Genome-Wide Association Analysis of Protein-Coding Variants in IgA Nephropathy

Visual Abstract

Abstract

Significance Statement

Genome-wide association studies have identified nearly 20 IgA nephropathy susceptibility loci. However, most nonsynonymous coding variants, particularly ones that occur rarely or at a low frequency, have not been well investigated. The authors performed a chip-based association study of IgA nephropathy in 8529 patients with the disorder and 23,224 controls. They identified a rare variant in the gene encoding vascular endothelial growth factor A (VEGFA) that was significantly associated with a two-fold increased risk of IgA nephropathy, which was further confirmed by sequencing analysis. They also identified a novel common variant in PKD1L3 that was significantly associated with lower haptoglobin protein levels. This study, which was well-powered to detect low-frequency variants with moderate to large effect sizes, helps expand our understanding of the genetic basis of IgA nephropathy susceptibility.

Background

Genome-wide association studies have identified nearly 20 susceptibility loci for IgA nephropathy. However, most nonsynonymous coding variants, particularly those occurring rarely or at a low frequency, have not been well investigated.

Methods

We performed a three-stage exome chip–based association study of coding variants in 8529 patients with IgA nephropathy and 23,224 controls, all of Han Chinese ancestry. Sequencing analysis was conducted to investigate rare coding variants that were not covered by the exome chip. We used molecular dynamic simulation to characterize the effects of mutations of VEGFA on the protein's structure and function. We also explored the relationship between the identified variants and the risk of disease progression.

Results

We discovered a novel rare nonsynonymous risk variant in VEGFA (odds ratio, 1.97; 95% confidence interval [95% CI], 1.61 to 2.41; P = 3.61×10⁻¹¹). Further sequencing of VEGFA revealed twice as many carriers of other rare variants in 2148 cases compared with 2732 controls. We also identified a common nonsynonymous risk variant in PKD1L3 (odds ratio, 1.16; 95% CI, 1.11 to 1.21; P = 1.43×10⁻¹¹), which was associated with lower haptoglobin protein levels. The rare VEGFA mutation could cause a conformational change and increase the binding affinity of VEGFA to its receptors. Furthermore, this variant was associated with the increased risk of kidney disease progression in IgA nephropathy (hazard ratio, 2.99; 95% CI, 1.09 to 8.21; P = 0.03).

Conclusions

Our study identified two novel risk variants for IgA nephropathy in VEGFA and PKD1L3 and helps expand our understanding of the genetic basis of IgA nephropathy susceptibility.

Introduction

IgA nephropathy is the most common form of primary GN, and a large proportion of patients eventually progress to kidney failure requiring kidney transplantation or dialysis.^1,2 The etiology of IgA nephropathy is complex and influenced by multiple factors. The large difference in prevalence across world populations, together with the evidence of familial clustering and renal abnormalities among relatives of patients with IgA nephropathy, strongly suggests that genetic factors have significant contribution to IgA nephropathy risk.^3,4

Genome-wide association studies (GWASs) performed in large populations have identified nearly 20 IgA nephropathy susceptibility loci and provided valuable insights into the genetic architecture of the disease.^5–9 However, GWAS-implicated variants are generally common variants that confer relatively small effects and lie within non–protein-coding DNA sequences, and the cumulative effects of these variants can only explain a moderate amount of overall risk for IgA nephropathy. The unexplained heritability of the disease may be attributable to low-frequency (minor allele frequency [MAF], 1%–5%) or rare (MAF <1%) coding variants with larger effect sizes. Up to today, few studies have systematically evaluated the contribution of coding variation in IgA nephropathy. Previous GWASs have identified three common nonsynonymous variants in the genes TNFSF13 (p.Asn96Ser), CARD9 (p.Ser12Asn), and ITGAX (p.Pro517Arg).^5–9 Recently, a GWAS performed in 3363 IgA nephropathy and 9879 controls by using the Illumina exome chip also identified a common missense variant in FBXL21 gene for IgA nephropathy.¹⁰ However, most nonsynonymous coding variations, particularly the rare and low-frequency ones, have not been well investigated for their potential effects on IgA nephropathy risk. Power analyses indicated that discovery of such variants would require significant expansion in sample size.

To systematically evaluate the role of nonsynonymous coding variation in IgA nephropathy, we conducted a three-stage exome chip–based association study in 8529 cases and 23,224 controls of Han Chinese ancestry.

Methods

Study Design

To detect novel coding variants conferring susceptibility to IgA nephropathy, we performed a three-stage GWAS. In the discovery phase, we conducted an exome-wide association study in 18,020 Han Chinese participants (2378 IgA nephropathy cases and 15,642 controls) using the Illumina Human Exome BeadChip array. After quality control (QC), to evaluate the effects of mixing Northern and Southern Chinese samples in the discovery analysis, we split the discovery cohort into two clusters on the basis of the first two principal components, one of which is predominantly Northern Chinese (Discovery cluster 1 consisting of 360 cases and 1761 controls) and the other predominantly Southern Chinese (Discovery cluster 2 consisting of 2007 cases and 13,628 controls). For the validation stage, two independent case–control samples were recruited from China as validation 1 (4362 cases and 5925 controls) and validation 2 (1800 cases and 1910 controls). The definition of Northern and Southern in the validation stage was based on the clinical centers where the samples were recruited. The validation 1 cohort consisted of 1509 cases and 1704 controls recruited from Northern China (validation 1 Northern) and 2853 cases and 4221 controls recruited from Southern China (validation 1 Southern). The validation 2 cohort consisted of 704 cases and 805 controls recruited from Northern China (validation 2 Northern) and 1096 cases and 1105 controls recruited from Southern China (validation 2 Southern).

