Revealing the genetic basis of multiple sclerosis: Are we there yet?

Abstract

For more than 30 years the only genetic factor associated with susceptibility to multiple sclerosis (MS) was the HLA region. Recent advancements in genotyping platforms and the development of more effective statistical methods, resulted in the identification of 16 more genes by genome-wide association analyses (GWAS) in the last 3 years alone. While the effect of each of these genes is modest compared to that of HLA, this list is expected to grow significantly in the near future, thus defining a complex landscape in which susceptibility may be determined by a combination of allelic variants in different pathways according to ethnic background, disease sub-type, and specific environmental triggers. A considerable overlap of susceptibility genes among multiple autoimmune diseases is becoming evident and integration of these genetic variants with our current knowledge of affected biological pathways will greatly improve our understanding of mechanisms of general autoimmunity and of tissue specificity.

Introduction

Autoimmune disorders arise when physiological tolerance to “self” antigens is lost. Although several mechanisms may be involved in this pathogenic process, dysregulation of T-cell and B-cell activation and of pathways leading to inflammation are logical candidates. Multiple sclerosis (MS) is a common inflammatory disorder of the central nervous system (CNS) characterized by myelin loss, gliosis, varying degrees of axonal pathology, and progressive neurological dysfunction. A large body of research supports a multifactorial etiology, with an underlying genetic susceptibility likely acting in concert with undefined environmental exposures [1–3]. Although the exact pathogenic mechanisms underlying MS remains unknown, it has been proposed that lymphocytes activated in the periphery by a microbial mimic home to the CNS, become attached to receptors on endothelial cells, and then proceed to cross the blood-brain barrier (BBB) directly into the interstitial matrix. T cells are then reactivated in-situ by fragments of myelin antigens exposed in the context of human leukocyte antigen (HLA) molecules on the surface of antigen presenting cells (macrophages, microglia, and perhaps, astrocytes). Reactivation induces the release of proinflammatory cytokines that open further the BBB and stimulate chemotaxis, resulting in a second, larger wave of inflammatory cell recruitment and leakage of pathogenic antibodies and other plasma proteins into the nervous system.

Susceptibility to MS has been associated with multiple factors including genetics, epigenetics, and the environment. While the modest concordance rate in monozygotic twins (30–35%) suggests that environmental factors are major players in MS, increased heritability within families and the decrease in risk with degree of relatedness all argue in favor of genetic factors. Modern genomics developments have made available miniaturization and automation of genotyping platforms, and more than 200 genome-wide association studies (GWAS) have been performed in different diseases to date [4,5] including 31 studies in 7 common AID (7 in MS alone). In this chapter the findings on these studies will be summarized, and hypotheses about the possible pathogenic mechanisms implicated in MS will be elaborated.

The role of the HLA system in MS

The HLA region has been associated with hundreds of human diseases, including most autoimmune diseases (e.g. B27 with Ankylosing spondylitis, DR3 with Grave’s disease, Myasthenia Gravis, systemic lupus erythematosus (SLE), and Type 1 Diabetes (T1D); DR15 with MS) (for a review, see [6]). For most of these diseases, however, it has not yet been possible to show the molecular mechanisms underlying disease association with a particular HLA molecule(s). One possible explanation for this shortcoming is that it has often been difficult to unequivocally ascertain the primary disease-risk HLA gene(s) due to the remarkably strong linkage disequilibrium across the HLA [7]. There is still debate, for example, as to whether the HLA-DRB1*15:01 association explains the entire HLA–class II genetic linkage signal in MS [8–12] and whether susceptibility genes also exist within the class III region[13] and/or are telomeric to the class I region[14–16]. In addition, MS could result from the combination of different HLA molecules expressed at various loci (class-I and/or class-II) rather than the result of one HLA variant only. The situation has been complicated by the fact that susceptibility to MS is clearly polygenic and particular HLA allele(s), in combination with other genetic variants and environmental factors, may be required to develop the disease. Moreover, disease-relevant autoantigen(s) are largely unknown, which prevents a thorough three-dimensional analysis of HLA–peptide interactions. Last, but not least, MS is phenotypically very heterogeneous, in terms of clinical presentation, age at onset, co-morbidities with other autoimmune disorders, and severity, and it is likely that different alleles, or allelic combinations at different loci, will predispose to different forms of the disease [17].

