Exome analysis links kidney malformations to developmental disorders and reveals causal genes

Abstract

Congenital anomalies of the kidneys and urinary tract (CAKUT) are developmental disorders that commonly cause pediatric chronic kidney disease and mortality. We examine here rare coding variants in 248 CAKUT trios and 1742 singleton CAKUT cases and compare them to 22,258 controls. Diagnostic and candidate diagnostic variants are detected in 14.1% of cases. We find a significant enrichment of rare damaging variants in constrained genes expressed during kidney development and in genes associated with other developmental disorders, suggesting phenotype expansion. Consistent with these data, 18% of CAKUT patients with diagnostic variants have neurodevelopmental or cardiac phenotypes. We identify 40 candidate genes, including CELSR1, SSBP2, XPO1, NR6A1, and ARID3A. Two are confirmed as CAKUT genes: ARID3A and NR6A1. This study suggests that many yet-unidentified syndromes would be discoverable with larger cohorts and cross-phenotype analysis, leading to clarification of the genetic and phenotypic spectrum of developmental disorders.

Fig. 1. Diagnostic analysis in 1990 individuals.

a Percentages of cases with diagnostic findings (green) and a candidate diagnosis (dark blue: VUS-high in a known gene, P/LP in a candidate gene or candidate structural variant) and those without a diagnostic or candidate variant (black). b Number of cases with an LP/P variant diagnostic or candidate finding (cases with two variants were counted once, with the variant most likely associated with their form of CAKUT) per gene and CAKUT (red: cystic kidney disease, black: renal hypoplasia and light blue: renal agenesis). c number of cases with a diagnostic or candidate structural variant per genomic area and CAKUT (red: cystic kidney disease, black: renal hypoplasia and light blue: renal agenesis). d Proportion of cases with diagnostic/candidate findings based on clinical characteristics (green: diagnostic findings and dark blue: candidate diagnosis). Diagnostic rate comparisons were performed using two-sided chi-square tests: (i) Bilateral vs. unilateral CAKUT: 20% vs. 7% (Chi-square p-value = 1.8 × 10⁻⁷; n = 1249). (ii) Females vs. males: 11% vs. 9% (Chi-square p-value = 3.8 × 10⁻³; n = 1710). (iii) Extra-renal anomalies vs. isolated CAKUT: 12% vs. 9% (Chi-square p-value = 0.02; n = 1710). (iv) Kidney phenotype: 13% (cystic dysplastic), 7% (agenesis), 12% (hypodysplasia) (Chi-square p-value = 1.9 × 10⁻³; n = 1710). Asterisks (*) indicate statistical significance. No adjustments for multiple comparisons were applied.

Fig. 2. Gene-burden and gene-set enrichment analyses.

a The gene-burden analysis was performed by extracting the number of cases and controls with and without a qualifying variant (QV) per gene. The exact two-sided Cochran–Mantel–Haenszel (CMH) test was used to test for enrichment of qualifying variants in cases vs controls, while controlling for cluster. Quantile–Quantile probability plot of the p-values generated by the gene-burden analysis focused on qualifying variants (LoF and predicted deleterious missense). Red indicates case-enriched p-values. The green lines represent the 95% confidence interval. Empiric confidence interval distributions created by permutations (n = 1000). b List of the ten most significant genes.

Fig. 3. Gene-set enrichment analyses and distribution of the number of qualifying variants per individual.

a Forest plot depicting the gene-set analysis using six gene sets based on three gene burden analyses. b Number of cases (black) and controls (Ctrl: gray) with 0, 1, 2, or at least 3 qualifying variants (LoF and predicted deleterious missense variants in constrained genes). Statistical significance of the difference between cases and controls was assessed using a two-sided Wilcoxon rank sum test. A significant difference was observed (OR = 1.1, p-value = 1.2 × 10⁻²⁸). c Number of cases with extra-renal phenotypes (dark gray) and cases without known extra-renal phenotypes (light gray) with 0, 1, 2, or at least 3 qualifying variants (LoF and predicted deleterious missense variants in constrained genes). Statistical significance of the difference between cases with and without known extra-renal phenotypes was assessed using a two-sided Wilcoxon rank sum test (OR = 1.1; p-value = 3.03 × 10⁻²). No adjustment for multiple comparisons was applied. CHD congenital heart disorder, ID/ASD intellectual delay and autism spectrum disorder, LoF loss-of-function variants, mis missense variants, NPCs nephron progenitor cells, OR odds ratio, syn synonymous variants. Predicted deleterious missense variants: subRVI domain score percentile < 50% and predicted damaging based on at least 2 of the following criteria: (1) REVEL > 0.5; (2) PrimateAI > 0.8; (3) AlphaMissense score > 0.6; and (4) CADD score > 25. Constrained genes had pLI ≥ 0.9 and LoeF < 0.35 and/or misZ > 3.09.

Fig. 4. Missense variants in NR6A1.

a Four independent Columbia CAKUT cases with predicted deleterious missense variants in NR6A1. b Family carrying the p.Arg92Trp (R92W) variant in NR6A1. (I: 1 pelvic and small kidney, II: 1 renal agenesis and Eye coloboma, III: 3 renal agenesis and eye coloboma, and IV: 2 renal agenesis and unknown genetic status). c Crystal structure of mouse NR6A1 and location of the three amino acids in which missense variants were identified in the four cases. d Modeling the potential impact of two of the three missense variants on the protein structure (R92W and R116C are modeled on the NR6A1 crystal structure PDB ID: 5KRB), the potential structural consequences of R66H could not be determined from the published NR6A1 structure.