The defining DNA methylation signature of Kabuki syndrome enables functional assessment of genetic variants of unknown clinical significance

Abstract

Kabuki syndrome (KS) is caused by mutations in KMT2D, which is a histone methyltransferase involved in methylation of H3K4, a histone marker associated with DNA methylation. Analysis of >450,000 CpGs in 24 KS patients with pathogenic mutations in KMT2D and 216 controls, identified 24 genomic regions, along with 1,504 CpG sites with significant DNA methylation changes including a number of Hox genes and the MYO1F gene. Using the most differentiating and significant probes and regions we developed a "methylation variant pathogenicity (MVP) score," which enables 100% sensitive and specific identification of individuals with KS, which was confirmed using multiple public and internal patient DNA methylation databases. We also demonstrated the ability of the MVP score to accurately reclassify variants of unknown significance in subjects with apparent clinical features of KS, enabling its potential use in molecular diagnostics. These findings provide novel insights into the molecular etiology of KS and illustrate that DNA methylation patterns can be interpreted as 'epigenetic echoes' in certain clinical disorders.

Keywords: DNA methylation; KDM6A; KMT2D; Kabuki syndrome; variant classification.

Figure 1.

Hypomethylation of MYO1F gene body in KS patients with KMT2D loss of function mutations. The top track represents the gene transcript. The middle track (rectangle) shows a CpG island. The lower track shows the average methylation levels ranging 0–1 for controls (purple) and KS patients with loss of function (red). Every breakpoint in the lines shows a CpG probe.

Figure 2.

Hypomethylation of AGAP2 body and 3’UTR in KS patients with KMT2D loss of function mutations. The top track represents the gene transcript. The gray rectangle denotes a CpG island. The middle and lower tracks show the average methylation levels ranging 0–1. Every breakpoint in the lines represents a CpG probe. The top methylation track shows that the average methylation level in subjects with benign variants (blue) is comparable to the healthy controls (purple), while the mean methylation levels of those with VUS (gray) are located in between the two groups. Individual examinations of the seven VUS variants in the lower methylation track reveal that, indeed, four of the patients (red) have a methylation level similar to the KMT2D loss of function group, while the remaining three (purple) resemble the normal cohort.

Figure 3.

Hierarchical clustering of subjects with KMT2D pathogenic, benign and VUS variants and controls. Rows represent 1,504 CpG probes. Dark blue shows higher methylation, while light blue represents lower levels of methylation. The color is indexed by the rows. (A) columns represent 24 KS samples with KMT2D loss of function mutations (blue bar), 216 controls (red bar), and 11 individuals with benign or likely benign variants (green bar) (B) columns represent 24 KS samples with KMT2D loss of function mutations (blue bar), 216 controls (red bar), and 7 individuals with variants of unknown significance (green bar). All carriers of benign and three VUS variants are clustered with the controls, whereas the other four VUS carriers are grouped with the patients.

Figure 4.

Probability scores generated by the classification model. Control T: 216 controls used to describe the signature and train the model. Control V1: 288 controls used to measure the specificity of the model. Control V2: 125 controls downloaded from the GEO to confirm the specificity of the model. Kabuki T: 24 Kabuki patients with pathogenic mutations in KMT2D used for identification of the signature and training of the model. Kabuki V: 19 Kabuki patients with pathogenic mutations in KMT2D downloaded from GEO to measure the sensitivity of the model. KMT2D B: 11 individuals with benign variants in KMT2D. KMT2D VUS: 7 samples with variants of unknown significance in KMT2D. Other syndromes: 241 patients with confirmed clinical and molecular diagnosis of various Mendelian disorders resulting from defects in epigenomic machinery. Y-axis represents scores 0–1, with higher scores indicating a higher chance of carrying a pathogenic mutation leading to Kabuki syndrome as stratified by categories in X-axis. By default, the SVM classifier defines a cut-off of 0.5 for predicting the class; however, the vast majority of the tested individuals received a score close to 0 or 1. Therefore, for the purpose of better visualization, the points are jittered.