Study Participants

A total of 2378 patients with biopsy-diagnosed IgA nephropathy and 15,642 controls of Han Chinese ancestry were genotyped on the Illumina exome chip. The same inclusion and exclusion criteria were used in all cohorts: Patients older than 14 years with IgA nephropathy diagnosed by renal biopsy were eligible for case inclusion. The diagnosis of IgA nephropathy was based on the presence of dominant IgA deposition in the mesangial area by immunofluorescence microscopy and confirmed by electron microscopy.^5,8 The exclusion criteria were as follows: (1) a biopsy specimen containing less than eight glomeruli; (2) secondary IgA nephropathy, such as Henoch–Schönlein purpura nephritis, systemic lupus erythematosus, hepatitis B-associated GN, or diabetic nephropathy; and (3) individuals with cirrhosis, cancer, or HIV infection. The controls were collected among healthy blood donors older than 14 years from the same self-reported geographic region. The cases and controls were matched for geographic origin and ethnicity. The control participants included 8562 healthy nondisease controls and 7080 cases of other diseases, excluding all samples with IgA nephropathy–related diseases, which could serve as population-based controls for the identification of IgA nephropathy–specific loci. For the validation stage, two independent case–control samples were recruited from China as validation 1 and validation 2 (Supplemental Methods), which were also analyzed in our previous GWAS.⁸

This study was approved by the Institutional Review Board at The First Affiliated Hospital of Sun Yat-Sen University and at the National University of Singapore according to the Declaration of Helsinki. Written informed consent was obtained from all the participants.

Exome Chip Genotyping

Genomic DNA was isolated from whole blood using the commercial DNA extraction kit (Qiagen). Genotyping analysis of the discovery samples was conducted using the Infinium HumanExome BeadChip (v1.0) array (Illumina), with customized add-on content enriched for 27,089 rare, recurrent (found in >1 sample) variants detected by whole-exome sequencing of 1998 Chinese participants (Exome Asian consortium). These samples included 198 Southern Chinese exomes from Singapore,¹¹ 300 Southern Chinese exomes from Guangzhou, and 1500 Northern Chinese exomes from Anhui.¹² The genotyping cluster plots of top-associated single nucleotide polymorphisms (SNP) were visually inspected to ensure clear separation of the rare allele heterozygous and homozygous calls (Supplemental Figure 1).

QC and Association Analysis

A total of 274,125 SNPs were genotyped on the Illumina Human Exome BeadChip array. We excluded SNPs from the X, Y, and mitochondrial chromosomes and focused all further analyses on autosomal SNPs. We then filtered for SNPs with call rate >95% and Hardy–Weinberg equilibrium in controls P > 1×10⁻⁶, leaving 267,095 autosomal SNPs passing QC filters. Before association analysis, 158 SNPs with bad genotyping clusters were removed. In addition, we implemented strict QC analyses for the samples, removing duplicates and samples with cryptic relatedness (pi-hat >0.125), ancestry outliers, high cross-sample contamination (heterozygosity rate >mean+3 SD), and samples with a detected sex mismatch. We also applied a filter of MAF >0.05% in cases and MAF >0.05% in controls. We performed identity by descent analysis using PLINK,¹³ and first-degree relative pairs were identified. For each relative pair, the sample with lower genotype call rate was removed. Principal components analysis (PCA) was performed using Eigensoft v3.0¹⁴ by a set on 21,201 independent SNPs (MAF >1%). These SNPs were pruned to remove SNPs in linkage disequilibrium (LD) (r² >0.2) after exclusion of SNPs in the five conserved long-range LD regions in Chinese as previously described.^5,8 The samples were divided into two clusters, and outliers on the basis of the first five principal components were excluded. A total of 17,756 samples (2367 IgA nephropathy cases and 15,389 controls) and 60,676 SNPs were remained for further analysis.

Logistic regression analyses were performed separately for each cluster using PLINK.¹³ To control for population stratification, the first five principal components were included as covariates. After this, the results from the two clusters were combined using an inverse variance-weighted meta-analysis (METAL software).¹⁵ Stepwise conditional logistic regression was performed using the conditioning SNP genotype as a model covariate, using primary genotype data within each cohort.¹³

Validation Stage Genotyping and Meta-analysis

The validation 1 cohort included 1509 cases and 1704 controls recruited from North China and 2853 cases and 4221 controls recruited from South China. The validation 2 cohort included 704 cases and 805 controls recruited from North China and 1096 cases and 1105 controls recruited from South China. Genotyping of the eight SNPs selected for validation was performed using the MassARRAY system from Sequenom locus-specific PCR. The second validation of the three associated SNPs was performed using TaqMan assays. We visually inspected genotype clustering patterns and confirmed that the genotypes were of good quality (Supplemental Figures 2 and 3). After removing samples with a call rate <90%, all SNPs passed QC filters (call rates >95%, Hardy–Weinberg equilibrium P > 0.001).

In the two validation cohorts, samples from Northern and Southern regions of China were analyzed separately using logistic regression models, and the association results were combined by meta-analysis. To combine the association results across cohorts (discovery and two validation samples), we conducted an inverse variance-weighted meta-analysis (METAL software).¹⁵

Haplotype Analysis

Haplotypes were generated using rs185218985, and eight common SNPs (MAF >5%) directly genotyped on Illumina 610/660/1M arrays in 5088 samples (1397 IgA nephropathy cases and 3691 controls) from our previous GWAS.⁸ The eight SNPs were located within 16 kb flanking rs185218985 (chr6:43,732,669–43,764,359, hg19) and spanned the entire vascular endothelial growth factor A (VEGFA) coding region. Haplotype phasing was conducted using the PHASE tool.¹⁶ Recombination rates were inferred based on the 1000 Genomes Asian data¹⁷ using LocusZoom.¹⁸

Sequencing of VEGFA Coding Exons

To identify other rare coding variants in VEGFA that were not covered by the exome chip, we sequenced all coding exons in 600 cases and 1000 controls from the discovery cohort by PCR and Sanger sequencing. We designed ten primer pairs to amplify the eight exons by PCR (Supplemental Table 1). Targeted enrichment of the coding exons was performed on 1728 cases and 1728 controls from validation 2 cohort using molecular inversion probes. The amplified products were sequenced on the Illumina MiSeq platform.^19,20 A total of 384 samples were barcoded and pooled for sequencing in each MiSeq run for a mean coverage of 1409 × per sample across 1270 bp of VEGFA exonic sequence. After alignment, we performed realignment and recalibration followed by multisample SNP and indel calling and filtering with Genome Analysis Toolkit Unified Genotyper.²¹ Samples without 100% coverage of all coding bases by at least 15 reads were excluded from further analysis, and genotypes with quality scores <20 were set to missing. All variants were visually inspected for proper read alignment using the Integrative Genomics Viewer tool.²² The numbers of cases and controls carrying rare coding variants in VEGFA were compared using two-tailed Fisher exact and chi-squared tests.

