Interaction between GALNT12 and C1GALT1 Associates with Galactose-Deficient IgA1 and IgA Nephropathy

Significance Statement

Galactose-deficient IgA1 plays a key role in the pathogenesis of IgA nephropathy. Although variability in serum levels of galactose-deficient IgA1 has a strong genetic component, the genetic link between this molecule and IgA nephropathy has not yet been clearly determined. The authors performed a genome-wide association study of serum galactose-deficient IgA1 levels among 1127 patients with IgA nephropathy in a Chinese population, identifying two genome-wide significant loci, of which one is novel. They also observed potential associations between galactose-deficient IgA1 loci and susceptibility to IgA nephropathy. In addition, they found genetic interactions between the two loci associated with both serum levels of galactose-deficient IgA1 and susceptibility to developing IgA nephropathy. This study provides novel insights into the genetic link between galactose-deficient IgA1 and IgA nephropathy.

Visual Abstract

Abstract

Background

Galactose-deficient IgA1 plays a key role in the pathogenesis of IgA nephropathy, the most common primary GN worldwide. Although serum levels of galactose-deficient IgA1 have a strong genetic component, the genetic link between this molecule and IgA nephropathy has not yet been clearly established.

Methods

To identify novel loci associated with galactose-deficient IgA1, we performed a quantitative genome-wide association study for serum galactose-deficient IgA1 levels, on the basis of two different genome-wide association study panels conducted in 1127 patients with IgA nephropathy. To test genetic associations with susceptibility to IgA nephropathy, we also enrolled 2352 patients with biopsy-diagnosed IgA nephropathy and 2632 healthy controls. Peripheral blood samples from 59 patients and 27 healthy controls were also collected for gene expression analysis.

Results

We discovered two loci, in C1GALT1 and GALNT12, that achieved genome-wide significance, explaining about 3.7% and 3.4% of variance in serum galactose-deficient IgA1 levels, respectively. We confirmed the previously reported association of C1GALT1 with serum galactose-deficient IgA1 levels, but with a different lead single-nucleotide polymorphism (rs10238682; β=0.26, P=1.20×10⁻⁹); the locus we identified at GALNT12 (rs7856182; β=0.73, P=2.38×10⁻⁹) was novel. Of more interest, we found that GALNT12 exhibits genetic interactions with C1GALT1 in both galactose-deficient IgA1 levels (P=1.40×10⁻²) and disease risk (P=6.55×10⁻³). GALNT12 mRNA expression in patients with IgA nephropathy was significantly lower compared with healthy controls.

Conclusions

Our data identify GALNT12 as a novel gene associated with galactose-deficient IgA1 and suggest novel genetic interactions. These findings support a key role of genetically conferred dysregulation of galactose-deficient IgA1 in the development of IgA nephropathy.

IgA nephropathy (IgAN) is the most common primary GN worldwide and the most common cause of ESKD among Asian populations.^1–3 The prevalence of IgAN shows marked interethnic differences, being more prevalent in East Asian ancestry and less prevalent in African ancestry compared with Europeans.⁴ The pathogenesis of IgAN is not well understood, but both genetic and environmental factors contribute to its development.^5,6 Genome-wide association studies (GWAS) have identified several genetic factors, mostly associated with mechanisms of defense against infection.^7–9 The largest GWAS meta-analysis of IgAN explained 7.6% of the risk in the Chinese cohorts and 6.2% in the European cohorts, suggesting a number of genetic factors for IgAN have not been identified.

Aberrantly glycosylated IgA1 molecules, mainly circulating galactose-deficient IgA1 (Gd-IgA1), play a key role in the pathogenesis of IgAN.^10,11 The Gd-IgA1 can be recognized by antiglycan autoantibodies to form nephritogenic immune complexes that deposit in the glomeruli and induce renal injury.¹² Recently, we have reported that the Gd-IgA1/C3 ratio was associated with disease progression independent of clinical and biopsy characteristics in IgAN.¹³ Although the heritability of serum Gd-IgA1 levels is high (estimated at 54%–80%),^14–16 the genetic association between Gd-IgA1 and IgAN has not been clearly determined. Two quantitative trait GWAS for Gd-IgA1 levels have identified two genome-wide significant loci, in C1GALT1 and C1GALT1C1.^17,18 These two loci explained approximately 7.1% of variability in Gd-IgA1 in Europeans, but only 2.0% in East Asians. In addition, neither of the identified Gd-IgA1 loci have been associated with IgAN.

