Advertisement

Features and factors that dictate if terminating ribosomes cause or counteract nonsense-mediated mRNA decay

Open AccessPublished:October 12, 2022DOI:https://doi.org/10.1016/j.jbc.2022.102592
      Nonsense-mediated mRNA decay (NMD) is a quality control pathway in eukaryotes that continuously monitors mRNA transcripts to ensure truncated polypeptides are not produced. The expression of many normal mRNAs that encode full-length polypeptides is also regulated by this pathway. Such transcript surveillance by NMD is intimately linked to translation termination. When a ribosome terminates translation at a normal termination codon, NMD is not activated, and mRNA can undergo repeated rounds of translation. On the other hand, when translation termination is deemed abnormal, such as that on a premature termination codon, it leads to a series of poorly understood events involving the NMD pathway, which destabilizes the transcript. In this review, we summarize our current understanding of how the NMD machinery interfaces with the translation termination factors to initiate NMD. We also discuss a variety of cis-acting sequence contexts and trans-acting factors that can cause readthrough, ribosome reinitiation, or ribosome frameshifting at stop codons predicted to induce NMD. These alternative outcomes can lead to the ribosome translating downstream of such stop codons and hence the transcript escaping NMD. NMD escape via these mechanisms can have wide-ranging implications on human health, from being exploited by viruses to hijack host cell systems to being harnessed as potential therapeutic possibilities to treat genetic diseases.

      Keywords

      Abbreviations:

      AS (alternative pre-mRNA splicing), eIF (eukaryotic translation initiation factor), EJC (exon junction complex), IP6 (inositol hexakisphosphate), HTLV-1 (human-T-cell leukemia virus type-1), NMD (nonsense-mediated mRNA decay), NTC (normal termination codon), PABP (poly(A)-binding protein), poly(A) (polyadenylate), PRF (programmed ribosome frameshifting), PTC (premature termination codon), SR (serine–arginine), uORF (upstream ORF)
      Cellular life depends on the accurate decoding of genetic information into functional molecules. As instructions in mRNA are translated into proteins, ribosomes engaged in every step of translation are closely monitored to ensure that genetic information is decoded with high fidelity. For example, during translation elongation, ribosome-linked quality control mechanisms monitor for abnormally paused, stalled, or colliding ribosomes (
      • Joazeiro C.A.P.
      Mechanisms and functions of ribosome-associated protein quality control.
      ). These mechanisms perform the dual role of degrading problematic mRNAs and rescuing the ribosomes stuck on these mRNAs. Similarly, ribosomes undergoing translation termination are monitored by a mechanism called nonsense-mediated mRNA decay (NMD), which differentiates normal termination events from those that occur at premature termination sites. NMD serves two vital functions in the cell; it acts as a transcript quality control monitor and as a gene regulation mechanism (
      • Yi Z.
      • Sanjeev M.
      • Singh G.
      The branched nature of the nonsense-mediated mRNA decay pathway.
      ). Both functions begin when translation termination is determined to be abnormal, and both result in rapid mRNA degradation. The primary difference between the quality control and gene regulation functions of NMD is if the abnormal termination arises from a mutation or a programmed feature of the transcript. In this review, we will focus on how the interplay between the ribosome, translation termination factors, and NMD machinery leads to NMD, and how certain behaviors of terminating ribosomes can sometimes suppress NMD activation.
      Several scenarios can lead to stop codons that can induce NMD (
      • Lejeune F.
      Nonsense-mediated mRNA decay, a finely regulated mechanism.
      ). As a quality control mechanism, NMD monitors translation termination to identify mRNAs that may have acquired premature termination codons (PTCs) because of mutations or errors during RNA biogenesis (
      • Celik A.
      • Baker R.
      • He F.
      • Jacobson A.
      High-resolution profiling of NMD targets in yeast reveals translational fidelity as a basis for substrate selection.
      ,
      • Celik A.
      • He F.
      • Jacobson A.
      NMD monitors translational fidelity 24/7.
      ). As the name suggests, PTCs are stop codons that occur within the normal ORF of a transcript. By rapidly degrading PTC-containing mRNAs, NMD limits the expression of truncated polypeptides, which may otherwise be detrimental to cells (
      • Lykke-Andersen S.
      • Jensen T.H.
      Nonsense-mediated mRNA decay: an intricate machinery that shapes transcriptomes.
      ). In vertebrates, a major class of NMD substrates arises from regulated alternative pre-mRNA splicing (AS) that introduces a PTC because of a variety of AS events (e.g., exon skipping or inclusion, intron retention). In this capacity, NMD serves as a gene regulatory function. NMD-targeted mRNAs also consist of normal protein coding transcripts bearing features such as upstream ORFs (uORFs), long 3′-UTRs or 3′-UTR introns, which make normal termination events resemble NMD-inducing termination events (
      • Yi Z.
      • Sanjeev M.
      • Singh G.
      The branched nature of the nonsense-mediated mRNA decay pathway.
      ,
      • Celik A.
      • He F.
      • Jacobson A.
      NMD monitors translational fidelity 24/7.
      ). This group of NMD targets also includes a special set of naturally occurring mRNAs containing internal stop codons that can be recoded in a regulated manner to specify selenocysteine, the 21st amino acid (
      • Moriarty P.M.
      • Reddy C.C.
      • Maquat L.E.
      Selenium deficiency reduces the abundance of mRNA for Se-dependent glutathione peroxidase 1 by a UGA-dependent mechanism likely to be nonsense codon-mediated decay of cytoplasmic mRNA.
      ). NMD-induced degradation of such normal mRNAs serves to fine-tune levels of encoded proteins. As NMD links abnormal translation termination to RNA degradation, it can also be employed to control levels of long noncoding RNAs and muffle pseudogene expression (
      • Lykke-Andersen S.
      • Jensen T.H.
      Nonsense-mediated mRNA decay: an intricate machinery that shapes transcriptomes.
      ). To generalize, NMD occurs when translation termination is deemed to be abnormal. To understand why termination in a variety of contexts is sensed as abnormal by NMD, and why some such events may go undetected leading to escape from NMD, we must first establish what occurs when a ribosome terminates translation on a normal stop codon.