Molecular Dynamics Simulation

Molecular dynamics simulation was performed with GROMACS (version 2020.6)²³ for all proteins. AMBER14SB²⁴ Force Field was used for all the molecular dynamics simulations. The 3D structure of VEGFA, VEGFR1, and VEGFR2 was predicted by AlphaFold2.²⁵ The molecular dynamics simulation was performed in the constant-pressure, constant-temperature ensemble within 100 ns for VEGFA protein and 200 ns for VEGFA with VEGFR1 as well as VEGFA with VEGFR2. The binding free energy was calculated with gmx_MMPBSA (version 1.4.3).²⁶ The 3D molecular visualizations were performed using PyMOL software, and 2D interaction diagrams were created with Molecular Operating Environment v2018.01 program. Details are provided in the Supplemental Methods.

Measurement of Serum Haptoglobin and VEGFA Levels by ELISA

Patients older than 14 years with IgA nephropathy diagnosed by renal biopsy were randomly selected for haptoglobin and VEGFA detection. Haptoglobin levels in serum were determined using ELISA according to the manufacturer's instructions (Assaypro, Winfield, MO). Serum levels of VEGFA were measured by ELISA from Raybiotech (ELH-vascular endothelial growth factor-1) according to the manufacturer's instructions. Duplicate determination was performed. Absorbance was measured at 450 nm using the SpectraMax Plus 384 Microplate reader (Molecular devices).

Analysis of Clinical Phenotypes and Disease Progression

We used two-tailed independent samples t tests to compare quantitative phenotypes in VEGFA V167I carriers versus noncarriers. Linear regression tests were used to assess the association of quantitative traits with PKD1L3 T429S genotypes. Ordinal logistic regression was performed for the correlation analyses of clinical subtype, chronic kidney disease stage, and the Oxford classification (T and C). Binary logistic regression was performed for the correlation with the presence of hypertension, non-nephrotic range proteinuria, hematuria, and the Oxford classification (M, E, and S). Non-nephrotic range proteinuria was defined as urinary protein excretion 0.3 g–3.5 g/24 hours, normal serum albumin, and no clinical symptoms. Pearson correlation coefficients were used to explore relationships between continuous variables. To explore the role of genetic variants in disease progression, a total of 293 patients with biopsy-proven IgA nephropathy diagnosed between 2008 and 2014 were selected from the prospective IgA nephropathy database at the First Affiliated Hospital of Sun Yat-sen University. Patients in the database were followed up regularly every 3–6 months. Unadjusted and multivariable-adjusted Cox proportional hazards models were adopted to evaluate the relationship between genotypes and risk of end point. Demographic characteristics (age, sex), clinical characteristics (baseline eGFR, 24 hours proteinuria, BP, Oxford classification scores), and use of immunosuppression were adjusted in multivariable-adjusted Cox proportional hazards models. Analyses were performed using IBM SPSS statistics v20.

Results

Exome Chip Association Study

In the discovery stage, a total of 274,959 variants in 2378 IgA nephropathy cases and 15,642 controls were genotyped on the Illumina exome chip. Of these 274,959 variants, 256,525 were coding (93.3%), 254,449 were nonsynonymous coding (92.5%), and 18,434 were noncoding variants (6.7%). A large percentage of the variants in the original content of the exome chip (59.8%) was nonpolymorphic, while only 6.6% of the customized content was nonpolymorphic (Supplemental Figure 4). After stringent QC filtering, we evaluated the association of 60,676 variants in 2367 IgA nephropathy cases and 15,389 controls. Of these, 43,977 SNPs were coding variants, with 26,330 SNPs found to be rare (MAF <1%) in our samples. To control for potential confounding by genetic heterogeneity between Northern and Southern Chinese, we performed a PCA¹⁴ and divided the samples into two clusters on the basis of the first two principal components (Supplemental Figures 5–7). We observed little evidence of genome-wide inflation of the association statistics in the discovery samples (λ=1.011), suggesting minimal bias resulting from population stratification (Supplemental Figure 8).

In the discovery stage, we identified eight novel loci with suggestive associations (P = 5×10⁻⁶ to 1×10⁻⁴) and good genotyping clusters, three of which were common variants and five were low-frequency or rare variants (Figure 1, Supplemental Table 2). Consistent with previous GWAS data, we also observed the strong evidence of replication at the four reported common nonsynonymous variants (Supplemental Table 3). We further examined the coding variants from ten other reported loci (HORMAD2, CFH, DEFA, VAV3, ST6GAL1, ACCS, ODF1-KLF10, PADI4, FCRL3, DUSP22.IRF4; ±1 Mb of reported SNP),^5–9 but there was no evidence for rare or common coding variants that could explain the GWAS signals at these loci.

Figure 1.

Manhattan plot showing significance of the association of each SNP allele with IgA nephropathy risk in the discovery cohort. The x axis shows the chromosome positions. The y axis indicates the level of statistical significance expressed as −log10 (P value). The horizontal red and blue lines indicate the genome-wide significant threshold (P = 5×10⁻⁸) and suggestive threshold (P = 1×10⁻⁴), respectively. Black, new genome-wide significant coding variants associated with IgA nephropathy; green, previously reported coding variants in GWASs for IgA nephropathy; the numbers represent variants with suggestive associations but did not exceed genome-wide significance in combined analyses (① SSUH2 p.C372S; ② ADAMTS9 p.R209C; ③ RPGRIP1L p.G1025S; ④ TCF3 p.T531M; ⑤ CD177 p.A3P; ⑥ MCM3AP p.V1363L). GWAS, genome-wide association study.

Follow-up Replication and Meta-analysis

We further evaluated the eight suggestive signals in an independent validation cohort (validation 1) of 4362 IgA nephropathy cases and 5925 controls. In the combined analysis of the discovery and validation 1 cohorts, we identified three of the eight loci selected for follow-up exceeding genome-wide significance: rs185218985 on 6p21.1 (P = 8.03×10⁻¹⁰; odds ratio [OR], 1.95), rs7185272 on 16q22.2 (P = 8.78×10⁻¹¹; OR, 1.17), and rs187316941 on 19p13.3 (P = 1.31×10⁻⁸; OR, 1.99). Each signal showed consistent ORs in each individual cohort, with little evidence of heterogeneity (Supplemental Table 2).

