Association of Systemic Lupus Erythematosus Susceptibility Genes with IgA Nephropathy in a Chinese Cohort

Abstract

Background and objectives

One hypothesis states that IgA nephropathy (IgAN) is a syndrome with an autoimmune component. Recent studies strongly support the notion of shared genetics between immune-related diseases. This study investigated single-nucleotide polymorphisms (SNPs) reported to be associated with systemic lupus erythematosus (SLE) in a Chinese cohort of patients with IgAN and in controls.

Design, setting, participants, & measurements

This study investigated whether SNP markers that had been reported to be associated with SLE were also associated with IgAN in a Chinese population. The study cohort consisted of 1194 patients with IgAN and 902 controls enrolled in Peking University First Hospital from 1997 to 2008.

Results

Ninety-six SNPs mapping to 60 SLE loci with reported P values <1×10⁻⁵ were investigated. CFH (P=8.41×10⁻⁶), HLA-DRA (P=4.91×10⁻⁶), HLA-DRB1 (P=9.46×10⁻⁹), PXK (P=3.62×10⁻⁴), BLK (P=9.32×10⁻³), and UBE2L3 (P=4.07×10⁻³) were identified as shared genes between IgAN and SLE. All associations reported herein were corroborated by associations at neighboring SNPs. Many of the alleles that are risk alleles for SLE are protective alleles for IgAN. By analyses of two open independent expression quantitative trait loci (eQTL) databases, correlations between genotypes and corresponding gene expression were observed (P<0.05 in multiple populations), suggesting a cis-eQTL effect. From gene-expression databases, differential expressions of these genes were observed in IgAN. Additive interactions between PXK rs6445961and HLA-DRA rs9501626 (P=1.51×10⁻²), as well as multiplicative interactions between CFH rs6677604 and HLA-DRB1 rs9271366 (P=1.77×10⁻²), and between HLA-DRA rs9501626 and HLA-DRB1 rs9271366 (P=3.23×10⁻²) were observed. Disease risk decreased with accumulation of protective alleles. Network analyses highlighted four pathways: MHC class II antigen presentation, complement regulation, signaling by the B-cell receptor, and ubiquitin/proteasome-dependent degradation.

Conclusion

From this “systems genetics” perspective, these data provide important clues for future studies on pleiotropy in IgAN and lupus nephritis.

Introduction

Over the past two decades, considerable progress has been made in unraveling the complex pathogenesis of IgA nephropathy (IgAN). However, the exact pathogenesis remains poorly determined. Current data suggest that genetic factors combined with environmental factors lead to increased synthesis of aberrantly galactosylated IgA1, formation of glycan-specific antibodies to IgG and IgA, and mesangio-podocytic-tubular cross-talk in the occurrence and development of the disease (1–6). Whether IgAN should be termed an “autoimmune disease” is controversial. However, recent genome-wide association studies (GWAS) strongly indicate that many of its associated loci also affect other autoimmune and infectious diseases (5,7–9), further supporting the notion of shared genetics of immune-related diseases (10). Recent estimates suggest that the identified loci collectively explain <10% of the genetic risk for IgAN, highlighting the fact that much of the heritable basis for IgAN has yet to be identified.

Previously, we reported that genetic factors have an appreciable influence on the production of under-galactosylated IgA1 and that GWAS data strongly implicate new clues as to the pathogenesis of IgAN (7,8,11–15). We also reported on the overlap between several autoimmune diseases: systemic lupus erythematosus (SLE), rheumatoid arthritis, ANCA-associated small vasculitis, and anti–glomerular basement membrane disease (16–25). Thus, we hypothesized that refinement of GWAS data or identification of IgAN susceptibility genes could be underpinned by investigation of the genetic variants reported to be associated with other immune-related diseases. Identification of novel IgAN genes and shared genetic pathways could improve understanding of common genetic mechanisms and eventually the development of improved methods of diagnosis, prognosis, and targeted therapies.