      Normal translation termination and ribosome recycling

      The basic translation termination reaction

      Translation termination begins upon recognition of one of the three stop codons: UAA, UAG, or UGA (Fig. 1A). While sense codons are recognized by tRNAs, stop codons are recognized by eRF1 (
      • Brown A.
      • Shao S.
      • Murray J.
      • Hegde R.S.
      • Ramakrishnan V.
      Structural basis for stop codon recognition in eukaryotes.
      ), a protein translation termination factor that structurally mimics tRNA. Like tRNAs, eRF1 enters the empty amino-acyl (A)-site on the ribosome where it forms hydrogen bonds with stop codons on mRNA (
      • Brown A.
      • Shao S.
      • Murray J.
      • Hegde R.S.
      • Ramakrishnan V.
      Structural basis for stop codon recognition in eukaryotes.
      ,
      • Matheisl S.
      • Berninghausen O.
      • Becker T.
      • Beckmann R.
      Structure of a human translation termination complex.
      ). eRF1 binding to the A-site is just as rapid as aminoacyl-tRNA binding, but eRF1 appears to reside in this position in the ribosome for a longer period (
      • Lawson M.R.
      • Lessen L.N.
      • Wang J.
      • Prabhakar A.
      • Corsepius N.C.
      • Green R.
      • et al.
      Mechanisms that ensure speed and fidelity in eukaryotic translation termination.
      ). Importantly, unlike sense codon recognition, where tRNA pairing is limited to the triplet nucleotides of a codon, the nucleotide (nt) following a stop codon (+4 nt) plays a key role in the termination reaction (Fig. 1A) (
      • Mccaughan K.K.
      • Brown C.M.
      • Dalphin M.E.
      • Berry M.J.
      • Tate W.P.
      Translational termination efficiency in mammals is influenced by the base following the stop codon.
      ). After recognition of the stop codon by eRF1 within the ribosome decoding center, the mRNA is compacted and the +4 nt is pulled into the A-site (
      • Brown A.
      • Shao S.
      • Murray J.
      • Hegde R.S.
      • Ramakrishnan V.
      Structural basis for stop codon recognition in eukaryotes.
      ,
      • Shao S.
      • Murray J.
      • Brown A.
      • Taunton J.
      • Ramakrishnan V.
      • Hegde R.S.
      • et al.
      Decoding mammalian ribosome-mRNA states by translational GTPase complexes.
      ). This mRNA conformation results from the flipping of the 18S rRNA adenine (A) base at position 1825 (rRNA numbering is as in Ref. (
      • Brown A.
      • Shao S.
      • Murray J.
      • Hegde R.S.
      • Ramakrishnan V.
      Structural basis for stop codon recognition in eukaryotes.
      ) based on rabbit/mammalian 18S rRNA sequences), which stacks with the +2 nt of the stop codon, which in turn stacks with the +3 nt. These base interactions further allow G626 and C1698 of the rRNA to stack with +4 and +5 nt of the mRNA, respectively, thereby extending the recognition beyond the triplet stop codon (
      • Brown A.
      • Shao S.
      • Murray J.
      • Hegde R.S.
      • Ramakrishnan V.
      Structural basis for stop codon recognition in eukaryotes.
      ,
      • Shao S.
      • Murray J.
      • Brown A.
      • Taunton J.
      • Ramakrishnan V.
      • Hegde R.S.
      • et al.
      Decoding mammalian ribosome-mRNA states by translational GTPase complexes.
      ). Importantly, stacking between rRNA and mRNA is more stable with a purine in the +4 nt position (
      • Brown A.
      • Shao S.
      • Murray J.
      • Hegde R.S.
      • Ramakrishnan V.
      Structural basis for stop codon recognition in eukaryotes.
      ), which could explain why purines are common in the +4 position, with UAA(A/G) and UGA(A/G) being the two most common tetra-nt stop codons in eukaryotes (
      • Brown C.M.
      • Stockwell P.A.
      • Trotman C.N.A.