Then, we further genotyped these three variants in an additional independent validation cohort (validation 2) comprising 1800 cases and 1910 controls of Han Chinese ancestry. In the combined analysis, we confirmed two of the three variants that surpassed the threshold for significant genome-wide association: VEGFA V167I (rs185218985, combined P = 3.61×10⁻¹¹; OR, 1.97; 95% confidence interval [95% CI], 1.61 to 2.41) and PKD1L3 T429S (rs7185272, combined P = 1.43×10⁻¹¹; OR, 1.16; 95% CI, 1.11 to 1.21) (Table 1, Figure 2, Supplemental Tables 4–6). The two signals also showed significant association in the combined validation samples (P < 0.05 after correction for the testing eight SNPs). Conditional analyses indicated that both SNPs were the only independent SNPs in the respective locus (Supplemental Figure 9). Furthermore, we observed consistent associations across Northern and Southern Chinese samples in the discovery, validation 1, and validation 2 cohorts (Supplemental Figure 10). We also observed consistent effects of the VEGFA variant (OR, 2.37 versus 1.86; P_{heterogeneity} = 0.33; I²=0) and the PKD1L3 variant (OR, 1.12 versus 1.18; P_{heterogeneity} = 0.35; I²=0) between the Northern and Southern Chinese samples, without evidence of heterogenicity (Supplemental Table 7).

Table 1.

Association results of the two novel IgA nephropathy–associated nonsynonymous variants that reach genome-wide significance in combined analyses

SNP/Locus	Sample	Frequency Cases, %	Frequency Controls, %	P Value	OR (95% CI)	P het	I2, %
rs185218985	Discovery cluster 1	1.528	0.568	0.075	2.111 (0.927 to 4.809)
chr6:43748545 A/G	Discovery cluster 2	1.420	0.991	1.97×10−4	1.748 (1.303 to 2.345)
VEGFA	Discovery combined			4.07×10−5	1.786 (1.354 to 2.355)	0.672	0
	Validation 1 Northern	1.292	0.646	0.0086	2.028 (1.197 to 3.437)
	Validation 1 Southern	0.964	0.415	1.05×10−4	2.313 (1.514 to 3.532)
	Discovery+validation 1			8.03×10−10	1.945 (1.573 to 2.405)	0.346	0
	Validation 2 Northern	1.375	0.125	0.00113	11.32 (2.628 to 48.79)
	Validation 2 Southern	0.970	0.638	0.221	1.531 (0.774 to 3.027)
Meta-analysis	Validation 1 and 2 only			1.22×10−7	2.195 (1.640 to 2.937)	0.993	0
	All			3.61×10−11	1.970 (1.611 to 2.407)	0.602	0
rs7185272	Discovery cluster 1	75.83	74.84	0.266	1.126 (0.914 to 1.386)
chr16:72013797 C/G	Discovery cluster 2	74.61	70.90	9.61×10−6	1.188 (1.101 to 1.282)
PKD1L3	Discovery combined			5.66×10−6	1.180 (1.099 to 1.268)	0.634	0
	Validation 1 Northern	76.47	74.27	0.045	1.124 (1.003 to 1.261)
	Validation 1 Southern	76.09	72.84	2.14×10−5	1.184 (1.095 to 1.279)
	Discovery+validation 1			8.78×10−11	1.172 (1.117 to 1.229)	0.788	0
	Validation 2 Northern	76.46	74.41	0.199	1.114 (0.945 to 1.313)
	Validation 2 Southern	75.89	73.75	0.107	1.117 (0.977 to 1.278)
Meta-analysis	Validation 1 and 2 only			4.83×10−7	1.151 (1.090 to 1.216)	0.802	0
	All			1.43×10−11	1.162 (1.112 to 1.213)	0.910	0

OR, odds ratio; 95% CI, 95% confidence interval; VEGFA, vascular endothelial growth factor A; PKD1L3, polycystin 1 like 3.

Figure 2.

Regional association plots for two distinct genome-wide significant loci. (A) Regional plot for IgA nephropathy association at the VEGFA locus. (B) Regional plot for IgA nephropathy association at the PKD1L3 locus. The x axis shows the chromosome positions. The y axis indicates the level of statistical significance expressed as −log10 (P value). The gray dashed line corresponds to the genome-wide significance threshold. The blue line corresponds to the suggestive significance threshold (P = 1×10⁻⁴). Dot color indicates LD of each variant with the highlighted lead variant (purple diamond). LocusZoom¹⁸ was used to provide regional visualization of results with the East Asian panel of 1000 Genomes Phase 3 v5 (GRCh37/hg19). V1, validation 1; v2, validation 2; VEGFA, vascular endothelial growth factor A; PKD1L3, polycystin 1 like 3.

Haplotype Analysis and Sequencing of VEGFA

The tag SNP rs185218985 induces a valine to isoleucine change (V167I) within the alternatively spliced exon 6 in VEGFA (Supplemental Figure 11, Supplemental Table 8).²⁷ The VEGFA V167I variant is present at low frequencies in East Asians (0.5% in 1000 Genome samples¹⁷ and 0.86% in Exome Aggregation Consortium data²⁸). We, therefore, investigated the haplotypes in the VEGFA V167I carriers in the Chinese population. We phased the haplotypes of V167I and eight common variants spanning the VEGFA gene in 5088 samples.^8,16 Of the 120 V167I carriers detected, 102 (85%) were found on the same haplotype of GACTGCCC, with a frequency of 5.4% in the Chinese population (Table 2, Supplemental Figure 12). The other 18 V167I carriers occurred on four different haplotypes that are likely generated by one or two recombination events. These results suggested that the VEGFA V167I variant most likely arose from a single common ancestor rather than multiple recurrent mutations. We further evaluated the association of VEGFA V167I with clinical phenotypes of IgA nephropathy and observed the significant association with non-nephrotic range proteinuria (P = 0.04) (Supplemental Table 9). Furthermore, we detected the serum expression of VEGFA in 180 patients with IgA nephropathy and further evaluated its correlation with the clinical characteristics of IgA nephropathy. Consistently, the serum VEGFA levels were negatively correlated with proteinuria (r=−0.15; P = 0.045) and hematuria (r=−0.26; P = 5.67×10⁻⁴) levels and positively correlated with C4 levels (r=0.34; P = 8.27×10⁻⁴) (Supplemental Table 10). However, no significant association between the VEGFA V167I variant and VEGFA serum levels was observed in 180 patients with IgA nephropathy (Supplemental Figure 13), which may be due to the low frequency of VEGFA V167I and the moderate number of samples.

Table 2.

