Variants in tubule epithelial regulatory elements mediate most heritable differences in human kidney function

Abstract

Kidney disease is highly heritable; however, the causal genetic variants, the cell types in which these variants function, and the molecular mechanisms underlying kidney disease remain largely unknown. To identify genetic loci affecting kidney function, we performed a GWAS using multiple kidney function biomarkers and identified 462 loci. To begin to investigate how these loci affect kidney function, we generated single-cell chromatin accessibility (scATAC-seq) maps of the human kidney and identified candidate cis-regulatory elements (cCREs) for kidney podocytes, tubule epithelial cells, and kidney endothelial, stromal, and immune cells. Kidney tubule epithelial cCREs explained 58% of kidney function SNP-heritability and kidney podocyte cCREs explained an additional 6.5% of SNP-heritability. In contrast, little kidney function heritability was explained by kidney endothelial, stromal, or immune cell-specific cCREs. Through functionally informed fine-mapping, we identified putative causal kidney function variants and their corresponding cCREs. Using kidney scATAC-seq data, we created a deep learning model (which we named ChromKid) to predict kidney cell type-specific chromatin accessibility from sequence. ChromKid and allele specific kidney scATAC-seq revealed that many fine-mapped kidney function variants locally change chromatin accessibility in tubule epithelial cells. Enhancer assays confirmed that fine-mapped kidney function variants alter tubule epithelial regulatory element function. To map the genes which these regulatory elements control, we used CRISPR interference (CRISPRi) to target these regulatory elements in tubule epithelial cells and assessed changes in gene expression. CRISPRi of enhancers harboring kidney function variants regulated NDRG1 and RBPMS expression. Thus, inherited differences in tubule epithelial NDRG1 and RBPMS expression may predispose to kidney disease in humans. We conclude that genetic variants affecting tubule epithelial regulatory element function account for most SNP-heritability of human kidney function. This work provides an experimental approach to identify the variants, regulatory elements, and genes involved in polygenic disease.

Figure 1:. Kidney candidate cis-regulatory elements are enriched for kidney function heritability.

a) Manhattan plot for the estimated GFR calculated from serum creatinine and cystatin C levels (eGFR_Cr-Cys) genome wide association study (GWAS) performed within the UK Biobank. Significant loci are colored to indicate whether they are unique to the eGFR_Cr-Cys GWAS or are found within indicated single biomarker GWAS. Labels at the top indicate the closest gene to the index variant. The y-axis is a −log₁₀(P) scale up to 30, after which it switches to a −log₁₀[−log₁₀(P)] scale. b) ATAC-seq of human kidney cortex identifies kidney noncoding candidate cis-regulatory elements (cCREs). The contribution of kidney cCREs to human traits is evaluated by calculating trait heritability cCRE enrichment using stratified LD score regression. c) Fold enrichment of the SNP heritability of fifteen human traits within kidney cCREs calculated with stratified LD score regression. Error bars denote jackknife standard errors for the enrichment estimate. These errors were used to calculate P values, which were Bonferroni corrected. * P<0.05. d) Several kidney cCREs lie in introns of PDILT 5’ to UMOD, which encodes a mendelian kidney disease gene. rs77924615 is associated with kidney function (eGFR_Cr-Cys), is in a kidney cCRE, and has low LD with all other variants at the locus. Degree of LD to rs77924615 is indicated by variant color.

Figure 2:. Single-cell ATAC-seq of human kidneys identifies cell type-specific cCREs.

a) Schematic of the nephron, colored by major epithelial cell types. b) scATAC-seq UMAP of 34,240 kidney cells from 3 donors. c) Motifs of transcription factors expressed in the kidney are enriched in cCREs of specific kidney cell types. Chromatin accessibility for bolded transcription factors is shown in d-f. d-f) Chromatin accessibility maps for three cell type-specific genes: HNF4A, expressed in the proximal tubule; TFAP2B, expressed in the loop of Henle and distal tubule; and SPI1, expressed in immune cells.

Figure 3:. Kidney tubule cCREs account for the majority of SNP heritability of kidney function biomarkers.

a) Clustering of cCREs distinguishes cell type-specific and ubiquitously accessible cCREs. Each vertical line represents one cCRE. A total of 526,273 cCREs are depicted. IC, intercalated cell. b-c) Fold enrichment of SNP heritability of eGFR_Cr-Cys and albuminuria within cell type-specific and ubiquitously accessible cCREs calculated with stratified LD score regression. Error bars denote jackknife standard errors for the enrichment estimate. These errors were used to calculate P values, which were Bonferroni corrected for multiple hypothesis testing. * P<0.05. d) Fraction of heritability of eGFR_Cr-Cys explained by kidney tubule epithelial cell cCREs (blue), coding exons (beige), and the remainder of the human genome (gray). e) rs77924615, a variant associated with eGFR_Cr-Cys level, lies with a tubule epithelial-specific cCRE (gray).

