Mechanisms of nonsense-mediated mRNA decay target selection

National Heart, Lung, And Blood Institute

Investigators

Linked publications, trials & patents

Abstract

The nonsense-mediated mRNA decay pathway recognizes and degrades transcripts containing premature translation termination codons, as well as 5-10% of apparently normal cellular mRNAs. Despite intensive interest in the mechanisms of NMD, how the pathway selects its targets remains poorly understood. To gain insight into the process and physiological consequences of NMD target selection, we have pursued biochemical and functional screening approaches to identify cellular factors that promote or inhibit decay. 1. Understanding the specificity and kinetics of nonsense-mediated mRNA decay. Extensive efforts to catalog and understand the mechanistic basis for degradation of NMD substrate mRNAs have yielded several "rules" for NMD (Supek, Lehner and Lindeboom, 2021). However, a significant fraction of the observed variability in turnover of mRNAs with PTCs remains unexplained (Hoek et al., 2019; Lindeboom et al., 2019; Kolakada et al., 2024), and the mechanistic basis for several proposed rules is unknown. This knowledge gap is even more pronounced for mRNAs that do not contain PTCs but are instead targeted due to other features such as long 3’UTRs or upstream open reading frames (Muñoz, Lore and Jagannathan, 2023). This current gap in knowledge not only impairs mechanistic studies of NMD but also prevents accurate prediction of the consequences of nonsense mutations in human disease. An important reason for the inability of current models to explain NMD variability is the lack of adequate methods to accurately and comprehensively measure isoform-specific RNA degradation rates in RNA-seq experiments. We have developed approaches for high-resolution kinetic analysis of transcript stability, for the first time measuring the isoform-level kinetic landscape of NMD in human cells. TimeLapse-seq and other recently developed nucleotide-recoding based methods (NR-seq) have enabled transcriptome-wide measurements of RNA synthesis and degradation (Herzog et al., 2017; Riml et al., 2017; Schofield et al., 2018). TimeLapse-seq involves metabolic labeling of cells with the nucleoside analog 4-thiouridine (s4U), followed by chemical modification to alter s4U to mimic the hydrogen bonding of cytosine. RNA-seq reads corresponding to labeled RNAs contain diagnostic U-C mutations, and the ratio of labeled to unlabeled reads is then used to infer RNA synthesis and degradation rates. A major shortcoming of current analysis pipelines is that they are only capable of analyzing RNA kinetics at the gene level (Vock et al., 2024). Because NMD and other RNA decay pathways work on transcripts, not genes, much information is lost when performing gene-level analyses. To address this shortfall, we are collaborating with the laboratory of Matthew Simon (Yale University), inventors of TimeLapse-seq. Together, we have developed a new approach for NR-seq analysis that allows flexible assignment of RNA-seq reads to any genomic features of interest, implemented in the user-friendly EZbakR-suite (Vock et al., 2024). Our approach also incorporates improved modeling of decay rate constants and a workflow for annotation refinement to provide more rigorous analysis of NR-seq data. We anticipate wide usage of this new tool, as it for the first time permits transcript-level quantification of RNA dynamics based on NR-seq data. 2. Understanding how RNP assembly and disassembly modulate NMD efficiency. For almost 30 years, it has been known that ATPase-deficient UPF1 mutants lack activity in NMD (Weng, Czaplinski and Peltz, 1996a, 1996b). Numerous models have been proposed, but there is no clear consensus on why UPF1 ATPase and helicase activities are required for decay. Existing models fall in two broad classes. The first, which we have called the bulldozer model, proposes that UPF1 uses mechanical work generated by ATP hydrolysis to remodel RNA-protein complexes, either to prepare the mRNP for degradation or to recycle proteins from decay intermediates (Franks, Singh and Lykke-Andersen, 2010; Fiorini et al., 2015; Kanaan et al., 2018). The second, which we term the butterfly model, proposes that a central role of UPF1 ATPase activity is to regulate UPF1-RNA interactions (Chapman et al., 2022). In this model, ATPase-stimulated dissociation allows UPF1 to circulate throughout the transcriptome; sustained UPF1-RNA interaction promotes decay while dissociation prior to assembly of a productive decay complex prevents decay (Lee et al., 2015; Fritz et al., 2020). In close collaboration with the laboratory of Keir Neuman (NHLBI), we have developed quantitative assays to systematically test the biochemical capabilities of wild-type and mutant UPF1 proteins. In total, our suite of high-throughput kinetic assays measures substrate binding dynamics, helicase translocation and unwinding, protein displacement, and helicase positioning on substrates (Chapman et al., 2022, 2024). These new tools complement established ATPase and equilibrium binding assays (Chakrabarti et al., 2011). To test the butterfly and bulldozer models, we focused on mutants that retained ATPase activity but had varying degrees of mechanochemical impairment. The E797R mutation caused ~5-fold slower ATPase, helicase, and ATP-stimulated dissociation rates but did not impair processivity or protein displacement. A546H