Research ArticleNEUROLOGICAL DISORDERS

Novel multiple sclerosis susceptibility loci implicated in epigenetic regulation

See allHide authors and affiliations

Science Advances  17 Jun 2016:
Vol. 2, no. 6, e1501678
DOI: 10.1126/sciadv.1501678
  • Fig. 1 Genome-wide representation of MS associations in the pooled analysis of German data sets.

    Manhattan plot showing strength of evidence for association (P value). Each variant is shown as a dot, with alternating shades of blue according to chromosome. Green dots represent established MS-associated variants and their proxies, as listed by Sawcer et al. (3) (except for rs2812197, which was not covered by that review). Top variants at the 15 non-MHC loci associated at the genome-wide significance threshold in our study are shown as diamonds. Novel variants showing genome-wide significance are plotted as red diamonds; their names are shown in bold font. Variants in high LD (r2 ≥ 0.7) with these novel variants are shown as red dots. Variants replicating in the Sardinian cohort are shown in red font. MA, minor allele. The OR is relative to the MA. Gene names for known loci are indicated as listed by Sawcer et al. (3). The plot is truncated at −log10p = 16 for better visibility; all truncated variants map to the MHC region. The lowest P value (rs3104373, *) was 1.3 × 10−234.

  • Fig. 2 Comparison of results from the pooled analysis of Germans to associations found in an IMSGC study.

    One hundred and four of the 108 variants showing genome-wide significant or suggestive associations with MS in the study published by the IMSGC in 2013 (4) were present in the pooled results of DE1 and DE2. All 104 variants showed the same direction of effect (P = 5 × 10−32, binomial sign test). Fifty-eight variants had lower ORs and 35 higher ORs compared to the published data set. P value–based categories labeled with different dots represent exclusive bins that add up to 104.

  • Fig. 3 Fine-mapping analysis results of locus rs4925166.

    (A) Regional plot for the rs4925166/SHMT1 locus. Color of dots indicates LD with the lead variant (rs4925166; pink). Gray dots represent signals with missing r2 values. cM, centimorgan. (B) Mediation analysis results in MPIP/GTP controls. Mediation effect: rs4925166→CpG cg26763362→SHMT1 expression. Direct effect: rs4925166→SHMT1 expression. Data have been calculated using the R package mediation (30), except for total effect (*), which was calculated by linear regression. Results were obtained using 1 million simulations. Effects and P values shown here differ from Table 5, as a lower number of samples contained both expression and methylation data than expression data alone. (C) Relationship between cg26763362 methylation, SHMT1 expression, and rs4925166 genotype in MPIP controls.

  • Table 1 Clinical characteristics of German MS cases.

    PPMS, primary progressive MS (as opposed to bout-onset MS).

    Cohort DE1Cohort DE2
    Number of cases3934954
    Age [mean (range)]39 (13–79)40 (17–82)
    Female [n (%)]2723 (69.2)695 (72.9)
    Male [n (%)]1211 (30.8)259 (27.1)
    PPMS [n (%)]105 (2.7)63 (6.6)
  • Table 2 Genome-wide significant HLA alleles.

    Alleles are in order of stepwise logistic regression. For each row, alleles from the rows above have been used as covariates in the model. AF (allele frequency of controls in %) is calculated from a joint set of DE1 and DE2. ORs and P values are from a fixed-effects pooled analysis of DE1 and DE2.

    HLA alleleAFOR (95% CI)PHLA alleles in LD (r2 > 0.9)
    DRB1*15:0114.82.85 (2.66–3.06)1.03 × 10−191DQB1*06:02
    A*02:0128.60.68 (0.64–0.73)3.68 × 10−29
    B*38:012.00.36 (0.27–0.49)2.09 × 10−11
    DRB1*13:031.51.96 (1.60–2.40)6.42 × 10−11
    DPB1*03:0110.31.33 (1.22–1.46)4.35 × 10−10
    DRB1*03:0112.21.29 (1.18–1.40)1.85 × 10−8DQA1*05:01, DQB1*02:01
    DRB1*08:013.01.63 (1.39–1.91)2.36 × 10−9DQA1*04:01, DQB1*04:02
  • Table 3 Genome-wide significant loci outside the MHC region and the top variant within the MHC region.

    Bold font in the left half of the table indicates novel loci, whereas bold font in the right half indicates variants that replicated in Sardinians. All P values shown are two-sided. Gene names of known loci are as listed by Sawcer et al. (3). C, chromosome. For additional details, see table S4.

