• Open Access

Role of SMG-1-mediated Upf1 phosphorylation in mammalian nonsense-mediated mRNA decay

Authors

  • Akio Yamashita

    Corresponding author
    1. Advanced Medical Research Center, Yokohama City University, Yokohama, Kanagawa, Japan
    • Department of Molecular Biology, Yokohama City University School of Medicine, Yokohama, Kanagawa, Japan
    Search for more papers by this author

  • Communicated by: Mitsuhiro Yanagida

Correspondence: yamasita@yokohama-cu.ac.jp

Abstract

SMG-1, a member of the PIKK (phosphoinositide 3-kinase-related kinase) family, plays a critical role in the mRNA quality control system known as nonsense-mediated mRNA decay (NMD). NMD protects cells from the accumulation of aberrant mRNAs with premature termination codons (PTCs) which encode nonfunctional or potentially harmful truncated proteins. SMG-1 directly phosphorylates Upf1 helicase, another key component of NMD, upon recognition of PTC on postspliced mRNA during the initial round of translation. Phosphorylated-Upf1 recruits the SMG-5/SMG-7 complex to induce ribosome dissociation and decapping-mediated decay. Phospho-Upf1 also recruits the SMG-6 endonuclease which might be involved in endo-cleavage. Upf1 ATPase/helicase activities are likely required for the activation of other mRNA decay enzymes and the mRNA-protein complex dissociation to complete NMD. At present, a variety of tools are available that can specifically suppress NMD, and it has become possible to examine the contribution of NMD in a variety of physiological and pathological conditions.

Introduction

Eukaryotic gene expression is exquisitely regulated at several points between transcription and protein synthesis to ensure fidelity in the conversion of genetic information into biological functions. Nonsense-mediated mRNA decay (NMD) is a post-transcriptional surveillance pathway responsible for the recognition and degradation of mRNAs containing premature termination codons (PTCs; Huang & Wilkinson 2012; Schoenberg & Maquat 2012). An estimated one-third of known genetic disease and cancer-associated mutations generate PTCs (Bhuvanagiri et al. 2010; Keeling et al. 2012). PTCs can be generated by mutations including point mutations that create a nonsense codon, frameshift mutations, and splice site mutations which induce intron inclusion (Bhuvanagiri et al. 2010; Huang & Wilkinson 2012; Keeling et al. 2012). PTC-containing mRNAs (PTC-mRNAs) may also arise from physiological processes such as V(D)J recombination during T- and B-cell maturation (Frischmeyer-Guerrerio et al. 2011; Huang & Wilkinson 2012). Such PTC-mRNAs encode aberrant C-terminally truncated proteins that can manifest gain-of function or dominant-negative effects which may result in disease or cancer (Bhuvanagiri et al. 2010; Huang & Wilkinson 2012; Keeling et al. 2012). Hence, NMD is a cellular defence mechanism eliminating PTC-mRNAs to suppress the production of potentially harmful or nonfunctional polypeptides and ensures the accuracy of gene expression (Fig. 1). However, there are many cases in which NMD down-regulates mutant proteins with residual activity, augmenting the defects caused by the original mutation (Usuki et al. 2004; Bhuvanagiri et al. 2010; Zarraga et al. 2011). In these cases, NMD suppression can lead to the phenotypic rescue of certain PTC-related mutations (Usuki et al. 2004, 2006). However, NMD suppression could also facilitate the production of non-natural polypeptides, putative tumor-specific antigens encoded by frameshift mutations that could act as signals for the immune system (Pastor et al. 2010). Thus, determining the exact mechanisms of NMD will be critical for the development of pharmacological interventions that could be used as therapies for genetic diseases or cancer.

Figure 1.

Nonsense-mediated mRNA decay modulates some disease phenotypes. (A) Nonsense-mediated mRNA decay (NMD) degrades premature termination codon (PTC)-containing mRNAs (when detected, see below) produced by mutations to prevent potentially harmful protein production. (B) NMD can modulate nonsense-associated disease phenotypes. Normally, nonsense mRNA is absent or its expression level is very low due to NMD, resulting in a clinically asymptomatic phenotype. NMD is triggered if the mutation is located more than 55 nts upstream of the last exon–exon junction. In some cases, a subset of individuals will be heterozygous for a PTC mutation within 55 nts of the last exon–exon junction. In these cases, the PTC-containing mRNA is decay resistant, its expression levels are not decreased, and it produces a truncated, potentially dominant-negative polypeptide, possibly resulting in a clinical phenotype in the heterozygote.