Haplotype analysis of V167I carriers on the basis of eight common SNPs in 5088 samples

Background Haplotypea	V167I Carriers			Noncarriers	Overall Background Haplotype Frequency, %
	Total	Frequency, %	% Carriers
GACTGCCC	102	1.002	85.0	447	5.40
GACTGCTT	2	0.020	1.7	3	0.049
GACTACCC	5	0.049	4.2	2003	19.7
AGTCGCCC	3	0.029	2.5	529	5.2
AGTCACCC	8	0.079	6.7	1258	12.4

^aHaplotypes phased from rs699946, rs833068, rs833069, rs3025010, rs3025033, rs3025035, rs6900017, and rs12204488 in 5088 samples (1397 IgA nephropathy and 3691 controls) using the PHASE tool. V167I is located between the fourth and fifth SNPs (rs3025010 and rs3025033).

Nonsynonymous mutations in VEGFA seem to be poorly tolerated in human as most of them are extremely rare (Supplemental Table 11).²⁸ We next performed sequencing analysis of all coding exons of VEGFA to identify rare coding variants that are not covered by the exome chip (Supplemental Tables 12 and 13). Consistent with our findings, the V167I variant was identified by sequencing in 58 of 2148 IgA nephropathy cases and 25 of 2732 controls (P = 1.69×10⁻⁶) (Figure 3). We found no significant enrichment for other rare coding variants within VEGFA (excluding V167I, 22 in cases versus 17 in controls, P = 0.12). Notably, within the exon 6 on which the V167I variant is encoded, there were twice as many carriers of other rare coding variants in IgA nephropathy cases compared with controls (excluding V167I, six in cases versus three in controls, P = 0.19).

Figure 3.

Rare coding variants in VEGFA that were identified by sequencing in 2148 IgA nephropathy cases and 2732 controls. The number of heterozygotes in IgA nephropathy and controls are listed. The p.V167I variant is shown in bold font. Variants are annotated on the longest RefSeq isoform NM_001025366.2/NP_001020537.2²⁷ but numbered according to NM_001171623.1/NP_001165094.1 (VEGFA-206). VEGFA, vascular endothelial growth factor A.

Predicting the Effect of VEGFA V167I on Protein Function

To calculate the effect of VEGFA V167I on the protein structure and stability of VEGFA protein, molecular dynamic simulation and molecular docking were performed (Supplemental Figure 14). As shown in Supplemental Figure 15, the residues surrounding the number 167 residue in VEGFA V167I were more abundant than VEGFA V167, which resulted in the more stability and compactness of VEGFA V167I. In addition, the calculated folding energies for VEGFA V167 and VEGFA V167I were 100.045 and 99.496 kcal/mol, respectively. It indicates that the VEGFA V167I is more accessible to stable structure through folding. Then, the effect of V167I mutation on binding affinity of VEGFA to its receptors, VEGFR1 and VEGFR2, was further studied. Molecular dynamic simulation indicated that the mutant V167I of VEGFA had multiple and abundant interactions with VEGFR1 and VEGFR2, leading to higher binding stability of the complexes (Figure 4, Supplemental Tables 14–17). The binding free energy on V167I carriers with VEGFR1, noncarriers with VEGFR1, V167I carriers with VEGFR2, and noncarriers with VEGFR2 was computed to be −323.13, −252.69, −188.55, and −149.99 kcal/mol, respectively (Supplemental Table 18).

Figure 4.

The effect of V167I mutation on binding affinity of VEGFA to VEGFR1 and VEGFR2. The interaction mode between (A) VEGFA V167, (B) VEGFA V167I, and VEGFR1. VEGFA is colored with yellow; VEGFR1 is colored with marine. The interaction mode between (C) VEGFA V167, (D) VEGFA V167I, and VEGFR2. VEGFA is colored with yellow; VEGFR2 is colored with magenta. The key residues I167/V167 in VEGFA are shown as green sticks. The red dashes represent hydrogen bond interaction. The blue dashes represent salt bridge. VEGFA, vascular endothelial growth factor A.

A Novel Common Variant in PKD1L3 in IgA Nephropathy

The top variant rs7185272 in PKD1L3 is in complete LD (r²=1, D′=1) with two other coding variants in PKD1L3 (Supplemental Table 19),²⁹ and none of the other coding variants in PKD1L3 showed association with IgA nephropathy risk (Supplemental Table 20, Supplemental Figure 16). In particular, one common nonsense mutation PKD1L3 R789X (with a frequency of approximately 42% in Chinese populations) was not associated with IgA nephropathy (P = 0.39), although our discovery samples provided sufficient power (>80%) to detect a very moderate effect (OR≥1.1) at P value of 0.05 (Supplemental Table 21).³⁰

Further functional annotation revealed that the top variant rs7185272 tagged several variants (r²>0.8) intersecting the Encyclopedia of DNA Elements and Roadmap enhancers and promoters in multiple tissue/cells (Supplemental Tables 19 and 22). Moreover, rs7185272 exhibited a strong cis-expression quantitative trait loci effect on the nearby HP gene (encoding haptoglobin) expression in whole blood (P = 6.05×10⁻⁸) (Supplemental Table 23),³¹ with the risk allele C associated with the decreased expression of HP (Table 3). In addition, rs7185272 was not significantly associated with other clinical phenotypes of IgA nephropathy (Supplemental Table 24). To further determine the effect of rs7185272 genotypes on HP expression, we measured serum levels of haptoglobin in 124 IgA nephropathy patients and 75 healthy donors by ELISA. We found no significant difference in serum haptoglobin levels between the two groups (P = 0.91). Nevertheless, we confirmed that haptoglobin levels in homozygous carriers of the risk allele C were significantly lower compared with cases with GG genotype (CC versus GG, 722.7±61.79 versus 1243±76.00 μg/ml, P < 0.0001) and cases with GC genotype (CC versus GC, 722.7±61.79 versus 1121±113.6 μg/ml, P = 0.004) (Supplemental Figure 17).

Table 3.