SLE is an autoimmune disease. Lupus nephritis is characterized by multiple immune complexes depositing in the kidney, including IgA molecules. IgAN is an immune complex–mediated GN defined by the predominant IgA molecule that deposits in the kidney. A recent study showed the pathogenicity of anti-glycan antibodies in IgAN, which suggested that IgAN is a type of autoimmune disease (26). A new theory suggests that most types of GN are primarily autoimmune diseases. Certain pathogenic similarities between autoimmune diseases (e.g., greater prevalence among Asians than Europeans, chronic course, renal involvement, circulating immune complexes, complement activation, morphologic similarities, certain pathways being involved in ESRD) prompted us to investigate the overlap in genetic susceptibility between SLE and IgAN. Well established co-occurrences of SLE with IgAN suggest common etiologic factors (27–29). Little progress has been made regarding the identification of genetic factors specific to lupus nephritis, but a genetic cause in SLE has been substantiated. More than 40 genes have been robustly associated with SLE.

We investigated whether single-nucleotide polymorphism (SNP) markers that had been reported to be associated with SLE were also associated with IgAN in a Chinese population.

Materials and Methods

The protocol of this study complied with the Declaration of Helsinki. The protocol was approved by the Ethics Committee of Peking University First Hospital (Beijing, China). Written informed consent was obtained from each patient.

Study Population

The samples used in the present study have been described previously. Briefly, exclusion of duplicates and first-degree relatives yielded 1194 IgAN cases and 902 healthy controls recruited in the Renal Division of Peking University First Hospital from 1997 to 2008 (8). All the cases were confirmed by renal biopsy, and all the controls were healthy blood donors without indicators of renal disease. Quality control was undertaken as described (8). Unexpected relatedness was excluded with a PLINK pi-hat cutoff of 0.125. We included men and women of Northern Chinese Han ancestry.

Selection and Genotyping of SNPs

We systemically examined data from GWAS, as well as large-scale replications conducted in SLE genetics through December 1, 2012. The reported SNPs associated with SLE in the GWAS context with a P value <1×10⁻⁵ were selected for analysis (30–42). The reported risk variants for SLE using data from the Catalog of Published Genome-Wide Association Studies from the National Human Genome Research Institute (http://www.genome.gov/gwastudies) were also checked. Finally, a panel of 96 SNPs representative of 60 genes or loci was selected (Supplemental Table 1). Genotyping was undertaken using the Illumina Human 610-Quad BeadChip, which involved 498,322 SNPs with a mean call rate of 0.9992.

Statistical Analyses

Only SNPs meeting the quality-control criteria of <1% overall missing data as well as consistency with Hardy–Weinberg equilibrium genotype frequency expectations (P<0.05) were included. As reported previously, after adjustment for population substructure, the inflation factor using all SNPs was λ=1.02, indicating a minimal effect of residual population structure. Thus, no further genomic control corrections were applied.

Genotype frequencies between IgAN cases and controls were compared using the chi-squared trend test implemented in PLINK software to determine whether individual SLE susceptibility loci were also associated with IgAN. Genetic models were defined relative to the minor allele. To reduce the risk of false-positive findings, all positive associations were checked further by associations at neighboring SNPs.

To test for additive interactions, the methods were taken using a 2×2 factorial design to calculate the attributable proportion due to interaction, the relative excess risk due to interaction, and the synergy index (20,43). P values <0.05 for attributable portion due to interaction were considered to be indicators of additive interactions. Ninety-five percent confidence intervals (95% CIs) were calculated using the delta method (44). Multiplicative interaction was assessed by adding an interaction variable (SNP×SNP) to the regression models. P<0.05 was considered to be evidence for multiplicative interactions.

Analyses of carriage of SLE alleles in patients with IgAN were carried out to determine whether there was an overall enrichment of SLE susceptibility variants in IgAN cases. Analyses were also undertaken to determine whether combining those risk alleles conferred a higher risk of disease.

Analyses of Bioinformatics

To explore whether the identified SNPs had expression quantitative trait loci (eQTLs) effects, Genevar software was used to determine associations between sequence variation and gene expression (http://www.sanger.ac.uk/resources/software/genevar). The sequence variation and gene-expression profiling data were from lymphoblastoid cell lines of 726 HapMap3 individuals. Another global map of the effects of polymorphism on gene expression in 400 children from families recruited through a proband with asthma was also investigated to associate gene expression on the basis of imputed genotypes (45).

