Functional DNA methylation signatures for autism spectrum disorder genomic risk loci: 16p11.2 deletions and CHD8 variants

Abstract

Background: Autism spectrum disorder (ASD) is a common and etiologically heterogeneous neurodevelopmental disorder. Although many genetic causes have been identified (> 200 ASD-risk genes), no single gene variant accounts for > 1% of all ASD cases. A role for epigenetic mechanisms in ASD etiology is supported by the fact that many ASD-risk genes function as epigenetic regulators and evidence that epigenetic dysregulation can interrupt normal brain development. Gene-specific DNAm profiles have been shown to assist in the interpretation of variants of unknown significance. Therefore, we investigated the epigenome in patients with ASD or two of the most common genomic variants conferring increased risk for ASD. Genome-wide DNA methylation (DNAm) was assessed using the Illumina Infinium HumanMethylation450 and MethylationEPIC arrays in blood from individuals with ASD of heterogeneous, undefined etiology (n = 52), and individuals with 16p11.2 deletions (16p11.2del, n = 9) or pathogenic variants in the chromatin modifier CHD8 (CHD8^+/-, n = 7).

Results: DNAm patterns did not clearly distinguish heterogeneous ASD cases from controls. However, the homogeneous genetically-defined 16p11.2del and CHD8^+/- subgroups each exhibited unique DNAm signatures that distinguished 16p11.2del or CHD8^+/- individuals from each other and from heterogeneous ASD and control groups with high sensitivity and specificity. These signatures also classified additional 16p11.2del (n = 9) and CHD8 (n = 13) variants as pathogenic or benign. Our findings that DNAm alterations in each signature target unique genes in relevant biological pathways including neural development support their functional relevance. Furthermore, genes identified in our CHD8^+/- DNAm signature in blood overlapped differentially expressed genes in CHD8^+/- human-induced pluripotent cell-derived neurons and cerebral organoids from independent studies.

Conclusions: DNAm signatures can provide clinical utility complementary to next-generation sequencing in the interpretation of variants of unknown significance. Our study constitutes a novel approach for ASD risk-associated molecular classification that elucidates the vital cross-talk between genetics and epigenetics in the etiology of ASD.

Keywords: Autism spectrum disorder; DNA methylation; Epigenetics; Genetic stratification; Genomic variants; Heterogeneity.

Fig. 1

Whole blood DNAm comparison between heterogeneous ASD group (n = 52) and age-, sex-matched controls (n = 30). Following filtering by variance (40,550 sites), limma regression and Mann-Whitney U comparison, no CpG sites meet significance criteria of q ≤ 0.05. The 22 sites shown at uncorrected p < 0.001, |Δβ| ≥ 5% do not distinguish clearly between ASD cases and controls. Hierarchical clustering (Euclidian) and principal component analysis (PCA, first 3 principal components labeled) plot show that 16p11.2del (purple) and CHD8^+/− (orange) cases are mixed with both heterogeneous ASD cases (gray) and controls (green). Data are normalized for visualization (mean = 0, variance = 1). In heat map, yellow represents high methylation and blue represents low methylation

Fig. 2

a, b DNAm signatures identified in whole blood of individuals with 16p11.2del (600 kb risk locus) or CHD8^+/−. a Hierarchical clustering and PCA plot (first 3 principal components labeled) show that 16p11.2del training cases (purple; n = 9) are distinct from age-, sex-matched controls (green; n = 23) at the DNAm signature sites (115 CpG sites; q < 0.05, absolute methylation difference (|Δβ| ≥ 5%). b Hierarchical clustering and PCA plot (first 3 principal components labeled) show that CHD8^+/− training cases (orange; n = 7) are distinct from age-, sex-matched controls (green; n = 21) at DNAm signature sites (103 CpG sites; q < 0.01, |Δβ| ≥ 5%) used for classification. Only a single CpG site overlaps between the 16p11.2del and CHD8^+/− DNAm signatures: cg25970491 (CLTCL1), which is hypomethylated in both CHD8^+/− and 16p11.2del cases relative to controls. All data are normalized for visualization (mean = 0, variance = 1). In heat map, yellow represents high methylation and blue represents low methylation