Figure 4:. Identification of causal kidney function variants using functionally informed fine-mapping.

a) Workflow of functionally-informed fine-mapping of variants affecting eGFR_cr-cys. GWAS for eGFRcr-cys identified 462 lead variants. 385 of these variants had consistent effects on both serum creatinine and cystatin C levels. Fine-mapping of these loci incorporated annotation of kidney tubule cCREs, evolutionary conservation, and gene annotations including annotations indicating whether a variant is coding, affects protein sequence (nonsynonymous), is in an intron, or is in a UTR. PIP, posterior inclusion probability. b) eGFR_cr-cys index variants plotted by the estimated size of their effect on both serum creatinine and cystatin C levels. Variants colored blue affect both creatinine and cystatin C levels in the same direction. c) Comparison of the number of fine-mapped kidney function variants identified with or without functional annotations. Variants are separated into those with moderate posterior inclusion probability (0.3–0.7) and high posterior inclusion probability (>0.7). d) Fold enrichment of fine-mapped kidney function variants that cause nonsynonymous changes in a coding sequence, are evolutionarily conserved, or that lie within tubule epithelial-specific cCREs. Variants are separated into those with high (>0.7), moderate (0.3–0.7) and low (0.05–0.3) posterior inclusion probabilities. e) Functionally informed fine-mapping at the SHROOM3 locus. The functionally informed posterior causal probability is plotted for variants, which are colored according to the causal probability estimated without functional annotations. Chromatin accessibility from scATAC-seq for the indicated kidney cell types are plotted below. Gray boxes indicate tubule epithelial cCREs containing fine-mapped variants.

Figure 5:. Chromatin accessibility allelic imbalance and machine learning identify cell type-specific effects of variants on chromatin accessibility.

a) Workflow to detect chromatin accessibility allelic imbalance (CAAI). Single-cell ATAC-seq reads were analyzed at heterozygous alleles and accessibility at both alleles was compared to measure the effect of variants on chromatin accessibility. b) Enrichment of proximal tubule expressed transcription factor-binding motifs at cCREs exhibiting CAAI in the proximal tubule. c) Schematic of ChromKid, a convolutional neural network trained to predict cell type-specific chromatin accessibility for 10 kidney cell types from DNA sequence. ChromKid uses a 1344bp DNA sequence input to generate a prediction of the quantitative chromatin accessibility at the center of the input region for each kidney cell type. d) Measured versus ChromKid-predicted proximal tubule chromatin accessibility on chromosome 11, which was withheld from training. The best fit line is shown. e) Receiver operating characteristic (ROC) curves for the direction of proximal tubule chromatin accessibility imbalance, displaying the false positive rate (x-axis) versus true positive rate (y-axis) of cell-type specific predictions from ChromKid. The areas under the curves (AUCs) for ChromKid predictions for each cell type are shown in the legend. ChromKid predictions of variant effects in each cell type are compared to proximal tubule CAAI. f) Fine-mapped variant rs12509595 (indicated by a vertical line) maps to a proximal tubule cCRE near FGF5. CAAI analysis revealed the cCRE with the alternate allele exhibits increased chromatin accessibility in the proximal tubule. ChromKid also predicted that the alternate allele would show increased chromatin accessibility at this cCRE in the proximal tubule. A motif in this cCRE corresponds to a consensus binding motif for the proximal tubule transcription factor RXRA. The variant, indicated in red, affects this binding motif. The alternative allele sequence more closely matches the consensus binding motif for RXRA. Statistical significance was calculated with a two-sided binomial test.

Figure 6:. Genetic variants affecting kidney function alter tubule epithelial cis-regulatory element function.

a) Fine-mapped kidney function variants exhibit larger predicted effects on accessibility than other variants within tubule epithelial cCREs. ChromKid-generated predictions of CAAI at fine-mapped kidney function (eGFR_cr-cys) variants within tubule cCREs stratified by PIP. Predicted CAAI is shown as the $\mathrm{m}\mathrm{a}\mathrm{x}\left|0.5-\left(\frac{ref}{ref+alt}\right)\right|$ across tubule epithelial cell types. Box boundaries indicate the 1^st and 3^rd quartile and whiskers indicate the most extreme data point within 1.5 times the interquartile range. Statistical significance was calculated using the Mann Whitney U test. b-d) Left: Plots of scATAC-seq data near fine-mapped kidney function variants, which are depicted by vertical lines. cCREs (gray) containing a kidney function variant were selected for functional testing. Top right: ChromKid predicted proximal tubule chromatin accessibility for both alleles at these cCREs. Predicted chromatin accessibility for the reference allele is depicted in dark green, and for the alternative allele is depicted in light green. Bottom right: Activity of enhancers in human primary tubule epithelial cells. Enhancer activity of the reference allele is depicted in dark green, and of the alternative allele is depicted in light green. Results are representative of three independent experiments. Statistical significance was based on a two-sided t test, ** P<0.01.

Figure 7:. Kidney function enhancers regulate RBPMS and NDRG1.

a) Schematic of the approach to map genes regulated by kidney function regulatory elements using CRISPRi-mediated silencing. cCREs containing fine-mapped kidney function variants were targeted in human primary tubule epithelial cells with CRISPRi. Gene expression and guide expression were measured by scRNA-seq. b) UMAP of scRNA-seq for 24,563 primary tubular epithelial cells from 4 donors used for CRISPRi-mediated silencing of kidney function cCREs. c) Volcano plot of differentially expressed genes in cells with expression of guides targeting kidney function cCREs. d-e). Left: Plots of proximal tubule chromatin accessibility from scATAC-seq data with the CRISPRi-targeted cCRE depicted with a black rectangle. Right: Violin plots of gene expression in cells expressing (green) or not expressing (gray) guides targeting the indicated regulatory element. Superimposed box plots indicate the median, 25^th, and 75^th percentiles of expression. All P values are adjusted for multiple hypothesis testing using the Bonferroni correction.