Indexing Effects of Copy Number Variation on Genes Involved in Developmental Delay

Abstract

A challenge in clinical genomics is to predict whether copy number variation (CNV) affecting a gene or multiple genes will manifest as disease. Increasing recognition of gene dosage effects in neurodevelopmental disorders prompted us to develop a computational approach based on critical-exon (highly expressed in brain, highly conserved) examination for potential etiologic effects. Using a large CNV dataset, our updated analyses revealed significant (P < 1.64 × 10(-15)) enrichment of critical-exons within rare CNVs in cases compared to controls. Separately, we used a weighted gene co-expression network analysis (WGCNA) to construct an unbiased protein module from prenatal and adult tissues and found it significantly enriched for critical exons in prenatal (P < 1.15 × 10(-50), OR = 2.11) and adult (P < 6.03 × 10(-18), OR = 1.55) tissues. WGCNA yielded 1,206 proteins for which we prioritized the corresponding genes as likely to have a role in neurodevelopmental disorders. We compared the gene lists obtained from critical-exon and WGCNA analysis and found 438 candidate genes associated with CNVs annotated as pathogenic, or as variants of uncertain significance (VOUS), from among 10,619 developmental delay cases. We identified genes containing CNVs previously considered to be VOUS to be new candidate genes for neurodevelopmental disorders (GIT1, MVB12B and PPP1R9A) demonstrating the utility of this strategy to index the clinical effects of CNVs.

Figure 1. Schematic of the analysis framework to identify candidate genes from copy number variation in developmental delay, using genome-wide human (prenatal and adult) brain transcriptome (RNA sequencing) and proteome data (Fourier-transform mass spectrometry).

The spatio-temporal transcriptome data was used to compute ‘brain critical exon’ analysis for all the genes in the genome. Quantified protein expression for each gene in the genome was used for weighted gene co-expression network analysis. To identify a candidate set of phenotypically relevant genes, integrated (transcriptome and proteome) analysis was conducted for genes impacted by rare CNVs in cases (pathogenic/VOUS) and controls. CVH, Credit Valley Hospital; VOUS, variant of unknown significance; DGV, Database of Genomic Variants.

Figure 2. Ascertainment of pathogenic variants or variants of uncertain significance (VOUS) in 10,619 developmental delay cases.

The rare CNVs of 30 kb to 5 Mb were classified as pathogenic variants or of VOUS. (A) Bars indicate the total number of CNVs in each classification. Of all samples assayed, 4.19% carried a pathogenic deletion, 1.81% a pathogenic duplication, 18.28% a VOUS deletion and 31.97% a VOUS duplication. The green line represents the number of unique genes impacted by the corresponding variants. (B) The percentage of male and female cases in the cohort impacted by pathogenic deletion variants. P value shown is for the one-sided Fisher’s exact test.

Figure 3. The fraction of critical exons (over all exons) computed from human prenatal brain regions for the genes impacted by pathogenic, VOUS and rare control deletion and duplication variants.

(A,B) The critical exon fraction was computed using gene expression level quantified from RNA sequencing in 392 brain tissues (controls) from 31 postmortem donors in 2 developmental periods (prenatal and adult) for 16 brain regions (AMY, amygdaloid complex; CBC, cerebellar cortex; V1C, primary visual cortex; STC, posterior (caudal) superior temporal cortex; IPC, posterior inferior parietal cortex; A1C, primary auditory cortex; S1C, primary somatosensory cortex; M1C, primary motor cortex; STR, striatum; DFC, dorsolateral prefrontal cortex; MFC, medial prefrontal cortex; VFC, ventrolateral prefrontal cortex; OFC, orbital frontal cortex; MD, mediodorsal nucleus of thalamus; ITC, inferolateral temporal cortex; HIP, hippocampus). The critical exon fraction was computed using prenatal brain transcriptome for the genes impacted by pathogenic (red dots) or VOUS (orange dots) or rare control deletions (gray dots).

Figure 4. Genome-wide protein co-expression and enrichment of genes ascertained from pathogenic and VOUS CNVs using human prenatal and adult tissues.

(A) The protein modules generated by weighted gene coexpression network analysis (WGCNA) using high resolution genome-wide Fourier-transform mass spectrometry data from 30 histologically normal human samples (prenatal and adult). Each colour (bars dispersed) represents a module. (B) For the blue module, the 20 most significant results from quantitative association with 18,826 gene sets. (C) Representation of the ‘blue’ module as a functional network, where each node is a gene; the edge between genes represents the weighted Pearson distance. Red nodes represent genes ascertained through CNVs in the developmental delay cohort. (D) The top (25^th percentile) critical exon genes in the genome ascertained using prenatal (green) and adult brain (purple) transcriptomes, and their corresponding quantified enrichment in the protein module, through Fisher’s exact test (FET) and 100,000 permutations. The orange bar represents the original observation of overlaps between blue module and a gene set. Similarly, (E,F) show the enrichment of genes impacted by pathogenic and VOUS duplications (blue) and deletions (red) within the protein module.

Figure 5. Deletions within the GIT1 gene identified in developmental disorder cases or controls.

(A) The breakpoints of 12 VOUS deletions (red) and duplications (blue) impacting GIT1 and nearby genes. The dataset includes 5 de novo deletions/duplications (denoted as DN) reported from developmental cases; no rare CNV was found in controls. (B) The analysis of the human protein co-expression network revealed that GIT1 is within the blue module and is highly connected (1-degree neighbors) with genes enriched for ‘critical exons’ (red nodes) and putative ASD genes reported to have de novo mutations (red node with black outline). (C) Expression of GIT1 (primer targeting critical exons) from quantitative real-time PCR (qRT-PCR) relative to housekeeping gene, MED13 (replicated with another housekeeping gene ACTB) in 11 different tissues.

Figure 6. Deletions within MVB12B gene identified in developmental disorder cases and controls.

(A) The breakpoints of 18 VOUS deletions (red) and duplications (blue) impacting MVB12B and nearby genes. The breakpoints include 12 de novo VOUS reported from developmental delay cases, including 3 recurrent breakpoints (6D-DN, 4G-DN, 3G). All de novo VOUS impacted the smaller isoform (highlighted by vertical dashed lines) of the gene (NM_001011703.2), and this was not impacted by CNV in controls. (B) The human protein co-expression network revealed that the MBV12B gene is the within the blue protein module and enriched for ‘critical exons’ (red nodes) and putative ASD genes reported to have de novo mutations (red node with black outline). (C) Expression of MVB12B (primer targeting critical exons) from quantitative real-time PCR (qRT-PCR) relative to housekeeping gene, MED13 (replicated with another housekeeping gene ACTB) in 11 different tissues.