The differential expressions of suspected IgAN candidate genes were compared with those of healthy controls using publically available data from the ArrayExpress Archive database (http://www.ebi.ac.uk/arrayexpress/) using “IgA nephropathy” as the search term. Three experiments (E-GEOD-37460, E-GEOD-35489, and E-GEOD-14795) involving comparatively large samples were included in the current analysis. The former two experiments took kidney biopsy samples and the latter experiment took whole-blood samples for gene-expression analyses. The normalized data available on the public databases were tested as reported previously.

To integrate data in biologic networks, Cytoscape software (which allows visualization of data in the context of networks) was applied (46). Cytoscape is widely used open-source software for the analyses of bimolecular interaction networks. MiMI integrates data on 119,880 molecules, 330,153 interactions, and 579 complexes from multiple, well known protein-interaction databases. An MiMI plugin, version 3.1.1, installed within Cytoscape 2.8.3, was used to determine the genetic interactions in positional/functional networks. Direct query of genes and their nearest neighbors from all data resources was done, and no further modifications were made.

Results

Analyses of SLE Risk Alleles in IgAN Show Suggestive IgAN Protective Alleles

Among the selected 96 SNPs, 10 SNPs of the lupus risk alleles in the region of HLA, CFH (suggested to be a tagging SNP for CFHR1, 3Δ), PXK, BLK, UBE2L3, and LYST showed evidence for association at an allele-type level (P<0.05) (Table 1). Of note, the associations between alleles in the HLA region, CFH, and IgAN were the top signals in our previous reports on GWAS. Interestingly, in comparing odds ratio (OR) values for these alleles in SLE and IgAN, all the directions of association were opposite those observed in the SLE studies, except for UBE2L3. Control allele frequencies were similar to those reported in SLE GWAS data. Thus, these findings suggested that the SLE risk alleles may be protective for susceptibility to IgAN. However, only SNPs in CFH and HLA regions could retain statistically significant evidence for association (Table 1). Although nonsignificant after applying a Bonferroni correction, PXK, BLK, UBE2L3, and LYST remained interesting candidates for further investigation.

Table 1.

Association results for systemic lupus erythematosus risk variants in IgA nephropathy

Chr	Base Pair	Locus	SNP	Major/Minor Allele	MAF Case/Control (%)	Trend Test P Values	Allele OR (95% CI) by SLE Risk Allelea	Dominant P Values	Recessive P Values	Genotype P Values	SLE Risk Allele OR (Reference)
1	194953541	CFH	rs6677604	G/A	4.10/7.26	8.41×10−6	0.55 (0.42 to 0.72)	3.37×10−5b	1.80×10−2	5.47×10−5	1.19 (41)
1	234106500	LYST	rs9782955	C/T	12.87/10.71	3.31×10−2	0.81 (0.67 to 0.98)	6.31×10−3	0.18	3.15×10−3	1.18 (35)
3	58345217	PXK	rs6445975	T/G	23.79/19.79	2.01×10−3	0.79 (0.68 to 0.92)	3.78×10−3	0.07	9.00×10−3	1.20 (31)
6	32508322	HLA-DRA	rs9501626	C/A	11.39/16.26	4.91×10−6	0.66 (0.55 to 0.79)	4.68×10−6b	0.12	2.60×10−5	1.86 (32)
6	32694832	HLA-DRB1	rs9271366	A/G	12.60/18.65	6.96×10−8	0.63 (0.53 to 0.75)	4.37×10−10b	0.81	3.40×10−10	1.26 (36)
8	11377591	BLK	rs7812879	C/T	26.59/23.23	1.23×10−2	0.83 (0.72 to 0.96)	3.28×10−2	0.05	0.04	1.45 (34)
8	11381089	BLK	rs2254546	G/A	26.63/23.12	9.32×10−3	0.83 (0.72 to 0.95)	2.35×10−2	4.88×10−2	2.96×10−2	1.42 (32)
8	11381382	BLK	rs2736340	T/C	29.94/26.94	3.33×10−2	0.86 (0.75 to 0.99)	7.78×10−2	0.07	0.08	1.35 (36)
22	20247190	UBE2L3	rs131654	T/G	46.48/49.94	2.63×10−2	1.15 (1.02 to 1.30)	0.17	2.38×10−2	0.06	1.28 (34)
22	20269675	UBE2L3	rs5754217	G/T	47.32/43.74	2.11×10−2	1.16 (1.02 to 1.31)	2.36×10−2	0.15	0.06	1.20 (36)