Gene discovery in MS

Linkage studies

The strategy for a genetic linkage study requires the collection of family pedigrees with more than one affected member to track the inheritance and use highly polymorphic genetic markers to identify discrete chromosomal segments that deviate from independent segregation and co-segregate with the trait. Linkage screens with different levels of resolution and genome coverage have been completed in more than 30 datasets of familial MS cases [18]. Each of these studies suggested multiple chromosomal regions with potential involvement in disease susceptibility, consistent with the long-held view that MS is a polygenic disorder. However, only the HLA-class II locus on chromosome 6p21.3 has ever exceeded the threshold for formal statistical significance (a LOD>3 is considered proof of linkage) and no other region of statistically confident linkage has ever been observed. Using the values of HLA allele sharing by descent in sibships, it has been estimated that the HLA locus accounts for 20–60% of the genetic susceptibility in MS [19]. Even at the higher end of this estimate, much of the heritability of MS remains to be explained.

An International Multiple Sclerosis Genetics Consortium (IMSGC) reported in late 2005 the results of a linkage screen in 730 multi-case MS families using more than 4500 SNPs [20]. The peak logarithm of the odds (LOD) score of 11.7 found in the HLA illustrates the substantially greater power achieved by this high-density screen compared with earlier efforts. Strikingly, however, although numerous regions on multiple chromosomes revealed possible linkage signals of interest, no other locus reached genome-wide significance. These data indicate that any susceptibility allele common in the population and outside the HLA region is likely to increase MS risk by a factor of less than of 2. Similar findings have been reported for other autoimmune diseases.

Genome-wide association studies (GWAS)

The aim of a GWAS is to characterize the genetic architecture of a complex genetic trait through the identification of disease variants against the background of random variation seen in a population as a whole. In a typical study, hundreds of thousands of markers covering a significant portion of the common variation in the population are tested simultaneously in cases and controls and the allelic frequencies of each marker are compared between the two groups. Compared to linkage, association studies have greater statistical power to detect common genetic variants that confer a modest risk for a disease [21•]. Although population stratification and the introduction of phenocopies due to poor ascertainment of study participants are important confounding factors in association studies, inadequate sample size is undoubtedly the main reason why most published claims of association are suspected type I errors [22]. Hence, the identification of genes influencing the development of MS needs to rely primarily on association-based methods and must involve very large patient cohorts.

A total of seven positive GWAS have been reported to date in MS. The classic HLA-DRB1 risk locus stood out in all studies with remarkably strong statistical significance (e.g. P < 1×10⁻³² in [23]). These studies were followed by extensive replications efforts of the top hits [24–28] that together with a comprehensive replicated meta-analysis [29] provided robust evidence for approximately a dozen non-HLA loci affecting disease susceptibility (Table 1).

Table 1.

