Researchers at Massachusetts General Hospital (MGH) and the Broad Institute of MIT and Harvard have built a genome-wide catalog of 9,814 “poison exons” in human genes. The study maps these exons systematically across the genome and links more than 1,600 disease-causing variants to them. The peer-reviewed, open-access study was published on 14 October 2026 in Genome Biology. It shows how a little-studied layer of RNA control shapes gene activity in the brain, and how it can fail in neurological disease.
What Is a Poison Exon?
Most human genes are pieced together from several coding segments called exons. Through a process called alternative splicing, cells can include or skip particular exons to make different messenger RNAs (mRNAs) from the same gene. Splicing is tightly regulated, and its disruption is increasingly linked to disease.
A poison exon is a special case. When it is included, it introduces a premature “stop” signal into the mRNA. Cells detect these faulty messages and destroy them through nonsense-mediated decay (NMD), a built-in RNA quality-control pathway. Poison exons therefore work like a dimmer switch: the more often one is spliced in, the less protein the gene makes. According to the authors, there has been no systematic way to find them across the whole transcriptome, so most remain poorly annotated.
Building a Genome-Wide Catalog
The team, led by co-first authors Paolo Pigini and Huilin Xu with senior authors Dadi Gao and Elisabetta Morini, started with an evolutionary comparison. Across 954 eukaryotic transcriptomes, humans showed the strongest enrichment of transcripts destined for NMD.
They then designed a computational pipeline that tests whether short annotated exons create a premature stop codon when placed in the context of each protein-coding transcript of their gene. The result was a catalog of 9,814 predicted poison exons in the human genome.
A Brain-Specific Layer of Control
Using RNA-sequencing data from GTEx, the researchers found 83 poison exons used only in the brain and 973 with brain-specific splicing patterns. Data from BrainSpan, which tracks the developing human brain, revealed 1,284 poison exons whose use changes during brain development. This suggests that the brain relies on poison exons to fine-tune protein levels as it grows and matures.
From Variants to Disease
Next, the team combined ClinVar’s database of clinical variants with SpliceAI, a deep-learning model that predicts how DNA changes affect splicing. They identified 1,625 pathogenic variants associated with neurological disorders that are predicted to alter splicing of 743 poison exons.
To test these predictions in the lab, they introduced a subset of the variants into human cells using CRISPR prime editing. One example stood out: a deep intronic variant in NDUFAF6 associated with Leigh syndrome, a severe mitochondrial disorder. In neurons derived from induced pluripotent stem cells (iPSCs), the variant increased inclusion of the poison exon. This lowered NDUFAF6 protein levels and impaired mitochondrial membrane potential.
Why It Matters for Researchers
- Look beyond coding regions. Variants deep inside introns may contribute to disease by switching poison exons on, and standard exome-focused pipelines can miss them. Genome-wide resources such as the AlphaGenome Atlas of DNA variant effects are making such non-coding variants easier to assess.
- Combine prediction with validation. SpliceAI-based screening narrowed thousands of ClinVar variants to testable candidates, and prime editing then confirmed a subset of them.
- New therapeutic targets. The authors argue that poison exons could be targets for RNA-based therapies that modulate splicing. They did not test any therapy in this study.
- Mind the limits. The predictions rely on annotated exons and short-read RNA-seq data, and only a subset of variants was tested experimentally.
The Bigger Picture
This catalog gives geneticists, splicing researchers, and drug developers a starting point for reinterpreting unexplained variants in neurological patients. It also points to poison exons as candidates for therapies that adjust protein levels at the RNA stage.
Working on splicing, variant interpretation, or neurogenomics? Share your thoughts with the CBIRT community and follow cbirt.net for daily research news.
Frequently Asked Questions
A poison exon is a segment of a gene that, when spliced into the messenger RNA, introduces a premature stop codon. The cell then destroys that mRNA instead of using it to make protein, so poison exons act as a switch that turns gene output down.
Nonsense-mediated decay is a quality-control pathway that detects mRNAs carrying premature stop codons and breaks them down before they can produce shortened, potentially harmful proteins. Poison exons work by deliberately routing transcripts into this pathway.
The team predicted 9,814 poison exons in the human genome. Of these, 83 were used only in the brain, 973 showed brain-specific splicing patterns, and 1,284 changed their use during brain development.
By combining ClinVar with the splicing-prediction model SpliceAI, the researchers found 1,625 pathogenic variants associated with neurological disorders that are predicted to alter the splicing of 743 poison exons. In one example, a deep intronic variant in NDUFAF6, linked to Leigh syndrome, increased poison exon inclusion and reduced NDUFAF6 protein in iPSC-derived neurons.
The authors suggest that poison exons could be targeted by RNA-based therapies that adjust splicing to restore protein levels. However, this study did not test any therapy, so that potential still needs to be demonstrated.
Reference: Pigini P, Xu H, et al. Poison exon splicing in the human brain provides a new paradigm for understanding and targeting neurological disorders. Genome Biology (2026). https://doi.org/10.1186/s13059-026-04292-8
Disclaimer:
The research discussed in this article was conducted and published by the authors of the referenced paper. CBIRT has no involvement in the research itself. This article is intended solely to raise awareness about recent developments and does not claim authorship or endorsement of the research.
Follow Us!
Dr. Tamanna Anwar is a Scientist and Co-founder of the Centre of Bioinformatics Research and Technology (CBIRT). She is a passionate bioinformatics scientist and a visionary entrepreneur. Dr. Tamanna has worked as a Young Scientist at Jawaharlal Nehru University, New Delhi. She has also worked as a Postdoctoral Fellow at the University of Saskatchewan, Canada. She has several scientific research publications in high-impact research journals. Her latest endeavor is the development of a platform that acts as a one-stop solution for all bioinformatics related information as well as developing a bioinformatics news portal to report cutting-edge bioinformatics breakthroughs.
