GWAS for serum galactose-deficient IgA1 implicates critical genes of the O-glycosylation pathway

Abstract

Aberrant O-glycosylation of serum immunoglobulin A1 (IgA1) represents a heritable pathogenic defect in IgA nephropathy, the most common form of glomerulonephritis worldwide, but specific genetic factors involved in its determination are not known. We performed a quantitative GWAS for serum levels of galactose-deficient IgA1 (Gd-IgA1) in 2,633 subjects of European and East Asian ancestry and discovered two genome-wide significant loci, in C1GALT1 (rs13226913, P = 3.2 x 10-11) and C1GALT1C1 (rs5910940, P = 2.7 x 10-8). These genes encode molecular partners essential for enzymatic O-glycosylation of IgA1. We demonstrated that these two loci explain approximately 7% of variability in circulating Gd-IgA1 in Europeans, but only 2% in East Asians. Notably, the Gd-IgA1-increasing allele of rs13226913 is common in Europeans, but rare in East Asians. Moreover, rs13226913 represents a strong cis-eQTL for C1GALT1 that encodes the key enzyme responsible for the transfer of galactose to O-linked glycans on IgA1. By in vitro siRNA knock-down studies, we confirmed that mRNA levels of both C1GALT1 and C1GALT1C1 determine the rate of secretion of Gd-IgA1 in IgA1-producing cells. Our findings provide novel insights into the genetic regulation of O-glycosylation and are relevant not only to IgA nephropathy, but also to other complex traits associated with O-glycosylation defects, including inflammatory bowel disease, hematologic disease, and cancer.

Fig 1. Longitudinal measurements of serum immunoglobulin levels and Gd-IgA1 levels over 4 years of follow-up.

Initial and 4-year follow-up levels of (a) serum total IgG, (b) serum total IgA, and (c) serum Gd-IgA1 normalized for serum total IgA. Panels (d, e, f) represent scatter plots of initial (x-axis) versus follow-up (y-axis) values. P-values correspond to the Pearson’s test of correlation; r²: squared correlation coefficient.

Fig 2. Combined meta-analysis of serum Gd-IgA1 levels in 2,633 individuals of European and East-Asian ancestry.

Manhattan plot (a), and regional plots for two distinct genome-wide significant loci: the C1GALT1 locus (b) and the C1GALT1C1 locus (c). The physical distance in kilobases (kb) is depicted along the x-axis, while–log(P-value) for association statistics is presented on the y-axis. The genome-wide significance threshold (P = 5×10⁻⁸) is depicted as a dotted horizontal line in a. The regional plots contain all genotyped and imputed SNPs in the region meta-analyzed between the discovery and replication cohorts.

Fig 3. siRNA knock-down of C1GALT1, COSMC and COSMC+C1GALT1 in IgA1-secreting cell lines increases Gd-IgA1 production.

(a) knock-down in IgA1-secreting cell lines from healthy controls; mock-control (n = 5), non-targeting siRNA (n = 7), C1GALT1 siRNA (n = 5), COSMC siRNA (n = 7), and COSMC+C1GALT1 siRNA (n = 2); (b) knock-down in IgA1-secreting cell lines from IgAN patients; mock-control (n = 5), non-targeting siRNA (n = 7), C1GALT1 (n = 5), COSMC siRNA (n = 7), and COSMC+C1GALT1 siRNA (n = 2); (c) relative change in mRNA in IgA1-secreting cell lines after siRNA knock-down of C1GALT1 (n = 5), COSMC (n = 7), and COSMC+C1GALT1 (n = 2) compared to non-targeting siRNA control.

Fig 4. Genotypic effects and worldwide allelic frequency distribution for the two top genome-wide significant loci.

(a) Mean trait values (+/- standard errors) by rs13226913 genotype at the C1GALT1 locus. (b) The distribution of rs13226913 alleles across the Human Genome Diversity Panel (HGDP) populations. (c) Mean trait values (+/- standard errors) by rs5910940 genotype at the C1GALT1C1 locus. (d) The distribution of rs5910940 alleles across HGDP populations. The allelic distribution plots were modified from the HGDP Selection Browser. The trait values were expressed as standard normal residuals of log-transformed serum Gd-IgA1 levels after adjustment for age, serum total IgA levels, case-control status and cohort membership.