Neuromyelitis optica is an HLA associated disease different from Multiple Sclerosis: a systematic review with meta-analysis

Neuromyelitis Optica and Multiple Sclerosis are idiopathic inflammatory demyelinating diseases of the central nervous system that currently are considered distinct autoimmune diseases, so differences in genetic susceptibility would be expected. This study aimed to investigate the HLA association with Neuromyelitis Optica by a systematic review with meta-analysis. The STROBE instrument guided research paper assessments. Thirteen papers published between 2009 and 2020 were eligible. 568 Neuromyelitis Optica patients, 41.4% Asians, 32.4% Latin Americans and 26.2% Europeans were analyzed. Only alleles of the DRB1 locus were genotyped in all studies. Neuromyelitis Optica patients have 2.46 more chances of having the DRB1*03 allelic group than controls. Ethnicity can influence genetic susceptibility. The main HLA association with Neuromyelitis Optica was the DRB1*03:01 allele in Western populations and with the DPB1*05:01 allele in Asia. Differences in the Multiple Sclerosis and Neuromyelitis Optica genetic susceptibility was confirmed in Afro descendants. The DRB1*03 allelic group associated with Neuromyelitis Optica has also been described in other systemic autoimmune diseases.

Multiple Sclerosis (MS) and Neuromyelitis optica (NMO) are inflammatory and neurodegenerative diseases of the central nervous system, that preferentially affect young woman causing neurological dysfunctions and disability 1 .
MS is the most frequent Idiopathic Inflammatory Demyelinating Diseases (IIDD), disseminated in time and space and a typical relapsing remitting clinical course. It has a peculiar geographical distribution, with a high prevalence in Caucasian of the Northern Hemisphere, and a very low prevalence in populations living in tropical regions 2,3 . NMO is a rare disease that occurs more frequently in Asians and Afro-descendants and is characterized, in most cases, by selective but not exclusive involvement of the optic nerve and spinal cord, also evolving with a relapsing remitting clinical course 4 .
A meta-analysis with the results of the thirteen studies that investigated the association of the DRB1*03:01 allele with NMO is summarized at the forest plot (Fig. 2), indicating that patients with NMO are 2.46 times more likely to have the DRB1*03 allele group than controls (95% CI 2.01-3.01). Table 4 describes the results of case controls studies showing the association of the HLA class II alleles (others than DRB1*03 allele group) and HLA class I alleles.
The association of class I HLA A*01 and B*08 with the NMO has only been described in Caucasians from the Netherlands 36 . HLA susceptibility in other CNS immune mediated diseases. Eight of the 13 case-controls studies elegible for this review also investigated the frequency of DRB1 alleles in MS patients and Controls. An associa-  Comparison between the NMO genetic susceptibility versus MS. The frequency of the DRB1 alleles associated with MS or NMO was compared in eight populations as shown at Table 5 and illustrated in Fig. 3. Two studies showed a significant difference between the frequency of the DRB1*03 allele group and the DRB1*15:01 allele (Ribeirão Preto (SP) 25

