Multiple sclerosis (MS) is a chronic immune-mediated and neurodegenerative disease of the central nervous system (CNS) (1, 2). Approximately 2.8 million people worldwide suffer from MS, with women being affected two to three times more often than men and with an average age at diagnosis of 32 years (3, 4). MS is classified into three different clinical courses: relapsing–remitting MS (RRMS) as the most common form (~85% of initial diagnoses), secondary progressive MS (SPMS), and primary progressive MS (PPMS) (~15% of initial diagnoses) (5–7). Clinically, RRMS is characterized by episodes of disease (relapses) followed by a partial recovery of symptoms (remissions). As the neurological deficits worsen with disease progression, approximately 80% of the RRMS cases convert to SPMS within 25 years after the diagnosis (6, 8, 9). PPMS and SPMS are characterized by a continuous worsening of symptoms without significant recovery. The symptoms of MS include, among others, limited mobility, impaired vision, and cognitive deficits (10). The severity of disability is usually determined by the Expanded Disability Status Scale (EDSS) (11).

The immune system plays a key role in the pathophysiology of MS. Immune cells infiltrate the CNS across the blood–brain barrier, leading to demyelination, neuroaxonal damage, the loss of synapses, and reactive gliosis (1, 8, 12). The disruption of neuronal signal transmission then results in clinical symptoms. Approaches to the management of MS include the treatment of acute relapses (13, 14), symptomatic therapies (15), and therapies to prevent relapses and slow the progression of disability (16–18). B cells are central players in the pathogenesis of MS as they mediate cytokine production, antigen presentation, intrathecal antibody synthesis, and the formation of ectopic follicles (19). As new research findings on MS immunopathology further underlined the functional role of B cells, disease-modifying drugs that mediate the depletion of B cells are now commonly used (20–24).

The etiology of MS is still unclear. However, environmental and lifestyle factors, such as smoking, adolescent obesity, and Epstein–Barr virus (EBV) infection, as well as genetic predisposition have been identified as risk factors contributing to the development of MS (25–29). Single-nucleotide polymorphisms (SNPs), the variations of single base pairs at specific genome locations, are the most common type of genetic risk factors (30, 31). Genome-wide association studies (GWASs) have been used to identify associations between SNP alleles and disease. In the most recent GWAS of MS, 233 SNPs have been associated with a significantly increased risk of developing MS [MS-associated lead SNPs (MS SNPs)] (32). However, considering the tendency of proximal SNPs to be inherited together (33), SNPs that are in linkage disequilibrium (LD) with an MS SNP are also associated with MS. Most disease-associated SNPs are considered to have regulatory implications, which means that they are colocalized with quantitative trait loci (QTLs) and thus can affect, e.g., gene expression (eQTL) or alternative splicing (sQTL) (34–38).

Precursor messenger RNA (pre-mRNA) splicing is a physiological process in the cell nucleus by which the introns (intragenic regions) of a pre-mRNA are cut out and the remaining exons (expressed regions) are joined together to form a mature mRNA molecule (39). The cotranscriptional splicing process is coordinated by a complex interplay of cis-elements, trans-acting factors, and the spliceosome complex, which consists of five small nuclear ribonucleoproteins (snRNPs) (40, 41). The important sequences within the pre-mRNA are 5’ and 3’ splice sites (donor and acceptor, respectively), the branch point, the polypyrimidine tract, and exonic or intronic motifs to enhance or silence splicing (42–45). The RNA-binding proteins (RBPs) that recognize these sequences are important for the recruitment of the spliceosome complex. The regulation of the splicing process enables the use of different splice sites, which, in turn, leads to alternative splicing and thus to an altered exon usage compared to the canonical splicing. This allows for the generation of various mRNAs from one pre-mRNA, resulting in a broad transcriptome diversity.

