Epigenotype-genotype-phenotype correlations in SETD1A and SETD2 chromatin disorders

Abstract

Germline pathogenic variants in two genes encoding the lysine-specific histone methyltransferase genes SETD1A and SETD2 are associated with neurodevelopmental disorders (NDDs) characterized by developmental delay and congenital anomalies. The SETD1A and SETD2 gene products play a critical role in chromatin-mediated regulation of gene expression. Specific methylation episignatures have been detected for a range of chromatin gene-related NDDs and have impacted clinical practice by improving the interpretation of variant pathogenicity. To investigate if SETD1A and/or SETD2-related NDDs are associated with a detectable episignature, we undertook targeted genome-wide methylation profiling of > 2 M CpGs using a next-generation sequencing-based assay. A comparison of methylation profiles in patients with SETD1A variants (n = 6) did not reveal evidence of a strong methylation episignature. A review of the clinical and genetic features of the SETD2 patient group revealed that, as reported previously, there were phenotypic differences between patients with truncating mutations (n = 4, Luscan-Lumish syndrome; MIM:616831) and those with missense codon 1740 variants [p.Arg1740Trp (n = 4) and p.Arg1740Gln (n = 2)]. Both SETD2 subgroups demonstrated a methylation episignature, which was characterized by hypomethylation and hypermethylation events, respectively. Within the codon 1740 subgroup, both the methylation changes and clinical phenotype were more severe in those with p.Arg1740Trp variants. We also noted that two of 10 cases with a SETD2-NDD had developed a neoplasm. These findings reveal novel epigenotype-genotype-phenotype correlations in SETD2-NDDs and predict a gain-of-function mechanism for SETD2 codon 1740 pathogenic variants.

Figure 1

Clustering of SETD2 and SETD1A based on methylation episignatures. Unsupervised PCA clustering results for SETD1A and SETD2 group. Sample name and group annotations were applied by each group after PCA. Dotted line: group names annotation, Solid line: clustering by ‘stat_ellipse’ function in R (assumes a multivariate t-distribution; applied except for the Type 2 (R1740Q) group since this function applies only when there are more than two samples in a group). (A) There were two distinct groups in SETD2 NDD patient samples. There was a significant difference between SETD2-1740 samples and controls but Type 2 (R1740Q) cases were closer to controls than Type 1 (R1740W). (B) Despite being distinct from controls, the distance of SETD2-LLS cases were much closer to the control group than SETD2-1740 Type2 (R1740Q) cases. (C) SETD1A cases were not distinguishable from the healthy controls. There are no detectable differences between the two groups of SETD1A splice acceptor variants (SETD1A-P1, P2, P3) or pathogenic variants (SETD1A-P4, P5, P6).

Figure 2

Methylation episignatures for SETD1A samples. SETD1A NDD did not exhibit a prominent methylation episignature. The hierarchical clustering was not able to separate SETD1A from control groups. A comparison of methylation profiles in 6 patients with a SETD1A NDD to those in 64 controls. Methylation analysis detected 7 significant differentially methylated CpG positions with 1 CpG island (chr12:49782966-49783 193) and no DMBs were detected.

Figure 3

Methylation episignatures for SETD2-LLS samples. The methylation episignatures for SETD2-LLS LoF variant patients (n = 4) displayed 778 DMPs (767 hypomethylated and 11 hypermethylated), including 34 CpG Islands and 8 DMBs. (A) Hierarchical clustering on the top annotation bar revealed that 4 SETD2-LLS patients have remarkable hypomethylated profiles that are distinguishable from control samples. (B) The majority of DMPs detected as significant are CpG islands (91.13%), whereas the rest of the DMPs (69 DMPs within 8 DMBs) are located in the Open Sea area (i.e. the rest of the genomic regions except Shelf, Shore or CpG Islands). (C) Normalized methylation values were used to determine whether methylated DMPs gained or lost methylation. The horizontal line (red) indicates a confidence interval of 3 standard deviations (3SD). As a result, all four patients (all of whom are frameshift variants) had similar LOM patterns of methylation. ^*DMB: Differentially methylated blocks ^*GOM: Gain of methylation ^*LOM: Loss of methylation.

Figure 4

Methylation episignatures for SETD2-1740 samples. The SETD2-1740 samples (n = 6) are revealed to have extensive methylation alterations with 7566 significant DMPs (789 hypomethylated and 6777 hypermethylated) containing 281 CpG Islands and 62 DMBs. (A) Hierarchical clustering dendrogram revealed that the majority of significant DMPs in SETD2-1740 cases exhibited very clear hypermethylation patterns compared with controls. A noteworthy observation is that 4 SETD2-1740 patients with Type 1 (R1740W) are distinguishable from Type 2 cases (R1740Q) and healthy controls. (B) Of the 7566 DMPs, 94.96% of DMPs are located on CpG islands, followed by Open Sea (4.85%), Shore (0.13%), and Shelf (0.05%). (C) Gain or loss of methylated DMPs were determined by the normalized methylation values. The horizontal line (red) indicates a confidence interval of 3 standard deviations (3SD). As a result, all four Type 1 (R1740W) cases showed a more severe hypermethylation pattern than in the two Type 2 (R1740Q) cases. ^*DMB: Differentially methylated blocks ^*GOM: Gain of methylation ^*LOM: Loss of methylation.

Figure 5

Methylation episignatures for overlapped CpGs in SETD2 patients. Common CpGs between SETD2-1740 and SETD2-LLS. Overlapped DMPs were identified from the data in Figure 2 (SETD2-1740 versus control) and Figure 3 (SETD2-LLS versus control) (NOT based on a comparative analysis from Supplementary Material, Fig. S3). Several overlapped DMPs are summarized in Supplementary Material, Figure S4A and SB. There are 25 CpG islands and 4 DMBs overlapped between the two groups. All DMPs show (A) hypermethylated across 6 SETD2-1740 cases [Type 2 (R1740Q) cases display slightly milder episignatures than Type 1 (R1740W)] whereas (B) all hypomethylated in 4 SETD2-LLS cases. The episignatures of both groups were completely opposite.

Figure 6

SETD2 codon 1740 structural predictions. (A) Domain architecture of human SETD2. (B) AlphaFold model of SETD2 DAS domain. (C) The structure of Iws1 conserved domain (PDB IDn2XPL).