Trans-acting factors of NMD in mammals

The phenomenon of PTC-mRNA down-regulation was initially discovered in humans and Saccharomyces cerevisiae in 1979 (Chang et al. 1979; Losson & Lacroute 1979) and later in other eukaryotes (Culbertson & Leeds 2003). The first link to disease came from genetic analysis of the dominant form of β-thalassemia in the late 1980s and early 1990s; and the subsequent molecular analysis that determined the responsible PTC which decreases β-chain mRNA abundance (Thein et al. 1990; Hall & Thein 1994; Thermann et al. 1998; Zhang et al. 1998). Generally, termination codons are recognized as premature if they are located more than 50–55 nucleotides (nts) 5′ of the last exon–exon junction (Fig. 1B). More than 98% of protein coding genes have termination codons in the last exon or termination codons within 50–55 nts of the last exon–exon junction (Nagy & Maquat 1998). Hence, NMD is a general and fundamental mechanism to survey gene mutations that generate PTC in mammals.

Nuclear splicing deposits an exon-junction complex (EJC) 20–24 nts 5′ of an exon–exon junction to mark ORFs (Le Hir et al. 2000). Transcriptome-wide analysis showed that about 80% of all exon–exon junctions are occupied by EJC (canonical EJC, cEJC; Sauliere et al. 2012; Singh et al. 2012). More than ten factors make up the EJC which is involved in various aspects of mRNA processing, such as inducing splicing, nuclear-cytoplasmic transportation, enhancement of translation, and NMD (Tange et al. 2004). Among the EJC components are RNPS1, Y14, Magoh/Magohb, eIF4A3, and MLN51 (also named CASC3 and Barentsz), each of which are also involved in NMD in mammals (Bono & Gehring 2011). Note that EJC can also deposit at noncanonical positions, where no splicing junctions exist (ncEJC), probably through SRSF1 and SRSF3 (Sauliere et al. 2012; Singh et al. 2012).

Genetic studies in S. cerevisiae and Caenorhabditis elegans have identified up-frameshift (UPF) and suppressor with morphogenetic effect on genitalia (smg) mutations, respectively, as trans-acting factors of NMD (Leeds et al. 1991; Pulak & Anderson 1993). Among them, Upf1 (SMG-2), Upf2 (SMG-3), and Upf3 (SMG-4) are ‘core’ factors of NMD (Anderson 2006; Culbertson 2007). These molecules are conserved in most, if not all, eukaryotes including mammals. In addition, genetic screens in C. elegans have identified SMG-1, SMG-5, SMG-6, and SMG-7 as multicellular organism specific trans-acting factors which are involved in the regulation of Upf1 phosphorylation (Page et al. 1999; Yamashita et al. 2005b; Anderson 2006). Furthermore, biochemical studies using mammalian cells have identified SMG-8 and SMG-9 as components of the SMG-1 complex (Yamashita et al. 2009). The details of Upf/SMG proteins are described in Table 1 and Box 1.

Table 1. Evolutionally conserved Upf/SMG genes are listed by model organism. The main structural feature(s) of each component is listed as well as its role in nonsense-mediated mRNA decay (NMD)
YeastNematodeFlyPlantVertebrateStructureFunction in NMD
  1. a

    Present in Oryza sativa and Vitis vinifera. Non orthologous in Arabidopsis thaliana.

SMG-1SMG1SMG-1aSMG-1Protein kinaseUpf1 phosphorylation
Upf1SMG-2Upf1Upf1Upf1RNA helicase, ATPaseCentral molecule of NMD
Upf2SMG-3Upf2Upf2Upf2Adaptor moleculeAdaptor for Upf1 and Upf3
Upf3SMG-4Upf3Upf3

Upf3a

Upf3b

Adaptor moleculeAdaptor for Upf2 and exon-junction complex
SMG-5SMG5SMG-5

14-3-3 like domain

PIN like domain

Upf1 dephosphorylation
SMG-6SMG6SMG-6

14-3-3 like domain

Endonuclease

Upf1 dephosphorylation, RNA endocleavage
Ebs1SMG-7

SMG7

SMG7L

SMG-71433 like domainUpf1 dephosphorylation, decapping
SMG-8SMG8SMG-8aSMG-8Subunit of SMG-1Regulation of SMG-1 kinase
SMG-9SMG9SMG-9aSMG-9Subunit of SMG-1Regulation of SMG-1 kinase

Box 1.