Discussion
The scientific evidence brought by the medical literature in this systematic review confirms that NMO is an associated HLA disease, thus classified because it occurs more frequently in individuals expressing certain DRB1 or DPB1 alleles. Evidence of the relationship between the HLA system and the genetic susceptibility has led to numerous studies concerning autoimmune etiopathogenesis.
The HLA complex maps to the short arm of chromosome 6 and provides instructions for making a group of related proteins known as HLA antigens. The human MHC is divided into three regions. The class I region contains the classical HLA-A, HLA-B, and HLA-C genes that encode the heavy chains of these class I molecules, expressed on the surface of most nucleated cells. The class II region contains HLA-DR, HLA-DQ, and HLA-DP genes, each encoding groups of antigens whose expression is limited to antigen-presenting cells (APC): B-lymphocytes, dendritic cells, monocytes, macrophages, endothelial cells, and activated T-lymphocytes. Class I molecules identify cells that are changed, bind to endogenous antigens in the target cells, and present the processed peptides from these antigens to CD8+ T cells (cytotoxic/suppressive), so the changed target cells can be killed by these lymphocytes. Class II molecules on the APC bind to extracellular exogenous proteins, and process and present them to CD4+ T lymphocytes (helper/inducer), initiating an immune response. The Class III region contains loci responsible for 21-hydroxylase, complement components, hormones, MIC molecules, and other signaling molecules such as tumor necrosis factors (TNFs) and heat shock proteins, and is not considered a part of the HLA complex. Nevertheless, it is located within the HLA region, and subject to similar genetic control mechanisms to the HLA genes 37 . As most of the genes located in the MHC complex encode molecules that have www.nature.com/scientificreports/ a high polymorphism, but low frequency of recombination, the allelic variation between them can make them good markers associated with either protection or susceptibility 17 . The discovery of the association between HLA allelic variants and susceptibility to MS was brought by studies conducted in Denmark in the 1970s 38 . It has been suggested that individuals could develop MS if they inherited certain HLA alleles that would make them vulnerable to environmental stimuli, initiating a chain of immunological events that would attack the myelin sheath. More than 500 studies worldwide using genotyping techniques confirmed a strong association of MS with the DRB1*15:01, DQA1*01:02 and DQB1*06:02 haplotype 17,39 .
This systematic review analyzed 13 case-control studies published from 2009 to 2020 that investigated the HLA association with NMO in populations with different ethnic background. Genotyping, with low, medium, or high resolution, was the method used in all eligible studies in the laboratory investigation of HLA class I alleles (A and B) and HLA class II alleles (DRB1, DQA1, DQB1, and DPB1). Only alleles of the DRB1 locus were genotyped in all the studies. All the studies genotyped alleles of the DRB1 locus ranging from seven to 30, the number of alleles investigated ( Table 2).
The DRB1*03/*03:01 were the most frequently found allelic group and allele, respectively in NMO groups with marked differences according to the ethnic background ( Table 3). The allelic frequency (2n) of the DRB1*03 or its subtype *03:01 varied from 10 to 26.2% and the phenotypic frequency (n) varied from 2 to 51.2%. The allelic frequency in Western populations was 14% in Mexican Mestizos 34 36 , what means that in Dutch, most NMO patients carried alleles of the DRB1*03 allele group.
To compare the DRB1*03 allele group's association, we used a meta-analysis based on the OR and the confidence interval (95% CI) described in the thirteen studies. The general evaluation showed heterogeneity of the OR among the studies of only 3.3% (I 2 = 3.28%; p = 0.41). The forest plot (Fig. 2) shows the summary measure of OR equal to 2.46 (95% CI 2.01-3.01). That is, patients with NMO are 2.46 times more likely to have the DRB1*03 allele group than controls. In the West, studies are not heterogeneous (I 2 = 0.00%; p = 0.92), with the measure of The results of the case-control studies comparing the allelic frequency of the DRB1*03 allele group in NMO with local controls also varied according to the ethnic background. In Caucasian populations, the association of DRB1*03 allelic group in NMO, firstly described in French Caucasians 24  In Asia, despite the low frequency of DRB1*03 allele group, an association with NMO was confirmed in India 31 (NMO 11% vs controls 2%, p = 0.00009). However, no association with this allele and NMO was founding in either Muslin Arabs from Israel 32 or in patients from Japan 30 or China 28 .
The strongest association with NMO in Asians was identified with the DPB1*05:01 allele in China 28 (NMO-90.0% vs controls-55.61%, p cB = 0.018) and in Japan 30 (NMO-85.7% vs controls-65.4%, p = 0.0074) confirming initial studies in Japanese patients with OSMS 21,22 . No association was found in Caucasians from Western Countries (France 24 and South Brazil 35 ) with DPB1*05:01 allele and NMO. There are no published data on the association of alleles of the DP locus in NMO patients with African ancestry.
The ethnicity can influence genetic susceptibility. The frequencies of DPB1*05:01 allele are higher in Asians (44.9-73.1%) than in Caucasians (2.6-5.3%) 22 . In two western populations, France 24 and the southern region of Brazil with strong European ancestry 35 , the frequency of class II alleles of the DP locus in NMO patients and in local controls was also investigated; however, the association of the DPB1*05:01 allele with NMO was not found. These results might be due to the so low frequency of the DPB1*05:01, limiting the statistical power to detect the association 28 .
The association of the DRB1*03 allele group in NMO stratified according to the NMO-IgG status (positive or negative) was also investigated in five studies (Table 3). In French Caucasians 24 , the DRB1*03 allelic group was associated only with the NMO IgG-positive subgroup. A combined analysis in cases from Spain and France 27    In the absence of a biological marker, the subgroup NMO IgG-negative may unduly include cases of classic MS, cases of spinal optic MS and cases of MOGAD, so the results on genetic susceptibility in these series need to be interpreted carefully. The second goal was to verify possible differences between the genetic susceptibility of NMO and other immune-mediated diseases of the CNS.

