele and clinical manifestations has been found. The present NMOSD patients had highly comparable demographic and clinical features, including sex ratio, age of onset, and frequency of OCB-positivity with previous Japanese clinical studies 48,49 . One of the susceptibility alleles, HLA-DPB1*05:01, was associated with lower PI compared with HLA-DPB1*05:01-negative patients, even though EDSS scores did not differ between them. Because the disability of NMOSD patients is mainly caused by acute attacks, and a progressive disease course is uncommon in NMOSD compared with MS 50-52 , the PI decreases as the disease duration increases, particularly when treatments such as corticosteroids reduce relapses 53 . Indeed, NMOSD patients with DPB1*05:01 had a significantly longer disease duration than those without the allele in our cohort, which confirms this idea. Multivariable analyses also supported the idea that the HLA-DPB1*05:01 allele has no significant impact on disability.
We found that the susceptibility alleles, HLA-DRB1*08:02 and DPB1*05:01, increased the frequencies of concomitant autoimmune diseases and autoantibodies such as Sjögren syndrome and anti-SS-A/SS-B antibodies, while the resistance allele, DRB1*09:01, was also associated with increased frequencies of co-existing autoimmune diseases, such as Sjögren syndrome. Thus, how associations of either the susceptibility or protective alleles with autoimmune traits modulate NMOSD susceptibility or resistance through autoimmune backgrounds remain elusive.
In our multivariable analyses, while higher age at onset, longer disease duration, and more relapses were independent risks for disability, the susceptibility allele, HLA-DPB1*05:01, was not an independent risk factor for disability, in contrast to MS. HLA-DPB1*05:01 could increase the susceptibility to NMOSD but may not confer disability. Associations between older age at onset and higher EDSS scores and between death and high ARR have been reported in European and Asian cohorts 52,54-56 . Collectively, older age at onset and relapse number are regarded as risk factors for disability, underscoring the importance of early introduction of efficacious preventative therapies in elderly-onset NMOSD patients. Although chronic progression is not considered to contribute to NMOSD, the longer disease duration was also an independent risk for disability in our NMOSD patients. We have reported a persistent increase of serum neurofilament light chain in the remission phase in a fraction of NMOSD patients, indicating continuous damage to axons following acute relapses 57 . Long term follow-up studies are required to elucidate whether, and if so how, a longer disease duration influences disability in NMOSD.
In terms of disability in NMOSD patients, our cohort showed relatively mild disability compared with the clinical course of NMOSD patients described by Wingerchuk et al. 27 . However, NMOSD patients without LESCL were reported to have a milder disease course than those with LESCL 58 . Because patients without LESCL occupied 57.4% of NMOSD patients in our cohort, overall disability in our NMOSD cases appeared to be mild. Moreover, a recent study showed that Japanese NMOSD patients with AQP4-IgG had a lower risk of relapse, especially transverse myelitis and optic neuritis attacks, compared with Caucasian patients 52 . This might explain why our cohort had lower EDSS scores and lower frequencies of LESCL than expected values based on the results of Caucasian NMOSD patients. In addition, more than 95% of NMOSD patients received prednisolone treatment, which can decrease disease severity by preventing relapses 53 . Therefore, we think that our cohort showed normal attributes for Japanese NMOSD patients.
In this study, we compared the demographic and clinical features of MS and NMOSD patients and assessed their distinct genetic backgrounds and the differential effects of HLA class II alleles on clinical features between these two diseases. Although MS and NMOSD are distinct diseases, both are inflammatory diseases affecting CNS tissues and their clinical findings sometimes overlap. Moreover, it is occasionally difficult to distinguish between these two diseases, particularly for cases of seronegative NMOSD in clinical settings. Thus, we consider that the collection of MS and NMOSD information and biosamples by the Japan MS/NMOSD biobank will be very useful for the characterization of clinical and laboratory findings to differentiate between the two conditions. By describing the differences in clinical and laboratory tests between MS and NMOSD concurrently, the differential aspects of these two diseases will become clearer.
There are several limitations in our study. First, susceptibility HLA alleles for MS and NMOSD, other than HLA-DRB1 and -DPB1, were recently reported for Japanese patients using next-generation sequencing 17 . These include HLA-B*39:01 and HLA-B*15:01 for MS and HLA-DQA1*05:03 for NMOSD 17 . We genotyped only two classical HLA class II genes (DRB1 and DPB1) using sequence-specific oligonucleotide hybridization. By applying new methodology, such as next-generation sequencing, to this newly established Japan MS/NMOSD Biobank of samples, new risk genes of MS or NMOSD could be identified in the future. Second, healthy controls were not included in the MS/NMOSD Biobank; therefore, we used data of healthy controls collected by the Japan Multiple Sclerosis Genetics Consortium 13 . Therefore, similar risk and resistance alleles to those in the previous study may tend to be obtained; however, the HLA genotype-clinical phenotype correlations were newly discovered in the present study. Third, in this MS/NMOSD biobank, the number of enrolled MS patients from southern Japan were approximately 3 times higher than that from northern Japan despite the higher prevalence rate of MS in northern Japan 59 . However, the general population of southern Japan is about 7 times higher than that of northern Japan. Therefore, although potential enrolment bias cannot be completely excluded, we do not think that the 3:1 ratio of enrolled MS patients between southern and northern Japan distorted the results of the present study. Fourth, as we did not systematically examine anti-myelin oligodendrocyte glycoprotein (MOG) antibodies, some patients with anti-MOG antibodies may be included in our MS cohort. However, most adult patients with typical MS, including Japanese, are negative for anti-MOG antibodies 60,61 ; therefore, possible contamination of MOG antibody disease would not severely distort our results.
In conclusion, based on the Japan MS/NMOSD Biobank data, we show HLA genotype-clinical phenotype correlations concerning the confirmed and newly identified susceptibility and resistance DRB1 and DPB1 genes Genotyping. The participants' genotypes of HLA-DRB1 and -DPB1 alleles were determined by the sequencespecific oligonucleotide probe method (Luminex method) using DNA samples at the National Institutes of Biomedical Innovation, Health and Nutrition and the HLA Laboratory (Kyoto, Japan), as described previously 16,63 . The carrier frequencies of Japanese healthy controls from a Japan Multiple Sclerosis Genetics Consortium study were used 13 .

Statistical analysis.
To compare the demographic characteristics and clinical information between two groups, Fisher's exact test was used for categorical data, whereas the Wilcoxon rank sum test was used for continuous data. To identify the susceptibility and protective alleles of HLA-DRB1 and -DPB1, the phenotypic frequencies of these alleles were compared using Fisher's exact test. Uncorrected p values (p uncorr ) were multiplied by the number of comparisons (Bonferroni-Dunn's correction) to calculate p corr values. To calculate the risk of optic neuritis and FS scores (≥ 1) of visual function for MS patients with HLA-DRB1*04:05, logistic regression analyses were used with adjustment for disease duration. When comparing the proportions of disease course between groups, a likelihood ratio chi-square test was employed as it includes more than two categories. To identify independent factors that correlate with disease disability, EDSS scores were rank-normalized and multivariable linear regression analyses were performed. All analyses were carried out using JMP version 14.1.0 (SAS Institute, Cary, NC, USA). Statistical significance was set at p < 0.05.

Data availability
The datasets generated and/or analysed during the present study are available from the corresponding author based on the guidelines of the ethics committees upon reasonable request from any qualified investigator.