The reported SLE risk alleles are set in boldface. Chr, chromosome; SNP, single-nucleotide polymorphism; MAF, minor allele frequency; OR, odds ratio; 95% CI, 95% confidence interval; SLE, systemic lupus erythematosus.

^aORs were calculated on the basis of SLE risk alleles for comparison. Reported ORs were derived from the references listed.

^bOnly SNPs in CFH and HLA regions could retain statistical significance after multiple correction.

Analyses of Neighboring SNPs Support Disease Effects

To reduce the chance of false-positive findings using a single marker by chance, all the positive associations were checked further by analyzing the neighboring SNPs. Multiple significant association signals were observed (Figure 1). In CFH, HLA-DRA, and BLK, the top signals were from the SNPs selected. In HLA-DRA, rs2027856 (protective allele T; P=4.91×10⁻⁶; OR, 0.66; 95% CI, 0.55 to 0.79) also showed the same significance compared with rs9501626 in association with IgAN (r² value between rs2027856 and rs9501626 is 1.00). In HLA-DRB1, rs9270984 (protective allele G; OR, 0.63; 95% CI, 0.53 to 0.74) and rs9271055 (protective allele T; OR, 0.63; 95% CI, 0.53 to 0.74) showed the same most significance with P=9.46×10⁻⁹ (r²=1.00 between rs9270984 and rs9271055; r²=0.84 between rs9270984 and rs9271366). HLA-DRB1 rs9270984 and rs9271055 showed no better fit (P=3.1×10⁻⁹ for both) in association at the genotype level than that of rs9271366 (P=3.40×10⁻¹⁰) as well as their incomplete information in eQTL analyses; rs9271366 was still selected as tag SNPs in further analyses. In PXK, rs6445961 (r²=0.92 between rs6445961 and rs6445975) showed the most significant association, with P=3.62×10⁻⁴ (protective allele A; OR, 0.77; 95% CI, 0.66 to 0.89). In UBE2L3, rs2298428 (r²=0.95 between rs2298428 and rs5754217) showed the most significant association, with P=4.07×10⁻³ (risk allele T; OR, 1.21; 95% CI, 1.06 to 1.37). These findings suggested that associations with PXK, BLK, and UBE2L3 may be true associations because all the associations reported herein were corroborated by associations at neighboring SNPs. Nevertheless, the significances were too weak to meet the threshold in multiple testing. However, the associations between SNPs within LYST and IgAN were not convincing.

Figure 1.

Regional plots of identified loci in Chinese patients with IgA nephropathy. Genotyped SNPs are plotted with their P values (as–log₁₀[P values]) as a function of genomic position (Human Genome Build 18) within a region surrounding the reported systemic lupus erythematosus risk alleles. SNP, single-nucleotide polymorphism.

eQTL Analyses Provides Functional Clues

We investigated whether the most associated SNPs were expression SNPs because they affected the abundance of a protein or gene product by altering transcription. In lymphocyte cell lines from HapMap individuals, rs2298428 and rs9271366 were correlated consistently with UBE2L3 expression (P=0.01–5.0×10⁻⁵) and HLA-DRB1 expression (P=4.7×10⁻¹³–3.9×10⁻¹⁹), respectively, without population restrictions (Table 2). This conclusion was confirmed by data from lymphoblastoid cell lines from 405 siblings in the United Kingdom (45). These findings suggested that the associations between rs6677604 and CFH, rs2254546, and BLK were more pronounced in Asian populations (although similar trends between genotypes and gene expressions could be observed among different populations). For rs6445961 and PXK, the correlation was marginally significant in white patients living in Utah and Han Chinese from Beijing, China. However, different association patterns appeared to exist between white and Chinese individuals.