GWAS in MS

Study	Year	Platform [SNPs passing QC]	Initial Sample Size	Replication Sample Size	Region	Gene(s)	SNPs	p-Value	OR	Reference
Nischwitz et al.	2010	Illumina [~300,000]	590 cases 825 controls	NR	10q11.21 19p13.2 9q34.2 6p21.32	Intergenic ZNF433 VAV2 DQA1	rs2503875 rs3745672 rs3780792 rs9271366	2.00E-07 1.00E-06 1.00E-06 4.00E-17	1.66 7.39 1.6 2.62	[61]
Sanna et al.	2010	Affymetrix [6,607,266] (imputed)	882 cases 872 controls	1,775 cases 2,005 controls	6p21.32 3q13.11	HLA-DRB, HLA-DQB1 CBLB	rs2040406 rs9657904	1.00E-20 2.00E-10	2.05 1.4	[62]
Jakkula et al.	2010	Illumina [297,343]	68 cases 136 controls	4,570 cases 10,143 controls	17q21.2 6p21.32	STAT3 HLA	rs744166 rs3135338	3.00E-10 2.00E-25	1.15 3.43	[63]
Bahlo et al.	2009	Illumina [302,098]	1,618 cases 3,413 controls	2,256 cases 2,310 controls	10p15.1 20q13.12 16p12.1 1p22.1 1p13.1 6p21.32 8q24.21 12q14.1	IL2RA CD40 NR EVI5, RPL5 CD58 HLA-DRB1 ASAP1, DDEF1 METTL1, CYP27B1	rs2104286 rs6074022 rs8049603 rs6604026 rs1335532 rs9271366 rs6984045 rs703842	7.00E-06 1.00E-07 1.00E-06 3.00E-06 1.00E-07 7.00E-184 2.00E-06 5.00E-11	1.16 1.2 1.19 1.17 1.28 2.78 1.59 1.23	[64]
De Jager et al.*	2009	Affymetrix & Illumina [~2.56 million] (imputed)	2,624 cases 7,220 controls	2,215 cases 2,116 controls	6p21.32 1p13.1 6p22.1 3q25.33 10p15.1 12p13.31 2q22.1 12p13.31 16p13.13 5p13.2 16q24.1 11q12.2 5p13.1 12q24.31 10q22.3	HLA-DRB1 CD58 HLA-B IL12A IL2RA TNFRSF1A CXCR4 TNFRSF1A CLEC16A IL7R IRF8 CD6 PTGER4 MPHOSPH9 ZMIZ1	rs3135388 rs2300747 rs2523393 rs4680534 rs2104286 rs4149584 rs882300 rs1800693 rs11865121 rs6897932 rs17445836 rs17824933 rs6896969 rs1790100 rs1250540	4.00E-225 3.00E-10 1.00E-17 6.00E-06 9.00E-08 5.00E-06 1.00E-07 2.00E-11 2.00E-07 2.00E-06 4.00E-09 4.00E-09 2.00E-07 7.00E-07 2.00E-06	2.75 1.3 1.28 1.12 1.15 1.58 1.19 1.2 1.15 1.12 1.25 1.18 1.1 1.11 1.12	[29]
Baranzini et al.	2008	Illumina [551,642]	978 cases 883 controls	NR	13q31.3 9p22.2 8p23.2 12q12 3q23 4q35.1 20p13 2p25.1 3q24 2q14.2	GPC5 SH3GL2 CSMD1 PDZRN4 SLC25A36 MGC45800 C20orf46 DDEF2 ZIC1 EN1	rs9523762 rs1755289 rs1529316 rs1458175 rs908821 rs7672826 rs397020 rs1109670 rs1841770 rs651477	1.00E-06 3.00E-06 2.00E-06 2.00E-06 3.00E-06 8.00E-06 8.00E-07 9.00E-06 8.00E-06 7.00E-06	1.36 1.35 1.36 1.34 1.37 1.37 1.41 1.38 1.34 1.38	[23]
Comabella et al.	2008	Affymetrix[428,867] (pooled)	242 cases 242 controls	375 cases 375 controls	6p21.32	HLA-DRB1	rs3129934	9.00E-11	3.3	[65]
Hafler et al..	2007	Affymetrix [334,923]	931 trios 2,431 controls	609 trios 2,322 cases 2,987 controls	9q33 6p21.32 16p13.13 10p15.1 5p13.2 1p22.1	DBC1 HLA-DRA KIAA0350 IL2RA IL7RA RPL5	rs10984447 rs3135388 rs6498169 rs12722489 rs6897932 rs6604026	8.00E-06 9.00E-81 4.00E-06 3.00E-08 3.00E-07 8.00E-06	1.17 1.99 1.14 1.25 1.18 1.15	[66]

^*Meta-analysis

It is important to note that these markers may not represent necessarily the causal disease variant themselves, explaining in part the very modest independent odds values for each allele. Additional follow-up experiments refined some of the association signals and revealed early mechanistic insights into the functional role of the identified genes, most notable a change in the soluble vs. membrane-bound ratio for the IL2 and IL7 receptors [30–32••] and diminished expression of CD58 mRNA [31]. It is noteworthy that some of the MS allelic variants were also proposed to be involved in other autoimmune diseases, suggesting common mechanisms underlying different autoimmune conditions [28,33–35••]. For example, the IL2RA-mediated susceptibility effect is shared among MS, T1D, Graves’ disease, and rheumatoid arthritis (RA) [32]. Interestingly, the direction of the association is not consistent across diseases as the IL2RA allelic variant associated with susceptibility to MS appears to confer resistance to T1D. A second allele confers susceptibility to both diseases, whereas yet a third allele is associated with susceptibility to T1D only [32,36]. Similarly, while the minor allele of the R620W polymorphism in PTPN22 has been associated with susceptibility to T1D, RA [37–40], and SLE [41,42], it appears to confer protection to CD [43]. Functional studies aimed at detecting tissue (or cell) specific variation in the expression or function of the target gene may contribute to elucidate the pathogenic mechanisms operating in each AID [36].