In this study, we performed a quantitative GWAS for serum Gd-IgA1 levels among 1127 patients with IgAN in a Chinese population. We successfully mapped two loci with significant contributions to the variability in Gd-IgA1 levels. We also observed genetic interactions between the two loci in both Gd-IgA1 levels and disease risk.

Methods

Study Participants

A total of 1162 participants were enrolled in our study: cohort 1 comprised 849 patients with IgAN genotyped with the Illumina Global Screening Array, and cohort 2 comprised 313 patients with IgAN genotyped with the Illumina Human 610-Quad BeadChip. All participants were recruited from the renal division of Peking University First Hospital. The diagnosis of IgAN was on the basis of the presence of dominant IgA deposition in the mesangial area by immunofluorescence microscopy and confirmed by electron microscopy. SLE, cirrhosis, Henoch-Schonlein purpura, or other autoimmune diseases were excluded. The serum samples of patients with IgAN were collected at the time of renal biopsy.

This study was approved by the Ethics Committee of Peking University First Hospital and informed written consent was obtained from all patients (Institutional Review Board 2013–548).

Measurement of Gd-IgA1 by ELISA

The serum levels of Gd-IgA1 were determined by ELISA as previously reported.^13,19 Briefly, 96-well plates were coated with F(ab’)₂ polyclonal fragments of goat anti-human IgA1 at 2.5 μg/ml at 4°C overnight. Duplicates of two-fold dilutions of samples and standards in blocking solution were incubated overnight at 4°C. To remove terminal sialic acid, 1 mU/well of sialidase A (ProZyme) in phosphate buffer (pH=6.0) was added and incubated for 3 hours at 37°C. After washing, Helix pomatia (Sigma) was added and incubated for 3 hours at 37°C. Then, each well was incubated with horseradish peroxidase–conjugated-ExtrAvidin (Sigma) for 1 hour and detected by incubation with 3,3′,5,5′-Tetramethylbenzidine substrate for 35 minutes. The absorbance was measured at 450/570 nm.

Quality Control and Imputation

We applied quality controls on the genotyping data as described previously.^9,20 Samples with low call rates (<95%) were excluded for further analysis. The sex of each individual was imputed on the basis of the analysis of sex chromosome markers, and individuals with mismatched sex were excluded. Potential genetic relatedness in terms of pairwise identity-by-state was examined using PLINK v1.9.²¹ Close relatives were identified as having an estimated identity-by-descent proportion of genome-wide shared alleles >0.185. Sample outliers were checked on the basis of a principal component analysis method using Eigensoft v7.2.1.²² Single-nucleotide polymorphisms (SNPs) were excluded if they showed either a call rate <95%, minor allele frequency <0.01, or significant departure from Hardy-Weinberg equilibrium (P<10⁻⁴). A total of 362,250 SNPs in cohort 1 and 478,771 SNPs in cohort 2 passed the quality control filters.

Genotype imputation was performed independently in each cohort using the Michigan Imputation Server v1.0.2.²³ The 1000 Genomes Project Phase 3 data (v5, build 37) were used as reference panel.²⁴ For technical validation, we reimputed our cohorts using IMPUTE2²⁵ with Han Chinese reference from 1000 Genomes (208 individuals). SNPs with imputation INFO scores <0.8 were removed for further analysis. Postimputation quality control was applied to each cohort (SNP call rate >95%, minor allele frequency >0.01, and Hardy-Weinberg equilibrium [P>1×10⁻⁴]).

Statistical Analyses

Genome-wide linear regression association analysis was performed for each cohort with PLINK v1.9, using Gd-IgA1 data standardized to a mean of 0 and a SD of 1. Serum IgA levels and age were included as covariates in the linear model (Supplemental Figure 1). An inverse variance-weighted meta-analysis was performed across cohorts (METAL software).²⁶ We used genome-wide complex trait analysis to perform approximate conditional analyses to detect distinct association signals at each of the genome-wide significant loci.²⁷

To test genetic associations with susceptibility to IgAN, 2352 patients with biopsy-diagnosed IgAN and 2632 healthy controls were enrolled in our study. Study cohorts and genotyping platforms are provided in Supplemental Table 1. Genetic association with susceptibility to IgAN was performed with logistic regression analysis under three genetic models. The results were combined by a meta-analysis using METAL. Bonferroni correction was applied for multiple testing and P<0.01 (0.01=0.05/5 tests) was considered significant. Power calculation was performed within the framework of Mendelian randomization as previously proposed.²⁸