Box 1 Biochemical characterization of Upf/SMG gene products

Upf1 binds RNA, has RNA-dependent ATPase activity and 5′–3′ RNA helicase activity (Czaplinski et al. 1995; Chamieh et al. 2008). All of these Upf1 activities are essential for NMD in both S. cerevisiae (Weng et al. 1996a,b) and mammals (Mendell et al. 2002; Franks et al. 2010; Kashima et al. 2006). Upf1 has an N-terminal zinc-knuckle domain (the cysteine-histidine-rich CH domain), followed by its helicase domain (Czaplinski et al. 1995; Chamieh et al. 2008). Structural analysis showed that the CH domain, which binds Upf2, is in direct contact with the helicase domain to regulate its activity (Clerici et al. 2009; Chakrabarti et al. 2011). Biochemical/structural analysis indicates that the Upf2/Upf3 complex binds to the CH domain and causes a large conformational change, activating Upf1 ATPase/helicase activity (Chamieh et al. 2008; Chakrabarti et al. 2011). Upf1 binds directly to eRF1 and eRF3 through the N-terminal CH domain containing region (Wang et al. 2001; Ivanov et al. 2008). Both eRF1 and eRF3 can inhibit yeast Upf1 ATPase activity in vitro (Czaplinski et al. 1998). Upf2 has three MIF4G (middle portion of eIF4G) domains and RNA-binding activity (Kadlec et al. 2004). Upf2 functions as an adaptor molecule that links Upf1 and Upf3 (Czaplinski et al. 1995; Chamieh et al. 2008; Melero et al. 2012). Yeast Upf2 binds directly to eRF3 through an acidic region present between its Upf1 and Upf3-binding regions (Wang et al. 2001). Upf2 also directly links the SMG-1 and Upf3b in a Upf1-independent manner in mammals (Kashima et al. 2006). Mammalian Upf3 is encoded by two genes, Upf3a and Upf3b. Both genes have redundant function and link Upf2 and the EJC (Kunz et al. 2006; Chamieh et al. 2008; Melero et al. 2012). The Upf3a splice variant, Upf3aS, does not bind Upf2 but preferentially binds phospho-Upf1 and the SMG-5:SMG-7 complex (Ohnishi et al. 2003; Kunz et al. 2006). A significant role for Upf3 is not known in other organisms.

SMG-1, SMG-5, SMG-6, SMG-7, SMG-8, and SMG-9 are multicellular organism-specific trans-acting factors of NMD. SMG-1, SMG-8, and SMG-9 form the SMG-1 kinase complex (SMG1C) and catalyze Upf1 phosphorylation, a rate-limiting step of NMD (Yamashita et al. 2001, 2009; Kashima et al. 2006; Arias-Palomo et al. 2011). Distant N- and C-terminal regions of SMG-1 are required for its intrinsic kinase activity (Morita et al. 2007). The kinase activity of SMG-1 is suppressed by SMG-8 in the SURF complex, and the SMG-1-mediated Upf1 phosphorylation is induced by DECID complex formation (Yamashita et al. 2009; Arias-Palomo et al. 2011). SMG-8 is also required for the recruitment of SMG1C to SURF (Yamashita et al. 2009). SMG-5, SMG-6, and SMG-7 are evolutionally conserved, related proteins with 14-3-3 like domains, but each is required for NMD (Anders et al. 2003; Gatfield et al. 2003). The SMG-7 14-3-3-like domain can directly bind in vitro to a Upf1 phosphopeptide (Fukuhara et al. 2005). SMG-5, SMG-6, and SMG-7 are involved in the dephosphorylation of Upf1, probably through the recruitment of protein phosphatase 2A (PP2A; Ohnishi et al. 2003; Okada-Katsuhata et al. 2012). The majority of SMG-5 and SMG-7 forms a complex that preferentially binds to phosphorylated S1096 of Upf1 in vivo (Ohnishi et al. 2003; Okada-Katsuhata et al. 2012), whereas phospho-independent binding of the SMG-5:SMG-7 complex with the N-terminus of Upf1 is observed during NMD (Ohnishi et al. 2003). SMG-6 preferentially binds to phosphorylated T28 of Upf1 in vivo (Okada-Katsuhata et al. 2012). SMG-6 is an endonuclease and its activity is required for NMD and is proposed to cleave the PTC-mRNA near the PTC (Huntzinger et al. 2008; Eberle et al. 2009). SMG-6 binds the EJC and mRNA (Redon et al. 2007; Kashima et al. 2010; Okada-Katsuhata et al. 2012). SMG-5, SMG-6, and SMG-7 are highly diverse among higher eukaryotes. For instance, vertebrates and nematodes have all these molecules, whereas flies do not have SMG-7, and plants have two nonredundant SMG-7s, but not SMG-5 or SMG-6 (Anders et al. 2003; Gatfield et al. 2003; Ohnishi et al. 2003; Benkovics et al. 2011). These differences might reflect the difference of mRNA decay mechanisms among these species.

The list of trans-acting factors of NMD in mammals has grown recently to include the Ruvbl1/Ruvbl2 complex, RPB5, NAG, DHX34, Int6 (also named eIF3e), and possibility HSP90 (Longman et al. 2007; Morris et al. 2007; Izumi et al. 2010, 2012a).