Altogether, the GWAS data seems to support the long-held view that MS susceptibility is conferred by the action of common (i.e. those with a risk allele frequency of >5%) sequence allelic variants in multiple genes [44]. Even with this expanding roster of risk loci, much of the heritability of MS remains unexplained. For example, the sorting and classification of duplications or deletions of genomic segments generating copy number variants (CNVs) lag behind. CNVs are a major source of human genetic diversity and have been shown to influence rare genomic disorders as well as complex traits and diseases [45–47]. Recent studies have reported associations between CNV and several autoimmune conditions such as SLE, psoriasis, CD, RA and T1D [48].

The second phase of the Wellcome Trust Case Control Consortium (WTCCC2), a massive collaborative project that is genotyping a total of 120,000 samples in MS and 12 other diseases (including ankylosing spondylitis, ulcerative colitis, and psoriasis) is now near completion [49]. The MS component of this project includes 10,000 cases genotyped with a high-SNP/CNV density platform. Such study will be adequately powered to identify common risk alleles with odds ratio as modest as 1.2 [50,51] thus practically guaranteeing identification of the complete set of common variants involved in MS.

Comparative analysis of susceptibility genes for MS and other autoimmune diseases

In most GWAS the number of markers in which the evidence for association exceeds the genome-wide significance threshold is small, and markers that do not exceed this threshold are generally neglected. A plausible strategy to increase the prior odds of finding a true significant marker is to incorporate prior biological knowledge into the GWAS data in the form of gene ontologies or pathways [52–56•]. The advantage of these methods is that even if markers in individual genes do not reach genome-wide significance, several modest associations in genes from the same biological pathway may highlight collectively its involvement in a disease process. Building on this rationale we merged statistical evidence from GWA analyses with experimental evidence of protein interaction from yeast two hybrid or chromatin immunoprecipitation (ChIP) studies to discover sub-networks (or modules) of interacting proteins associated with MS susceptibility [57]. Through this approach we were able to identify novel susceptibility pathways in MS, such as axon guidance and glutamate metabolism. We also uncovered genetic overlaps between MS and Alzheimer’s disease and bipolar disorder. In addition, the presence of common variants in the MHC region between MS, RA, and T1D, but not T2D or CD with MHC alleles was highlighted.

In order to systematically study genetic commonalities among autoimmune disorders we used publicly available information [5,58] to extract all moderately significant (p<10⁻⁴) associations from studies in celiac disease (CeD), Crohn’s disease (CD,) MS, psoriasis, RA, SLE, and T1D (plus those in T2D). Altogether, 1,201 genes with modest evidence for association in at least one of these autoimmune disorders were identified. We then used a network-based approach to visualize all the reported associations at once and identified 71 non-MHC genes shared by at least two diseases (Figure 1), 7 by three diseases, and only 2 by four diseases (Table 2) [33••]. This select list of potentially “general” autoimmunity genes includes PTPN22, a tyrosine phosphatase strongly associated with T1D (aggregate p<10⁻²²⁶), RA (aggregate p<10⁻⁹⁰), and to a lesser extent with CD (aggregate p<10⁻⁸). TNFAIP3 is highly associated with RA (aggregate p<10⁻²⁰), SLE (aggregate p<10⁻¹¹), Ps (aggregate p<10⁻¹¹), and moderately associated with CD (aggregate p<10⁻⁵). Other general autoimmunity genes are IL23R, KIAA1109 and CTLA4. Altogether, the data presented here suggests that genes involved in activation, proliferation, and homeostasis of cells involved in adaptive immune responses are more likely to represent general autoimmunity genes. This is further supported by the observation that a large proportion of these genes physically interact among each other (S. Baranzini, unpublished observation), thus possibly taking part in the same or highly overlapping biological pathways. A corollary to this observation indicates that associations unique to each disease would be responsible for attracting immune responses to specific tissues, although more functional studies in animal models will be needed to confirm this.

Figure 1. Autoimmune disease-gene network.

Top genetic associations in 7 autoimmune diseases and T2D. The most significant SNP per gene was selected. Only associations with significance of at least p<10⁻⁴ are visualized. If a given gene was identified in more than one disease, multiple lines connecting it with each disease were drawn. Lines are colored using a “heat” scheme according to the evidence for association. Thus “hot” edges (e.g. red, orange) represent more significant associations than “cold” edges (e.g. purple, blue). Diseases are depicted by circles of size proportional to the number of associated genes, non-MHC genes by grey triangles. To facilitate visualization, only genes shared by at least two diseases are shown.

Table 2.