Functional Annotation

Functional annotation was performed by HaploReg v4.1 on the basis of Roadmap Epigenomes and ENCODE data.²⁹ Long-range interactions between genetic variants and gene promoter and regulatory regions were identified using the CHiCP browser.^30,31 We surveyed the expression quantitative trait locus (eQTL) effects by querying the eQTLGen Consortium data (an eQTL meta-analysis performed in blood samples from 31,684 individuals).³² Protein interaction network analysis was performed by STRING v11.0.³³

Gene Expression Analysis

We collected peripheral blood samples from 59 patients with IgAN and 27 healthy donors. PBMCs were isolated from 10 ml EDTA blood by Ficoll-Paque density gradient centrifugation (GE Healthcare). Total RNA was extracted from PBMC using TRIzol reagent (Invitrogen). Double-stranded cDNA was synthesized using an Illumina TotalPrep RNA Amplification Kit (Invitrogen). Relative gene expression was quantified using the Affymetrix PrimeView Human Gene Expression Array (Beijing Compass Biotechnology). Differential gene expression (natural log-transformed) between groups was analyzed using multivariable linear regression adjusted for age and sex.

Characteristics of Study Population

After quality control, a total of 1127 patients with IgAN (819 patients with IgAN in cohort 1 and 308 patients with IgAN in cohort 2) were included in this study. The clinical characteristics are summarized in Supplemental Table 2. The mean Gd-IgA1 level in patients with IgAN at biopsy was 324.5±50.9 U/ml (Supplemental Figure 2). The Gd-IgA1 levels showed positive correlations with IgA1 levels (r=0.26; P=1.0×10⁻⁶), total IgA levels (r=0.14; P=4.0×10⁻⁶), and the severity of tubular atrophy/interstitial fibrosis (r=0.09; P=4.1×10⁻³). The Gd-IgA1 levels were negatively correlated with C3 levels (r=−0.06; P=0.04) and eGFR (r=−0.07; P=0.02) (Supplemental Table 3).

Identification of Gd-IgA1 Loci in Patients with IgAN

By stringent quality control, we evaluated 4,320,314 variants in cohort 1 and 4,891,292 variants in cohort 2. We observed minimal effects of population stratification in cohort 1 (λ=1.009), cohort 2 (λ=0.983), and a combined sample set (λ=1.003) (Supplemental Figure 3). We identified two loci with genome-wide significance (P<5×10⁻⁸) on chromosomes 7p22.1 and 9q22.33 (Figure 1A, Table 1, Supplemental Table 4). Jointly, these two loci explain approximately 6.6% of trait variability.

Figure 1.

The meta-analysis results for serum Gd-IgA1 levels in 1127 patients with IgAN. (A) Manhattan plot showing significance of the association of each SNP allele with serum Gd-IgA1 levels. The horizontal red and blue lines indicate the genome-wide significant threshold (P=5×10⁻⁸) and suggestive threshold (P=1×10⁻⁵), respectively. Regional plots for two distinct genome-wide significant loci at (B) the C1GALT1 locus and (C) the GALNT12 locus. (D) Regional plot for the C1GALT1 locus after conditioning on the top SNP (rs10238682). (E) Regional plot for the GALNT12 locus after conditioning on the top SNP (rs7856182). Chr., chromosome; Mb, mega base pairs.

Table 1.

Associations of two genome-wide significant signals with Gd-IgA1 levels

Chr.	Location (bp)a	SNP	Alleleb	Cohort	Frequencyc	Effect	P	I2	Q	Gene
7	7215386	rs10238682	A	1	0.51	0.27	6.11×10−8			C1GALT1
				2	0.54	0.22	7.33×10−3
				Overall	0.51	0.26	1.20×10−9	0	0.58
9	98871030	rs7856182	T	1	0.04	0.68	9.50×10−7			GALNT12
				2	0.03	0.96	6.08×10−4
				Overall	0.04	0.73	2.38×10−9	0	0.36

Chr., chromosome.

^aOn the basis of version 38 (hg38) of the National Center for Biotechnology Information genome assembly.

^bGd-IgA1–increasing allele is provided as reference.

^cGd-IgA1–increasing allele frequency is provided.