There are five basic types of alternative splicing events (ASEs). While during exon skipping, an exon is excised and not inserted into the mRNA, during intron retention, an intron is not removed and remains in the mRNA molecule. The use of different splice sites can also result in mutually exclusive exons, where only one of two possible exons occurs in the mRNA, or in exons with different lengths due to the use of different acceptor or donor splice sites (46). In addition to the physiological role of alternative splice sites, genetic variants, such as SNPs, can alter the splicing pattern and thereby contribute to the risk of developing diseases (47). As the majority of ASEs in the human EST database are not conserved in mice (48), investigations on the splicing pattern in the experimental autoimmune encephalomyelitis (EAE) mouse model for MS are limited, and thus studies with MS patients are needed. We previously reviewed studies in which ASEs in association with MS have been investigated and found that alternative splicing in MS has been little studied so far (49). The most prominent example is exon 6 skipping in the transcript for the interleukin-7 receptor (IL7R) dependent on SNP rs6897932 (50).

In this study, we investigated ASEs related to SNPs in genetic loci associated with the risk of MS. For this purpose, we used a bioinformatic approach to identify SNPs that potentially alter splicing in MS. We then measured the expression of genes and of individual exons and exon–exon junctions in B cells from MS patients and healthy individuals and analyzed whether the expression is related to MS and/or the SNP. We further used splicing reporter minigene assays to verify alternative pre-mRNA splicing dependent on the genotype of the SNPs. Our study provides new insights into the molecular pathomechanisms of MS by exploring the putative functional role of genetic variants associated with disease susceptibility.

This study is divided into in silico, ex vivo, and in vitro parts ( Figure 1 ). A detailed description of all methods is provided in the supplement ( Supplementary file ).

In this study, we combined in silico evaluations to identify SNPs that may alter pre-mRNA splicing with expression analyses of B cells and with cell culture experiments. We demonstrate that the genotype of SNPs in LD with MS-associated genetic variants can affect pre-mRNA splicing and thus the expression of splice isoforms. We observed an association of the splice SNP genotype with the expression of exons and exon–exon junctions for six SNP–gene pairs (EFCAB13 | rs3851808, GSDMB | rs11078928, HLA-C | rs1131123, SP140 | rs28445040, TSFM | rs10783847, and TSFM | rs2014886) in the microarray data. The differential alternative splicing could be verified by qPCR analyses for EFCAB13, GSDMB, SP140, and TSFM. With our findings for GSDMB and SP140, we could support previous results in the literature showing that the MS-associated SNPs affect alternative splicing (49).

As a starting point, we used various bioinformatic tools to prioritize genetic variants that are likely to alter the pre-mRNA splicing of MS risk genes. We here focused on SNPs located in an exon or within 400 bp of the adjacent intronic regions of these genes. According to previous studies, most splicing factor motifs can be found within this selected 400 bp window (59–61). For the prediction of splicing events due to genetic variants, different tools and databases are available (62–64). We used the Human Splicing Finder to investigate whether a SNP may affect a cis-element such as a branch point, a splice site, or an exonic/intronic splicing enhancer or silencer, and we used the POSTAR2 database to identify SNPs in experimentally determined RBP-binding sites. Finally, we determined 10 SNP–gene pairs (10 SNPs and 8 different genes) for the further event-focused investigations. The reliability of our splice SNP selection procedure was supported by the identification of ASEs for GSDMB | rs11078928, IL7R | rs6897932, and SP140 | rs28445040 as an aberrant genotype-dependent splicing in MS has been previously described for these three SNP–gene pairs (49), and the SNPs are also reported as sQTL SNPs for whole-blood and EBV-transformed lymphocyte samples in the GTEx portal (34). Most of the eight prioritized genes are expressed with a low immune cell type specificity according to the Human Protein Atlas (65). However, two of the genes are expressed more specifically in certain immune cell types: IL7R is expressed mainly in the subsets of T cells and natural killer cells, and SP140 is expressed mainly in memory B cells (66).

