Glycosylation of immunoglobulin G is regulated by a large network of genes pleiotropic with inflammatory diseases

Abstract

Effector functions of immunoglobulin G (IgG) are regulated by the composition of a glycan moiety, thus affecting activity of the immune system. Aberrant glycosylation of IgG has been observed in many diseases, but little is understood about the underlying mechanisms. We performed a genome-wide association study of IgG N-glycosylation (N = 8090) and, using a data-driven network approach, suggested how associated loci form a functional network. We confirmed in vitro that knockdown of IKZF1 decreases the expression of fucosyltransferase FUT8, resulting in increased levels of fucosylated glycans, and suggest that RUNX1 and RUNX3, together with SMARCB1, regulate expression of glycosyltransferase MGAT3. We also show that variants affecting the expression of genes involved in the regulation of glycoenzymes colocalize with variants affecting risk for inflammatory diseases. This study provides new evidence that variation in key transcription factors coupled with regulatory variation in glycogenes modifies IgG glycosylation and has influence on inflammatory diseases.

Fig. 1. Gene prioritization in loci associated with IgG N-glycosylation.

The Manhattan plot was created by taking the lowest P value at every genomic position from all 77 GWAS. For simplicity, the plot was trimmed at the equivalent of P = 10⁻⁵⁰. The lowest observed P value in this analysis was 4.65 × 10⁻²⁷⁶ at ST6GAL1. Known loci, loci detected in previous IgG N-glycosylation GWAS; replicated, UPLC replication GWAS. Bottom: Summary of support for prioritization of every gene in the Manhattan plot. DEPICT, genes from enriched gene sets; expression, genes whose expression is pleiotropic with IgG N-glycosylation; CD19, B cells; PB, peripheral blood; coding, genes for which IgG N-glycosylation–associated SNP results in a changed amino acid sequence.

Fig. 2. Functional network of loci associated with IgG N-glycosylation.

Correlation estimates are computed on the basis of squared pairwise Spearman’s correlation of SNP effects. The loci are denoted with names of genes that were prioritized in regions tagged by the given lead SNP. (A) Functional network of loci associated with IgG N-glycosylation. In this network, each node represents a lead SNP in the locus, and each edge represents the squared correlation of glycome-wide effects of the two nodes. Only significant correlations after multiple testing correction (P ≤ 1.4 × 10⁻⁴) are shown. The thickness and intensity of edges depend on variation in one locus explained by the effect estimates in the second locus. Round-edged rectangular nodes denote genes that are, according to GO, involved in glycosylation; purple-edged nodes denote genes involved in immune system processes; green nodes denote loci containing genes involved in transcription regulation; orange nodes denote glycosyltransferases; and blue rectangles indicate diseases pleiotropic with IgG glycans in the given locus (Table 3). (B) Hierarchical clustering of pairwise Spearman’s locus-effect correlations.

Fig. 3. Results of in vivo validation of the functional links between IKZF1 and FUT8.

(A) IKZF1 binds to a regulatory region upstream of FUT8. (B) Knockdown of IKZF1 leads to decreased expression of IKZF1 (n = 3). (C) Representative Western blot showing that IKZF1 is depleted at the protein level (protein reduced by around 50% compared with a random shRNA line or a non-IKZF1 targeting shRNA line). WB, Western blot. (D) Knockdown of IKZF1 leads to decreased expression of IKZF3 (n = 3). (E) Knockdown of IKZF1 leads to increased expression of FUT8 in the LCL (n = 3). (F) There is a small but significant increase in fucosylation of IgG secreted by the LCLs in which IKZF1 is knocked down (n = 3 lines, measured at two time points). All P values are t tests, and error bars show SD from the mean. Asterisk indicates P value <0.05.