NEW YORK – Noncoding de novo mutations may contribute to congenital heart disease about as much as coding de novo mutations, a new genomic analysis has found.
Congenital heart disease has genetic origins in about a third of cases. In particular, a previous analysis of congenital heart disease (CHD), which affects about 1 percent of live births, found that about 8 percent of cases could be traced to coding de novo variants (DNVs). As many of these coding DNVs altered proteins involved in chromatin modification, transcription regulation, and RNA processing, researchers led by Mount Sinai School of Medicine's Bruce Gelb sought to uncover whether noncoding DNVs might also contribute to CHD.
Gelb and his colleagues applied three strategies to uncover noncoding DNVs from within whole-genome sequencing data from parent-child trios in which the child has CHD. As they reported Monday in Nature Genetics, the researchers homed in on an excess of noncoding DNVs within individuals with CHD, including variants predicted to influence RNA processing or post-transcriptional regulation during cardiac development.
"These data systematically associate human CHD with cardiac regulatory DNVs," the team wrote.
The researchers sequenced the whole genomes of 763 individuals with CHD and their unaffected parents to 30X coverage. Previous whole-exome sequencing had been unable to identify missense or loss-of-function coding variations in CHD genes in these samples.
After excluding probands for whom WGS identified a probable genetic cause of CHD, the researchers analyzed 749 probands for noncoding DNVs. In all, they identified a mean 71 de novo SNVs and five de novo indels per CHD proband, and noted an increased burden of DNVs in people with CHD, as compared to controls.
Using a neural network they dubbed HeartENN — an extension of the deep-learning algorithm DeepSEA — the researchers predicted the functional effects of the identified noncoding DNVs. More than 96 percent of DNVs had HeartENN scores that suggested they had little functional impact.
However, more individuals with CHD had variants with scores indicating an increased likelihood of a functional effect. Such high-scoring variants were also enriched among known human CHD genes.
At the same time, the researchers explored the effect of the DNVs they uncovered on regions thought to regulate gene expression during cardiac development in humans. They uncovered 27 genes that were enriched for DNVs among individuals with CHD. Ten of these genes were highly expressed in mouse hearts and 12 are expected to be intolerant to loss-of-function variants.
The researchers noted a significant overlap between these two approaches and confirmed that five variants lead to differences in gene expression. Genes associated with these five DNVs include JPH2 and SEMA4B. This, the researchers noted, begins to suggest how these DNVs might have their effects, as JPH2 encodes a membrane protein needed for T-tubule formation, while SEMA4B is expressed in developing hearts and encodes a semaphorin that signals through plexin receptors. Alterations to semaphorin-plexin signaling have been linked to CHD.
In their third analytical approach, the researchers examined how the DNVs they identified might influence RNA processing or post-transcriptional regulation. They noted an increased number of DNVs within RNA-binding-protein regulatory sites among individuals with CHD, suggesting that noncoding DNVs contribute to CHD by their influence on transcriptional and post-transcriptional regulation during cardiac development.
They further estimated that the number of CHD cases with noncoding DNVs contributing to disease is at least as high as the portion of CHD cases in which damaging coding DNVs have been found.