      • Tate W.P.
      Sequence analysis suggests that tetra-nucleotides signal the termination of protein synthesis in eukaryotes.
      ). Similarly, a purine at the +5 position can strengthen a weak stop with pyrimidine at the +4 position (
      • Mccaughan K.K.
      • Brown C.M.
      • Dalphin M.E.
      • Berry M.J.
      • Tate W.P.
      Translational termination efficiency in mammals is influenced by the base following the stop codon.
      ). Thus, translation termination is dictated by interactions beyond eRF1–stop codon recognition and includes mRNA–rRNA interactions.
      Figure thumbnail gr1
      Figure 1Key steps in normal translation termination and recycling. mRNA is shown in gray with the thicker line depicting the coding sequence and the thinner line depicting the “normal” 3′ UTR. The ribosome is shown in blue, and other termination factors are labeled. A, normal translation termination begins when eRF1 recognizes a termination codon (shown as UGA) and the following nucleotide (+4 position; preferred nucleotides are shown) in the A-site along with the GTP-bound form of eRF3 to form the termination complex. B, GTP hydrolysis by eRF3 causes a conformational change in eRF1 that positions it to mediate hydrolysis and release of the nascent polypeptide. C, eRF3-GDP dissociates from the termination complex, and ABCE1, bound to ATP, joins to begin ribosome recycling. D, ATP hydrolysis by ABCE1 leads to ribosome splitting followed by dissociation of eRF1, ABCE1, and 60S from the mRNA. E, the 40S recycling factor heterodimer MCT-1/DENR binds the terminated ribosome. F, the 40S subunit and P-site tRNA are then removed from the mRNA by MCT-1/DENR, freeing all ribosome components to engage in another round of translation.
      eRF1 does not act alone at the terminating ribosome; it functions in tandem with the GTPase eRF3 (
      • Stansfield I.
      • Jones K.M.
      • Kushnirov V.V.
      • Dagkesamanskaya A.R.
      • Poznyakovski A.I.
      • Paushkin S.V.
      • et al.
      The products of the SUP45 (eRF1) and SUP35 genes interact to mediate translation termination in Saccharomyces cerevisiae.
      ). In humans, two genes, GSPT1 and GSPT2, encode translation termination GTPases eRF3a and eRF3b, respectively (
      • Jakobsen C.G.
      • Segaard T.M.
      • Jean-Jean O.
      • Frolova L.
      • Justesen J.
      Identification of a novel termination release factor eRF3b expressing the eRF3 activity in vitro and in vivo.
      ). eRF3a is ubiquitously expressed and considered the predominant source of eRF3 activity, whereas eRF3b expression in most human tissues is lower, with higher expression in certain tissues such as the brain and testis where it may have a more specialized function (
      • Chauvin C.
      • Salhi S.
      • Le Goff C.
      • Viranaicken W.
      • Diop D.
      • Jean-Jean O.
      Involvement of human release factors eRF3a and eRF3b in translation termination and regulation of the termination complex formation.
      ). We will generically refer to these two GTPases as eRF3. Binding of eRF1 to eRF3 stabilizes the GTP-bound form of eRF3 (Fig. 1) (
      • Hauryliuk V.
      • Zavialov A.
      • Kisselev L.
      • Ehrenberg M.
      Class-1 release factor eRF1 promotes GTP binding by class-2 release factor eRF3.
      ,
      • Mitkevich V.A.
      • Kononenko A.V.
      • Petrushanko I.Y.
      • Yanvarev D.V.
      • Makarov A.A.
      • Kisselev L.L.
      Termination of translation in eukaryotes is mediated by the quaternary eRF1∗eRF3∗GTP∗Mg2+ complex. The biological roles of eRF3 and prokaryotic RF3 are profoundly distinct.
      ). The eRF1–eRF3–GTP complex can either form by itself or assemble in association with the ribosome (
      • Taylor D.
      • Unbehaun A.
      • Li W.
      • Das S.