The top signal was rs10238682 on chromosome 7p22.1, an intronic SNP within C1GALT1 (β=0.26; P=1.20×10⁻⁹) (Figure 1B), which explains 3.7% of the variability in Gd-IgA1 levels. We also confirmed reported associations of C1GALT1 variants, including rs1008897 (β=0.27; P=5.36×10⁻³) and rs13226913 (β=0.25; P=1.09×10⁻²), which are more common in Europeans than in Chinese populations (Supplemental Figure 4). Conditioning on rs10238682 abolished the associations at rs1008897 (P_conditional=0.09) and rs13226913 (P_conditional=0.06) (Figure 1D). CHiCP analysis showed interaction between rs10238682 and the promoter of C1GALT1 (Supplemental Figure 5A). Moreover, rs10238682 exhibits a strong cis-eQTL effect on C1GALT1 expression in peripheral blood cells (P=2.22×10⁻²³¹), with the allele A associated with lower mRNA levels (Supplemental Figure 6).³²

Consistent with previous data, we also replicated the reported association of C1GALT1C1 with serum Gd-IgA1 levels (rs5910940; β=0.10, P=4.04×10⁻³).

For novel observations, we discovered a novel genome-wide significant locus on chromosome 9q22.33. The top signal at this locus was rs7856182, a SNP 3′ downstream from GALNT12 (β=0.73; P=2.38×10⁻⁹) (Figure 1C, Supplemental Figure 7), which explains 3.4% of the variability in Gd-IgA1 levels. Conditional analysis showed the association at rs7856182 was independent (Figure 1E). rs7856182 is located in the enhancer region defined by H3K4me1 histone modifications in blood cells. CHiCP analysis showed an interaction between rs7856182 and the promoter of GALNT12 (Supplemental Figure 5B).

With transcriptome data from 59 patients with IgAN and 27 healthy donors, we confirmed GALNT12 mRNA expression in patients with IgAN was significantly lower than those in healthy controls (IgAN versus controls, 4.94±0.39 versus 5.21±0.34, P=4.65×10⁻³) (Supplemental Figure 8). C1GALT1 mRNA also showed lower expression in patients with IgAN, but only with marginal significance (IgAN versus controls, 5.79±0.50 versus 5.95±0.25, P=5.21×10⁻²).

Gd-IgA1-Associated Variants and Susceptibility to IgAN

We also examined known IgAN susceptibility loci with serum Gd-IgA1 levels, but none of these loci reached the significant threshold after Bonferroni correction (P<3.13×10⁻³, 0.05/16 tests) (Supplemental Table 5).

For further checking of Gd-IgA1–associated variants with susceptibility to IgAN, we refined the genetic association under different genetic models by logistic regression analysis (Table 2, Supplemental Table 6). The variant rs10238682 in C1GALT1 was associated with IgAN under the additive and recessive model (additive odds ratio [OR], 1.09, 95% confidence interval [95% CI], 1.01 to 1.18, P=2.66×10⁻²; recessive OR, 1.17, 95% CI, 1.03 to 1.34, P=1.69×10⁻²). The variant rs7856182 in GALNT12 also showed an association with IgAN under the dominant model (OR, 1.31, 95% CI, 1.01 to 1.69, P=4.34×10⁻²). However, these associations with disease risk would not be statistically significant after multiple-testing correction.

Table 2.

Gd-IgA1–associated variants and susceptibility to IgAN

SNP	Allelea	Cases		Controls		Genetic Model	OR	P	I2	Q
		n	RAF	N	RAF
rs10238682	A	2352	0.52	2632	0.50	Additive	1.09	2.66×10−2	0	0.53
						Dominant	1.09	2.08×10−1	0	0.33
						Recessive	1.17	1.69×10−2	0	0.96
rs7856182	T	2352	0.03	2632	0.02	Additive	1.26	6.77×10−2	0	0.75
						Dominant	1.31	4.34×10−2	0	0.98
						Recessiveb	—	—	—	—

RAF, risk allele frequency.

^aRisk allele for IgAN is provided as reference.

^bThe minor allele frequency of rs7856182 is low for a recessive model.

Genetic Interactions in Gd-IgA1 Levels and IgAN Risk

Additive interaction analysis indicated that GALNT12 (rs7856182) had an interactive effect with C1GALT1 (rs10238682) on serum Gd-IgA1 levels (Figure 2A). The linear regression model was significantly improved by adding an additional multiplicative pairwise interaction term for those two SNPs (P=1.40×10⁻²). By adding the interaction term, the total variance explained jointly by C1GALT1 and GALNT12 loci increased from 6.6% to 7.2%.