HLA
Eight studies selected for this review, while focusing primarily on the HLA association with NMO, also looked at MS's genetic susceptibility (data shown in Table 5). The strongest association of the DRB1*15 allelic group with MS worldwide was confirmed in six of the eight studies (Caucasians from France 24 and Denmark 29 , Latin Americans from the Caribbean 26 , Brazil-SP 25 and Brazil-RJ 33 and Asians from India 31 ). The DRB1*15 allelic group was not associated with MS in Spanish Caucasians 27 and Asians from South China 28 .
Differences in the frequency of the DRB1*15 allelic group (MS) and the DRB1*03 allelic group (associated with NMO) were investigated in these eight populations, as illustrated in Fig. 3. In Non-Caucasian populations from Caribbean Islands 26 and India 31 , a significant difference was found in the frequency of the HLA DRB1*15:01 but not in the frequency of the DRB1*03. Only in two populations living in the Southeast region of Brazil 35 , with strong African ancestry, it was shown that the distribution of both HLA DRB1 allele group (DRB1*03 and DRB1*15) in NMO is different from that observed in MS. Caucasians do not differ in the frequency of those associated alleles in the groups NMO and MS. As shown in Fig. 3, the DRB1*15 allele group in Denmark 29 , was practically similar in NMO and MS (30% and 35% respectively).
Only one study in the Dutch population with European ancestry 36 investigated HLA class I and class II alleles in NMOSD and MOGAD diseases. The susceptibility for NMO was strongly associated with the HLA-A*01, B*08, and DRB1*03 but no significant HLA association was found in MOG-IgG-seropositive patients.
Lincoln et al. 40 investigating the epistatic effect between the DQA1, DRB1 and DQB1 alleles and their association with MS drew attention to the possibility that the HLA-associated diseases are more haplotypical than allelic. DR/DQ haplotypes in NMO, MS, and controls were only investigated in the population of Rio de Janeiro 33 (data shown in Table 6). Among 29 haplotypes, eight were associated with either NMO or MS. The DRB1*03:01-DQA1*05:01/3/5-DQB1*02:01 was the most frequent haplotype (20%) associated with NMO. The haplotype DRB1*15:01-DQA1*01:02-DQB1*06:02 was associated with MS. Therefore, the significant difference in the NMO and MS groups confirmed haplotypic differences in the genetic susceptibility.
Other shreds of evidence link NMO with other autoimmune diseases. Acute events of optic neuritis and transverse myelitis in SLE and SS's raised the following question: would they occur due to a genetic influence on the autoimmunity shared between these diseases? Would they be complications of rheumatic diseases affecting the CNS 41 ? In SLE, inflammation damages the lungs, kidneys and CNS membranes, which express the AQP4 protein. Autoantibodies typically associated with SLE bind to DNA and RNA proteins, ribosomal proteins, and phospholipids. NMO-IgG antibodies have been detected in the serum of patients with SS or SLE and concomitant NMOSD, but not in the serum of patients with SLE or SS who do not have NMO spectrum diseases 56 . Based on these data, Pittock et al. 57 suggested that the occurrence of SLE/SS or autoantibodies in association with diseases of the NMO spectrum combined with seropositivity for the NMO-IgG antibody indicates that there is an association of these diseases. For this reason, they were included among the NMO spectrum syndromes 11 .
One GWAS study analyzing exclusively NMO genetic risk factors in Caucasians showed an association with the DRB1*03:01-DQA1*05:01/3/5-DQB1*02:01 haplotype and the class I, HLA-B*08:01 and HLA-C*07:01alleles in NMO subgroup positive for the NMO-IgG. Additionally, a reduced copy number variation (CNV) in the region of complement component C4 encoded in the MHC class III region was found. Estrada et al. 58 suggested that the C4 deletions could be the functional driver of the NMO association and call the attention that the same C4 CNV and DRB1*03:01-DQA1*05:01/3/5-DQB1*02:01 haplotype were risk factors for SLE.

