Skip to main content

Conference Program

Subpage Hero

Loading

Validating Fetal RNA Sequencing to Improve Classification of Splicing Variants in Prenatal Diagnosis

Laboratory Genetics and Genomics
  • Primary Categories:
    • Laboratory Genetics
  • Secondary Categories:
    • Laboratory Genetics
Introduction:
Genome sequencing (GS) often identifies rare variants with in silico predictions suggestive of aberrant splicing, which are often classified as Variants of Uncertain Significance (VUS) due to lack of functional evidence. RNA Sequencing (RNA-seq) has the ability to evaluate mRNA splicing, but typical splicing patterns for fetal specimens are not well established. We performed RNA-seq on cultured amniocytes from 60 fetal specimens which were negative for reportable Pathogenic/Likely Pathogenic variants by GS for a severe early-onset disorder. We established quantitative splicing ranges in prenatal transcriptomes, creating a resource to resolve VUSs with damaging splicing predictions, and enabling results to be returned based on observed splicing patterns.

Methods:
Total RNA was extracted from 60 cultured amniotic fluid samples collected between 16 and 28 weeks gestation, for which GS had been previously performed. Library preps underwent poly-A selection and were sequenced with 150bp paired end reads. Reads were aligned to GRCh38, and splice junctions (SJ) were extracted. Variants were manually inspected in IGV, and aberrant splicing events were confirmed as real if total read depth exceeded 5 and aberrant events were supported by >2 reads. Percent spliced in (PSI) scores were calculated for 3’ and 5’ ends of all SJs,and PSI ranges were established for all samples.

Results:
RNA-seq data was interrogated at SJs predicted by spliceAI to be aberrant and compared with samples lacking the DNA variant of interest. The splicing impact of 895 rare DNA variants from 60 fetal samples was investigated. These variants were (1) in OMIM disease genes (2) had SpliceAI scores >0.2. Predicted splicing impact included donor gain (DG, 437 variants), acceptor gain (AG, 345), donor loss (DL, 67), and acceptor loss (AL, 46 variants). Average spliceAI scores for variants with aberrant splicing were 0.492 (DG), 0.726 (DL), 0.485 (AG), and 0.565 (AL). Scores for variants with aberrant splicing were significantly higher for DG, AG, and DL predictions (p < 0.05), but not for AL. Among variants with confirmed splicing effects in RNA-seq data, 97/192 (51%) generated SJs present in three or fewer samples, typically only in those with the DNA variant. Other variants were associated with splicing patterns observed at some level in individuals without the variant and confirmed as outliers (p < 0.05).

Conclusion:
We have generated the first dataset establishing quantitative splicing patterns during fetal development. This dataset enabled objective assessment of effects of DNA variants on mRNA splicing, aiding in the classification of splicing-related variants. Our findings reveal that many aberrant splicing events occur at low levels in the absence of obvious DNA splicing variants but are pronounced in the presence of deleterious DNA variants. This approach offers an objective method to identify splicing-variants. While in silico tools like SpliceAI are valuable for identifying candidate variants, our data reveal that it is difficult to distinguish functional splicing variants based on predictions alone. This discrepancy likely reflects the broad scope of SpliceAI predictions, which encompasses splicing probabilities across diverse tissue types and developmental states. These findings underscore the importance of tissue-specific and developmental-stage-specific datasets for interpreting splicing effects. Importantly, RNA-seq provides a robust, universal platform for functionally evaluating splicing variants genome-wide, enabling timely and improved variant classification in the prenatal setting.

 

Agenda

Sponsors