Figure 2.

Genetic interactions between C1GALT1 and GALNT12. (A) Gd-IgA1 levels according to SNP and genotype combinations. The mean Gd-IgA1 level in patients with the AA/TC+TT genotype combination for rs10238682/rs7856182 was significantly higher than in patients with the GG/CC genotype combination (389.6±121.9 versus 310.1±38.9 U/ml; P=3.38×10⁻³). (B) IgAN risk according to SNP and genotype combinations. The combination of risk genotypes of the two SNPs (rs10238682/rs7856182: AA/TC+TT) conferred a 1.92-fold risk of IgAN compared with the GG/CC genotype combination (P=6.55×10⁻³). (C) Interactive network analysis of C1GALT1 and GALNT12 and their ten most confident interactors. Each node represents a protein and each edge represents a high confidence interaction.

We then explored the possible genetic interaction between C1GALT1 and GALNT12 in the development of IgAN (Supplemental Table 7). As shown in Figure 2B, the risk of IgAN occurrence was most pronounced in the combination of risk genotypes (rs10238682/rs7856182: AA/TC+TT), which conferred a 1.92-fold risk (95% confidence interval, 1.19 to 3.09) compared with either protective genotype at both loci (GG/CC) (P=6.55×10⁻³) (Supplemental Table 8).

Finally, we performed an interactive network analysis of the proteins encoded by C1GALT1 and GALNT12 and their ten most confident interactors, which indicated a protein-protein interaction enrichment with P value of 1.0×10⁻¹⁶ (Figure 2C). The most significant biologic process was O-glycan processing (GO:0016266, P _FDR =1.02×10⁻²⁴), and the most significant KEGG pathway was Mucin type O-glycan biosynthesis (hsa00512, P _FDR =5.61×10⁻²⁹) (Supplemental Figures 5–8).

In large Chinese cohorts with new quantitative GWAS data, our study provides novel insights into the genetic link between Gd-IgA1 and IgAN.

Firstly, we confirmed the reported association of C1GALT1, but with a different lead SNP (rs10238682), which explains 3.7% of trait variability. The previously reported rs13226913 in C1GALT1 (r² between rs10238682 and rs13226913=0.004) explains 4.2% of trait variability in Europeans, but only 0.9% in Chinese populations.¹⁷ Our data suggested an ethnicity-specific effect at C1GALT1. In addition, we identified a novel locus for serum Gd-IgA1 levels at GALNT12. This locus was not found in previous GWAS,¹⁷ probably due to the genetic heterogeneity among populations or the relatively small sample size in East Asian ancestry.

Secondly, we observed potential genetic associations between Gd-IgA1-associated variants and IgAN risk, which, however, did not survive the Bonferroni correction for multiple testing. These potential associations with disease risk have not been reported in previous GWAS, probably due to the weak effects, inadequate power, or ethnic differences. Power calculation indicates we would need at least 11,600 patients and 11,600 controls to have >80% power to detect the effect of C1GALT1 and GALNT12 loci in a GWAS for IgAN, assuming the odds of IgAN increased by a ratio of 1.56/SD increase in Gd-IgA1.¹³

Thirdly, we identified genetic interactions between GALNT12 and C1GALT1 in both Gd-IgA1 levels and disease risk. GALNT12 encodes the enzyme polypeptide N-acetylgalactosaminyltransferase 12, which catalyzes the transfer of N-acetylgalactosamine (GalNAc) to a serine or threonine residue on the protein receptor in the initial step of O-glycosylation. N-acetylgalactosaminyltransferase 12 belongs to a large family of GalNAc-transferases. IgA1 hinge region O-glycosylation begins with attachment of GalNAc to serine or threonine residues catalyzed by GalNAc-transferases, and followed by the addition of galactose catalyzed by core 1 beta1,3-galactosyltransferase 1 (Supplemental Figures 9 and 10).³⁴ Recent studies indicate that downregulation of C1GALT1 and GALNT12 may contribute to aberrant intestinal O-glycosylation. C1galt1 deficiency in mice results in changes in the intestinal microbiota and the presence of intestinal inflammation.³⁵ A variant in linkage disequilibrium with rs7856182 at GALNT12 was associated with the survival outcome of colorectal cancer.³⁶ Consistent with this, rs10238682 exhibits a strong cis-eQTL effect on C1GALT1 expression, with the Gd-IgA1–increasing allele associated with lower mRNA levels. In addition, we observed reduced expression of GALNT12 in patients with IgAN.