Molecular mechanism of NMD in mammals

Mechanism of PTC recognition

Ribosomes are molecular machines that read triplet codons and recognize translation termination sites in the cytoplasm (Jackson et al. 2012). Ribosomes also recognize PTCs because NMD is inhibited by general translation inhibitors (Carter et al. 1995), the suppression of translation initiation (Chiu et al. 2004), or suppressor tRNAs which have an anticodon that binds the termination codon (Belgrader et al. 1993). NMD occurs during the initial round of translation (Thermann et al. 1998; Ishigaki et al. 2001), also called the pioneer round of translation, which is distinguished from the subsequent rounds of translation by different messenger ribonucleoprotein (mRNP) complexes, including the nuclear cap-binding protein complex (CBC; the CBP80/CBP20 complex; Ishigaki et al. 2001; Chiu et al. 2004; Choe et al. 2012). At present, the most critical question in the NMD research field is ‘How do cells distinguish PTCs from normal translation termination codons?’ Although our understanding is far from complete, recent progress has provided two mammalian models, the ‘'Faux 3′-UTR model'’ and the ‘'Post-Upf1 phosphorylation to mRNA decay'’. Neither of these models is able to explain all PTC-recognition scenarios, thus, both models are likely incomplete.

Downstream marker model

As described earlier, exon–exon junctions located 3′ of termination codons can trigger NMD as downstream cis-acting sequence elements and EJCs function as downstream markers for NMD (Fig. 2A top). In the initial round of translation, if a translation termination codon is located 5′ of an EJC, the SMG-1 kinase complex (SMG1C: SMG-1/SMG-8/SMG-9), Upf1 helicase, and translation termination factors (eRF1 and eRF3) form the SMG-1/Upf1/eRF1/eRF3 (SURF) complex on the ribosome/mRNP complex which includes CBC and PABPC1/C4 (Fig. 2A middle; Kashima et al. 2006; Yamashita et al. 2009). There is supporting evidence for the active involvement of CBP80 in PTC recognition (Hosoda et al. 2005; Hwang et al. 2010). The SMG-1-mediated phosphorylation of Upf1, a rate-limiting step of NMD, is allosterically suppressed by SMG-8 within the SURF complex at this stage (Yamashita et al. 2009; Arias-Palomo et al. 2011; Fernandez et al. 2011). The association of the ribosome, SURF, and Upf2/Upf3/EJC complexes forms the decay inducing complex (DECID; ribosome/SURF/EJC complex; Fig. 2A bottom; Kashima et al. 2006; Yamashita et al. 2009; Izumi et al. 2010). The DECID complex formation induces SMG-1-mediated Upf1 phosphorylation on mRNP (Fig. 2C top; Kashima et al. 2006; Yamashita et al. 2009; Izumi et al. 2010; Arias-Palomo et al. 2011; Fernandez et al. 2011). The transition from the ribosome/SURF complex to the DECID complex is supported by the Ruvbl1/Ruvbl2 complex (AAA+ family ATPase), RPB5, and HSP90 probably as putative molecular chaperon complex (Izumi et al. 2010, 2012a,b). Note that caffeine, an SMG-1 inhibitor can cause the accumulation of PTC-mRNA and the truncated proteins they encode (Yamashita et al. 2001; Usuki et al. 2004), and the removal of caffeine, which restores SMG-1 activity will restore NMD and degrade the PTC-mRNAs (Ivanov et al. 2007; A. Yamashita, unpublished data). This indicates that PTC-mRNAs keep susceptibility to NMD after multiple rounds of translation. Thus, there is no NMD immunity of PTC-mRNA; that is, they escape further susceptibility to this decay pathway (Stephenson & Maquat 1996; Maderazo et al. 2003), before Upf1 phosphorylation in mammals.

Figure 2.

Molecular mechanism of nonsense-mediated mRNA decay (NMD) in mammals. (A) Premature termination codon (PTC)-recognition by downstream marker model. PTC recognition is established by the formation of mRNA surveillance complexes called ‘SURF’ and ‘DECID’ that occur at sites containing a PTC-recognizing ribosome and a downstream exon-junction complex (EJC) during the initial round of translation. During translation termination, the transiently formed SURF complex composed of SMG1C (SMG-1/SMG-8/SMG-9 complex), Upf1, and eRF detects the downstream EJC and forms a DECID complex, which induces SMG-1-mediated Upf1 phosphorylation. The SMG-5/SMG-7 complex binds to phospho-S1096 of Upf1 to dissociate the ribosome and release factor from Upf1. SMG-6 binds to phospho-T28 of Upf1 to induce Upf1 dissociation from the mRNA. (B) PTC recognition by faux 3′UTR model. Translation termination is induced when mRNAs have an ‘aberrant’ 3′UTR. An aberrantly terminated ribosome might recruit Upf proteins to prevent the binding of eRF3 and PABP which forms in normal translation termination. (C) mRNA degradation by NMD. mRNA degradation is initiated by bi-phasic deadenylation in mammals. After the first phase of deadenylation to approximately 110 nt, mRNA degrades from both ends. SMG-6 endonuclease-mediated endo-cleavage also takes place to degrade mRNA. AUG, start codon; Ter, termination codon; eRF, eRF1/eRF3 complex; 2, Upf2; 3, Upf3; S8, SMG-8; S9, SMG-9; eRF, eRF1/eRF3 complex; P, phosphate group; S5:7, the SMG-5/SMG-7 complex, S6, SMG-6.