Table 2.

Correlation between genotypes of identified IgA nephropathy–associated single-nucleotide polymorphisms with gene expression in Epstein–Barr virus–transformed lymphoblastoid cell lines from an open database

SNP	Gene	HapMap 3 Unrelated Individuals (P Value)				Children Siblings of British Descent (n=405)
		CEU (n=165)	CHB (n=137)	JPT (n=113)	YRI (n=203)
rs6445961-A	PXK	0.27 (4.10×10−3)a	−0.20	−0.18	0.02	ND
			0.07	0.10	0.83
rs2298428-C	UBE2L3	−0.28 (3.30×10−3)a	−0.28 (0.01)a	−0.43 (5.00×10−5)a	–	−0.390 (8.50×10−5)a
rs6677604-A	CFH	0.12 (0.22)	0.02 (0.84)	0.26 (0.03)a	0.11 (0.26)	–
rs9501626-A	HLA-DRA	–	–	–	–	–
rs9270984-G	HLA-DRB1	0.59 (1.00×10−11)a	0.72 (1.30×10−13)a	0.68 (1.40×10−12)a	0.68 (4.90×10−16)a	–
rs9271366-G	HLA-DRB1	0.63 (4.70×10−13)a	0.74 (3.80×10−15)a	0.75 (3.10×10−16)a	0.73 (3.90×10−19)a	0.878 (4.00×10−17)a
rs2254546-G	BLK	0.02 (0.82)	−0.43 (8.20×10−5)a	−0.51 (1.10×10−6)a	−0.06 (0.57)	ND

With the function of each increase of the risk allele, the table depicts the correlation between genotypes and gene expressions. Protective alleles were regarded as reference alleles in the correlation. Pearson correlation coefficients are presented with P values in brackets. CEU, Caucasians living in Utah who were of northern and western European ancestries; CHB, Han Chinese from Beijing, China; JPT, Japanese in Tokyo, Japan; YRI, Yoruba in Ibadan, Nigeria; ND, no data could be derived from the database.

^aP<0.05.

Differential Gene-Expression Analyses Suggest Gene Involvement in IgAN

We ascertained whether the associated genes described above were expressed differently in patients with IgAN and healthy controls. Except for PXK (for which data were not available), all of the genes were differentially expressed from IgAN than those of controls, with elevated expressions of CFH, HLA-DRA, and HLA-DRB1 in renal biopsy specimens as well as BLK and UBE2L3 in whole-blood samples (Table 3). Only HLA genes were significantly differentially expressed when subjected to multiple testing, but only in renal biopsy specimens rather than in blood samples.

Table 3.

Differential candidate gene expressions in patients with IgA nephropathy compared with healthy controls from an open database

Candidate Gene	Samples
	Renal Biopsies						Whole Blood: Experiment E-GEOD-14795
	Experiment E-GEOD-37460			Experiment E-GEOD-35489
	IgAN (n=27)	Controls (n=27)	P Value	IgAN (n=25)	Controls (n=6)	P Value	IgAN (n=12)	Controls (n=8)	P Value
CFH	9.41±0.94	8.95±0.64	4.09×10−2a	5.72±0.32	5.51±0.14	0.14	96.90±56.10	88.11±61.04	0.74
HLA-DRA	11.59±0.33	10.89±0.54	6.56×10−7a,b	9.42±0.76	8.62±0.27	2.56×10−4a,b	8576.43±2251.01	8638.24±2355.87	0.95
HLA-DRB1	13.10±0.26	12.52±0.51	4.22×10−6a,b	11.31±0.65	10.43±0.28	5.58×10−5a,b	16661.58±5086.23	15779.10±3730.21	0.68
PXK	–	–	–	–	–	–	–	–	–
BLK	4.91±0.25	4.82±0.17	0.14	4.48±0.13	4.44±0.13	0.53	372.31±148.09	245.60±104.07	3.75×10−2a
UBE2L3	9.58±0.18	9.66±0.29	0.21	7.94±0.13	7.75±0.16	3.24×10−3a	492.78±94.12	362.57±132.65	1.90×10−2a

Data are the means±SD. IgAN, IgA nephropathy.

