Mendelian pathway analysis of laboratory traits reveals distinct roles for ciliary subcompartments in common disease pathogenesis

Abstract

Rare monogenic disorders of the primary cilium, termed ciliopathies, are characterized by extreme presentations of otherwise common diseases, such as diabetes, hepatic fibrosis, and kidney failure. However, despite a recent revolution in our understanding of the cilium's role in rare disease pathogenesis, the organelle's contribution to common disease remains largely unknown. Hypothesizing that common genetic variants within Mendelian ciliopathy genes might contribute to common complex diseases pathogenesis, we performed association studies of 16,874 common genetic variants across 122 ciliary genes with 12 quantitative laboratory traits characteristic of ciliopathy syndromes in 452,593 individuals in the UK Biobank. We incorporated tissue-specific gene expression analysis, expression quantitative trait loci, and Mendelian disease phenotype information into our analysis and replicated our findings in meta-analysis. 101 statistically significant associations were identified across 42 of the 122 examined ciliary genes (including eight novel replicating associations). These ciliary genes were widely expressed in tissues relevant to the phenotypes being studied, and eQTL analysis revealed strong evidence for correlation between ciliary gene expression levels and laboratory traits. Perhaps most interestingly, our analysis identified different ciliary subcompartments as being specifically associated with distinct sets of phenotypes. Taken together, our data demonstrate the utility of a Mendelian pathway-based approach to genomic association studies, challenge the widely held belief that the cilium is an organelle important mainly in development and in rare syndromic disease pathogenesis, and provide a framework for the continued integration of common and rare disease genetics to provide insight into the pathophysiology of human diseases of immense public health burden.

Keywords: Mendelian disease; bioinformatics; ciliopathies; cilium; complex trait; genomics.

Figure 1

Schematic of analysis workflow for the data presented Phenotype and genotype data from the UKBB release version 2 was extracted and processed as indicated. Linear association studies between ciliary gene variants and laboratory traits were performed, and significant associations (p < 2.5e−7) were studied further by tissue-specific expression, eQTL, and replication meta-analysis. Using this data, we performed a DiCE/pathway analysis to identify ciliary subcompartments associated with specific traits.

Figure 2

Association study and meta-analysis results for lipid-related traits Hudson plot illustrating the results of the multi-ancestry discovery association analysis (top) and replication meta-analysis (bottom) of common variants within 122 ciliary genes with lipid-related traits. Data for cholesterol are in teal, HDL are in red, LDL are in purple, and triglycerides are in orange. The study-wise Bonferroni-adjusted significance threshold for all analyses performed in our manuscript (p < 3.4e−8) is shown as a red dashed line, while the experiment-wise Bonferroni-adjusted significance threshold (p < 2.5e−7 for the discovery analysis, p < 5e−7 for the meta-analysis) is shown as an orange dashed line. Each analyzed gene is given equal space along the horizontal axis, and all tested genetic variants for a given gene are plotted at the midline of the gene block.

Figure 3

Association study and meta-analysis results for kidney-related traits Hudson plot illustrating the results of the multi-ancestry discovery association analysis (top) and replication meta-analysis (bottom) of common variants within 122 ciliary genes with kidney-related traits. Data for creatinine are in teal, and data for urea are in red. The study-wise Bonferroni-adjusted significance threshold for all analyses performed in our manuscript (p < 3.4e−8) is shown as a red dashed line, while the experiment-wise Bonferroni-adjusted significance threshold (p < 2.5e−7 for the discovery analysis, p < 5e−7 for the meta-analysis) is shown as an orange dashed line. Each analyzed gene is given equal space along the horizontal axis, and all tested genetic variants for a given gene are plotted at the midline of the gene block.

Figure 4

Association study and meta-analysis results for liver-related traits Hudson plot illustrating the results of the multi-ancestry discovery association analysis (top) and replication meta-analysis (bottom) of common variants within 122 ciliary genes with liver-related traits. Data for AlkPhos are in teal, ALT are in red, AST are in purple, and GGT are in orange. The study-wise Bonferroni-adjusted significance threshold for all analyses performed in our manuscript (p < 3.4e−8) is shown as a red dashed line, while the experiment-wise Bonferroni-adjusted significance threshold (p < 2.5e−7 for the discovery analysis, p < 3.2e−7 for the meta-analysis) is shown as an orange dashed line. Each analyzed gene is given equal space along the horizontal axis, and all tested genetic variants for a given gene are plotted at the midline of the gene block.