    VariantCMAGeneMAF DEOR (CI) DE1 + DE2P DE1 + DE2P SardiniaOR (CI) DE + SardiniaP DE + Sardinia
    rs107974311TMMEL134.10.84 (0.80–0.89)1.81 × 10−10
    rs66894701AEVI514.21.24 (1.16–1.33)3.93 × 10−10
    rs23007471GCD5812.40.75 (0.69–0.81)1.74 × 10−12
    rs75358181GRGS119.20.76 (0.71–0.82)1.51 × 10−15
    rs26814243CCD8649.70.86 (0.82–0.90)9.51 × 10−10
    rs68592195AANKRD5522.20.84 (0.79–0.89)8.06 × 10−9
    rs31043736THLA-DRB113.62.90 (2.72–3.09)1.34 × 10−234
    rs43645066AL3MBTL326.40.84 (0.80–0.89)4.06 × 10−90.830.89 (0.85–0.93)1.99 × 10−6
    rs218241010TIL2RA38.10.84 (0.79–0.88)1.15 × 10−11
    rs189162110GIntergenic46.70.87 (0.83–0.91)2.94 × 10−8
    rs180069312CTNFRSF1A42.11.17 (1.11–1.23)1.06 × 10−9
    rs281219713TDLEU138.40.86 (0.82–0.91)9.95 × 10−96.86 × 10−30.87 (0.83–0.91)2.83 × 10−10
    rs649816816TCLEC16A35.51.23 (1.17–1.29)1.98 × 10−15
    rs3428659216TMAZ14.21.21 (1.13–1.30)4.58 × 10−80.441.16 (1.09–1.23)4.79 × 10-7
    rs492516617TSHMT134.50.85 (0.81–0.90)2.69 × 10−95.63 × 10−40.86 (0.82–0.90)7.40 × 10−12
    rs283642521TERG12.71.22 (1.14–1.31)2.84 × 10−80.351.18 (1.11–1.25)1.54 × 10−7
  • Table 4 eQTL and mQTL analysis for rs4925166.

    Direction of effect is relative to the minor allele T. Note that the effect sizes cannot be directly compared because normalization methods and covariates partly differ between studies. Additional eQTLs and mQTLs are described in the Supplementary Materials. Because only the single eQTL rs4925166/SHMT1 was examined in GTEx data, no FDR is indicated here. NA, not applicable.

    Expression
    Data setTranscriptEffectPFDR
    DE1SHMT10.364.42 × 10−132.99 × 10−10
    MPIPSHMT10.194.26 × 10−121.28 × 10−11
    GTPSHMT10.113.12 × 10−41.25 × 10−3
    GTExSHMT10.5609.2 × 10−28NA
    Methylation
    Data setCpGEffectPFDR
    MPIPcg26763362−0.033.21 × 10−205.04 × 10−18
    GTPcg26763362−0.031.98 × 10−141.58 × 10−13
  • Table 5 Fine-mapping of the DLEU1 locus.

    MAF (controls in %) and r2 (with rs2812197) are calculated from a joint set of DE1 and DE2, ORs and P values from the pooled analysis of DE1 and DE2. Second and third P value columns are from conditional analysis.

    VariantMAFOR (CI)PP (rs2812197)P (rs9591325)r2Reference
    rs281219738.40.86 (0.82–0.91)9.95 × 10−94.79 × 10−51.00
    rs80632148.50.89 (0.85–0.94)6.36 × 10−60.812.02 × 10−30.66(5)
    rs80634946.01.10 (1.04–1.15)2.73 × 10−40.990.0190.41(4, 21)
    rs95913258.10.78 (0.70–0.85)2.26 × 10−79.13 × 10−40.14
    rs95962708.10.78 (0.71–0.86)4.45 × 10−71.49 × 10−30.270.14 (0.99*)(6)

    *r2 with rs9591325.

    Supplementary Materials

    • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/2/6/e1501678/DC1

      Supplementary Results

      Supplementary Materials and Methods

      table S1. QC of data set DE1.

      table S2. QC of data set DE2.

      table S3. Genomic inflation.

      table S4. Genome-wide significant loci.

      table S5. eQTLs with FDR <0.05 in data set DE1.

      table S6. Replicated mQTLs of rs4925166 and CpG sites in SHMT1.

      table S7. Mediation analysis.

      table S8. Causal mediation analysis.

      fig. S1. Substructure analysis results in DE1.

      fig. S2. Substructure analysis results in DE2.

      fig. S3. GWAS with age at onset.

      fig. S4. Forest plots of all non-MHC top genome-wide significant variants.

      fig. S5. Locus-specific Manhattan plots.

      fig. S6. Forest plots of novel variants replicated in a Sardinian cohort.

      fig. S7. eQTL and mQTL analysis for rs4925166.

      fig. S8. Transcription factor binding sites.

      References (7075)

    • Supplementary Materials

      This PDF file includes:

      • Supplementary Results
      • Supplementary Materials and Methods
      • table S1. QC of data set DE1.
      • table S2. QC of data set DE2.
      • table S3. Genomic inflation.
      • Legends for tables S4 and S5
      • table S6. Replicated mQTLs of rs4925166 and CpG sites in SHMT1.
      • table S7. Mediation analysis.
      • table S8. Causal mediation analysis.
      • fig. S1. Substructure analysis results in DE1.
      • fig. S2. Substructure analysis results in DE2.
      • fig. S3. GWAS with age at onset.
      • fig. S4. Forest plots of all non-MHC top genome-wide significant variants.
      • fig. S5. Locus-specific Manhattan plots.
      • fig. S6. Forest plots of novel variants replicated in a Sardinian cohort.
      • fig. S7. eQTL and mQTL analysis for rs4925166.
      • fig. S8. Transcription factor binding sites.
      • References (70–75)

      Download PDF

      Other Supplementary Material for this manuscript includes the following:

      • table S4 (Microsoft Excel format). Genome-wide significant loci.
      • table S5 (Microsoft Excel format). eQTLs with FDR <0.05 in data set DE1.

      Download Tables S4 and S5

      Files in this Data Supplement:

    Navigate This Article