Core NMD factors

A central player in the NMD pathway is the ATP‐dependent RNA helicase of the SF1 superfamily, Upstream Frameshift 1 (UPF1), which harbors an amino‐terminal cysteine‐and histidine‐rich (CH) domain and a carboxy‐terminal RNA helicase domain (Fig 3; Kim & Maquat, 2019).

Open in a separate window

Figure 3

UPF1 structure, binding sites, and known mutations

Diagram depicting the domain structure of the canonical UPF1_SL isoform, highlighting key residues for its phosphorylation and ATP binding capacity. An isoform generated via alternative splicing, UPF_LL, includes an extra 11 amino acid extended regulatory region, while all other structural elements remain the same. Mutations and their effect on UPF1 function are indicated by the colored boxes. Some mutations in UPF1 have the capacity to eliminate NMD function, whilst some mutations sustain functional NMD. All residue numbers relate to the canonical UPF1_SL isoform.

UPF1 is activated by phosphorylation carried out by the SMG1c complex, comprised of the phosphoinositide 3‐kinase (PI3K)‐like kinase, SMG1, and two additional subunits, SMG8 and SMG9, that negatively regulates its activity (Yamashita et al, 2001, 2009; Fernández et al, 2011; Langer et al, 2021). SMG1 phosphorylates UPF1 at multiple SQ and TQ motifs located in the amino‐ and carboxy‐terminal domains (Yamashita et al, 2001). Until recent years it was widely accepted that initially, the association of UPF1 with SMG1 and the eukaryotic release factors eRF1 and eRF3 form the surveillance complex (SURF) in the vicinity of the PTC (Kashima et al, 2006) (Fig 4). Subsequently, interaction of the SURF complex with UPF2 and UPF3B and an EJC downstream of the PTC leads to the assembly of a decay‐inducing complex (DECID), where UPF1 is phosphorylated and eRF1 and eRF3 are released (Kashima et al, 2006; Chamieh et al, 2008; López‐Perrote et al, 2016). In addition to this, a central role for UPF3B in translation termination has been highlighted. A fully reconstituted in vitro translation system showed the predominance of the interaction of UPF3B with ribosome release factors, to delay translation termination and dissociate post‐termination ribosomal complexes that are devoid of the nascent peptide. UPF1 was shown to interact transiently with the termination factors and UPF3B to initiate the subsequent mRNA decay (Neu‐Yilik et al, 2017). Despite the lack of mammalian in vivo data, a recent in vivo study in yeast re‐established the central role of the UPF1:80S interaction for translation termination and NMD initiation (Ganesan et al, 2022).

Open in a separate window

Figure 4

Mechanism of NMD activation

Schematic depicting the widely accepted molecular events leading to the assembly of the surveillance complex (SURF), its transition to the decay‐inducing complex (DECID) leading to UPF1 phosphorylation and recruitment of SMG6 and/or SMG5/SMG7 that elicit mRNA degradation.

Under normal conditions UPF1 has low basal helicase activity; however, upon recognition of a PTC and subsequent binding of UPF1 to UPF2, this activity is greatly increased. Alongside this, a large conformational change of the CH inhibitory domain modifies the RNA‐binding properties and the catalytic activity of UPF1, causing a switch from an RNA‐clamping mode to an RNA‐unwinding mode (Chakrabarti et al, 2011). The active UPF1 helicase functions as an RNPase translocating along the mRNA with a 5′ to 3′ polarity, acting to resolve secondary structures, remove proteins from mRNA, and provide access to nucleases (Franks et al, 2010; Fiorini et al, 2015). Importantly, NMD can also be elicited on mRNAs that do not have a downstream EJC, although the mechanism of non‐EJC dependent NMD is less well defined (Bühler et al, 2006; He & Jacobson, 2015). Interestingly, alternative branches of the NMD pathway that act independently of UPF2, UPF3B, or the EJC have also been described (Gehring et al, 2005; Chan et al, 2007; Ivanov et al, 2008). In mammals, there are two highly related UPF3 paralogs, UPF3A and UPF3B (also called UPF3X due to its location in chromosome X). Recent studies have shown that UPF3A and UPF3B perform redundant functions and can activate NMD without EJC binding, suggesting that UPF3 paralogs play a more active role in NMD than simply bridging the EJC and the UPF complex. UPF1 almost exclusively associates with UPF3B and only minimally with UPF3A; however, when UPF3B is mutated or removed, the association of UPF1 with UPF3A is enhanced 4‐6 times, independent of RNA. Thus, UPF3A seems almost dispensable for NMD; however, it performs a compensatory role and can maintain NMD in the absence of UPF3B (Wallmeroth et al, 2022; Yi et al, 2022; Chen et al, 2023). In cells lacking both UPF3 paralogs, although NMD is not completely abrogated, its activity is reduced significantly (Bufton et al, 2022; Yi et al, 2022).

