High-throughput minigene screening maps pseudoexon risk in NF1
Deep intronic variation can be invisible to standard variant interpretation until it forces a hidden exon, or pseudoexon, into the transcript. In NF1, a gene already known for complex splicing disruption, that problem becomes more than a molecular curiosity; it is a direct obstacle to diagnosis, classification, and downstream counseling. This preprint describes a specialized minigene strategy built to sort signal from noise across a large set of deep intronic candidates, with the goal of identifying which variants actually alter splicing and which remain biologically inert.
Investigators focused on intron 3 of NF1 and examined 501 deep intronic variants using a high-throughput mutagenesis and next-generation sequencing workflow. The design is notable for its scale and for its logic: instead of inferring splice effects from sequence alone, the assay tests transcript behavior experimentally in a controlled system. Eleven variants proved capable of activating pseudoexon inclusion, and seven of those had not been described previously. The activated variants were not randomly scattered across the intron; they clustered into four activation-prone hotspots, suggesting that the local sequence environment is helping define where aberrant splice recognition becomes possible.
Pseudoexon activation is not a subtle event. Once an intronic segment acquires or strengthens splice-like features, the splicing machinery can mistake it for a true exon and incorporate it into the mature transcript. In a gene such as NF1, this can have immediate functional consequences if the inserted sequence disrupts the reading frame or introduces premature termination. The study also identified a rare wild-type pseudoexon that is normally out of frame, yet became significantly more included after three specific mutations. Even when the sequence is already present in the transcript pool at low levels, small changes can tilt the balance toward clinically relevant mis-splicing.
A small number of variants drove the strongest splice defects
The headline result is not simply that pseudoexon activation occurs in NF1; that much was already suspected from prior work. The more useful finding is that the majority of tested intronic variants did not produce this effect, while a small subset consistently did. For diagnostic laboratories, this kind of functional partitioning is invaluable. Deep intronic changes are often hard to classify because they sit outside canonical splice junctions and may evade standard prediction rules. An experimentally verified map of activating sites gives variant interpreters a way to separate plausible pathogenic candidates from background sequence variation.
The clustering into four hotspots also matters mechanistically. It implies that pseudoexon formation in NF1 is not distributed evenly across the intron but is concentrated in sequence neighborhoods where splice motifs, regulatory elements, and transcript context converge. Such hotspot behavior is consistent with the idea that aberrant splicing emerges when a region crosses a threshold of splice-site strength and enhancer or silencer balance. In practice, this means one deep intronic region may harbor several nearby variants with very different functional outcomes, while adjacent segments remain comparatively quiet.
The dataset was also used to benchmark SpliceAI and ESRseq scores. That comparison is clinically relevant because computational tools are often the first line of triage when a novel intronic variant appears in testing. Prediction algorithms can accelerate prioritization, but they cannot replace direct functional evidence when the sequence context is unusual or when a variant sits in a region already prone to pseudoexon use. By aligning algorithmic scores with experimentally validated outcomes, the study offers a practical way to refine how in silico tools are deployed in the laboratory workflow.
False reassurance is one of the central hazards in intronic variant interpretation.
Why this matters for diagnostic genetics and transcript-based testing
For patients and clinicians dealing with suspected NF1-related disease, the central challenge is often not whether a variant exists, but whether it changes splicing enough to matter biologically. A deep intronic change that creates a pseudoexon can produce a loss-of-function allele without touching the coding sequence itself. That makes transcript-aware testing essential when genomic sequencing alone leaves the answer ambiguous. Functional assays such as this minigene system can bridge the gap between DNA-level findings and RNA-level consequences, especially in genes with a known propensity for cryptic splicing.
The study’s integrated design is also a reminder that variant annotation is moving toward multiplex, experimentally grounded pipelines rather than one-variant-at-a-time analysis. High-throughput mutagenesis paired with sequencing allows many candidate changes to be assessed in parallel, which is particularly important in large introns where manual confirmation would be slow and incomplete. The authors present the assay as a scalable model for other disease-associated genes, and that claim follows naturally from the data structure: once a platform can identify activating sequence features and localize them into hotspots, it becomes useful beyond a single locus.
From a mechanistic standpoint, the work reinforces a familiar but often underappreciated principle in medical genetics: sequence context governs pathogenicity. A variant does not need to alter an amino acid to destabilize gene function. It can instead rewrite the splicing instructions embedded in intronic DNA, create a pseudoexon, and derail the transcript before translation ever begins. The fact that three mutations were enough to enhance inclusion of a rare out-of-frame pseudoexon illustrates how narrow the margin can be between cryptic silence and measurable aberrant expression.
This preprint does not claim to settle every diagnostic question in NF1. It does, however, provide an experimentally validated framework for prioritizing deep intronic variants and for understanding why some intronic regions are disproportionately vulnerable to pseudoexon activation. The assay also creates a benchmark set that can be used to test whether computational splicing scores are sufficiently sensitive for clinical triage, or whether certain intronic configurations still require direct functional confirmation before a variant can be called pathogenic.