Functional annotation of novel IgA nephropathy susceptibility loci and genes

Locus	Proposed Function	Known Disease Associations
VEGFA 6p21.1	• Promotes angiogenesis, vasculogenesis, endothelial cell growth, cell migration32 • Inhibits apoptosis • Stimulates kidney epithelial cells • Can induce monocyte or macrophage migration • VEGFA-206 is sequestered in the extracellular matrix32–34	• Mutations: proliferative and nonproliferative diabetic retinopathy35 • Common SNPs: serum creatinine and kidney function, metabolic traits, age-related macular degeneration36,37
PKD1L3-HP 16q22.2	• PKD1L3: sour taste receptor38 • Although PKD1-like, no known role in the kidney • HP: prevents loss of iron through the kidneys and protects the kidneys from damage by hemoglobin39–41 • Identified by proteomic studies as a potential serum biomarker for steroid-resistant nephrotic syndrome42 • Considered to be related with proteinuria, a major progression factor of kidney disease43	• Mutations in HP and/or its regulatory regions cause ahaptoglobinemia or hypohaptoglobinemia44 • Candidate gene studies in diabetic nephropathy,45 Crohn disease,46 and several other diseases but not supported by GWAS findings

VEGFA, vascular endothelial growth factor A; PKD1L3, polycystin 1 like 3; GWAS, genome-wide association study.

Association between Risk Genotypes and CKD Progression in IgA Nephropathy

Then, we explored the relationship between two genome-wide significant signals and risk of CKD progression in 293 patients with IgA nephropathy. After a median follow-up of 55.5 months (IQR, 16–95 months), 96 (32.8%) participants reached the CKD progression event, which was defined by a 50% decline in eGFR or ESKD that was confirmed by a second evaluation obtained at least 4 weeks later. The results showed that rs185218985-A in VEGFA was associated with a greater risk of CKD progression (hazard ratio [HR], 2.99; 95% CI, 1.09 to 8.21; P = 0.033) (Supplemental Figure 18). After adjusted for traditional risk factors, rs185218985-A was associated with the risk of CKD progression with a HR of 2.46 (95% CI, 0.86 to 7.00) and a P_adjusted of 0.093. In addition, rs7185272 in PKD1L3 was not associated with the risk of CKD progression (P = 0.382; P_adjusted = 0.320).

Associations of HLA Variants

In addition, our discovery analysis identified extensive associations within the major histocompatibility complex region. The strongest HLA signal was in the known region of HLA-DQB1 represented by a common variant rs9275390 (OR, 1.36; P = 5.14×10⁻¹⁹) (Supplemental Figure 19, Supplemental Table 25). This SNP was in strong LD with the previously reported rs9275424 (D′=1.00, r²=0.93).^6,7 The previously reported lead variants rs9275596 and rs9357155 also showed significant associations in our cohorts (OR, 1.36; P = 8.32×10⁻¹⁰ and OR, 1.25; P = 1.80×10⁻⁶, respectively) and conditioning on rs9275390 abolished the associations at rs9275596 (P_conditioned = 2.88×10⁻⁵) and rs9357155 (P_conditioned = 8.64×10⁻⁴). Conditioning for the effect of rs9275390 eliminated evidence for association for most SNPs in close proximity; however, two distinct loci maintained genome-wide significance. The second independent locus is defined by rs2523882 (OR, 1.27; P = 7.50×10⁻¹⁴, P_conditioned = 4.34×10⁻¹¹) showing D′=0.12 and r²=0.004 with rs9275390, which was in high LD with a stop-gained variant rs1265054 (D′=0.98, r²=0.95) in C6orf15. C6orf15 is a protein-coding gene related to collagen V–binding activity and fibronectin-binding activity. After we conditioned for the effects of both rs9275390 and rs2523882, we found that a third locus within major histocompatibility complex, defined by rs12722039 (OR, 1.42; P = 3.21×10⁻¹², P_conditioned = 6.87×10⁻¹¹), which showed r²=0.006 and r²=0.008 with rs9275390 and rs2523882, respectively. This variant was a missense variant in HLA-DQA1 (p.Val17Met). Further conditioning on the effect of rs12722039 eliminated evidence for all associations in HLA region.

Discussion

In this study, we systematically evaluated the role of rare and common protein-coding variants in IgA nephropathy susceptibility in large Chinese cohorts. This study is the largest exome-wide study for IgA nephropathy to date, where the coverage of genome-wide coding variants by the exome array have been improved by adding 27,089 coding variants discovered by genome or exome sequencing analyses of Asian populations.

Recent studies have demonstrated that protein-coding variants, particularly the rare and low-frequency ones, are a major source of the still missing heritability of complex traits and diseases.^47,48 Protein-coding variants could be a precise target of functional analysis for understanding disease mechanisms, a favorable target for drug development, and a genetic marker with high disease risk for personalized medicine. These variants have distinctive features, including lower linkage disequilibrium with flanking variants and a larger population specificity, resulting in unsatisfactory imputation accuracy in multiethnic GWAS studies. Variants with larger effect sizes are likely to be rare, either recent mutations that have not been removed by purifying selection or older variants that remain at low frequencies in the population. As a result, protein-coding variants, particularly low-frequent and rare ones, have not been well studied by published GWASs. By performing a three-stage exome chip–based association study of IgA nephropathy in a large sample, our study is well-powered to detect low-frequency variants with moderate to large effect sizes. The PCA of the final dataset indicated that cases and controls were evenly distributed across the ancestry axes, reflecting successful implementation of genetic matching. Since the southern sample size was about five times larger than the northern sample, we believe that the differences in the significance level between the two regional samples is likely due to the difference in sample size rather than a population specific effect. Power calculations demonstrate that, with a total of 18,020 discovery samples, our study provided sufficient power (>80%) to detect a risk variant with moderate effect (OR≥1.8) and a low frequency (MAF, 1%–5%) at a suggestive significance (P < 1×10⁻⁴). Given that >90% of variants with a MAF of 1% were estimated to have an intermediate effect size (2<OR<6),^47,49 this study is effective enough to detect low-frequency variants with moderate to large effect sizes. Nevertheless, to detect the effect of rare variants with relatively small effects, the sample size requirement is more than three-fold greater than the largest GWAS for IgA nephropathy published to date. Furthermore, it may be informative to investigate whether the two newly identified coding variants segregate in families or cosegregate with disease phenotypes.