A current model postulates that UPF1 binding to mRNAs does not inevitably mark mRNAs for NMD‐mediated degradation. Use of CLIP (cross‐linking and immunoprecipitation) revealed that UPF1 binds target RNAs prior to mRNA translation. Once translating ribosomes are engaged, they displace UPF1 from coding sequences, leading to UPF1 enrichment at 3′UTRs (Hurt et al, 2013; Zünd et al, 2013). By contrast, binding of phosphorylated UPF1 (P‐UPF1) marks mRNAs for NMD‐mediated degradation, since P‐UPF1 is enriched on endogenous transcripts degraded by NMD, whereas unphosphorylated UPF1 is released from non‐targeted transcripts in an ATP‐dependent manner (Kurosaki et al, 2014; Lee et al, 2015). Interestingly, UPF1 mutants with substantially impaired processing and slower unwinding rates are still functional in NMD and still have the capacity to restore NMD functionality upon loss of WT UPF1 (Fig 3) (Chapman et al, 2022).

In addition to a rapidly increasing number of NMD regulators, it was recently shown that UPF1_LL, an alternative mammalian‐specific isoform of the core NMD factor UPF1 (UPF1_SL), harbors a regulatory loop that is 11‐residues longer and preferentially binds and down‐regulates a different subset of NMD targets through a reduction in the RNA dissociation potential (Fig 3) (Gowravaram et al, 2018; Fritz et al, 2022). Canonical NMD is mediated by UPF1_SL that represents >75% of the total UPF1 pool and is inhibited by moderate translational repression. By contrast, the UPF1_LL isoform triggers NMD in response to activation of the integrated stress response (ISR), and repression of translation, targeting novel mRNA, including stress response genes (Fritz et al, 2022). Interestingly, UPF1_LL requires translation termination events; however, due to its improved RNA‐binding capacity in the presence of ATP compared to UPF1_SL, its sustained RNA interaction favors a reduced frequency of termination events in conditions of attenuated translation. Therefore, under conditions of moderate translation inhibition where NMD is inhibited, UPF1_LL activity is enhanced, changing the specificity of NMD in response to stress conditions (Fritz et al, 2022).

UPF1 phosphorylation represents a non‐reversible point in NMD progression during which the transcript is committed for degradation, leading to repression of further translation initiation, a key step in the NMD pathway (Isken et al, 2008). This phosphorylation event leads to the recruitment of additional NMD factors, namely the phospho‐binding proteins SMG6 and/or SMG5/SMG7, which function in two independent, yet overlapping pathways, and lead to further recruitment of nucleases to elicit mRNA degradation. The SMG6 endonuclease cleaves NMD targets in the vicinity of the PTC (Huntzinger et al, 2008; Eberle et al, 2009) and generates a 5′ cleavage product that is most likely degraded 3′ to 5′ by the exosome and/or DIS3L2 (DIS3‐like exonuclease 2; Kurosaki et al, 2018). The resulting 3′ cleavage product is cleared of protein components by UPF1 to provide access to the exoribonuclease XRN1 (Kurosaki et al, 2019). Alternatively, the heterodimer SMG5/SMG7 binds to P‐UPF1 and recruits the CCR4/NOT complex to promote deadenylation, leading to 3′ to 5′ decay (Loh et al, 2013) and decapping, resulting in XRN1‐catalyzed 5′ to 3′ degradation (Unterholzner & Izaurralde, 2004). Transcriptome profiling revealed that SMG6 and SMG7 act on essentially the same transcripts, suggesting extensive redundancy between the endo‐ and exonucleolytic decay pathways (Colombo et al, 2017). There is a tight regulation between these two RNA degrading pathways, since inactivation of the SMG5‐SMG7 pathway also abrogates the SMG6‐dependent mechanism, supporting a mechanism that involves UPF1 phosphorylation and SMG5‐SMG7 recruitment to access SMG6 activity (Boehm et al, 2021).

