Cytoplasmic Intron Retention As a Regulatory Mechanism of Mitochondrial Homeostasis in Erythroid Cells

Introduction: Erythroid development features a dynamic intron retention (IR) program throughout all stages of erythropoiesis (Pimentel et al, NAR 2016). Transcripts with retained introns (RI) are predicted to be degraded by nonsense-mediated decay (NMD), but a subset of RI-transcripts evade NMD and persist stably in the cytoplasm (Boutz et al, Genes Dev 2015). One such transcript in erythroid cells is SLC25A37 (encoding the critical mitochondrial (Mt) iron transporter, Mitoferrin-1). Up to 50% of Cy SLC25A37 transcripts retain the second intron (SCL25A37-IR). Analysis of patient transcriptomes (Pellegatti et al, Blood 2018) reveals that SCL25A37-IR is reduced across all splicing factor (SF) mutant clonal myeloid states. Mechanisms that confer stability to cytoplasmically retained intron (cRI) transcripts, and their functional roles during erythropoiesis remain unexplored.

Methods: To determine the cytoplasmic (Cy) half-life (ct_1/2) and stability of Cy IR (cIR) in erythroid cells, we conducted 4-thiouridine (4sU)-labeling based time-lapse sequencing (Schoefield et al, Nat Methods 2018) in HUDEP-2 cells. Mature polyadenylated RNA, at 0, 1, 2, and 4 hours of exposure to 500 µM 4sU, from nuclear (Nu) and Cy compartments, was analyzed with with RNA-seq to determine C>T conversions after time-lapse chemistry. Intronic t_1/2 were estimated with GRAND-SLAM (Jurges, Bioinformatics 2018) with a custom-built intronic GTF. 6886 introns were further filtered, for uniform read coverage and excluding overlapping annotated features, to a total of 607 IR-RNAs. t_1/2 and turnover rates for Nu and Cy compartments were kinetically modeled (Letswaart et al, Mol Cell 2024), and the fractional Cy turnover (Φ) metric was used to measure relative Cy RNA abundance (Φ = [Nu abundance - Cy abundance]/Nu abundance); a higher Φ indicates greater processing (by splicing or decay) of unspliced RNA within the Nu. The cytoplasmic IR ratio (cIR-ratio) was calculated as the ratio of unspliced to total (spliced + unspliced) reads spanning an intron.

Results: To determine if SLC25A37-IR is translated to peptides, we over-expressed the construct using plasmid constructs, but found no evidence of non-canonical protein isoforms. To determine kinetics of cIR-ratio, we plotted ct_1/2 against Φ for the 607 introns which revealed a strong sigmoidal pattern (ρ = 0.63). Highly stable cRI transcripts (ct_1/2 > 15 hours ) always had high Φ values, indicating that such RI transcripts are largely spliced within nucleus and only a fraction is exported unspliced. Absence of introns with low Φ and high t_1/2 suggest that nuclear processing primarily determines levels of cRI transcripts. Weighted K-means clustering identified three intron categories: (A) low Φ (<0.4) and low ct_1/2 (<15 hrs), (B) high Φ (>0.4) and low ct_1/2 (<15 hrs), and (C) high Φ (>0.4) and high ct_1/2 (>15 hrs). t_1/2 did not correlate with the cIR suggesting that cIR-ratio is not a valid measure of stability. Category C (n=166, which included SLC25A37) was especially enriched in genes regulating metal-ion binding, Mt homeostasis, acyltransferase activity, and phospholipid metabolism. Given that IR is largely regulated by its binding proteins (or RBPs), we interrogated functional genomics data from ENCODE consortium to determine potential RBPs that maybe regulating cIR. This dataset included knockdown (KD) of 175 RBPs by shRNA and 180 RBPs by CRISPR deletion in K562 cells. Hierarchical clustering of IRFinder data from these datasets showed several RBPs whose loss caused a unidirectional increase in splicing across the 607 cRI transcripts (Ward distance (d): 15, max d: 340), suggesting potential roles in maintaining cIR in physiological conditions. These included SRSF3, SF3B1, and CPSF factors, which have roles in speckle splicing, 3' mRNA processing, nuclear export, and m6A binding.

Conclusion: Many cRI transcripts encoding critical regulators of Mt iron delivery and metabolism appear to be highly stable in the cytoplasm but not serve as templates for peptide translation. By intersecting public datasets, we identify multiple candidate RBPs mediating cRI. Ongoing studies exploring: i) cRI transcript ct_1/2 using long-read sequencing, ii) regulators of cIR (using CRISPR/dCas13 pull-down and proteomics) and iii) functional significance of cIR (e.g., SLC25A37) using Cas13-NES targeting are ongoing, and will offer insights into the role of cIR in physiology and SF mutant disease.