Identification of the DNA methylation signature of Mowat-Wilson syndrome

Abstract

Mowat-Wilson syndrome (MOWS) is a rare congenital disease caused by haploinsufficiency of ZEB2, encoding a transcription factor required for neurodevelopment. MOWS is characterized by intellectual disability, epilepsy, typical facial phenotype and other anomalies, such as short stature, Hirschsprung disease, brain and heart defects. Despite some recognizable features, MOWS rarity and phenotypic variability may complicate its diagnosis, particularly in the neonatal period. In order to define a novel diagnostic biomarker for MOWS, we determined the genome-wide DNA methylation profile of DNA samples from 29 individuals with confirmed clinical and molecular diagnosis. Through multidimensional scaling and hierarchical clustering analysis, we identified and validated a DNA methylation signature involving 296 differentially methylated probes as part of the broader MOWS DNA methylation profile. The prevalence of hypomethylated CpG sites agrees with the main role of ZEB2 as a transcriptional repressor, while differential methylation within the ZEB2 locus supports the previously proposed autoregulation ability. Correlation studies compared the MOWS cohort with 56 previously described DNA methylation profiles of other neurodevelopmental disorders, further validating the specificity of this biomarker. In conclusion, MOWS DNA methylation signature is highly sensitive and reproducible, providing a useful tool to facilitate diagnosis.

Fig. 1. Mowat-Wilson syndrome (MOWS) is characterized by a specific DNAm signature.

A Euclidean hierarchical clustering heatmap, each column represents one MOWS case or selected control, each row represents one probe selected for this episignature. The heatmap shows a clear separation between cases (in red) and controls (in blue), and properly classifies all validation samples (in orange) with the MOWS cases of the discovery cohort. B Multidimensional scaling (MDS) plot shows segregation of MOWS cases and controls. C Support Vector Machine (SVM) classifier model. The model was trained using the selected MOWS episignature probes, 75% of controls and 75% of other neurodevelopmental disorder samples (blue). The remaining 25% controls and 25% of other disorder samples were used for testing (grey). Plot shows that all MOWS cases have methylation variant pathogenicity (MVP) scores close to 1.

Fig. 2. Differentially methylated probes (DMPs) annotated in the context of CpG islands and genes.

A DMPs in relation to genes. B DMPs in CpG islands. Promoter, 0–1 kb upstream of the transcription start site (TSS); Promoter+, 1–5 kb upstream of TSS; CDS, coding sequence; Intergenic, all other regions of the genome. Island, CpG islands; Shore, within 0–2 kb of a CpG island boundary; shelf, within 2–4 kb of a CpG island boundary; Inter_CGI, all other regions in the genome. The “Probes” column in both A and B represents the background distribution of all array probes determined in the study by Levy et al. [24], considered after initial filtering and used as input for DMP analysis.

Fig. 3. Overlap between the MOWS episignature and the 56 other disorders included in the EpiSign v3 classifier.

A Global methylation profiles of all differentially methylated probes (DMPs, false discovery rate <0.05) for each cohort, sorted by mean methylation. Each circle represents a single probe, red lines show the mean methylation. B Heatmap showing the percentage of probes shared between each paired cohort. Colors indicate the percentage of the y-axis cohort’s probes that are also found in the x-axis cohort’s probes.

Fig. 4. Tree and leaf visualization of Euclidean clustering of all 57 cohorts using the top n DMPs for each group, where n = min (# of DMPs, 500).

Cohort samples were aggregated using the median value of each probe within a group. A leaf node represents a cohort, with node sizes illustrating relative scales of the number of selected DMPs for the corresponding cohort, and node colors are indicative of the global mean methylation difference.