      • Lei J.
      • Liao H.Y.
      • et al.
      Cryo-EM structure of the mammalian eukaryotic release factor eRF1-eRF3-associated termination complex.
      ,
      • Des Georges A.
      • Hashem Y.
      • Unbehaun A.
      • Grassucci R.A.
      • Taylor D.
      • Hellen C.U.T.
      • et al.
      Structure of the mammalian ribosomal pre-termination complex associated with eRF1.eRF3.GDPNP.
      ,
      • Preis A.
      • Heuer A.
      • Barrio-Garcia C.
      • Hauser A.
      • Eyler D.E.
      • Berninghausen O.
      • et al.
      Cryoelectron microscopic structures of eukaryotic translation termination complexes containing eRF1-eRF3 or eRF1-ABCE1.
      ). The eRF1–eRF3–GTP complex then binds the A-site of the terminating ribosome much faster than eRF1 can alone (
      • Lawson M.R.
      • Lessen L.N.
      • Wang J.
      • Prabhakar A.
      • Corsepius N.C.
      • Green R.
      • et al.
      Mechanisms that ensure speed and fidelity in eukaryotic translation termination.
      ,
      • Kryuchkova P.
      • Grishin A.
      • Eliseev B.
      • Karyagina A.
      • Frolova L.
      • Alkalaeva E.
      Two-step model of stop codon recognition by eukaryotic release factor eRF1.
      ). Together, the ribosome, eRF1, and GTP-bound eRF3 form the termination complex (
      • Zhouravleva G.
      • Frolova L.
      • Le Goff X.
      • Le Guellec R.
      • Inge-Vechtomov S.
      • Kisselev L.
      • et al.
      Termination of translation in eukaryotes is governed by two interacting polypeptide chain release factors, eRF1 and eRF3.
      ). Stop codon recognition by eRF1 accelerates GTP hydrolysis by eRF3 (
      • Des Georges A.
      • Hashem Y.
      • Unbehaun A.
      • Grassucci R.A.
      • Taylor D.
      • Hellen C.U.T.
      • et al.
      Structure of the mammalian ribosomal pre-termination complex associated with eRF1.eRF3.GDPNP.
      ), and the GTP hydrolysis function of eRF3 accelerates its release from the complex (
      • Lawson M.R.
      • Lessen L.N.
      • Wang J.
      • Prabhakar A.
      • Corsepius N.C.
      • Green R.
      • et al.
      Mechanisms that ensure speed and fidelity in eukaryotic translation termination.
      ). Subsequently, a conformational change occurs in eRF1, whereby its catalytic GGQ motif is accommodated in the peptidyl transferase center of the ribosome. eRF1 thus positioned catalyzes the hydrolysis of the ester linkage between the tRNA and the nascent polypeptide, releasing it from the peptidyl (P) site of the terminating ribosome (Fig. 1B) (
      • Brown A.
      • Shao S.
      • Murray J.
      • Hegde R.S.
      • Ramakrishnan V.
      Structural basis for stop codon recognition in eukaryotes.
      ,
      • Shao S.
      • Murray J.
      • Brown A.
      • Taunton J.
      • Ramakrishnan V.
      • Hegde R.S.
      • et al.
      Decoding mammalian ribosome-mRNA states by translational GTPase complexes.
      ,
      • Taylor D.
      • Unbehaun A.
      • Li W.
      • Das S.
      • Lei J.
      • Liao H.Y.
      • et al.
      Cryo-EM structure of the mammalian eukaryotic release factor eRF1-eRF3-associated termination complex.
      ,
      • Des Georges A.
      • Hashem Y.
      • Unbehaun A.
      • Grassucci R.A.
      • Taylor D.
      • Hellen C.U.T.
      • et al.
      Structure of the mammalian ribosomal pre-termination complex associated with eRF1.eRF3.GDPNP.
      ). The hydrolysis reaction favors the release of eRF1 although it is energetically costly. Subsequently, eRF1 affinity to the terminated ribosome is substantially decreased (
      • Lawson M.R.
      • Lessen L.N.
      • Wang J.
      • Prabhakar A.
      • Corsepius N.C.
      • Green R.
      • et al.
      Mechanisms that ensure speed and fidelity in eukaryotic translation termination.
      ).

