Introduction

Multiple sclerosis (MS) is an inflammatory disease of the central nervous system characterised by demyelination and axonal loss.[1] Studies conducted in mono- and dizygotic twin pairs and siblings have shown that genetics plays a role in MS susceptibility.[2] Linkage was effective in identifying the locus exerting the single strongest genetic effect in MS, namely, the human leukocyte antigen (HLA) class II region.[3] The risk associated with this region has since been shown to be determined by epistatic interactions between different HLA alleles,[4] and is thought to be responsible for approximately 50% of the genetic risk of MS.[5] Beyond this powerful determinant of MS genetic susceptibility, research has taken considerably longer to bear useful fruit. Finally, after the genotyping of hundreds of thousands of single nucleotide polymorphisms (SNPs) in many thousands of MS patients and controls, we are beginning to establish a network of loci outside of the HLA region involved in determining MS susceptibility.[6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17] It is worth considering that even the most strongly associated of these with MS is still a significantly weaker determinant of MS susceptibility than HLA alleles. For some of these loci, functional studies have been undertaken.[11], [18], [19] However, such studies are rarely carried out in an unbiased manner since these generally correlate genetic variations with the expression of a candidate gene. A recent study of mRNA levels in MS patients and healthy controls showed a great multitude of differentially expressed genes however it is uncertain to what extent this reflects the aetiology of disease as opposed to the disease process or adaptive biological pathways.[20]

A recent investigation has performed whole genome expression analysis in lymphoblastoid cell lines (LCLs) from healthy volunteers who were also genotyped for a large number of SNPs.[21] We used the data from this study to examine the effects of current susceptibility loci in MS on gene expression.

Methods

Results

SNP selection

We chose to look at a set of 38 SNPs which were the top loci to reach genome-wide significance selected from currently reported genome wide association studies (GWAS) of which 14 had been independently replicated in 2 studies. 17 of these were not present in the genome-wide association mRNA expression library and so when possible proxy SNPs in strong-to-moderate LD were used instead. The SNPs and proxy SNPs used are detailed in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. SNPs and proxy SNPs analysed.

https://doi.org/10.1371/journal.pone.0010142.t001

Discussion

Our findings show that some, but by no means all, susceptibility SNPs in MS are associated with changes in gene expression. Some of these (CD58, had already been noted by previous investigators.[11] We were unable to find supporting evidence in this dataset for the previously reported allelic effect of the susceptibility SNP in IL7R on expression of the gene.[18] Similarly, SNPs in the IL2RA gene did not correlate with expression of IL2RA mRNA, despite previously finding altered levels of this in MS patients relative to controls.[19]

We also found several novel effects of susceptibility SNPs. Two SNPs in tight LD with susceptibility SNPs in the HLA region correlated with expression of several HLA class II mRNAs. However, measuring gene expression in the HLA is a complex task. There is haplotype specificity for some genes (HLA-DRB4 and HLA-DRB5) and thus we are not sure whether differential expression of HLA genes measured by microarray reflects different probe affinity for different HLA class II alleles and thus further work is needed to fully understand this association. Our identification that a SNP in the HLA class I region was associated with altered expression of ZFP57 is an interesting observation as this gene has been linked with DNA methylation changes across the genome resulting in transient neonatal diabetes.[22] There is some epidemiological evidence that MS may be partly determined by epigenetic alterations and this would be an ideal candidate functionally linking MS to the epigenome.[23] A SNP in IL7 recently confirmed as associated with MS was shown to correlate with the expression of several genes: IL7 and FAM164A. Naturally the most compelling candidate of these is IL7 due to its probable role in autoimmunity. However, the advantage of an unbiased screen is that it raises the possibility of candidate genes that would otherwise not be considered. This is especially so since the SNP is far more strongly associated with FAM164A expression than with IL7. FAM164A is a hypothetical protein encoded in the reverse direction to IL7 and its functional importance is largely unknown.[24] The susceptibility region on chromosome 12 was previously linked with the expression of FAM119B.[20] We feel that the relationship of all three major susceptibility SNPs with the expression of TSFM suggests this as a strong candidate. This is a plausible candidate in terms of function too as it is involved in the translation of mitochondrial proteins, providing a potential link with other susceptibility genes linked to mitochondrial function, such as KIF21B.[8], [25] Further functional work will be needed to better assess these candidates.

The limited cross-over between known and suspected susceptibility genes in the whole genome expression analysis of Gandhi and colleagues is likely due to a number of differences including the use of whole blood mRNA and individuals with established disease in the Gandhi study.[20] It is possible that future whole genome analyses of expression conducted using RNA-seq in cell-sorted samples of patients with very early disease may reveal alterations in the level of susceptibility gene mRNA.

The advantage of an unbiased approach to linking the expression of genes with genetic variation associated with disease susceptibility is that there is no a priori hypothesis to blind investigators to the presence of other genes. There are several limitations to the approach we used. The mRNA screen was conducted in transformed LCLs and so it would not be informative about tissue-specific gene expression.[21] Also, SNP coverage across the genome was not complete and so the functional effects of some SNPs for which no proxy was available will be concealed. Furthermore, despite using expression data from 400 LCLs, we may have been underpowered to detect relevant effects. However, our finding of several novel associations between MS SNPs and gene expression is worthy of further investigation and also raises the hypothesis that some disease associated SNPs may not exert their effects on MS susceptibility through simple effects on gene expression.

Supporting Information