RNA Splicing Events in Circulation Distinguish Individuals with and without New-Onset Type 1 Diabetes

J Clin Endocrinol Metab. 2024 Sep 10:dgae622. doi: 10.1210/clinem/dgae622. Online ahead of print.

Abstract

Context: Alterations in RNA splicing may influence protein isoform diversity that contributes to or reflects the pathophysiology of certain diseases. Whereas specific RNA splicing events in pancreatic islets have been investigated in models of inflammation in vitro, how RNA splicing in the circulation correlates with or is reflective of T1D disease pathophysiology in humans remains unexplored.

Objective: To use machine learning to investigate if alternative RNA splicing events differ between individuals with and without new-onset type 1 diabetes (T1D) and to determine if these splicing events provide insight into T1D pathophysiology.

Methods: RNA deep sequencing was performed on whole blood samples from two independent cohorts: a training cohort consisting of 12 individuals with new-onset T1D and 12 age- and sex-matched nondiabetic controls and a validation cohort of the same size and demographics. Machine learning analysis was used to identify specific isoforms that could distinguish individuals with T1D from controls.

Results: Distinct patterns of RNA splicing differentiated participants with T1D from unaffected controls. Notably, certain splicing events, particularly involving retained introns, showed significant association with T1D. Machine learning analysis using these splicing events as features from the training cohort demonstrated high accuracy in distinguishing between T1D subjects and controls in the validation cohort. Gene Ontology pathway enrichment analysis of the retained intron category showed evidence for a systemic viral response in T1D subjects.

Conclusions: Alternative RNA splicing events in whole blood are significantly enriched in individuals with new-onset T1D and can effectively distinguish these individuals from unaffected controls. Our findings also suggest that RNA splicing profiles offer the potential to provide insights into disease pathogenesis.

Keywords: alternative RNA splicing; biomarkers; machine learning; type 1 diabetes.