The ribosome and translation termination factors are involved in both premature and normal translation termination events. However, fundamental differences exist between the two types of termination events. The premature translation termination complexes, namely the ribosome/SURF and the DECID complex, do not appear to communicate with the poly(A) tail or the 5′ cap of the mRNA (Kashima et al. 2006; Yamashita et al. 2009; Izumi et al. 2010), whereas the normal translation termination complex communicates with both of these elements via protein–protein interactions for efficient translation termination (Fig. 2B-a; Clerici et al. 2009) and translation initiation (Uchida et al. 2002).

Faux 3′-UTR model

In mammals, PTC recognition generally depends on splicing and the EJC (Huang & Wilkinson 2012; Schoenberg & Maquat 2012; Fig. 2B). However, these features are nearly dispensable for PTC discrimination in S. cerevisiae, C. elegans, and Drosophila melanogaster (Losson & Lacroute 1979; Pulak & Anderson 1993; Gatfield et al. 2003; Longman et al. 2007). In addition, the recognition of PTCs, independent of downstream exon-junction, has also been demonstrated in mammals (Zhang et al. 1998; Buhler et al. 2006). The molecular mechanism behind this process is still debated (Matsuda et al. 2007; Eberle et al. 2008; Ivanov et al. 2008; Singh et al. 2008; Hogg & Goff 2010).

Ribosome release from a termination codon is dramatically less efficient in long 3′UTR compared with short 3′UTR, and this less efficient ribosome dissociation is dependent on NMD factors in yeast extract (Amrani et al. 2004). This distinct biochemical feature of premature termination and normal termination is also reported in reticulocyte lysate (Peixeiro et al. 2012). This suggests that cells can detect ‘normal’ 3′UTRs and aberrantly ‘long’ 3′UTRs to discriminate PTCs from normal translation termination codons. The 3′UTR is defined by a translation termination codon, which is recognized by the ribosome and release factors, and a poly(A) tail which is coated by PABPs (PABPC1, PABPC4, and PABPN1). The position within the mRNA could also be marked by 3′ end processing factors such as Hrp1p which is involved in NMD in S. cerevisiae (Kessler et al. 1997; Gonzalez et al. 2000). Nucleophosmin is a factor deposited at the 3′ end of a mRNA during polyadenylation in mammals, although this modification has not been shown to impact NMD (Sagawa et al. 2011). As PABPC1 (Pab1p in S. cerevisiae) can compete for binding of Upf1 to eRF3, a simple competition model between Upf1 and PABPC1 is proposed as the trigger for ‘long’ 3′UTR-mediated PTC recognition (Fig. 2B; Eberle et al. 2008; Ivanov et al. 2008; Silva et al. 2008; Singh et al. 2008). The less efficient ribosome dissociation from PTC might be induced by the Upf1-mediated inhibition of the communication of eRF3 and PABPC1. Consistently, PTC-recognition complexes do not contain PABPC1 (see 'Faux 3′-UTR model').

Intriguingly, the transcriptome-wide profile of the N6-methyladenosine (m6A), the most common internal mRNA modification in adenosine found in eukaryotes (Jia et al. 2011), in mRNA revealed that the m6A is highly enriched near the translation termination codon and in the 3′UTR in a large number of mRNAs in mammals (Dominissini et al. 2012; Meyer et al. 2012). This epigenetic modification of a translation termination codon can be a marker to distinguish a ‘normal’ termination codon from a ‘premature’ termination codon. In S. cerevisiae, there is little modification of m6A existing in coding regions, although this is not near the translation termination codon (Clancy et al. 2002).

A large number of mammalian genes have ‘normal’ long 3′UTRs which do not undergo NMD. Therefore, the importance of RNA secondary structure and the existence of putative NMD inhibiting elements/factors have been discussed. Further studies are needed to resolve the mechanisms of PTC discrimination triggered by ‘long’ 3′UTRs. Aberrantly long 3′UTRs can be generated by mutations in the polyadenylation signal sequence (Danckwardt et al. 2008). Although no studies have been reported in mammals, NMD-mediated degradation of ‘long’ 3′UTR mRNA, caused by polyadenylation signal sequence deletion, is reported in C. elegans (Pulak & Anderson 1993).

Exon-junction complex components are also involved in downstream exon junction independent NMD for the β-globin reporter and the glutathione peroxidases 1 (GPx1) reporter mRNA in mammals (Matsuda et al. 2007; Singh et al. 2008; A.Y., unpublished data). Noncanonical EJCs might be recruited to these mRNAs, although the presence of ncEJCs in long 3′UTR cannot be predicted (Sauliere et al. 2012; Singh et al. 2012). In addition, sequence specific but splicing independent recruitment of the EJC component eIF4A3 to GPx1 mRNA and Upf1 to various mRNA has been reported (Kaygun & Marzluff 2005; Kim et al. 2005; Budiman et al. 2009).

