Genome-wide association meta-analyses and fine-mapping elucidate pathways influencing albuminuria

Abstract

Increased levels of the urinary albumin-to-creatinine ratio (UACR) are associated with higher risk of kidney disease progression and cardiovascular events, but underlying mechanisms are incompletely understood. Here, we conduct trans-ethnic (n = 564,257) and European-ancestry specific meta-analyses of genome-wide association studies of UACR, including ancestry- and diabetes-specific analyses, and identify 68 UACR-associated loci. Genetic correlation analyses and risk score associations in an independent electronic medical records database (n = 192,868) reveal connections with proteinuria, hyperlipidemia, gout, and hypertension. Fine-mapping and trans-Omics analyses with gene expression in 47 tissues and plasma protein levels implicate genes potentially operating through differential expression in kidney (including TGFB1, MUC1, PRKCI, and OAF), and allow coupling of UACR associations to altered plasma OAF concentrations. Knockdown of OAF and PRKCI orthologs in Drosophila nephrocytes reduces albumin endocytosis. Silencing fly PRKCI further impairs slit diaphragm formation. These results generate a priority list of genes and pathways for translational research to reduce albuminuria.

Fig. 1

Genome-wide association results. The circos plot provides an overview of the association results: Red band: –log₁₀(p) for association in the trans-ethnic meta-analysis of urinary albumin-to-creatinine ratio (UACR), ordered by chromosomal position. The blue line indicates genome-wide significance (p = 5 × 10⁻⁸). Black gene labels indicate novel loci, blue labels indicate known loci (known index SNP within ± 500 kb region of current index SNP), gray labels indicate loci not associated with UACR at the nominal significance level (p ≥ 0.05) in the 53 CKDGen cohorts without UKBB. Blue band: –log₁₀(p) for association with microalbuminuria (MA), ordered by chromosomal position. The red line indicates genome-wide significance (p = 5 × 10⁻⁸). Green band: measures of heterogeneity related to the UACR-associated index SNPs, where the dot sizes are proportional to two measures of heterogeneity, I² and the –log₁₀(p) for heterogeneity attributed to ancestry (pA)

Fig. 2

Internal concordance of the urinary albumin-to-creatinine ratio (UACR) results, and association with microalbuminuria, urinary creatinine and albumin. a Comparison of effect estimates of the 59 genome-wide significant trans-ethnic UACR index SNPs in the UKBB (x-axis) and in the CKDGen cohorts without UKBB (y-axis). Blue dots indicate nominal significance (p < 0.05) in the CKDGen cohorts without UKBB, and loci at genome-wide significance (p < 5 × 10⁻⁸) in that meta-analysis are labeled with the closest gene. b Comparison of effect estimates of the 59 trans-ethnic UACR index SNPs (x-axis) with their corresponding estimate from the GWAS of microalbuminuria (MA; y-axis). Blue dots indicate significance in the MA results after multiple testing correction (p < 0.05/59 = 8.5 × 10⁻⁴), and loci that achieved genome-wide significance (p < 5 × 10⁻⁸) for MA are labeled. In both panels, the dashed line represents the line of best fit through the effect estimates. c Comparison of effect estimates of the 59 genome-wide significant trans-ethnic UACR index SNPs for their effect on urinary creatinine (x-axis) and urinary albumin levels (y-axis) in the UKBB sample. Blue, red, and purple color indicate significant associations after multiple testing correction (p < 0.05/59 = 8.5 × 10⁻⁴) with urinary creatinine, urinary albumin, and both, respectively. Significant associations are labeled with the closest gene name. The dashed line represents the median y = x. In all panels, error bars indicate 95% confidence intervals (CIs), and the Pearson correlation coefficient r between the effect estimates is shown. The effect directions correspond to the effect allele of the trans-ethnic UACR meta-analysis results

Fig. 3

Phenome-wide association scan of a genetic urinary albumin-to-creatinine ratio (UACR) risk score. PheWAS association results were obtained from EA participants of the Million Veteran Program. Association test -log₁₀(p-values) are plotted on the y-axis, and the corresponding trait or disease category on the x-axis. Significant results, after correcting for the 1422 phenotypes tested (p < 0.05/1422 = 3.5 × 10⁻⁵), are labeled in the figure

