A genome-wide association study in autoimmune neurological syndromes with anti-GAD65 autoantibodies

Abstract

Autoimmune neurological syndromes (AINS) with autoantibodies against the 65 kDa isoform of the glutamic acid decarboxylase (GAD65) present with limbic encephalitis, including temporal lobe seizures or epilepsy, cerebellitis with ataxia, and stiff-person-syndrome or overlap forms. Anti-GAD65 autoantibodies are also detected in autoimmune diabetes mellitus, which has a strong genetic susceptibility conferred by human leukocyte antigen (HLA) and non-HLA genomic regions. We investigated the genetic predisposition in patients with anti-GAD65 AINS. We performed a genome-wide association study (GWAS) and an association analysis of the HLA region in a large German cohort of 1214 individuals. These included 167 patients with anti-GAD65 AINS, recruited by the German Network for Research on Autoimmune Encephalitis (GENERATE), and 1047 individuals without neurological or endocrine disease as population-based controls. Predictions of protein expression changes based on GWAS findings were further explored and validated in the CSF proteome of a virtually independent cohort of 10 patients with GAD65-AINS and 10 controls. Our GWAS identified 16 genome-wide significant (P < 5 × 10-8) loci for the susceptibility to anti-GAD65 AINS. The top variant, rs2535288 [P = 4.42 × 10-16, odds ratio (OR) = 0.26, 95% confidence interval (CI) = 0.187-0.358], localized to an intergenic segment in the middle of the HLA class I region. The great majority of variants in these loci (>90%) mapped to non-coding regions of the genome. Over 40% of the variants have known regulatory functions on the expression of 48 genes in disease relevant cells and tissues, mainly CD4+ T cells and the cerebral cortex. The annotation of epigenomic marks suggested specificity for neural and immune cells. A network analysis of the implicated protein-coding genes highlighted the role of protein kinase C beta (PRKCB) and identified an enrichment of numerous biological pathways participating in immunity and neural function. Analysis of the classical HLA alleles and haplotypes showed no genome-wide significant associations. The strongest associations were found for the DQA1*03:01-DQB1*03:02-DRB1*04:01HLA haplotype (P = 4.39 × 10-4, OR = 2.5, 95%CI = 1.499-4.157) and DRB1*04:01 allele (P = 8.3 × 10-5, OR = 2.4, 95%CI = 1.548-3.682) identified in our cohort. As predicted, the CSF proteome showed differential levels of five proteins (HLA-A/B, C4A, ATG4D and NEO1) of expression quantitative trait loci genes from our GWAS in the CSF proteome of anti-GAD65 AINS. These findings suggest a strong genetic predisposition with direct functional implications for immunity and neural function in anti-GAD65 AINS, mainly conferred by genomic regions outside the classical HLA alleles.

Keywords: autoimmune encephalitis; genome-wide association study; glutamic acid decarboxylase; limbic encephalitis; stiff-person syndrome.

Figure 1

Overview of findings for the anti-GAD65 AINS GWAS. (A) Manhattan and (B) quantile-quantile plots of the anti-GAD65 AINS GWAS. The red dotted line in A depicts the threshold of genome-wide significance (P < 5 × 10⁻⁸). The total number of variants reaching this threshold was 191. The top variant was rs2535288 (chr6:31064007, P = 4.42 × 10⁻¹⁶, OR = 0.26, 95%CI = 0.187–0.358). (C) Functional consequences of all genome-wide and genomic loci variants in LD (r2 ≥ 0.6 and ±500 kb from lead, P < 0.05). The greatest proportion of variants locates to intergenic and intronic regions of the genome.

Figure 2

Investigation of GWAS findings at the protein level. (A) Protein-protein interaction network of GWAS coding genes. The network shows six functionally related gene clusters. Important connectors are SP1, FOS, TP53, JUN and SRC. (B) Volcano plot illustrating differential protein expression profiles in the CSF of anti-GAD65 AINS patients (red) compared to healthy controls (blue). HLA-A and C4A are differentially (P < 0.05) downregulated in anti-GAD65 AINS patients.

Figure 3

Analysis of the classical HLA alleles and haplotypes. (A) Manhattan and (B) disentangler plots of the analysis of the classical HLA alleles and haplotypes. In A, the GWAS variants are shown in light blue, the classical alleles in dark blue and the amino acid variants in orange. The yellow and red dotted lines indicate the commonly used suggestive (P < 1 × 10⁻⁵) and genome-wide (P < 5 × 10⁻⁸) significance thresholds, respectively. In B, the relationships of the allele of highest significance in our study, HLA-DRB1*04:01, with other classical alleles in the whole GWAS dataset are shown as disentangler plot.