^aP<0.05.

^bP values remained significant after multiple correction using Benjamini and Hochberg false-discovery rate methods.

Additive and Multiplicative Interaction Analyses Suggest Gene–Gene Interactions

Fifteen tests involving different combinations of six of the most significantly associated SNPs (n×[n−1])/2) within their respective loci (PXK rs6445961, UBE2L3 rs2298428, CFH rs6677604, HLA-DRB1 rs9271366, HLA-DRA rs9271366, and BLK rs2254546) were conducted in the Chinese population. Supplemental Table 2 shows the results of analyses for additive and multiplicative interactions between identified SNPs categorized by whether they had or did not have protective alleles. There was a modest additive (but not multiplicative) gene–gene interaction between PXK rs6445961 and HLA-DRA rs9501626, with the proportion of risk due to an additive interaction of 2.86 (0.55–5.16), interaction P=1.51×10⁻² for IgAN. Significant multiplicative interactions were observed between FH rs6677604 and HLA-DRB1 rs9271366 (P=1.77×10⁻²), as well as for HLA-DRA rs9501626 and HLA-DRB1 rs9271366 (P=3.23×10⁻²).

Analyses of Joint Effects Suggest Cumulative Effects on the Risk of Disease

To determine the cumulative effect of six SNPs, disease risk was assessed according to the number of protective alleles they had. Individuals with more protective alleles seemed to be less prone to IgAN (whole model P=5.96×10⁻¹³) (Table 4). With each increase in the number of protective alleles, the disease risk decreased by approximately 7% (r²=–0.97; P=1.38×10⁻³). The disease risk decreased up to seven-fold in individuals with eight or more protective alleles compared with those with fewer than 2.

Table 4.

Joint effects of newly identified loci stratified by the number of protective alleles

Protective Alleles (n)	Frequency in Cases/Controls (%/%)	Odds Ratio (95% CI)	P Value
≤2	5.4/1.9	1.00 (Reference)
3	13.5/10.3	0.46 (0.25 to 0.83)	9.11×10−3
4	25.7/19.5	0.46 (0.26 to 0.82)	6.68×10−3
5	26.3/25.4	0.36 (0.21 to 0.64)	2.73×10−4
6	19.0/21.4	0.31 (0.18 to 0.55)	3.06×10−5
7	6.4/13.3	0.17 (0.09 to 0.31)	1.44×10−9
≥8	3.7/8.0	0.16 (0.08 to 0.31)	8.77×10−9

Integrating Identifies Molecules in Cytoscape-Supported Network Involvement

The six identified molecules showed physical interactions between genes or through their products/neighbors (Figure 2). The network was divided mainly into four modules, representative of pathways: MHC class II antigen presentation, complement regulation, signaling by the B-cell receptor (BCR), and ubiquitin/proteasome-dependent degradation. Several cellular interrelated genes have been suggested to participate in the pathogenesis of IgAN as well as SLE: HLA, ITGAM, C3, CFI, FCGR, and PTEN (26,47–50). We also checked the differential expression of those interrelated genes: great enrichment of differences in gene expression between IgAN and healthy controls was observed (Supplemental Table 3). Whole genome-wide expression data were just from tens of samples, but C3 (it was linked with CFH), ITGAM (CFH), CD74 (HLA-DRB1), HLA-DMA (HLA-DRB1), HLA-DMB (HLA-DRA), EGFR (BLK), SMAD7 (UBE2L3), and PTEN (UBE2L3) still produced significant associations in the context of multiple testing.

Figure 2.

A network plot of connections between identified loci in Chinese patients with IgA nephropathy.