Then, we examined the association between the genotype of splice SNPs with the expression of the genes as well as with the levels of individual exons/junctions of the distinct splice isoforms of these genes in B cells from MS patients and healthy controls. Apart from the fact that we did not include SPMS patients, the group of MS patients resembled the typical characteristics of MS patients in European MS registries in terms of age, disease status, and sex (67). In line with the literature (50, 68–70), we observed a significant differential expression of exon 6 of IL7R in MS patients as compared to healthy controls. In our data, the levels of transcripts containing exon 6 were lower in MS patients, but we could not find the previously described association to the MS risk allele C of the non-synonymous splice SNP rs6897932 (T244I). However, the latter might result from the fact that we studied the expression in B cells and not in T cells, in which IL7R is more strongly expressed (66). IL7R encodes for a cell surface receptor for interleukin-7, which plays an essential role for the development and survival of T cells (71). Gregory et al. reported that the C allele of SNP rs6897932 augments an exonic splicing silencer and thus promotes exon 6 skipping, leading to a splice isoform that encodes a soluble form of the protein (50). This is of relevance as increased levels of soluble interleukin-7 receptor have been shown to exacerbate the disease severity in an EAE mouse model, presumably by increasing the activity or bioavailability of interleukin-7 (72). Our analyses of B-cell RNA samples by microarrays and qPCR indicated a genotype-dependent skipping of GSDMB exon 6 and SP140 exon 7. Consistent with our findings, Cardamone et al. (68), Garrido-Martín et al. (73), and Morrison et al. (74) found that the MS risk allele C of SNP rs11078928 affects the acceptor splice site of GSDMB exon 6, resulting in increased exon 6 skipping. The encoded protein Gasdermin-B mediates pyroptosis (75) and, in addition to MS, genetic variants in the GSDMB gene have also been associated with susceptibility to other multifactorial autoimmune diseases like rheumatoid arthritis (76) and ulcerative colitis (77). With regard to the genotype-dependent splicing of SP140, Cardamone et al. (78) and Matesanz et al. (79) could demonstrate via minigene assays that the MS risk allele T of SNP rs28445040 leads to the skipping of exon 7. The function of the protein encoded by SP140 is only partially known. However, the presence of chromatin-related protein domains indicates a role in the chromatin-mediated regulation of gene expression (80). In addition, Karaky et al. reported that SP140 regulates the expression of immune-related genes that are associated with MS (81).

For four other SNP-gene pairs (EFCAB13 | rs3851808, HLA-C | rs1131123, TSFM | rs2014886, and TSFM | rs10783847), we could detect differential alternative splicing in B cells in relation to the MS risk allele. We observed increased EFCAB13 expression and preferential inclusion of exons 9 and 10 in the presence of the MS risk allele C of splice SNP rs3851808. The genotype dependency of this ASE is supported by an sQTL association that is reported for EFCAB13 | rs3851808 for EBV-transformed lymphocytes and other cell types and tissues in the GTEx portal (34). The protein encoded by EFCAB13 contains a calcium-binding domain that is shared by a variety of calcium sensor proteins, which play a role in neuronal function and plasticity (82, 83). Diseases implicated with calcium sensor proteins are, for instance, Alzheimer’s disease (84) and various cancer types (85, 86).

For HLA-C | rs1131123, we observed a trend toward preferential HLA-C intron 2 retention in the presence of the MS risk allele T of the non-synonymous splice SNP rs1131123 (D114A). This genotype dependency was also observed with the minigene assay. HLA-C encodes a class I major histocompatibility complex antigen. Class I molecules play a central role in the immune system and have repeatedly been demonstrated to contribute to the genetic susceptibility to MS (87–89). However, there were challenges in examining the ASE of pre-mRNA from HLA-C: first, in the microarray data, we could only evaluate the expression of the transcript variant in which intron 2 is retained in the mRNA because there are no PSR/JUC probe sets for HLA-C intron 2 exclusion on the employed chip model. Second, only one of the two qPCR assays used to measure transcript splice isoforms of HLA-C provided valid data, which might be due to a sensitivity of the primer pair toward HLA-C subtypes. Since the SNP rs1131123 is not recorded in the GTEx portal (34), further investigations, e.g., with RNA sequencing, could be helpful to ascertain the presumed genotype-dependent splicing of HLA-C intron 2.