      Enhancers of translation termination

      Several other trans-acting factors can further promote the termination activity of eRF1 and eRF3, chief among these is the polyadenylate (poly(A))-binding protein (PABP), which binds to the poly(A) tails at the 3′ ends of mRNA. A C-terminal domain in PABP binds with the PAM2 motif in the N terminus of eRF3 (
      • Ivanov A.
      • Mikhailova T.
      • Eliseev B.
      • Yeramala L.
      • Sokolova E.
      • Susorov D.
      • et al.
      PABP enhances release factor recruitment and stop codon recognition during translation termination.
      ). It was observed using in vitro reconstitution assays that under low release factor concentrations, eRF3–PABP interaction enhances the termination reaction by stabilizing the binding of eRF3 to the ribosomal complexes, which leads to stimulation of stop codon recognition and peptidyl-tRNA hydrolysis (
      • Ivanov A.
      • Mikhailova T.
      • Eliseev B.
      • Yeramala L.
      • Sokolova E.
      • Susorov D.
      • et al.
      PABP enhances release factor recruitment and stop codon recognition during translation termination.
      ). The effect of 3′ poly(A)-tail–bound PABP on termination may also be maximal when stop codons occur close to poly(A) tails (
      • Singh G.
      • Rebbapragada I.
      • Lykke-Andersen J.
      A competition between stimulators and antagonists of Upf complex recruitment governs human nonsense-mediated mRNA decay.
      ,
      • Ivanov P.V.
      • Gehring N.H.
      • Kunz J.B.
      • Hentze M.W.
      • Kulozik A.E.
      Interactions between UPF1, eRFs, PABP and the exon junction complex suggest an integrated model for mammalian NMD pathways.
      ,
      • Silva A.L.
      • Ribeiro P.
      • Inácio A.
      • Liebhaber S.A.
      • Romão L.
      Proximity of the poly(A)-binding protein to a premature termination codon inhibits mammalian nonsense-mediated mRNA decay.
      ,
      • Amrani N.
      • Ganesan R.
      • Kervestin S.
      • Mangus D.A.
      • Ghosh S.
      • Jacobson A.
      A faux 3′-UTR promotes aberrant termination and triggers nonsense-mediated mRNA decay.
      ,
      • Wu C.
      • Roy B.
      • He F.
      • Yan K.
      • Jacobson A.
      Poly(A)-binding protein regulates the efficiency of translation termination.
      ) (also see later). Interestingly, the stimulating effect of PABP on termination in vitro is seen both with and without its binding to the poly(A) tail, which suggests that PABP may enhance termination via multiple mechanisms (
      • Ivanov A.
      • Mikhailova T.
      • Eliseev B.
      • Yeramala L.
      • Sokolova E.
      • Susorov D.
      • et al.
      PABP enhances release factor recruitment and stop codon recognition during translation termination.
      ). One such mechanism could be through interaction between PABP and translation initiation factor eukaryotic translation initiation factor 4G (eIF4G), a connection that can suppress premature termination and promote normal termination and ribosome recycling (
      • Fatscher T.
      • Boehm V.
      • Weiche B.
      • Gehring N.H.
      The interaction of cytoplasmic poly(A)-binding protein with eukaryotic initiation factor 4G suppresses nonsense-mediated mRNA decay.
      ,
      • Joncourt R.
      • Eberle A.B.
      • Rufener S.C.
      • Mühlemann O.
      Eukaryotic initiation factor 4G suppresses nonsense-mediated mRNA decay by two Genetically separable mechanisms.
      ) While PABP can enhance translation termination in vitro under certain conditions, it remains to be seen if PABP plays a similar role in translation termination in vivo.
      In addition to PABP, other factors are known to enhance the termination reaction. Elongation factor eIF5A, which binds to the ribosome such that it partially occupies the empty exit (E)-site of the elongating ribosome to stabilize peptidyl-tRNA for peptide bond formation, also promotes peptidyl-tRNA hydrolysis by eRF1 (
      • Schuller A.P.
      • Wu C.C.-C.
      • Dever T.E.
      • Buskirk A.R.
      • Green R.
      eIF5A functions globally in translation elongation and termination.
      ) and eRF1 release (
      • Lawson M.R.
      • Lessen L.N.
      • Wang J.
      • Prabhakar A.
      • Corsepius N.C.
      • Green R.
      • et al.
      Mechanisms that ensure speed and fidelity in eukaryotic translation termination.
      ), thereby enhancing the termination reaction (
      • Schuller A.P.
      • Wu C.C.-C.
      • Dever T.E.
      • Buskirk A.R.
      • Green R.
      eIF5A functions globally in translation elongation and termination.
      ,
      • Pelechano V.
      • Alepuz P.
      eIF5A facilitates translation termination globally and promotes the elongation of many non polyproline-specific tripeptide sequences.
      ). Dbp5, a DEAD-box ATPase involved in mRNA export, and its activator Gle1 and inositol hexakisphosphate (IP6), have also been implicated in translation termination in yeast (
      • Alcázar-Román A.R.
      • Bolger T.A.
      • Wente S.R.
      Control of mRNA export and translation termination by inositol hexakisphosphate requires specific interaction with Gle1.
      ,
      • Gross T.
      • Siepmann A.
      • Sturm D.
      • Windgassen M.
      • Scarcelli J.J.
      • Seedorf M.
      • et al.
      The DEAD-Box RNA helicase Dbp5 functions in translation termination.
      ,
      • Beißel C.
      • Grosse S.
      • Krebber H.
      Dbp5/DDX19 between translational readthrough and nonsense mediated decay.
      ). Dbp5 and Gle1 physically interact with termination factors, enhance eRF3 recruitment to the ribosome, and the loss of Dbp5, Gle1, or IP6 production increases stop codon readthrough (
      • Alcázar-Román A.R.
      • Bolger T.A.
      • Wente S.R.
      Control of mRNA export and translation termination by inositol hexakisphosphate requires specific interaction with Gle1.
      ,
      • Gross T.
      • Siepmann A.
      • Sturm D.
      • Windgassen M.
      • Scarcelli J.J.
      • Seedorf M.
      • et al.
      The DEAD-Box RNA helicase Dbp5 functions in translation termination.
      ). DDX19B, the Dbp5 homolog in humans, increases the speed of termination by enhancing peptidyl-tRNA hydrolysis by binding to eRF1 (
      • Mikhailova T.
      • Shuvalova E.
      • Ivanov A.
      • Susorov D.
      • Shuvalov A.
      • Kolosov P.M.
      • et al.
      RNA helicase DDX19 stabilizes ribosomal elongation and termination complexes.
      ,
      • Beißel C.
      • Neumann B.
      • Uhse S.
      • Hampe I.
      • Karki P.
      • Krebber H.
      Translation termination depends on the sequential ribosomal entry of eRF1 and eRF3.
      ). NMD-activating UPF proteins can modulate premature termination (see later), but it is not clear if they also act on normal termination.