Discussion

In recent years, three GWASs in IgAN have been conducted. They uncovered several susceptibility loci and greatly broadened our understanding of the genetic architecture of the susceptibility to IgAN (5,7–9). Among these three GWAS, we took part in two of them (7,8). As reported, all the identified associations within the regions of MHC, 1q32, 8p23, 17p13, and 22q12 could be confirmed in our cohort (7,8), which proved to be the cornerstone of credibility of the present study. The findings of the present study added to the loci showing associations with IgAN, as well as overlap between IgAN and SLE: CFH, HLA-DRA, HLA-DRB1, PXK, BLK and UBE2L3. Although some of the associations did not remain significant after the Bonferroni correction was applied, all the associations reported herein were corroborated by associations at neighboring SNPs, suggesting that they are true associations.

It is widely accepted that initial GWAS can detect just the greatest effects rather than all the susceptibility variants. The ORs of all the novel variants were much weaker (0.8 or 1.2) than the ORs from variants within CFH and HLA (0.6), both of which were previously identified signals in GWAS. The observation that the associated allele was the reverse of that reported previously for SLE was in accordance with a report stating that the protective alleles within MHC, 1q32 and 22q12 regions for IgAN had been implicated as risk factors for other autoimmune disorders (8). Most of these associated loci showed the same tendency for disease susceptibility, so the result is not likely to be a coincidence. The different association directions of the same alleles nevertheless supported the notion of pleiotropy (effect of a single gene on multiple phenotypes), quantitative genetics (combination of the influences of multiple genes together with environmental variation resulting in continuous distributions of phenotypes), and the human “diseasome” (the synthesis of all human genetic disorders [“disease phenome”]) and all human disease genes [“disease genome”]). Ideally, GWAS testing for identifying the common or shared genetic influences on SLE and IgAN in the same population should be carried out and is underway.

GWAS have been used to identify multiple SNPs associated with disease risk, and attention has turned to explaining the underlying molecular mechanisms of action (5). One hypothesis is that a proportion of the causal variants tagged by these disease-associated markers may affect the abundance of a protein (or the relative abundance of its different isoforms) by altering transcription. Efficient identification of additional susceptibility loci with more modest effects might benefit from the integration of statistical evidence with some assessment of functional candidacy. Here, we investigated the positive correlations between identified SNPs and their corresponding gene expression, especially for HLA-DRB1, UBE2L3, and BLK. The data further supported the candidacy of those genes as causal factors in IgAN. Data from Epstein–Barr virus B cell–transformed lymphoblastoid cell lines should be more illustrative than data based on other cell lines in IgAN, because gene expression and eQTLs can be tissue-specific and because IgAN is a disease characterized by production of the nephritogesnic IgA1 molecule from B cells. Confirmation from a different gene-expression database strongly supported the probability of reliability (45). In addition, immortalized lymphoblasts that were clonal could more readily be studied without the environmental influences or transcriptome diversity found in mixed lymphocyte populations in vivo (51). Also, when differential gene expressions in IgAN patients were checked, the expression of all of those genes was upregulated in IgAN patients. However, the associations seemed to have tissue specific-characteristics because elevated expressions of CFH, HLA-DRA and HLA-DRB1 seemed to be restricted to renal biopsies and BLK and UBE2L3 to whole-blood samples. More widespread gene-expression analyses will be warranted, especially in specific cell clones. It seemed that HLA-DRA and HLA-DRB1 protective alleles corresponded to lower gene expressions, whereas PXK, BLK, and UBE2L3 protective alleles corresponded to higher gene expressions, which may indicate an abnormal balance between antigen presentation and lymphocyte signaling. Nevertheless, future studies linking alleles and differential gene expressions in specific tissues will be needed. In addition, rare variants, which may have a greater effect in conferring disease risk and may contribute to a substantial fraction of heritability, will need further evaluations in future genetic studies in IgAN.