Fig. 4

Genetic correlation of urinary albumin-to-creatinine ratio (UACR) with other traits and diseases. Significant (p < 9.7 × 10⁻⁵) genetic correlations based on the genome-wide summary statistics from the EA UACR GWAS and 517 pre-computed and publicly available GWAS summary statistics of UKBB traits and diseases, available through LDHub. Traits are shown on the x-axis, and colored according to broad physiological categories. Genetic correlations between traits and UACR are reported on the y-axis. Dot size is proportional to the –log₁₀(p) of the corresponding genetic correlation

Fig. 5

Fine-mapping and functional annotation of potentially causal variants. Overview of 995 SNPs with a posterior probability of association with urinary albumin-to-creatinine ratio (UACR) of >1%. The x-axis indicates the 99% credible set size and the y-axis the SNPs’ posterior probability of association. In panel a, missense SNPs are marked by triangles, with size proportional to the SNP CADD score. In panel b, SNPs are color-coded with respect to location in regulatory regions of specific kidney tissues. The labels show the closest gene, and are restricted to variants mapping to small credible sets (≤5 SNPs), or to variants with high individual posterior probability (>0.5) of driving the association signal. For the CUBN locus, a credible set was computed for each independent SNP

Fig. 6

Co-localization of associations signals for urinary albumin-to-creatinine ratio (UACR) and gene expression in kidney tissues. The plot shows the nine genes for which there is a high likelihood (posterior probability ≥ 80%) of a shared causal signal for gene expression in at least one of three kidney tissues and UACR. The loci are colored-coded and shown on the y-axis with the closest gene next to the index SNP. Co-localization with gene expression across all tissues (x-axis) is shown as dots, where the size of the dots (implying that eQTL data were available) corresponds to the posterior probability of the co-localization. The change in UACR is color-coded relative to the change in gene expression, or gray in case of a posterior probability < 80%

Fig. 7

Co-localization of association signals of the OAF locus. Regional association plots of the OAF locus in the European ancestry urinary albumin-to-creatinine ratio (UACR) GWAS (a), with OAF gene-expression levels in healthy kidney tissue sections (b), and with OAF plasma levels (c, d). The dots are colored according to their correlation r² with the index SNP estimated based on the 1000 Genomes EUR reference samples (gray for missing data). This locus has two independent pQTLs for OAF levels, where panel c shows the association between the index pQTL at the locus (rs117554512) conditioned on its secondary signal (indexed by rs508205), and panel d shows the association with a conditionally independent SNP (rs508205, r² < 0.01 in 1000 Genomes EUR). The secondary signal rs508205 has strong evidence of co-localization with the UACR association signal (posterior probability H₄ = 0.99, Methods), while the signal rs117554512 has not (posterior probability H₄ = 0). There was strong evidence of co-localization between the UACR association signal and OAF expression in kidney tissue (posterior probability H₄ = 0.97)

Fig. 8

In vivo results of Drosophila orthologs. The Drosophila orthologs of OAF and PRKCI (aPKC) are both required for nephrocyte function and aPKC-RNAi affects slit diaphragm formation. a Garland cell nephrocytes were exposed to FITC-albumin. Nephrocytes expressing control RNAi exhibit intense endocytosis, while expression of RNAi directed against oaf and aPKC (ortholog of PRKCI) decreases tracer uptake. b Quantitation of fluorescence intensity from FITC-albumin uptake is shown for the indicated genotypes. Values are presented as mean ± standard deviation of the ratio to a control experiment. Statistical significance was calculated using ANOVA and Dunnett’s post hoc analysis. A statistically significant difference (defined as p < 0.05) is observed for oaf-RNAi-1 (N = 4), oaf-RNAi-2 (N = 3), aPKC-RNAi-1 (N = 3), and aPKC-RNAi-2 (N = 4), where ** indicate p < 0.01 and ***p < 0.001. c Staining the slit diaphragm proteins Sns (ortholog of nephrin) and Kirre (ortholog of NEPH1) in control nephrocytes shows regular formation of slit diaphragms. Airyscan technology partially allows for distinguishing individual slit diaphragms (insets). d Tangential sections through the surface of control nephrocytes reveals the regular fingerprint-like pattern of slit diaphragm proteins. e, f Expression of oaf-RNAi-1 does not entail an overt phenotype, suggesting reduced nephrocyte function may be a consequence of impaired protein reabsorption while slit diaphragm formation is not affected. g, h Expression of aPKC-RNAi-1 results in a clustered and irregular pattern of slit diaphragm proteins (insets in g) and a complete loss of slit diaphragm protein distinct areas on the cell surface. This suggests the loss of nephrocyte function is a consequence of impaired slit diaphragm formation. All scale bars represent 10 µm