We found that exons 3 and 4 of the short transcript variant ENST00000417094 are more frequently skipped in the presence of the MS risk allele C of splice SNP rs2014886 and that this short transcript is only rarely expressed in B cells. In line with our B-cell transcriptome data and minigene assay data, an association of the C allele of SNP rs2014886 with TSFM exon 3 skipping was previously reported by Morrison et al. (74). However, they only studied a small study cohort of eight individuals per genotype. In a recently published report, which focused exclusively on the identification of potential cryptic exons based on literature reports and the dbSNP database, a genotype-dependent splicing of TSFM exon 3 was also postulated (90). Moreover, an sQTL that links the skipping of TSFM exon 3 and 4 with SNP rs2014886 is listed for EBV-transformed lymphocytes in the GTEx portal (34). For the second SNP–gene pair with TSFM, we found that the MS risk allele G of splice SNP rs10783847 showed a strong trend toward TSFM exon 6 and 7 inclusion of the transcript isoform ENST00000550559. TSFM encodes for a mitochondrial translation elongation factor, which catalyzes the exchange of GDP to GTP (91, 92). As the respiratory chain function relies on proper mitochondrial gene expression, differential TSFM expression is associated with various diseases such as encephalomyopathy, hypertrophic cardiomyopathy, and MS (93–96). Noteworthy, Alcina et al. (96) described that SNP rs10877013 affects TSFM expression in MS by altering the enhancer activity of a regulatory element. This SNP is in almost perfect LD with the two splice SNPs rs10783847 and rs2014886 in the European population (33). Further studies are needed to better understand the functional role of the different splice isoforms of TSFM in relation to the pathogenesis of the multifactorial disease MS. According to the Ensembl database, ENST00000417094 codes for an 89 amino acid long protein sequence (UniProt F8WCK2) but it is likely a target of nonsense-mediated decay. However, experimental evidence remains to be established.

The following limitations should be considered when interpreting the data of this study. First, due to the stringent restrictions on the selection of experimentally screenable SNP–gene pairs, it is possible that we have missed some MS-specific ASEs. For instance, we did not include rare variants (minor allele frequency < 1%) because the sample size would be insufficient to study such variants. In addition, we focused only on SNPs in or near exons and thus did not capture the potential influence of deep intronic SNPs on splicing. Such variants have been described for other diseases (61, 97–100). Second, some genomic regions are characterized by long-range LD. Hence, the observed effects on splicing may not represent the only effects underlying the genetic associations with MS. Third, in the analyses of differences in gene expression and alternative splicing, we cannot exclude the possibility of confounding variables, e.g., medical treatment and comorbidities. Specifically, we observed a shift in the proportions of B-cell subsets in patients treated with alemtuzumab or cladribine (53). This contributed to the variance in the gene expression data. Fourth, we conducted our measurements in B cells and therefore may have missed or underestimated the differential alternative splicing of genes that are more abundantly expressed in other cell types (101). Even though genetic effects on splicing are usually highly shared across tissues and cell types (34), further insights into the effects of genetic risk variants could be obtained by studying other cell types, e.g., other peripheral immune cells such as T cells. Fifth, our analysis of the microarray data relied on transcript isoforms as annotated in the reference genome. Thus, we studied known splice isoforms and may have missed novel splicing patterns, which can potentially be identified by using RNA sequencing (102, 103). Fifth, as we have previously described (46), there are issues regarding the use of the minigene assay system, such as a possible interference by the Gateway cloning attachment sites, an insufficient amount of an important splicing factor in the used cell line, or the fact that only a small and specific part of the gene is examined. The latter may lead to the misinterpretation of ASEs as the splicing of exons can depend on the correct splicing of other exons of the gene that are not included in the minigene construct.