Additional NMD factors

Core NMD factors SMG1‐7 were initially discovered using genetic screens in C. elegans and in S. cerevisiae and were later found by homology searches in other species, including Arabidopsis, Drosophila, and mammals. Since then, a variety of experimental approaches have been employed to identify novel regulators of this pathway, most of which are still being investigated and therefore do not feature in models of NMD (Table 1) (Hug et al, 2016).

We developed RNAi screens in C. elegans that resulted in the identification of novel NMD factors, including smgl‐2/DHX34 and smgl‐1/NBAS that are conserved throughout evolution and function in NMD in nematodes, zebrafish, and human cells (Longman et al, 2007, 2013; Anastasaki et al, 2011). DHX34, a DExH/D box RNA helicase, forms a complex with SMG1 and UPF1 and activates NMD by promoting the transition from the surveillance to the decay‐inducing complex (Hug & Cáceres, 2014; Melero et al, 2016). Heterozygous mutations in DHX34 were identified in four families affected with inherited acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS). These mutations map to different domains of DHX34, and all germline variants identified in these families abrogated NMD activity (Rio‐Machin et al, 2020). DHX34 is also associated with the human spliceosomal catalytic C complex and regulates a large number of alternative splicing (AS) events in mammalian cells in culture, establishing a dual role for DHX34 in both NMD and pre‐mRNA splicing (Hug et al, 2022).

Neuroblastoma amplified sequence (NBAS) encodes a protein localized to the endoplasmic reticulum (ER) that is a component of the Syntaxin 18 complex and fulfils a role in Golgi‐to ER retrograde transport, which is independent of its function in NMD (Aoki et al, 2009; Longman et al, 2020; discussed below in the “Localized NMD and the stress response” section).

Interactome studies, RNAi screens, and newer CRISPR screens in mammalian cells in culture have led to the identification of novel NMD regulators, which are still being characterized (Table 1). The RNA helicase MOV10, a member of the UPF1‐like group helicase superfamily 1 (SF1), interacts with UPF1 and promotes degradation of UPF1‐regulated mRNA transcripts (Gregersen et al, 2014). An interactome of the SMG1 protein kinase identified the AAA ATPases, RuVB‐like 1 (RUVBL1) and RuVB‐like 2 (RUVBL2). These proteins are involved in a variety of cellular functions, including transcription and DNA repair, and were previously shown to have a role in the early stages of NMD (Izumi et al, 2010). We recently showed that RUVBL1/2 also interact with DHX34, coupling their ATPase activity to the assembly of factors required to initiate the NMD response (López‐Perrote et al, 2020).

A genome‐wide RNAi screen in human cells identified ICE1, an EJC‐associated factor that promotes the interaction of UPF3B with the EJC and activates NMD (Baird et al, 2018). CRISPR screens identified several candidate NMD genes (Alexandrov et al, 2017) and highlighted a role for the ribosome recycling factor ABCE1 in NMD (Annibaldis et al, 2020; Zhu et al, 2020). A CRISPR screen in K562 cells identified the translational repressors GIGYF2 and EIF4E2, suggesting a model wherein recognition of a stop codon as premature leads to its translational repression mediated by GIGYF2 and EIF4E2, a process shared with the NGD pathway (Zinshteyn et al, 2021). Finally, a haploid‐cell genetic screen for NMD effectors identified several components of the AKT signaling pathway. It was shown that AKT‐mediated phosphorylation of the UPF1 CH domain at T151D overcomes auto‐inhibition of UPF1 helicase activity, which is critical for NMD and decreases the dependence of helicase activity on ATP. AKT also promotes formation of EJCs that contain AKT at the expense of UPF2, potentially facilitating a UPF2 independent branch of NMD (Cho et al, 2022). Interestingly, AKT1 had been independently shown to phosphorylate UPF1 and activate NMD (Palma et al, 2021; Table 1).