      Ribosome recycling

      Following nascent polypeptide hydrolysis, eRF3 bound to GDP is released from the termination complex. Ribosome recycling begins as the highly conserved recycling factor ABCE1 (Rli1 in yeast), an iron–sulfur cluster–containing ATPase, binds to eRF1 at the same domain that was earlier occupied by eRF3 (Fig. 1C) (
      • Brown A.
      • Shao S.
      • Murray J.
      • Hegde R.S.
      • Ramakrishnan V.
      Structural basis for stop codon recognition in eukaryotes.
      ,
      • Preis A.
      • Heuer A.
      • Barrio-Garcia C.
      • Hauser A.
      • Eyler D.E.
      • Berninghausen O.
      • et al.
      Cryoelectron microscopic structures of eukaryotic translation termination complexes containing eRF1-eRF3 or eRF1-ABCE1.
      ). ATP hydrolysis by ABCE1 promotes the separation of the 60S ribosome subunit from the 40S subunit, which remains bound to tRNA and mRNA (Fig. 1D) (
      • Pisarev A.V.
      • Skabkin M.A.
      • Pisareva V.P.
      • Skabkina O.V.
      • Rakotondrafara A.M.
      • Hentze M.W.
      • et al.
      The role of ABCE1 in eukaryotic posttermination ribosomal recycling.
      ). This first step in ribosome recycling is also dependent on eIF3j/Hcr1, which may act by either recruiting ABCE1 to the termination complex or by stimulating its ATPase activity (
      • Beznosková P.
      • Cuchalová L.
      • Wagner S.
      • Shoemaker C.J.
      • Gunišová S.
      • von der Haar T.
      • et al.
      Translation initiation factors eIF3 and HCR1 control translation termination and stop codon read-through in yeast cells.
      ,
      • Young D.J.
      • Guydosh N.R.
      Hcr1/eIF3j is a 60S ribosomal subunit recycling accessory factor in vivo.
      ). Following the release of the 60S subunit by ABCE1/eIF3j, the remaining 40S subunit–deacylated tRNA–mRNA complex is disassembled and recycled by the MCT-1/DENR heterodimer (Tma20/Tma22 in yeast) (Fig. 1, E and F) (
      • Young D.J.
      • Meydan S.
      • Guydosh N.R.
      40S ribosome profiling reveals distinct roles for Tma20/Tma22 (MCT-1/DENR) and Tma64 (eIF2D) in 40S subunit recycling.
      ). This 40S subunit recycling activity is also shared by eIF2D (Tma64 in yeast) (
      • Young D.J.
      • Guydosh N.R.
      Hcr1/eIF3j is a 60S ribosomal subunit recycling accessory factor in vivo.
      ,
      • Young D.J.
      • Meydan S.
      • Guydosh N.R.
      40S ribosome profiling reveals distinct roles for Tma20/Tma22 (MCT-1/DENR) and Tma64 (eIF2D) in 40S subunit recycling.
      ,
      • Young D.J.
      • Makeeva D.E.
      • Zhang F.
      • Anisimova A.S.
      • Stolboushkina E.A.
      • Ghobakhlou F.
      • et al.
      Tma64/eIF2D, Tma20/MCT-1, and Tma22/DENR recycle post-termination 40S subunits in vivo.
      ), whose N and C terminus are homologous to MCT-1 and DENR, respectively (
      • Skabkin M.A.
      • Skabkina O.V.
      • Dhote V.
      • Komar A.A.
      • Hellen C.U.T.
      • Pestova T.V.
      Activities of Ligatin and MCT-1/DENR in eukaryotic translation initiation and ribosomal recycling.
      ). This process of ribosome recycling is not only critical in ensuring that the ribosome does not resume translation in the 3′ UTR but also to maintain the pool of free ribosome subunits for another round of translation.