In conclusion, in this study, we focused on SNPs located in genetic risk loci for MS that presumably affect pre-mRNA splicing and thus may have an influential role in the pathogenesis of the disease. We were able to support findings from previous studies on MS-related ASEs for the pre-mRNAs of GSDMB, IL7R, and SP140. For four novel SNP–gene pairs, we found an association of the splice SNP genotypes with differential alternative splicing in the B-cell transcriptome data. Except for two SNP–gene pairs, we were able to validate the findings of the microarray data analysis with the qPCR assays. In addition, we were able to further substantiate our observations from the B-cell expression data on TSFM exon 3 skipping by using minigene assays. The MS risk allele C of the SNP rs2014886 almost always led to TSFM exon 3 skipping, whereas the alternative allele led to a low expression of ENST00000417094 transcripts. However, the potential functional impact of this ASE remains unclear. Further functional studies are needed to identify the disease-causing genetic variants and to explore their effects on splicing and the resulting consequences of an aberrant expression of splice isoforms to improve our understanding of the molecular pathomechanisms of MS.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary Material .

The studies involving human participants were reviewed and approved by the Ethics committee of the University of Rostock. The patients/participants provided their written informed consent to participate in this study.

EP performed the bioinformatic prioritization of SNP-gene pairs as well as the majority of the experiments. EP evaluated and interpreted the obtained data. Furthermore, EP drafted the manuscript and prepared the figures and tables. MH provided support for the bioinformatic and statistical analyses of the data and critically revised the manuscript. MS coordinated the sample collection and gathered the clinical-demographic data. EP, NB and BF processed the blood samples. NB performed the SNP genotyping. DK was responsible for performing the microarray measurements. PL provided valuable advice concerning the splicing reporter minigene assays. MH, BF and UKZ conceptualized the study. MH, EP, NB and UKZ secured research funding. UKZ provided important intellectual insights and supervised the research. All authors contributed to the article and approved the submitted version.

EP received a scholarship from the Landesgraduiertenförderung Mecklenburg-Vorpommern. NB was funded by the Stiftung der Deutschen Wirtschaft (sdw). Research funding for this work was provided by the Rostock University Medical Center (FORUN program, grant 889034). In addition, this study was partially funded by Sanofi Genzyme (grant GZ-2016-11560) and Merck Serono GmbH (Darmstadt, Germany, an affiliate of Merck KGaA, CrossRef Funder ID: 10.13039/100009945, grant 4501860307). The funder was not involved in the study design, collection, analysis, interpretation of data, the writing of this article or the decision to submit it for publication.

We thank Alexander Winkelmann, Stefanie Meister, and Ales Dudesek for patient care and Antje Bombor and Ina Schröder for coordinating the clinical visits and blood draws. We wish to thank Nele Retzlaff and Deborah Sonnenberg for reviewing the medical records. Our thanks go to Hans-Jürgen Thiesen, and Robert Jaster for providing access to technical equipment. We thank the Core Facility for Cell Sorting and Cell Analysis headed by Brigitte Müller-Hilke for providing the FACS equipment and technical support. We are grateful to Ildikó Tóth for laboratory support and to Felix Steinbeck for his input on computational tools for analyzing genetic data.

EP received travel funds from Novartis. MH received speaking fees and travel funds from Bayer HealthCare, Biogen, Merck, Novartis, and Teva. UKZ received research support as well as speaking fees and travel funds from Alexion, Almirall, Bayer HealthCare, Biogen, Bristol Myers Squibb, Janssen, Merck Serono, Novartis, Roche, Sanofi Genzyme, and Teva as well as EU, BMBF, BMWi, and DFG.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.