In general, natural selection shapes the joint distribution of effect size and allele frequency of genetic variants for complex genetic diseases in populations. Compared with rare variants, common variants tend to confer smaller effect sizes. Consistently, we discovered a rare variant in VEGFA with a large effect (OR, 1.97) and a common variant in PKD1L3 with a relatively moderate effect (OR, 1.16). Molecular dynamic simulation suggested that the rare VEGFA mutation could cause a conformational change that resulted in the more stability and compactness of VEGFA protein. Furthermore, VEGFA V167I leads to higher binding affinity of VEGFA to its receptors, VEGFR1 and VEGFR2. On the other hand, the common variant in PKD1L3 was annotated as a regulatory variant with score 1f in RegulomeDB, mapping to a region with transcriptional factor-binding signals for FOXA2/FOXA1 and TCF7L2 in the Encyclopedia of DNA Elements Consortium chromatin immunoprecipitation-sequencing data. Our data suggest that the potentially deleterious rare variant in VEGFA and the regulatory common variant in PKD1L3 together contribute to the genetic architecture of IgA nephropathy. Further functional experiments are required to determine the role of the identified variants and genes in IgA nephropathy pathogenesis. Consistent with previous GWAS data, we successfully replicated the HLA-DQB1 region represented by the top SNP rs9275390. In addition, we discovered two novel independent loci in C6orf15 and HLA-DQA1, which require further validation in other populations.

We identified a rare variant in VEGFA which was significantly associated with a two-fold increased risk of IgA nephropathy—this effect size was larger than most signals from common variants detected in previous GWASs.^5–10 Moreover, we found that the VEGFA V167I variant was associated with increased risk of kidney disease progression in IgA nephropathy, with a HR of 2.99. VEGFA encodes a member of the vascular endothelial growth factor family.³² VEGFA plays a key role in the establishment and maintenance of the glomerular filtration barrier, and dysregulation of its expression caused the occurrence of proteinuria.^35,50 Consistently, we observed a significant association between the VEGFA V167I variant and non-nephrotic range proteinuria. Neutralizing VEGFA antibodies as a common therapeutic strategy in oncology have been reported to cause glomerular endothelial injury with proteinuria.^51,52 Consistently, we found that the serum VEGFA levels were negatively correlated with proteinuria levels in IgA nephropathy. Notwithstanding the experimental and clinical evidence indicates that the inhibition of VEGFA causes the occurrence of proteinuria, VEGFA gene mutations have yet to be identified in humans with proteinuric glomerulopathies, such as FSGS and membranous nephropathy. Further functional experiments, especially with in vivo models, are required to determine the molecular mechanisms through which the VEGFA mutation is involved in the pathogenesis of IgA nephropathy. VEGFA is highly conserved across species, and the coding region of VEGFA appears to be under strong negative or purifying selection across evolution. Nonsynonymous mutations in VEGFA seems to be poorly tolerated—there were 141 variants presented cumulatively in <0.5% of the Exome Aggregation Consortium population. In this study, we discovered an IgA nephropathy–associated variant in VEGFA inducing a Val167 replacement with isoleucine in the alternatively spliced exon 6 (exon 6b) that is retained in only two of 18 Refseq isoforms.^33,53 This variant seems to affect only specific isoforms of the gene, which, together with its moderate effect on disease susceptibility, may explain how they may be tolerated at low frequencies in the population. The VEGFA p.V167I variant is completely absent in non-Asian populations in 1000 Genomes Project Phase 3 data. However, frequency data from projects with a larger sample size indicated that this variant was also present in other populations—the frequency was presented as 0.0024% in ALFA project European cohort, 0.003% in Prenatal Assessment of Genomes and Exomes study African American cohort, and 0.176% in Prenatal Assessment of Genomes and Exomes study Native Hawaiian cohort. Further sequencing analysis of VEGFA revealed twice as many carriers of other rare variants in IgA nephropathy compared with controls. However, larger sample sizes are needed to assess potential effects of these variants.

We also identified a novel common association at PKD1L3 T429S with a relatively moderate effect. Indeed, detecting variants with a moderate effect (OR<1.2), even for high-frequency variants, requires a large initial discovery sample size.⁴⁹ With a total of 18,020 discovery samples, our study provided sufficient power (>80%) to detect a very moderate effect (OR≥1.1) at a P value of 0.05. This may explain why this association was not found in previous GWAS studies. The common variant in PKD1L3 was significantly associated with lower haptoglobin protein levels in this study. Our previous proteomic study has identified haptoglobin as a potential serum biomarker for steroid-resistant nephrotic syndrome.⁴² Haptoglobin binds to free plasma hemoglobin, allowing degradative enzymes to gain access to hemoglobin, preventing loss of iron through the kidneys.^39,40 Haptoglobin levels were considered to be associated with proteinuria and progression of kidney diseases.^40,43,54 Nevertheless, there was no significant difference in serum haptoglobin levels between IgA nephropathy patients and healthy donors. Further investigation is needed to determine the target gene along with the molecular mechanisms underlying the statistical associations at this locus.

In summary, our study has identified two novel risk variants for IgA nephropathy in VEGFA and PKD1L3, which expanded our understanding on the genetic basis of IgA nephropathy susceptibility.

Supplementary Material

jasn-34-1900-s001.pdf ^{(3.1MB, pdf)}

Disclosures

J.-X. Bei reports Ownership Interest: Junshi Ltd.; and Patents or Royalties: Sun Yat-sen University Cancer Center. G.-R. Jiang reports Research Funding: Fibrogen. J. Lee reports Consultancy: Boehringer Ingelheim; and Advisory or Leadership Role: ThoughtFull World Pte. Ltd. E.S. Tai reports Consultancy: Farma Mondo S.A., Novartis Singapore Pte. Ltd., and Kowa Pharmaceutical Asia Pte. Ltd.; Ownership Interest: Abbott, Abbvie, and Novavax; Honoraria: As listed under consultancy agreements; and Speakers Bureau: as listed under consultancy agreements. E.-K. Tan reports Research Funding: YiQi Company; and Honoraria: Academic activities (Esai) and Editorial duties (Elsevier). H. Zhang reports Consultancy: Calliditas, Chinook, Novartis, OMEROS, and Ostuka; and Advisory or Leadership Role: Ostuka. All remaining authors have nothing to disclose.

Funding

This work was supported by the National Key Research and Development Project of China (No. 2016YFC0906100), Guangdong-Hong Kong Joint Laboratory on Immunological and Genetic Kidney Diseases (No. 2019B121205005), National Natural Science Foundation of China (No. 81920108008; No. 82170714; No. 82200785), and the Agency for Science, Technology, and Research (A*STAR) of Singapore (to J.-J.L.).

Author Contributions

Conceptualization: Jian-Jun Liu, Xue-Qing Yu.

