One hundred and forty-nine SNPs were genotyped and analyzed across the 16p13 region for MS susceptibility. The CEU LD data for this region of chromosome 16 suggests that there are three distinct LD blocks (Fig. ). After identifying the previously replicated SNP with the most significant association to MS (strongest P-value) from each of the three sub-regions, we further investigated the region for interaction effects. All of the nominally significant pairwise interaction models (Table ) had significant main effects and the significance of the interactions disappears with correction for multiple testing. In addition, we also observed significant correlation of expression probes between DEXI, CLEC16A and SOCS1 (Table ), which could indicate a common regulatory element. Although the HLA locus contributes more to MS risk than any other locus, our work again confirms that the chromosome 16p13 region independently contributes to MS risk. While the change in risk conferred by this region is small compared to that of HLA, further study of the 16p13 genes is likely to give insight to the less well understood, non-HLA-mediated cellular events that augment MS pathology.
The patterns of LD and the tests of independence in the logistic regression model (by adjusting for an interaction term) both suggest that each of these SNPs (and hence their tagged genes) has an independent effect(s) on MS susceptibility. Furthermore, these data suggest that there are interactive effects within this 600 kb region that also contribute to MS susceptibility.
We were surprised to discover that several SNPs were eQTLs for PRM1
(and one for PRM2
) in several different non-testes cell lines, despite the likely testes-specific expression and function of these in vivo
and obvious ‘repressive' chromatin marks on the PRM
genes in cell lines. We speculate that the region between CLEC16A
contains a regulatory domain that influences expression of multiple genes in this region and that genetic variants in this region alter MS risk through expression of more than one gene. The long-range nature of this cis
-regulation might influence low levels of ‘leaky' PRM1
transcription in cultured cell lines, explaining the eQTL effect on PRM1
. Alternatively, some of the eQTL signals could reflect indirect (trans) effects on gene expression. Other SNPs in this region are shown to be associated with other autoimmune disorders (22
Because of the differential associations with HLA strata, we have identified four sub-regions of independent association in this genomic region. The Subregion I variant(s) may be functioning through CIITA exclusively, as this area has little LD and no evidence for eQTL effects (Figs and ). Likewise, Subregion II is contained with the CLEC16A transcription unit and in an LD block associated with this gene. In contrast, Subregion IIIa shares little LD with the structural genes in the region, but contains SNPs with eQTL effects on several genes in the region. The role of Subregion IIIb is unclear, although it is doubtful that it acts via altering PRM gene cluster expression as expression is likely to be tightly restricted to testes in vivo. More likely, Subregion IIIb variants affect cis-regulation of or are simply in LD with one or more SNPs that act through SOCS1 or C16orf75. Therefore, further searches for undiscovered variants in these genes may be warranted.
Figure 4. Diagram of genetic signals and genes in the chromosome 16 MS complex. The four regions that harbor distinct genetic signals (black bars) are labeled with roman numerals. DEXI, CLEC16A and SOCS1 are colored green to reflect their correlated expression (more ...)
Because this is a small chromosomal region, it is reasonable to hypothesize that some regulatory mechanisms for separate genes have co-evolved. Sets of genes that are highly conserved across species could support a model for some degree of common regulatory control and thus, expression of several or all genes in this region.
In addition to comparing expression, searching for known regulatory regions could also be influential in the search for the mechanism of disease pathology. If any other SNPs with significant P
-values are in LD with a known regulatory region, it would suggest possible co-regulation of these genes. These effects could be located in regulatory elements that affect the transcription of the whole region, similar to the locus control regions that affect globin gene cluster expression (24
, the genes with the most significant SNPs in the LD block containing SOCS1
, do not appear to have a biological function relevant to known MS etiology (25
). However, these SNPs, or SNPs in LD with them, could be involved in regulating transcription of SOCS1,
which is a more logical candidate gene for MS pathogenesis. This is especially interesting given the previously shown role of SOCS1
in regulating inflammation. SOCS1
deficiency in T cells leads to hyperactivity in Th1 cells and a reduction in Th17 response, which has been linked to the pathogenesis of autoimmune disease (26
). In addition, SOCS1
deficiency causes the breakdown of self MHC-T cell receptor recognition during thymic differentiation. This process leads to a release of autoreactive T cells (27
). This mechanism could explain the breakdown in the normal control of T cell differentiation leading to T-cell hyper-reactivity, which may ultimately contribute to the attack on the myelin sheath.
Some differences in gene expression across this region argue against co-regulation. For example, CIITA
expression in lymphoblastoid cells does not correlate with that of other genes in the region. On the other hand, the expression-correlated genes DEXI, CLEC16A
lie in a contiguous ‘active' region of H3K27 methylation, and the eQTL data may support cis
-effects that act across gene boundaries (e.g. LITAF) (Supplementary Material, Table S1
The confluence of significant results on this arm of chromosome 16 has raised important questions. We have demonstrated that there are multiple independent signals in this genomic region. In various cell lines that are publicly available, these genes are not expressed together despite the conserved synteny of the 600 kb region. If regulatory elements located here (e.g. the SOCS1 region) overlap with MS-associated SNPs, they might influence expression of the nearby more biologically plausible gene, SOCS1. As there is no direct evidence for gene co-regulation, more studies are needed to confirm a model for SNPs that regulate multiple genes in cis. Further studies of expression and genotype correlations in additional lymphoblastoid cell lines and/or T cells from MS patients and controls should help refine these possibilities. These data suggest distinct functional mechanisms for these independent genetic effects in the etiology of MS.