Genes shared by at least three diseases at (aggregate) p< 10⁻⁴

Gene	Chr	Description	Crohn’s	RA	T1D	Celiac	MS	SLE	Ps
PTPN22*	1	Protein tyrosine phosphatase, non-receptor type	10−8	10−90	10−226			10−5
IL23R	1	Interleukin 23 receptor	10−102	10−4					10−7
NRXN1	2	Neurexin 1 isoform beta precursor	10−4	10−4			10−4
KIAA1109	4	Hypothetical protein LOC84162		10−5		10−12	10−5
EPHA7	6	Ephrin receptor EphA7		10−4	10−4	10−5
TRIM27	6	Tripartite motif-containing 27	10−5	10−6	10−11
TNFAIP3*	6	Tumor necrosis factor, alpha-induced protein 3	10−5	10−20				10−11	10−11
TNKS	8	Tankyrase	10−4				10−4	10−6
C20orf42	20	Fermitin family homolog 1			10−5	10−4	10−4

^*Genes involved in 4 diseases.

The next steps in MS genetics research

Despite the success of GWA studies in identifying novel susceptibility loci that withstood the challenge of independent replication, many questions remain concerning the genetic architecture of MS [59]. Noteworthy, all of the reported associations are derived from microarray-based studies, where only common variants (minor allele frequency > 10%) are interrogated in GWAS. Thus, the almost certain influence of evolutionarily younger (rare) variants has not been adequately evaluated. With the advent of next generation sequencing, more data is expected to be gathered in the near future to address this important question. The 1000 genomes project (www.1000genomes.org), is an international research consortium with the goal of finding most genetic variants that have frequencies of at least 1% in several populations around the world. Data emerging from this effort will undoubtedly uncover a multitude of private variants giving rise to a new catalog of human variation, an invaluable resource in the search for novel disease associations.

In a modest, but important first step to assess this hypothesis we recently sequenced the entire genomes of a female MS-discordant MZ twin pair, generating over one billion, high quality, shot gun whole-genome reads, corresponding to approximately 22 fold coverage of each genome [60•]. These are the first female, twin, and autoimmune genome sequences reported to date. Among ~3.6 million single nucleotide polymorphisms (SNPs), ~200,000 insertion-deletion polymorphisms (indels), 27 copy number variations (CNVs) and 1.1 million mCpG dinucleotides detected, no SNPs, indels or CNVs and methylation of cytosine residues differed between the twins. We also analyzed the full mRNA transcriptome and epigenome sequences of CD4+ lymphocytes from 3 pair MZ, discordant twins for MS. While 19,000 genes were expressed in each of the CD4+ T cell preparations, no reproducible transcriptional differences were identified between MS-affected and unaffected twins. Surprisingly, only 2–176 differences in methylation of ~2 million CpG dinucleotides were detected between siblings in three twin pairs, in contrast to ~800 methylation differences between T cells of unrelated individuals and several thousand differences between tissues or normal and cancerous tissues. The sequence of many more discordant twin pairs, patients and control genomes is necessary to realize the power of this approach. Given the impressive rate of advances in the field, the technology for whole genome sequencing will be accurate and inexpensive enough for large-scale application within the next few years.

Conclusions

The genetic bases of autoimmune diseases are just starting to be uncovered. While several bona-fide susceptibility genes have been identified in most common traits, technological advances in high throughput genotyping platforms and more affordable second generation sequencing methods will contribute to significantly expand these lists. In addition, structural variants (insertion/deletion polymorphisms, copy number variations, etc) are also likely to play a significant role in determining susceptibility to AID. Taking into account the known susceptibility loci for each trait, disease-specific custom genotyping chips will be designed so as to cover a wide spectrum of variants in larger cohorts of individuals. At the same time, deep sequencing of candidate regions will be carried out to identify rare (private) mutations and structural variants that affect only a few individuals. Together, these approaches will eventually discover most if not all of the genetic contribution to these diseases and allow for the systematic search of similarity and differences among them.

While the identification of the precise pathways involved in susceptibility to AID will clearly require additional time and effort, integration of data from multiple diseases represents the logical next step in discovering similarities and differences among them. While similarities will shed light into the general mechanism behind autoimmunity, genetic features private to a given disease will help pinpoint the basis for tissue specificity. One obvious benefit of this new knowledge would be the cross-utilization of drugs for diseases with similar genetic fingerprint. Ultimately, this high resolution genetic disease landscape will contribute to more accurate models of pathogenesis setting the bases for the development of more rational therapeutic approaches.