Data curation: Tin Aung, Jin-Xin Bei, Feng-Tao Cai, Xiu-Qing Dong, Shao-Zhen Feng, Wee-Yang Meah, Dian-Chun Shi, Xue-Qing Tang, Ling Wang, Pei-Ran Yin, Zhong Zhong.

Formal analysis: Jia-Nee Foo, Ming Li, Ling Wang, Yan-Na Wang.

Funding acquisition: Xue-Qing Yu, Jian-Jun Liu.

Methodology: Jimmy Lee, E. Shyong Tai, Eng-King Tan.

Resources: Siow-Ann Chong, Chiea-Chuen Khor, Wee-Yang Meah, Mythily Subramaniam.

Supervision: Jian-Jun Liu, Xue-Qing Yu.

Validation: Meng-Hua Chen, Nan Chen, Qin-Kai Chen, Ching-Yu Cheng, Jie-Ruo Gu, Khai-Koon Heng, Geng-Ru Jiang, Yao-Zhong Kong, Yun-Hua Liao, Zhi-Hong Liu, Jian-Xin Wan, Li Wang, Gang Xu, Hong Zhang.

Writing – original draft: Ming Li.

Writing – review & editing: Ming Li, Yan-Na Wang.

Data Sharing Statement

Summary statistics of the exome-wide association study are available from the National Omics Data Encyclopedia (http://www.biosino.org/node) under the accession number OEP003779 on reasonable request.

Supplemental Material

This article contains the following supplemental material online at http://links.lww.com/JSN/E523.

Supplemental Figure 1. Illumina exome chip cluster plots in the discovery study: VEGFA V167I and PKD1L3 T429S.

Supplemental Figure 2. Sequenom cluster plots in validation 1: VEGFA V167I and PKD1L3 T429S.

Supplemental Figure 3. Taqman cluster plot for rs185218985 (VEGFA V167I) in a subset of validation 2 samples.

Supplemental Figure 4. Allele frequency distribution of 267,095 SNPs that passed call rate filters on the exome chip in 18,020 samples.

Supplemental Figure 5. PCA plots of 2367 cases and 15,389 controls in the discovery samples along PC1-5.

Supplemental Figure 6. Discovery cluster 1 and cluster 2 classified based on PCs1 and 2.

Supplemental Figure 7. PCA plots of samples in Discovery cluster 1 and cluster 2 along PCs1 and 2.

Supplemental Figure 8. QQ plots of the discovery study before and after excluding SNPs in the major histocompatibility complex (MHC) region.

Supplemental Figure 9. Conditional analysis results at the two distinct genome-wide significant loci in the discovery samples.

Supplemental Figure 10. Forest plots showing consistent associations across different cohorts.

Supplemental Figure 11. Splice isoform and protein domain (Pfam) information on VEGF-206.

Supplemental Figure 12. Recombination plot of the VEGFA gene region in a previous genome-wide association study of IgA nephropathy.

Supplemental Figure 13. The serum levels of VEGFA in IgA nephropathy that are VEGFA V167I variant carriers versus noncarriers.

Supplemental Figure 14. The cartoon model for VEGFA, and its receptors, VEGFR1 and VEGFR2.

Supplemental Figure 15. The interaction mode of the residue of number 167 and surrounding residues within 5.0 Å of protein for VEGFA V167, VEGFA V167I.

Supplemental Figure 16. Recombination plot of the PKD1L3-HP locus in a genome-wide association study of IgA nephropathy.

Supplemental Figure 17. Scatter plot of serum haptoglobin levels.

Supplemental Figure 18. Survival curves of participants with IgA nephropathy according to the VEGFA V167I variant and the PKD1L3 T429S variant.

Supplemental Figure 19. Association results of the three independent loci in the HLA region in the discovery analysis.

Supplemental Table 1. Primers used for amplification of VEGFA exons.

Supplemental Table 2. Eight SNPs taken forward for validation 1.

Supplemental Table 3. Previously reported nonsynonymous SNPs associated with IgA nephropathy.

Supplemental Table 4. Association results of the two novel IgA nephropathy–associated nonsynonymous variants using alternative genetic models.

Supplemental Table 5. The frequency distribution of the risk alleles in diverse populations.

Supplemental Table 6. Nonvalidation of TCF3 T531M association in combined analysis.

Supplemental Table 7. Comparisons of association results in Northern (2573 cases, 4270 controls) and Southern (5956 cases, 18,954 controls) Chinese samples.

Supplemental Table 8. Effect of rs185218985 on each splice isoform of VEGFA.

Supplemental Table 9. Clinical phenotypes of IgA nephropathy cases that are VEGFA V167I variant carriers versus noncarriers.

Supplemental Table 10. Correlations between serum VEGFA levels and clinical characteristics in IgA nephropathy patients.

Supplemental Table 11. Rare coding variants in VEGFA in Exome Aggregation Consortium (ExAC) East Asian data.

Supplemental Table 12. Other rare and common nonsynonymous variants in VEGFA that are on the exome chip.

Supplemental Table 13. Rare coding variants in VEGFA that were identified by sequencing 2148 cases and 2732 controls.

Supplemental Table 14. The contact list between VEGFA V167 and VEGFR1.

Supplemental Table 15. The contact list between VEGFA V167I and VEGFR1.

Supplemental Table 16. The contact list between VEGFA V167 and VEGFR2.

Supplemental Table 17. The contact list between VEGFA V167I and VEGFR2.

Supplemental Table 18. The binding free energy on VEGFA V167, VEGFA V167I with VEGFR1, VEGFR2.

Supplemental Table 19. HaploReg v4.1 annotation of SNPs in novel and suggestive loci in LD (r²>0.8) with rs7185272.

Supplemental Table 20. Exome array variants in PKD1L3 and HP and the association results in the discovery cohort.

Supplemental Table 21. Power calculations.

Supplemental Table 22. RegulomeDB annotation of SNPs in novel and suggestive loci in LD (r²>0.8) with rs7185272.

Supplemental Table 23. Association between rs7185272 genotypes and mRNA expression levels in whole blood using an expression quantitative trait loci meta-analysis.

Supplemental Table 24. Clinical phenotypes of IgA nephropathy cases by rs7185272 (PKD1L3 T429S) genotype.

Supplemental Table 25. Association analysis results for the three independent loci in the HLA region in the discovery cohort.

Supplemental Methods.