      Abnormal translation termination triggers NMD

      When a ribosome terminates translation at a normal termination codon (NTC) at the end of a conserved ORF, it is often followed by several subsequent rounds of translation. However, if the ribosome terminates translation prematurely within an ORF, for example, because of the presence of a PTC, it results in translation repression (
      • Isken O.
      • Kim Y.K.
      • Hosoda N.
      • Mayeur G.L.
      • Hershey J.W.B.
      • Maquat L.E.
      Upf1 phosphorylation triggers translational repression during nonsense-mediated mRNA decay.
      ,
      • Zinshteyn B.
      • Sinha N.K.
      • Enam S.U.
      • Koleske B.
      • Green R.
      Translational repression of NMD targets by GIGYF2 and EIF4E2.
      ) and rapid mRNA destabilization (
      • Yi Z.
      • Sanjeev M.
      • Singh G.
      The branched nature of the nonsense-mediated mRNA decay pathway.
      ,
      • Celik A.
      • He F.
      • Jacobson A.
      NMD monitors translational fidelity 24/7.
      ,
      • Lykke-Andersen S.
      • Jensen T.H.
      Nonsense-mediated mRNA decay: an intricate machinery that shapes transcriptomes.
      ). Next, we summarize our understanding of transcript features and molecular interactions between translation termination and NMD factors that divert the mRNAs to the path of NMD.

      Transcript features that distinguish normal and NMD-inducing stop codons

      The signals that initiate NMD following translation termination at certain stop codons can be categorized into two general mechanistic themes.