Figure 5

Association study and meta-analysis results for glucose-related traits Hudson plot illustrating the results of the multi-ancestry discovery association analysis (top) and replication meta-analysis (bottom) of common variants within 122 ciliary genes with glucose-related traits. Data for A1c are in teal, and data for glucose are in red. The study-wise Bonferroni-adjusted significance threshold for all analyses performed in our manuscript (p < 3.4e−8) is shown as a red dashed line, while the experiment-wise Bonferroni-adjusted significance threshold (p < 2.5e−7 for the discovery analysis, p < 5e−7 for the meta-analysis) is shown as an orange dashed line. Each analyzed gene is given equal space along the horizontal axis, and all tested genetic variants for a given gene are plotted at the midline of the gene block.

Figure 6

Tissue-specific expression analysis of ciliary genes Heatmap depicting the results of the tissue-specific expression analysis for each of the 42 ciliary genes with significant associations in our discovery analysis. The phenotype domain(s) significantly associated with each gene are indicated by the colored bars at the top of the graph. Tissues are listed to the left of the graph, hierarchically clustered by expression profile across all 42 genes, as displayed by the dendrogram on the right. Blue color scale corresponds to the transcripts per million (TPM) for each gene, per tissue, as obtained from the GTEx v8 database, and any TPM value less than 1.5 is displayed in black. Pink color scale corresponds to the standard deviation in expression for each gene across all tissues.

Figure 7

BBS1 variants significantly associated with LDL levels also serve as eQTLs for the BBS1 gene eQTpLot analysis of the BBS1 locus association with LDL levels. (A) Colocalization between phenotype-significant variants and eQTLs for BBS1 is visualized in chromosomal space. For each variant, significance of association with LDL levels is displayed along the vertical axis, and significance of association with BBS1 expression is indicated by color gradient. Any variant with p_eQTL for BBS1 <0.05 is displayed in color. The genome-wide p value significance threshold of 2.5e−7 for LDL association is indicated by a red line. Note that the variants most significantly associated with LDL levels are also most significantly associated with BBS1 expression. (B) A depiction of the genomic region surrounding the BBS1 locus. (C) Visualization of enrichment of eQTLs for the BBS1 gene among variants significantly associated with LDL. The p value of enrichment was determined by Fisher’s exact test. (D) P-P plot illustrating correlation between p_eQTL and p_trait for BBS1. Correlation between the two probabilities is visualized by plotting a best-fit linear regression over the points with the line equation displayed on the plot. The Pearson correlation coefficient and p value of correlation are displayed on the plot as well.

Figure 8

Schematics illustrating the results of DiCE/pathway analysis and a depiction of the primary cilium (A) To integrate the multiple lines of evidence supporting each ciliary gene’s association with a given phenotype domain, we employed a Diverse Convergent Evidence (DiCE) analysis approach estimating the strength of available corroborating data (detailed in Table S16). The DiCE scores for each gene, illustrating the strength of evidence supporting each ciliary gene’s association with a given phenotype domain, are displayed here; the magnitude of the score is indicated by the size of each gene bubble. For each phenotype domain, gene bubbles are displayed at the same location for ease of comparison. Genes are colored by ciliary subcompartment—green for IFT-related genes, pink for the BBSome, orange for the transition zone, and gray for the basal body. A stacked bar graph is also displayed for each phenotype domain, illustrating the summed DiCE scores for all genes, by ciliary subcompartment, as a proportion of the total DiCE score per phenotype domain. (B) A schematic of the primary cilium. The basal body and centrioles are displayed in gray, the ciliary transition zone in orange, the IFT machinery in green, and the BBSome in pink. The core of nine microtubule doublets and the overlying membrane of the organelle are shown, and receptors being trafficked into/along the cilium are shown in blue.