SMG-1 and its co-factors are also needed for exon junction independent NMD (A.Y., unpublished data). Together with the essential role of EJC for Upf1 phosphorylation, DECID formation and subsequent SMG-1-mediated phosphorylation of Upf1 seems to be universally required in mammalian NMD.

Post-Upf1 phosphorylation to mRNA decay

SMG-1 phosphorylates more than seven serine (S)/threonine (T)/glutamine (Q) motifs of Upf1 in vitro (Yamashita et al. 2001), and at least four sites are phosphorylated in vivo (Yamashita et al. 2001; Ohnishi et al. 2003; Matsuoka et al. 2007; Okada-Katsuhata et al. 2012; Fig. 2C). Among them, T28 and S1096 are responsible for phospho-specific recruitment of SMG-6 to the N-terminal conserved region, and the SMG-5/SMG-7 heterodimer complex to the C-terminal SQ-rich region of Upf1, respectively (Fig. 2C bottom; Okada-Katsuhata et al. 2012). The SMG-5/SMG-7 complex and SMG-6 can simultaneously bind to phospho-Upf1 at different phosphorylation sites, and their binding is required for PP2A-mediated Upf1 dephosphorylation and dissociation from the mRNP before degradation (Ohnishi et al. 2003; Okada-Katsuhata et al. 2012). Because inhibition of Upf1 dephosphorylation by inactivation of SMG-5, SMG-6, SMG-7 or PP2A inhibitors suppresses NMD, Upf1 phosphorylation and dephosphorylation cycles are required for NMD (Ohnishi et al. 2003; Ivanov et al. 2007; Okada-Katsuhata et al. 2012).

The SMG-5/SMG-7 complex appears indispensable for the dissociation of the ribosome, release factors, Upf2, and EJC because inactivation of SMG-5 causes the accumulation of a phospho-Upf1 complex containing them, on CBC and PABPC1/C4 containing mRNPs (Okada-Katsuhata et al. 2012). In addition, SMG-7 is considered to be a mRNA decay mediator because its artificial tethering at either the 3′ or 5′UTR of mRNA induces Dcp2 (decapping enzyme) and Xrn1 (5′–3′ exonuclease) dependent mRNA decay (Fig. 2D-b,c; Unterholzner & Izaurralde 2004). Conversely, SMG-6 is dispensable for the dissociation of the ribosome, release factors, and Upf2 from Upf1 (Okada-Katsuhata et al. 2012). However, SMG-6 is an endonuclease and its activity is required for the degradation of the mRNA; therefore, SMG-6 is proposed to cleave PTC-mRNA at the PTC (Fig. 2D-d; Huntzinger et al. 2008; Eberle et al. 2009). The major mRNA degradation mechanism; that is, endo- or exo-ribonucleic decay, remains an open question in mammals.

Inactivation of SMG-5, SMG-6, or SMG-7 causes the accumulation of phospho-Upf1 and increases the amount of PTC-mRNA encoded truncated protein in mammals (Paillusson et al. 2005; Kashima et al. 2006), as reported in C. elegans (Pulak & Anderson 1993; Page et al. 1999). In addition, the NMD inhibitor NMDI1 does not suppress PTC-mRNA translation even though the treatment causes the accumulation of phospho-Upf1 in mammals (Durand et al. 2007). These observations indicate that Upf1 phosphorylation alone does not suppress translation of PTC-mRNAs. Conversely, phospho-Upf1 does suppress translation initiation through a direct interaction with eIF3a (Isken et al. 2008). Further studies are needed to reveal the details of phospho-Upf1-mediated inhibition of translation initiation.

The stability of PTC-mRNA-encoded aberrant protein products are allele dependent in human culture cells (Yamashita et al. 2001; Usuki et al. 2004; Anczukow et al. 2008). Some aberrant proteins became destabilized, whereas others are stabilized, and the stability of the aberrant protein is not affected by NMD activity in human cells (Yamashita et al. 2001; Usuki et al. 2004; Anczukow et al. 2008). In contrast, Upf1 mediates destabilization of PTC-mRNA encoded aberrant protein products in S. cerevisiae (Kuroha et al. 2009).

Kinetic analysis of PTC-mRNAs revealed that mRNA degradation is carried out in a bi-phasic deadenylation model and that in NMD is similar to normal mRNA decay (Fig. 2D-a; Yamashita et al. 2005a; Ezzeddine et al. 2007). Initially, mRNAs have poly(A) tails ranging from 200 to 250 nts in mammalian cells. In the first phase, which is PAN2/PAN3 mediated, the deadenylation rate of PTC-β-globin mRNA to approximately 110 nts is accelerated compared with normal mRNA deadenylation (Yamashita et al. 2005a; Ezzeddine et al. 2007). After first-phase deadenylation, Dcp2-mediated decapping and subsequent XrnI-mediated exoribonucleic degradation (from 5′ to 3′), or the CCR4-CAF1-mediated deadenylation and subsequent mRNA degradation from both ends (from decapping to 5′–3′ degradation and exosome-mediated 3′–5′degradation) take place to eliminate the PTC-mRNA (Yamashita et al. 2005a; Chen & Shyu 2011).

