UNDERPIN delivers a unified computational–experimental framework to reveal how untranslated regions (UTRs) tune RNA-binding protein (RBP) expression and shape stress granule (SG) formation and dynamics.
Firstly, we built a transcript-resolved resources of disease-relevant UTR variants by integrating DisGeNET and GWAS data (1,984 and 1,467 entries, respectively) and annotating each with RNA structure, predicted RNA–protein/RNA–RNA contacts, expression context and proximity to experimentally mapped binding sites. Two open tools were released to the community: PRALINE, a database that ranks single-nucleotide variants with condensate-relevant annotations (predicted and experimentally validated protein–protein, protein–RNA and RNA–RNA, for proteins liquid–liquid phase separation and liquid–solid phase separation propensities, and for RNA the predicted secondary structure content RNA structure), published in Bioinformatics with my controbution), and catRAPID 2.2 – RNA Modifications, which predicts how m6A, A-to-I and pseudouridine reshape protein–RNA binding (published in Molecular Therapy – Nucleic Acids, with my contribution). In addition, we are working on an interactome-based framework to define eQTLs and pQTLs, paving the way to rank UTR variants by their effect on protein production.
Secondly, after prioritising the TARDBP 3′ UTR—encoding TDP-43 and harbouring three variants likely linked to neurodegenerative disease—I profiled its protein interactome with and without each variant using RNA–protein interaction detection (RaPID) coupled to mass spectrometry. TDP-43 is a recurrent hallmark across neurodegeneration, with nuclear loss and cytosolic inclusions; this pathology occurs in ~97% of ALS, ~50% of FTD, and 40–60% of AD, while mutations explain ~5% of familial ALS. The RaPID datasets map how the variants reshape the interaction network and potentially alter post-transcriptional regulation; the interactome is defined and pathway-level analyses to pinpoint variant-affected processes are underway. Functionally, I found that nucleotide changes in the 3′ UTR modulate TDP-43 translation during recovery from stress, when SG disassemble, thereby shifting the timing of translation re-engagement. To generalise the observations, I combined AHARIBO with RNA-seq to interrogate the translatome during stress and recovery and to estimate genome-wide translation-restart kinetics for SG-enriched RBP mRNAs (ongoing work).
Lastly, I adopted an RBP-centric strategy to characterise the function and mode of action of SAM68 (KHDRBS1)—an RBP that partitions into SGs and is implicated in the pathogenesis of fragile X–associated tremor/ataxia syndrome. We refined its binding landscape by iCLIP2 and investigated SAM68-mediated post-transcriptional regulation via its interactions with 5′ UTRs of target mRNAs. Beyond this, SAM68 is a key regulator of alternative splicing, and our data indicate it promotes the formation of numerous circular RNAs (circRNAs), adding a further layer of post-transcriptional control (manuscript under revision).
Taken together, these findings illuminate how UTRs govern RBP levels and SG dynamics, and pave the way for targeted validation and UTR-aware therapeutic exploration.