Recently, ethnic disparities have been reported in levels of Gd-IgA1 between European and Chinese subjects, both healthy and with IgAN. Gd-IgA1 levels in Chinese patients with IgAN were lower than those in White patients with IgAN, and were comparable with healthy White subjects.¹⁸ To the best of our knowledge, this study is the largest GWAS for serum Gd-IgA1 levels in the Chinese population. Given the distributional differences in Gd-IgA1 between IgAN and healthy subjects, only patients with IgAN were included. The Gd-IgA1 level measurements were performed in the same laboratory, with the same batch of standards and Helix pomatia lectin run in all plates. We observed two significant signals with large effects in Chinese populations. Consistent with higher levels of Gd-IgA1 in Europeans, both the Gd-IgA1–increasing alleles of these two signals tend to be more common in Europeans than in Chinese populations. Potential limitations include limited sample size, lack of validation from Whites, and inadequate functional assay.

In summary, our data identify GALNT12 as a novel gene associated with Gd-IgA1, and suggest the genetic interactions between GALNT12 and C1GALT1 in Gd-IgA1 levels and susceptibility to IgAN. Our findings provide novel insights into the genetic link between Gd-IgA1 and IgAN.

Disclosures

H. Zhang reports having consultancy agreements with Janssen and Novartis; being a scientific advisor or member of the Board Committee of the Chinese Society of Nephrology, Board Committee of Nephrology in Chinese Medical Doctor Association, Vice-director of the Nephrology Committee in the Beijing Society of Medicine, a Member of the International Society of Nephrology Advancing Clinical Trials Committee, and a Member of the International Society of Nephrology Global Outreach Sister Renal Center Committee. All remaining authors have nothing to disclose.

Funding

This work was funded by the National Natural Science Foundation of China grants 81970613, 82070733, and 82022010, Natural Science Foundation of Beijing Municipality grant Z190023, Chinese Academy of Medical Sciences Research Unit grant 2019RU023, Clinical Medicine Plus X-Young Scholars Project of Peking University grant PKU2020LCXQ003, Fok Ying Tung Education Foundation grant 171030, Beijing Nova Program Interdisciplinary Cooperation Project grant Z191100001119004, and Beijing Youth Top-notch Talent Support Program grant 2017000021223ZK31.

Data Sharing Statement

GWAS summary statistics for Gd-IgA1 levels will be available from GWAS Catalog (Accession ID: GCST90011884, https://www.ebi.ac.uk/gwas) 6 months after the official publication.

Supplemental Material

This article contains the following supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2020060823/-/DCSupplemental.

Supplemental Figure 1. GWAS for IgA levels in 1,127 patients with IgAN.

Supplemental Figure 2. Gd-IgA1 levels in the Peking University First Hospital IgAN cohort were normally distributed.

Supplemental Figure 3. QQ-Plots.

Supplemental Figure 4. The heterogeneity of the C1GALT1 locus between European and East Asian ancestry.

Supplemental Figure 5. Results of CHiCP analysis.

Supplemental Figure 6. Colocalization of Gd-IgA1 association signals in C1GALT1 and blood eQTLs from eQTLGen Consortium data.

Supplemental Figure 7. Haplotype-based association analysis at the GALNT12 locus.

Supplemental Figure 8. C1GALT1 and GALNT12 mRNA expression in PBMCs from patients with IgAN and healthy controls.

Supplemental Figure 9. Structure of human IgA1 and biosynthesis of O-glycans on human IgA1.

Supplemental Figure 10. C1GALT1 and GALNT12 are expressed in IgA producing cells.

Supplemental Table 1. Summary of study cohorts and genotyping platforms.

Supplemental Table 2. Characteristics of participants in the Peking University First Hospital IgAN cohort.

Supplemental Table 3. Correlations between serum Gd-IgA1 levels and clinicopathological parameters in IgAN patients.

Supplemental Table 4. Variants in 95% fine-mapped credible sets at loci C1GALT1 and GALNT12.

Supplemental Table 5. Association of known IgAN susceptibility loci with serum Gd-IgA1 Levels.

Supplemental Table 6. Association of Gd-IgA1-associated variants with susceptibility to IgAN.

Supplemental Table 7. Additive interaction analysis of C1GALT1 and GALNT12 in genotype combinations by chi-square test.

Supplemental Table 8. Odds ratios for IgAN according to SNP and genotype Combinations.