mutants were previously observed to reduce UPF1 processivity (Kanaan et al., 2018), and we additionally identified ~5-10-fold defects in unwinding rate and impaired ATP-stimulated dissociation from RNA. In an attempt to rescue A546H, we combined it with a G619K mutation, which enhances ATPase and helicase activity in isolation. The resulting double-mutant had only minimally more effective helicase and protein displacement activities (~10 nt processivity, ~0.2 bp/second translocation rate—meaning approximately 5 seconds on average per nt traveled) but restored ATP-stimulated dissociation. UPF1 primarily exists in an enzymatically autoinhibited "closed" state, where its cysteine/histidine-rich (CH) domain docks against the RecA2 domain of the helicase core. When the NMD factor UPF2 binds to the CH domain, it triggers a dramatic conformational change, relocating the CH domain to the opposite side of the helicase and activating UPF1 ATPase and helicase activities (Chakrabarti et al., 2011). Although the stimulatory effect of UPF2 on UPF1 is well documented, the biological significance of UPF1 autoinhibition and subsequent activation remains unclear. Our finding that UPF1 retained NMD function even when mutations reduced its ATP hydrolysis rate by a factor exceeding UPF2's stimulatory effect led us to reexamine this relationship (Chapman et al., 2022). Combining our observations with recent evidence that UPF2 inhibits UPF1-RNA interactions (Xue et al., 2022), we investigated how the transition between open and closed conformational states affects UPF1 function, both independently and in concert with UPF2 binding. We hypothesized that UPF2 might alter the RNA binding kinetics of UPF1. To test this hypothesis, we established a bio-layer interferometry (BLI) assay to track the kinetics of UPF1 binding to and dissociation from a biotinylated RNA (Chapman et al., 2024). For these experiments, we used a previously characterized UPF1 construct containing the CH and helicase domains (UPF1 CH-HD) (Chakrabarti et al., 2011). To our surprise, BLI and complementary fluorescence-based assays revealed a dual effect of UPF2 binding on UPF1-RNA interactions. UPF2 both promoted ATP-dependent UPF1 dissociation, an expected consequence of ATPase rate enhancement, and reduced the rate of UPF1 association with RNA. Using mutants that disrupt the linkage between the CH and RecA2 domains (Chakrabarti et al., 2011), we found that these altered RNA binding kinetics are intrinsic properties of the UPF1 open state. These findings suggest that UPF2 has positive and negative roles in NMD, both suppressing UPF1 interactions with potential substrates and activating decay. In addition, we used agent-based modeling to predict that the UPF1 CH domain can spontaneously undock, allowing UPF2-independent relief from autoinhibition. Recent structural data support UPF2-independent dissociation of the CH domain from the RecA2 domain (Langer et al., 2024). The ability of UPF1 to operate independently of UPF2 thus provides a mechanism to explain how UPF1 can maintain effective ATP-dependent transcriptome surveillance despite substoichiometric UPF2 levels in the cell. 3. Altered NMD specificity under stress conditions. In mammalian cells, the central NMD regulator UPF1 is alternatively spliced to yield a protein with an 11nt extension in a critical regulatory loop (UPF1 “long loop” or UPF1LL) (Gowravaram et al., 2018). This alternative isoform accounts for ~15-20% of UPF1 mRNA in most human cell types (GTEx Consortium, 2013). In the more abundant canonical isoform (UPF1 “short loop” or UPF1SL), the regulatory loop protrudes into the UPF1 RNA binding channel, where it mediates ATPase-stimulated dissociation of the helicase from RNA (Cheng et al., 2007; Gowravaram et al., 2018). The UPF1LL loop extension, perfectly conserved among mammals, reduces the likelihood that UPF1 will dissociate from RNA upon ATP hydrolysis (Gowravaram et al., 2018). Prior to our investigation, the specific cellular functions of UPF1LL had been entirely unstudied. Using an siRNA that specifically depletes UPF1LL, we found that it regulates hundreds of target mRNAs (Fritz et al., 2022). UPF1LL is largely dispensable for degradation of canonical NMD targets containing premature termination codons but instead regulates a distinct set of mRNAs, many of which encode membrane or secreted proteins. A particularly exciting finding from our studies of UPF1LL was the isoform’s distinct activity in conditions of stress and translational repression. We have particularly examined the continued activity of UPF1LL upon induction of the integrated stress response (ISR). There is an extensive literature describing reciprocal repression between the ISR and NMD; the effect of the ISR on NMD is likely primarily due to the phosphorylation of eIF2α, which results in reduced global translation initiation (Gardner, 2010; Karam et al., 2015; Li et al., 2017). Consistent with previous studies, we found that canonical NMD substrates increased in abundance upon induction of the ISR with the ER stress inducer thapsigargin. Conversely, transcripts sensitive to UPF1LL depletion were not similarly up-regulated. In addition, we found that new mRNAs became susceptible to degradation in a UPF1LL-dependent manner, either upon ISR induction or treatment with the elongation inhibitor puromycin. A subset of these RNAs were among those normally protected from degradation by PTBP1 and/or hnRNP L.