Furthermore, to determine whether the identified genes cause effects in a joint manner or epistatic fashion, we conducted gene–gene interaction analyses as well as cumulative gene effect analysis. Investigating genetic interactions has proved difficult, and an optimal statistical approach is not available, so combining several analytical methods may be best for detecting epistatic interactions. Gene–gene interactions can be assessed with additive or multiplicative mathematical models. We demonstrated significant additive and multiplicative interactions among the identified SNPs: that is, additive interactions between PXK rs6445961and HLA-DRA rs9501626, as well as multiplicative interactions between CFH rs6677604 and HLA-DRB1 rs9271366, and between HLA-DRA rs9501626 and HLA-DRB1 rs9271366. However, because of the moderate effects of these alleles and a low incidence of IgAN (estimated incidence in the general population, 25–50 cases per 100,000 individuals), our study remained underpowered to detect epitasis with our sample size (calculated power for epistasis was approximately 0.1–0.2). In joint analyses, we observed that the disease risk decreased by about 7% with each increase in the alleles, and it decreased up to 7-fold in individuals with eight or more protective alleles compared with those who have fewer than two. These results repeatedly supported the notion that the identified genes were the susceptibility genes for IgAN.

One of the most compelling reasons for identifying the genetic underpinnings of common diseases is to generate new hypotheses about the mechanisms and pathogenesis of disease (5). Hence, we checked further the newly identified genes in a pathway-based manner. A molecular network using a correlation structure was produced in which all the identified genes were connected to each other by intermediary genes, and four modules were highlighted. Great enrichment of differences in gene expression between IgAN and healthy controls was observed even though the whole genome-wide expression data were just from tens of samples. The pathways were MHC class II antigen presentation, complement regulation, signaling by the BCR, and ubiquitin/proteasome-dependent degradation. The role of MHC and complement in IgAN has been supported strongly by several observational studies. BLK encodes a tyrosine kinase that is involved in the regulation of B-cell activation. B-cell signaling may have a key role in the pathogenesis of IgAN through elevation of IgA levels in serum, production of autoantibodies, antigen presentation to T cells, and cytokine production (52). Also, B-cell depletion has proved successful in the treatment of GN. UBE2L3 encodes a ubiquitin-conjugating enzyme involved in ubiquitin/proteasome-dependent degradation, which is important in the cell cycle, cell differentiation, apoptosis, sodium-channel function, and modulation of inflammatory responses. The ubiquitin/proteasome pathway has been suggested to be implicated in the development of multiple kidney diseases (53), and proteasome inhibitors have been efficacious in some forms of renal disorders, such as lupus nephritis (54), renal ischemia-reperfusion injury (55), and ANCA-induced GN (56). PXK encodes a multimodular protein composed of a phox homology domain, a protein kinase–like domain, and a Wiskott-Aldrich syndrome protein homology 2 domain. The gene product of PXK regulates the activity of Na-ATPase and K-ATPase ion transport, and is expressed in the kidney (57,58). Recent data suggest that PXK has a critical role in trafficking of the EGF receptor through modulation of ligand-induced ubiquitination of the receptor. Thus, the present study provided important clues for better elucidation of IgAN pathophysiology in the future and possible therapy optimization. Nevertheless, one must be cautious because most of the findings from the initial GWA studies were association signals rather than direct information about susceptibility genes.

The degree of shared genes between IgAN and SLE is substantial, but is likely to still be an underestimate. First, we analyzed only associated loci at the P<1×10⁻⁵ level, and the SNPs or genes meeting this criterion are increasing with enrollment of larger sample sizes. Second, because of different linkage disequilibrium between variants in cases and controls, a different variant in the same locus may be responsible for disease risk in a second phenotype. Third, GWASs directly conducted in patients with lupus nephritis are still underway. For these reasons, the gene overlap between the two diseases may be higher than that identified in the present study.

In conclusion, we identified CFH, HLA-DRA, HLA-DRB1, PXK, BLK, and UBE2L3 as shared loci between IgAN and SLE. Many of the alleles that are risk alleles for SLE are protective alleles for IgAN. Genotypes were correlated with the corresponding gene expression, suggesting a cis-eQTL effect. Positive gene–gene interactions were observed, and disease risk decreased with accumulation of protective alleles. Four pathways (MHC class II antigen presentation, complement regulation, signaling by the BCR, and ubiquitin/proteasome-dependent degradation) were highlighted. From the “systems genetics” perspective, our data represent important clues for future studies on pleiotropy in IgA nephropathy and lupus nephritis.

Disclosures

None.