Fig. 3

a–d Evaluating the sensitivity and specificity of the 16p11.2del and CHD8^+/− DNAm signatures. The classification models (details in Ref. [15]) are represented by the following classification plots: a The classification model based on the 16p11.2del DNAm signature sites (115 sites; q < 0.05, |Δβ| ≥ 5%) accurately classified CHD8^+/− signature cases, and independent test cases consisting of CHD8 sequence variants (orange box with cross) and heterogeneous ASD cases (open gray triangle) as more similar to controls (green C). The model also accurately classified 7 of 9 independent 16p11.2del test cases (solid purple circle) as more similar to 16p11.2del signature cases (open purple circle), demonstrating 100% sensitivity. Two variants (arrows) correctly received negative classification scores: hatched arrow indicates a mosaic 16p11.2del case (2-0088-003), solid arrow indicates a 16p11.2del distal to and not overlapping the 600 kb typical deletion region (1-0616-003). b The 16p11.2del DNAm signature was tested on an independent set of compiled blood control DNAm data (n = 162) extracted from the Gene Expression Omnibus (GEO) Database. All GEO controls (turquoise C) were properly classified with experimental controls (green C), as distinct from 16p11.2del signature training cases (open purple circle), demonstrating 100% specificity. c The classification model based on the CHD8^+/− DNAm signature sites for classification (103 sites; q < 0.01, |Δβ| ≥ 5%) accurately classified 16p11.2del signature cases and independent test cases consisting of 16p11.2del variants (purple circle with cross) and heterogeneous ASD cases (gray triangle) as more similar to controls (green C). The model also accurately classified 1 of 9 independent CHD8 test cases consisting of sequence variants (solid orange square) as more similar to CHD8^+/− signature training cases (open orange square) (unable to report sensitivity with one positive case, indicated by arrow). Eight variants correctly received negative classification scores (7 missense, 1 with an in-frame deletion in the last exon of CHD8, not predicted pathogenic). Classifications are in agreement with in silico predictions. d The CHD8^+/− DNAm signature was tested on the same GEO controls (turquoise C) as in b; all but one GEO control were properly classified with experimental controls (green C), as distinct from CHD8^+/− signature cases (open orange square), demonstrating >99.3% specificity

Fig. 4

Evaluating the sensitivity of the CHD8^+/− DNAm signature using additional CHD8 test cases (EPIC array). Additional independent test cases consisting of CHD8 sequence variants (n = 4; orange squares with cross) were run on the EPIC array. Of the 103 sites in the 450K-derived CHD8^+/− DNAm classification signature, 92 sites overlapped those of the EPIC array and were thus used to classify the additional EPIC test cases. One VUS case received a negative classification score. Three of the four additional CHD8 test cases were accurately classified as more similar to CHD8^+/− signature training cases (open orange square); these cases were known to have pathogenic CHD8 variants, demonstrating 100% sensitivity. The one sample classified as benign was an inherited missense variant, similar to those 450K test cases receiving negative scores. All other cases and controls (450K) were accurately classified as in Fig. 3

Fig. 5

Biological and cross-tissue functional significance of differentially methylated genes. a Differentially methylated genes associated with the 16p11.2del and CHD8^+/− DNAm signatures (q < 0.05, |Δβ| ≥ 5%) and the DNAm signature overlap with known ASD-risk genes (SFARI Gene). A single SFARI gene, clathrin heavy chain like 1 (CLTCL1), is significantly hypomethylated in both groups. An independent set of 9 SFARI genes are differentially methylated in the CHD8^+/− DNAm signature. b Our findings are further corroborated by an independent study (48) showing that CHD8^+/− human iPSC-derived neuronal precursor cells (NPCs) and differentiated neurons result in differentially expressed genes that overlap with some of the differentially methylated genes in CHD8^+/− DNAm signature, including known ASD-risk genes (SFARI Gene), and c another study (49) showing that cerebral organoids derived from the iPSCs in b also have differentially expressed genes that overlap our CHD8^+/− DNAm signature.