Both the PAN2/PAN3 complex and the CCR4/CAF1 complex are recruited to the mRNP by a direct interaction with PABPC1/C4 through PAN3 and TOB1/2, using the PAM2 motif within these proteins (Ezzeddine et al. 2007; Funakoshi et al. 2007). The PAM2-binding site of PABPC1/C4 is masked by the PAM2 motif of eRF3, which has a higher affinity, during normal translation termination (Kozlov & Gehring 2010; Osawa et al. 2012). However, as described in 'Faux 3′-UTR model', PABPC1/C4 is not part of the PTC-recognition complex (Kashima et al. 2006; Yamashita et al. 2009). Consistently, the binding of eRF3 and PABPC1 is inhibited by Upf1 in vitro (Singh et al. 2008; Kervestin et al. 2012). Based on these observations, the simplest interpretation is that the PAN2/PAN3 complex and the CCR4/CAF1/TOB1/2 complex are efficiently recruited to PABPC1/C4 when eRF3 is not bound to PABPC1/C4 (Chen & Shyu 2011). It is also possible that the PTC-recognized Upf1 containing complex directly recruits the poly(A) nuclease complex to the PTC-mRNP. Dcp2 forms a complex with Dcp1, Edc4, and PNRC2 to bind Upf1 through several direct interactions (Fenger-Gron et al. 2005; Cho et al. 2009). Therefore, Upf1 recruits Dcp2 to the NMD-induced complex in mammals.

ATPase-deficient Upf1, which cannot promote NMD, is highly phosphorylated (Kashima et al. 2006; Isken et al. 2008; Cho et al. 2009; Franks et al. 2010; Okada-Katsuhata et al. 2012) and binds efficiently with SMG-5, SMG-6, SMG-7 (Franks et al. 2010; Okada-Katsuhata et al. 2012), and other mRNA decay enzymes on the PTC-mRNPs (Isken et al. 2008; Cho et al. 2009; Franks et al. 2010). These observations indicate that the binding of SMG-5, SMG-6, and SMG-7 to phospho-Upf1 is not sufficient to promote dephosphorylation, even though they bind to PP2A (Chiu et al. 2003; Ohnishi et al. 2003) and that the binding of mRNA decay enzymes to phospho-Upf1 does not promote mRNA decay efficiently (Kashima et al. 2006; Franks et al. 2010). The activation of protein phosphatases and mRNA decay enzymes is likely inhibited in this situation even if they are recruited to the PTC-mRNP. The ATPase and helicase activity of Upf1 might be required for the activation of the phosphatase and mRNA decay enzymes in this context. Additionally, Upf1 ATPase-dependent mRNP disassembly might also be required for the completion of NMD (Franks et al. 2010).

Gene expression regulation through NMD

In addition to PTC-mRNAs derived from mutated genes, NMD targets inappropriately spliced transcripts (Fig. 3). Based on predictions, more than 60% of genes have alternatively spliced products and most of them have at least one PTC isoform (Green et al. 2003). The elimination of these PTC isoforms by NMD is experimentally confirmed in both mammalian cell line and mouse tissues (Pan et al. 2006; Weischenfeldt et al. 2012).

Figure 3.

Structure of physiological targets of nonsense-mediated mRNA decay (NMD) in mammals. Physiological targets of NMD. (A) Intron-retaining mRNA due to splicing error. (B) mRNA like noncoding RNA such as snoRNA host gene transcripts. (C) mRNA harboring an upstream open reading frame (uORF). (D) 3′UTR containing an intron. (E) Alternative splicing products undergo premature termination codon (PTC)-generating exon inclusion to regulate their own gene expression. (F) Selenocysteine codons, UAG, can be recognized as PTCs during selenium depletion and (G) mRNA having a NMD-sensitive long 3′UTR.

Nonsense-mediated mRNA decay also targets many physiological mRNAs to regulate their abundance (Mendell et al. 2004; Thoren et al. 2010) and stability (Tani et al. 2012) in mammals. In particular, NMD targets (i) mRNA like noncoding RNA such as SNHG1 and SNHG2, snoRNA host gene products (Ideue et al. 2007; Yamashita et al. 2009); (ii) the selenocysteine codon, UAG, can be recognized as a PTC during selenium depletion such as GPx1 (Sun et al. 2000; Usuki et al. 2011); (iii) mRNAs with upstream open reading frames (uORFs) such as ATF3 and NAT9 (Mendell et al. 2004; Viegas et al. 2007); (iv) 3′UTR intron containing gene products such as GADD45B and TBL2 (Viegas et al. 2007); (v) mRNA alternative splicing products that include PTCs to regulate their own gene expression such as ribosomal proteins and SR proteins (Cuccurese et al. 2005; Lareau et al. 2007); and (vi) mRNA having a NMD sensitive ‘long’ 3′UTR such as SMG-5 (Mendell et al. 2004; Singh et al. 2008).