NMD factors have also been shown to display additional cellular functions. This is particularly prominent in the case of UPF1 that not only functions in genome stability (Azzalin & Lingner, 2006), but also contributes to several RNA decay pathways (Kim & Maquat, 2019). These include the Staufen‐mediated decay pathway (SMD), which involves UPF1 recruitment to stem‐loops in the 3′UTRs of mRNAs bound by the RNA‐binding proteins STAUFEN1 and its paralog STAUFEN 2 (Kim et al, 2005; Park et al, 2013), and in the degradation of histone mRNAs, where UPF1 is recruited to target mRNAs via the stem‐loop binding protein (SLBP) (Kaygun & Marzluff, 2005). Interestingly, UPF1 has also been proposed to act as an E3‐ubiquitin ligase and promote degradation of the truncated polypeptide produced by translation of a PTC‐containing transcript (Takahashi et al, 2008; Feng et al, 2017), establishing a link between RNA degradation and protein decay (Kim & Maquat, 2019; Inglis et al, 2023). In S. cerevisiae, UPF1 facilitates proteasome degradation of truncated polypeptides in a ubiquitin dependent manner. The truncated protein product gets released from the ribosome following NMD‐mediated mRNA degradation, but remains associated with UPF1 which directs it to the proteasome for removal, thus leading to a protein and mRNA turnover coupled process (Kuroha et al, 2013).

NMD targets

Despite decades of research, it is still not entirely clear what constitutes a bona fide NMD target. The introduction of a mutation, such as a single nucleotide variant, which gives rise to a PTC, represents the most obvious target for NMD. However, transcriptome profiling to identify NMD targets in cells of different species revealed that the majority of NMD‐sensitive transcripts do not contain PTCs but are rather mRNAs coding for full‐length proteins. This led to the hypothesis that a combination of NMD‐inducing and NMD‐antagonizing features will contribute to determine NMD susceptibility for any given mRNA. Some of the common features that render mRNAs susceptible to NMD include the presence of a PTC located at least 50‐55 nucleotides upstream of an EJC (Colombo et al, 2017), mRNAs with upstream ORFs (uORFs; Calvo et al, 2009), and the presence of a long 3′UTR (Hogg & Goff, 2010). However, recent use of cDNA Nanopore sequencing combined with short RNA‐seq allowed the detection of full‐length NMD substrates that are highly unstable and only display an increase in RNA levels when NMD is inhibited. This analysis identified NMD target mRNAs derived mainly from alternative exon usage, yet it did not identify long 3′UTRs as a common feature for NMD regulated mRNAs (Karousis et al, 2021). RNA‐seq allows for the analysis of steady‐state changes, which are influenced by stability, degradation, and transcription rates. Recently, SH‐linked alkylation for the metabolic sequencing of RNA (SLAM‐seq) using 4‐thiouracil pulse‐chase labeling (Herzog et al, 2017) was used to accurately measure changes in RNA half‐lives and to identify new targets of the NMD pathway in S. cerevisiae (Alalam et al, 2022). SLAM seq analysis of Smg5‐7 genetic knockouts in mouse ESCs revealed that NMD controls expression levels of the translation initiation factor Eif4a2 and its alternative splicing isoform that harbors a PTC‐encoding isoform (Eif4a2 ^PTC ). Upon NMD inhibition, aberrant expression of the eIF4A2^PTC elicits increased mTORC1 activity and translation rates and causes differentiation delays, highlighting a role of RNA stability regulation in development (Huth et al, 2022).

NMD regulation

The NMD pathway is dynamic and subject to regulation and impacts several physiological processes, such as the stress and immune responses. A correct magnitude of NMD activity is particularly important for proper brain function and indeed, NMD activity is extensively regulated during neural development (Jaffrey & Wilkinson, 2018). It was shown that NMD activity in human neuroblastoma cells is attenuated by fragile X protein FMRP, which is recruited to NMD targets by UPF1. FMRP acts as an NMD repressor in neural cells and in its absence, NMD is hyperactivated, leading to widespread transcriptome changes which contributes to intellectual disability and autism (Kurosaki et al, 2021). Recently, a conditional Smg6 mutant mouse model revealed that the NMD pathway has a role in controlling circadian clock regulation (Katsioudi et al, 2023). The NMD response varies within and across cell lines (Sato & Singer, 2021), in different cell tissues (Zetoune et al, 2008), and even among individuals (Nguyen et al, 2014; Rivas et al, 2015). NMD is tightly regulated by a negative feedback network that leads to a large proportion of core NMD factors being regulated by NMD itself in several organisms, including mammalian cells, nematodes, zebrafish, and plants (Huang et al, 2011; Yepiskoposyan et al, 2011; Longman et al, 2013). NMD can occur with equal probability during each round of translation of an mRNA molecule; however, this probability is variable and linked to sequence features, including the exon sequence downstream of the PTC, the PTC‐to‐intron distance, and the number of introns both upstream and downstream of the PTC. Furthermore, a subpopulation of mRNAs can escape NMD, further contributing to variation in NMD efficiency (Hoek et al, 2019). Use of a single‐cell approach comprising a bi‐directional NMD reporter expressing two β‐globin genes with or without a PTC in the same cell, allowed the characterization NMD efficiency in individual cells. This revealed a broad range of NMD efficiencies in the population (where some cells degraded essentially all mRNAs and others escaped NMD almost completely) and was correlated to the differential level of SMG1 expression and P‐UPF1. Mechanistically, this escape occurred either by translational read‐through at the PTC or by inefficient mRNA degradation following translation termination at the PTC (Sato & Singer, 2021).

