A new blood DNA methylation signature for Koolen-de Vries syndrome: Classification of missense KANSL1 variants and comparison to fibroblast cells

Abstract

Pathogenic variants in KANSL1 and 17q21.31 microdeletions are causative of Koolen-de Vries syndrome (KdVS), a neurodevelopmental syndrome with characteristic facial dysmorphia. Our previous work has shown that syndromic conditions caused by pathogenic variants in epigenetic regulatory genes have identifiable patterns of DNA methylation (DNAm) change: DNAm signatures or episignatures. Given the role of KANSL1 in histone acetylation, we tested whether variants underlying KdVS are associated with a DNAm signature. We profiled whole-blood DNAm for 13 individuals with KANSL1 variants, four individuals with 17q21.31 microdeletions, and 21 typically developing individuals, using Illumina's Infinium EPIC array. In this study, we identified a robust DNAm signature of 456 significant CpG sites in 8 individuals with KdVS, a pattern independently validated in an additional 7 individuals with KdVS. We also demonstrate the diagnostic utility of the signature and classify two KANSL1 VUS as well as four variants in individuals with atypical clinical presentation. Lastly, we investigated tissue-specific DNAm changes in fibroblast cells from individuals with KdVS. Collectively, our findings contribute to the understanding of the epigenetic landscape related to KdVS and aid in the diagnosis and classification of variants in this structurally complex genomic region.

Fig. 1. Schematic representation of genotype information for all individuals included in this study.

A Schematic of the genomic location and type of single nucleotide variant in KANSL1 [Hg19, NM_001193466.4] generated using ProteinPaint. The number in each lollipop represents the number of individuals with that variant. B Schematic of the size of microdeletions in four individuals in this study generated using UCSC Genome Browser. Individuals EX1-EX8 are a part of the discovery cohort and individuals EX9-EX16 are part of the validation cohort.

Fig. 2. Haploinsufficiency in KANSL1 causing KdVS is associated with a distinct DNAm signature.

A Principal component analysis (PCA) and B heatmap showing clustering of the KdVS discovery subjects (n = 8; purple) and control discovery subjects (n = 21; grey) using DNAm values at the 456 CpG sites identified in the KdVS DNAm signature. The heatmap color gradient indicates the normalized DNAm value ranging from −2.0 (blue) to 2.0 (yellow). Euclidean distance metric is used in the heatmap clustering dendrograms.

Fig. 3. Classification of variants using SVM machine learning model based on the KdVS DNAm signature.

Sample groups were scored using the KdVS support vector machine (SVM) model. KdVS syndrome validation subject (n = 8) classified as KDVS-like demonstrating high sensitivity of the model. Whereas validation control subjects (n = 200) classified with low probabilities demonstrating high specificity of the model. One KANSL1 missense variant [p.(Gly900Glu); EX17] has a control-like probability score and the other missense variant [p.(Thr887Met); EX18] has a KdVS-like probability score. Four additional variants in KANSL1 and the 17q21.31 region in individuals [EX19-22] with atypical genotype-phenotype correlations all have control-like probability scores.

Fig. 4. KANSL1 protein tolerance landscape and predicted model structure.

A The diagram illustrates the landscape of KANSL1’s tolerance to missense changes according to MetaDome web server. Positions for missense variants analyzed in the present work are indicated in green. B The predicted structure which includes KAT8 shown in grey, acetylCoA in cyan, Zn in red, and the KANSL1 fragment in orange with Thr887 residue is show in dark blue, and Gly900 in dark green.

Fig. 5. Significant DNA methylation changes in fibroblast cells from individuals with KdVS.

A Principal component analysis (PCA) and B heatmap showing clustering of the KdVS discovery subjects (n = 3; pink) and control discovery subjects (n = 7; grey) using DNAm values at the 43 significant CpG sites identified. The heatmap color gradient indicates the normalized DNAm value ranging from −2.0 (blue) to 2.0 (yellow). Euclidean distance metric is used in the heatmap clustering dendrograms.