NSD1+/- DNA methylation (DNAm) signature: A novel functional diagnostic tool for Sotos syndrome

Abstract

Sotos syndrome (SS) is characterized by somatic overgrowth and intellectual disability. Most SS cases have mutations in NSD1 (nuclear receptor-binding SET domain protein 1), a histone lysine methyltransferase. To determine if NSD1 mutations impact stable epigenetic marks such as DNAm, we compared DNAm in peripheral blood from SS cases with NSD1 mutations (NSD1+/-; n=20) to controls (n=53) using the Illumina Infinium450methylation BeadChip (450k array). Differential DNAm analysis using non-parametric statistics (with correction for multiple testing) identified a surprisingly high number of differentially methylated (DM) CG sites between SS and controls. The majority (99.3%) of these sites demonstrated loss of DNAm and were distributed across the genome. Using unsupervised hierarchical clustering of the significant DM sites, all SS cases with loss of function mutations in NSD1 clustered as a distinct group. The specificity of this signature was 100%; in comparison to DNAm profiles of 450k data from the GEO database for blood samples (n= 1200). The sensitivity of the NSD1+/- signature was tested in an independent replication cohort of 19 SS cases from Hong Kong with NSD1 loss of function mutations. The NSD1+/- signature demonstrated a sensitivity of 100%; highlighting its exceptional power in defining pathogenicity for mutations in NSD1. The classification of single nucleotide substitutions into benign or disease causing (damaging) variants represents an ongoing challenge in clinical diagnostics. To test the hypothesis that the NSD1+/- DNAm signature will inform the functional classification of NSD1 variants into benign or disease causing, we analyzed 16 cases with reported missense variants in NSD1. Using hierarchical clustering we classified these variants as pathogenic (n=9) or benign (n=7). Clinical re-assessment of 11/16 of these cases, for whom photos and adequate clinical information were available, was conducted by two experienced dysmorphologists (RW and DC) who were blinded to the methylation results. There was 100%; concordance between clinical impression and DNAm data. In comparison, 4/5 algorithms (SIFT, Polyphen-2, etc) were inconsistent in their prediction of pathogenicity. Our data suggest that the NSD1+/- DNAm signature reflects the functional effect of NSD1 variants on the methylome and can be used as a more robust predictor than existing algorithms for the functional classification of NSD1 variants in overgrowth disorders.