Localized NMD and the stress response

The decay of NMD reporters in mammalian cells occurs in the cytoplasm (Trcek et al, 2013) and is closely linked to mRNA translation (Kervestin & Jacobson, 2012). We previously identified two novel NMD factors, NBAS and SEC13, which localize to the ER, raising the possibility that they could be involved in a localized NMD pathway (Longman et al, 2007; Casadio et al, 2015). There are precedents for a localized NMD response in neurons that regulate the expression of dendritic and axonal mRNAs upon the activation of their localized mRNA translation (Colak et al, 2013). It has been shown that mRNAs coding for secreted or transmembrane proteins are translated only when they encounter the ER (Wu et al, 2016). However, it remained largely unknown how NMD regulates the stability of RNAs translated at the ER, which due to their intrinsic localized translation, will not have sufficient exposure to cytoplasmic NMD surveillance. NBAS is a member of the Syntaxin 18 complex as part of the COPI vesicle and is involved in Golgi‐to‐ER retrograde transport (Aoki et al, 2009). We showed that NBAS fulfils a second, independent function and recruits the core NMD factor UPF1 to the membrane of the ER and activates a local NMD response that regulates RNAs associated with cellular stress and membrane trafficking (Fig 5; Longman et al, 2020). Loss‐of‐function mutations in NBAS have been found in diseases affecting bone, connective tissue, and acute liver failure (Hug et al, 2016; Staufner et al, 2020). We identified compound heterozygous variants in NBAS as a cause of atypical osteogenesis imperfecta (Balasubramanian et al, 2017). It remains to be seen whether the phenotype of NBAS mutations is due to faulty NMD response and/or defective Golgi‐to‐ER retrograde transport (Haack et al, 2015).

Open in a separate window

Figure 5

Localized NMD response at the ER

NBAS localizes to the outside membrane of the endoplasmic reticulum (ER), in the vicinity of the translocon, where it recruits the core NMD factor UPF1 to activate a local NMD response at the ER (Longman et al, 2020).

The Unfolded Protein Response (UPR) senses and responds to excessive amounts of misfolded proteins in the ER, that cause ER stress (Walter & Ron, 2011). Inappropriate UPR activation contributes to many human pathologies, most notably related to neurodegeneration (Hetz & Mollereau, 2014), therefore the fidelity of UPR activation must be tightly regulated. Accordingly, a role for NMD in regulating the UPR has been proposed (Karam et al, 2015; Goetz & Wilkinson, 2017). In particular, mRNAs encoding UPR components, including the UPR sensor IRE1α, as well as ATF‐4 and CHOP that are activated by PERK branch signaling, are regulated by NMD and thus control the threshold of cellular stress that is necessary to activate the UPR (Gardner, 2008; Karam et al, 2015; Sieber et al, 2016). A feed‐back loop mechanism operates, wherein NMD ensures a correct activation threshold for the UPR response and the activity of NMD is in turn down‐regulated by the UPR (Karam et al, 2015). It is likely that limiting the chronic activation of the UPR has a protective effect in neurodegenerative diseases. Along those lines, a protective role for UPF1 was shown in rodent primary neuronal models of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD; Barmada et al, 2015) and in a rat ALS paralysis model (Jackson et al, 2015).

It is also possible that modulation of ER‐NMD activity could be crucial to regulate ER stress (Longman et al, 2020). In such a scenario, NMD‐mediated degradation of RNA targets would decrease ER load by suppressing aggregation of truncated and/or misfolded proteins in the ER, limiting UPR activation. The newly identified UPF1_LL that, upon ISR activation and impaired translation, regulates a new subset of NMD targets, can be a key to the appropriate cellular stress response (Fritz et al, 2022).