Limitations.
We have identified some limitations in these studies, such as the low number of NMO cases analyzed in each study (ten studies with 45 or fewer NMO patients). This is explained by the fact that NMO is a rare disease and only recognized as a different condition from MS by specific diagnostic criteria after 1999. Furthermore, another limitation was the low resolution of the genotyping technique since it was limited, in most studies, only to the typing of HLA-DRB1*03 allele group, without specification of its subtypes; as well as the small number of studies genotyping DR/DQ alleles to identify the haplotypes associated with NMO.
Finally, the genetic susceptibility of the NMO group negative for AQP4-Ab needs to be analyzed with caution because optic spinal disease could be related to Asian type MS, Conventional Multiple Sclerosis, or MOG-IgG related disorders.

Conclusions
NMO is an HLA associated disease.
Patients with NMO are 2.46 times more likely to have the DRB1*03 allelic group than controls. Alleles of the DRB1*03 group, specifically the DRB1*03:01, conferred genetic susceptibility to NMO in most of Latin Americans, in half of the Caucasians and in one-quarter of the Asians. In Far East Asian, the genetic susceptibility for NMO is associated with the DPB1*05:01 allele.
Most of the studies confirmed the DRB1*03 allele group's association with NMO positive for the NMO-IgG antibody.
The genetic susceptibility for NMO differed from MS in Latin America populations with a high ethnic African background.
In the Netherlands, the DRB1*03:01 allele was associated with NMO, but no HLA association was found with MOGAD. Those findings bring new evidence that NMO, MS and MOGAD are different immune-mediated CNS conditions. It is recommended that new studies with a greater number of patients analyzed by the four-digit HLA DR/ DQ alleles immunophenotyping technique be performed in different populations to increase knowledge about genetic susceptibility in NMO.

Methods
Selection of the articles. A systematic review of the literature was carried out by a search in the MEDLINE (Medical Literature Analysis and Retrieval System Online) via PubMeb's updated version interface, LILACS (Scientific and Technical Literature of Latin America and the Caribbean) via VHL (Virtual Health Library) and Sci-ELO (Scientific Electronic Library Online) electronic databases. The search for publications in any of the three languages, English, Spanish or Portuguese was done by two independent evaluators (LFC and HAF). The period for inclusion was 2009 to March 31, 2020. The search strategy used the combined MeSH terms "Neuromyelitis Optica" and "HLA antigens", and the combined text words Neuromyelitis Optica and HLA association studies. Studies considered for this review: case-control studies (association studies) analyzing genetic susceptibility through genotyping of HLA genes in human subjects with NMO according to international diagnostic criteria 5,10,14 and only publications in English, Spanish, or Portuguese languages. Case reports, reviews, publications not related to the review's objectives, and publications in other languages were excluded. Articles identified in more than one database were considered only once. The papers which fulfilled the eligibility criteria were included in the qualitative and quantitative analyzes.
Outcomes. The primary outcome was the association of the HLA alleles with NMO. Secondary outcomes were a comparison of the genetic susceptibility in NMO and MS.
Study quality evaluation. The selected articles were submitted to the STROBE evaluation method (Strengthening the Notification of Observational Studies in Epidemiology), for case-control studies 59 . Two evaluators (LFC and HAF) addressed the questions, with a maximum score of 22, equivalent to the number of items presented in the STROBE instrument. We regarded studies that scored "15-22" as high quality, those that scored "7-14" as moderate quality, and those that scored "0-7" as low quality (Supplementary Table S1).
This review employed the guidelines indicated in the MOOSE (Meta-analysis Of Observational Studies in Epidemiology) and PRISMA (Preferred Reporting Items for Systematic reviews and meta-analyses) Consensus Statements 60,61 .
Statistical analysis. Several comparisons were noted in the included studies; NMO vs controls, NMO vs controls stratified by the NMO-IgG/AQP4-IgG status; NMO vs MS and NMO vs MOGAD. The statistical analysis applied allelic frequencies expressing number of alleles (2n) or phenotypic frequencies (n) indicating the number of participants carrying specific allele. Frequencies of HLA alleles were compared using the chi-square test (p) and corrected by Fisher´s exact test (p cF ), Bonferroni (p cB ), or Sidak (p cS ) methods. The level of significance was < 0.05. OR with 95% confidence interval (CI) was calculated for each comparison.
A meta-analysis by mixed-effects models was performed using the metaphor library (2010) 62 of software R version 3.3.2 (2016). To evaluate the studies' heterogeneity, the I 2 statistics of Higgins and Green 63 were used. The forest plot chart was used to present the results.