NMD research approaches the clinic

Nonsense-mediated mRNA decay normally protects cells from nonfunctional, potentially harmful, polypeptides encoded by PTC-mRNAs from mutations in the genome (Bhuvanagiri et al. 2010; Keeling & Bedwell 2011; Fig. 1). However, <10% of mutations cause PTCs because of a single nucleotide substitution. In these cases, PTC read-through may be an option to preserve some protein function. Some compounds, for example, aminoglycoside derivatives and ataluren, promote read-through and they can provide near-normal protein production (Keeling & Bedwell 2011; Fig. 4A). The read-through strategy might be further improved by NMD suppression in these cases (Linde et al. 2007).

Figure 4.

Application for premature termination codon (PTC)-associated genetic diseases and cancer. (A) Read-through molecules produce nearly normal proteins from PTC-mRNA. This method can be used for patients who have single nucleotide nonsense mutations. (B) Nonsense-mediated mRNA decay (NMD) inhibitory molecules produce truncated proteins from PTC-mRNAs. This method can be used for patients whose truncated protein has residual activity encoded by the PTC-mRNA. (C) PTC-mRNAs produced from frameshift mutations encode non-natural polypeptides. NMD suppression can increase the production of non-natural polypeptides and induce antitumor immune response to eliminate tumor cells.

Another situation when inhibiting NMD may be beneficial would be when some of the truncated proteins encoded by PTC-mRNAs retain normal function, at least partially. In these cases, NMD destroys the aberrant proteins exacerbating disease conditions (Usuki et al. 2004; Bhuvanagiri et al. 2010; Zarraga et al. 2011). This scenario has been validated in PTC-associated Ullrich's disease (Usuki et al. 2004, 2006; Fig. 4B). However, the pharmacological approaches using wortmannin or caffeine, both SMG-1 inhibitors, used in the study cannot be applied to patients because of toxicity and nonspecific side effects. Importantly, there are NMDI1 and new SMG-1-specific inhibitors (Durand et al. 2007; Gopalsamy et al. 2012). They might be useful for not only NMD-exacerbated diseases, but also read-through sensitive genetic disorders. It will also be important to test these new reagents and to identify appropriate target molecules for NMD suppression to reduce cytotoxicity in future studies.

Finally, it is predicted that some cancers are aided by NMD. Tumor cells frequently lose their DNA repair, DNA damage, and apoptosis pathways (Scholzova et al. 2007). Hence, point and frameshift mutations are thought to accumulate in tumors during malignant transformation. These mutations are expected to affect not only nonessential genes but also essential cell survival genes. Although aberrant truncated proteins that are encoded by mutations in essential genes could be harmful (reducing the tumor), NMD may eliminate them and might maintain cancer cell homeostasis (Scholzova et al. 2007). In addition, cancer cells with frameshift mutations that encode non-natural polypeptides could potentially induce the immune system if they were not eliminated by NMD (El-Bchiri et al. 2008). This hypothesis was tested in mice. The results showed that NMD suppression allowed the generation of polypeptides encoded by frameshift mutations, inducing the antitumor immune response to eliminate tumor cells (Pastor et al. 2010; Fig. 4C). Based on these observations, NMD repression may be a potential target for certain cancers.

Perspective

After the cloning of the human SMG-1 (Yamashita et al. 2001), a dozen years of research, side by side with many scientists, have gradually uncovered some of the mechanisms of PTC-recognition and mRNA degradation during NMD in mammals. However, many unresolved questions remain in the NMD research field. I introduced NMD as ‘a mRNA surveillance mechanism’ in this review of recent research progress. However, NMD is just being recognized as ‘a regulatory mechanism of gene expression’ (Huang & Wilkinson 2012; Schoenberg & Maquat 2012). In addition, most trans-acting factors of NMD have various cellular functions in addition to NMD such as telomere maintenance and histone mRNA decay (Huang & Wilkinson 2012; Schoenberg & Maquat 2012). Noteworthy, gene knockout analysis demonstrated that Upf1, Upf2, and SMG-1 are essential for mouse embryogenesis (Medghalchi et al. 2001; Weischenfeldt et al. 2008; McIlwain et al. 2010). Further analyses will reveal the physiological significance of NMD, trans-acting factors of NMD, and NMD intervention may become a new strategy for the treatment of genetic diseases and cancer.

Acknowledgements

I thank Professor Shigeo Ohno for all of his support. I also thank my current and previous colleagues and collaborators. This study was funded by the Ministry of Education, Culture, Sports, Science, and Technology of Japan [a Grant-in-Aid for Young Scientists (A), Scientific Research on Innovative Areas ‘RNA regulation’ and Scientific Research on Innovative Areas ‘Functional machinery for noncoding RNAs’], Takeda Science Foundation, and the Yokohama Foundation for Advancement of Medical Science.

Ancillary