      Abnormally long 3′ UTRs

      The occurrence of early stop codons (e.g., PTC, uORF stop codon) can turn the downstream coding sequence into a UTR, thereby extending the 3' UTR. According to current NMD models, when translation termination occurs in an altered 3′ UTR context, it can be recognized as abnormal because of multiple non–mutually exclusive reasons (
      • Singh G.
      • Rebbapragada I.
      • Lykke-Andersen J.
      A competition between stimulators and antagonists of Upf complex recruitment governs human nonsense-mediated mRNA decay.
      ,
      • Fritz S.E.
      • Ranganathan S.
      • Wang C.D.
      • Hogg J.R.
      The RNA-binding protein PTBP1 promotes ATPase-dependent dissociation of the RNA helicase UPF1 to protect transcripts from nonsense-mediated mRNA decay.
      ,
      • Lindeboom R.G.H.
      • Supek F.
      • Lehner B.
      The rules and impact of nonsense-mediated mRNA decay in human cancers.
      ). One reason could be the diminished stimulation of termination by PABP. When termination happens far upstream of the normal 3' UTR, the increased distance between the terminating ribosome and the poly(A) tail could reduce or compromise interaction between eRFs and PABP (
      • Ivanov A.
      • Mikhailova T.
      • Eliseev B.
      • Yeramala L.
      • Sokolova E.
      • Susorov D.
      • et al.
      PABP enhances release factor recruitment and stop codon recognition during translation termination.
      ,
      • Ivanov P.V.
      • Gehring N.H.
      • Kunz J.B.
      • Hentze M.W.
      • Kulozik A.E.
      Interactions between UPF1, eRFs, PABP and the exon junction complex suggest an integrated model for mammalian NMD pathways.
      ,
      • Silva A.L.
      • Ribeiro P.
      • Inácio A.
      • Liebhaber S.A.
      • Romão L.
      Proximity of the poly(A)-binding protein to a premature termination codon inhibits mammalian nonsense-mediated mRNA decay.
      ,
      • Amrani N.
      • Ganesan R.
      • Kervestin S.
      • Mangus D.A.
      • Ghosh S.
      • Jacobson A.
      A faux 3′-UTR promotes aberrant termination and triggers nonsense-mediated mRNA decay.
      ,
      • Fatscher T.
      • Boehm V.
      • Weiche B.
      • Gehring N.H.
      The interaction of cytoplasmic poly(A)-binding protein with eukaryotic initiation factor 4G suppresses nonsense-mediated mRNA decay.
      ). Another basis for extended 3′ UTRs to promote abnormal termination and NMD is by increasing the RNA occupancy of central NMD factor UPF1, which binds RNA before translation in a length-dependent and sequence-independent manner (
      • Hogg J.R.
      • Goff S.P.
      Upf1 senses 3’UTR length to potentiate mRNA decay.
      ,
      • Hurt J.A.
      • Robertson A.D.
      • Burge C.B.
      Global analyses of UPF1 binding and function reveal expanded scope of nonsense-mediated mRNA decay.
      ,
      • Zünd D.
      • Gruber A.R.
      • Zavolan M.
      • Mühlemann O.
      Translation-dependent displacement of UPF1 from coding sequences causes its enrichment in 3’ UTRs.
      ). The translating ribosome is thought to remove any bound UPF1 in its path, thereby reducing UPF1–mRNA binding when translation terminates in a normal context (
      • Hogg J.R.
      • Goff S.P.
      Upf1 senses 3’UTR length to potentiate mRNA decay.
      ). When translation terminates in an abnormal context, such as that of a PTC, it can lead to a higher local concentration and/or more likelihood of UPF1 binding to the extended 3' UTRs (Fig. 2A). Such transcript-bound UPF1, through its interactions with the 40S ribosome (
      • Ghosh S.
      • Ganesan R.
      • Amrani N.
      • Jacobson A.
      Translational competence of ribosomes released from a premature termination codon is modulated by NMD factors.
      ) as well as with eRF1 and eRF3 (
      • Ivanov P.V.
      • Gehring N.H.
      • Kunz J.B.
      • Hentze M.W.
      • Kulozik A.E.
      Interactions between UPF1, eRFs, PABP and the exon junction complex suggest an integrated model for mammalian NMD pathways.
      ,
      • Kashima I.
      • Yamashita A.
      • Izumi N.
      • Kataoka N.
      • Morishita R.
      • Hoshino S.
      • et al.
      Binding of a novel SMG-1-Upf1-eRF1-eRF3 complex (SURF) to the exon junction complex triggers Upf1 phosphorylation and nonsense-mediated mRNA decay.
      ,
      • Yamashita A.
      • Izumi N.
      • Kashima I.
      • Ohnishi T.
      • Saari B.
      • Katsuhata Y.
      • et al.
      SMG-8 and SMG-9, two novel subunits of the SMG-1 complex, regulate remodeling of the mRNA surveillance complex during nonsense-mediated mRNA decay.
      ), can engage with termination machinery to trigger NMD (
      • Singh G.
      • Rebbapragada I.
      • Lykke-Andersen J.
      A competition between stimulators and antagonists of Upf complex recruitment governs human nonsense-mediated mRNA decay.
      • Hogg J.R.
      • Goff S.P.
      Upf1 senses 3’UTR length to potentiate mRNA decay.
      ,
      • Gehring N.H.
      • Lamprinaki S.
      • Hentze M.W.
      • Kulozik A.E.
      The hierarchy of exon-junction complex assembly by the spliceosome explains key features of mammalian nonsense-mediated mRNA decay.
      ). The window of opportunity for UPF1 to interact with release factors and cause premature termination can further be increased because of reduced engagement between eRFs and PABP. Interestingly, normal error-free mRNAs with abnormally long 3′ UTRs (>800 nt in humans) can also be destabilized by UPF1 (
      • Hurt J.A.
      • Robertson A.D.
      • Burge C.B.
      Global analyses of UPF1 binding and function reveal expanded scope of nonsense-mediated mRNA decay.
      ), possibly because of termination on these mRNAs deemed to be abnormal as a result of higher UPF1 occupancy in 3′ UTRs.
      Figure thumbnail gr2
      Figure 2NMD can occur in an EJC-dependent or EJC-independent manner. mRNA is shown in gray. The introduction of premature termination codon in coding sequence (thicker line) results in an “extended 3' UTR” that also includes the normal 3′ UTR (thinner line). The ribosome is shown in blue, and other important proteins are labeled. Recognition of NMD-inducing translation termination codon occurs through one of two general pathways: the EJC-independent (A) or EJC-dependent (B) pathway. A, the EJC-independent pathway begins when UPF1, the core NMD factor, is bound downstream of the terminating ribosome at abnormally high levels. After interaction between UPF1 and the termination complex, other NMD factors are recruited to assemble the “SURF” complex, which includes UPF1 phosphorylation factors SMG1 (the kinase) and its regulatory SMG8/9 heterodimer. This complex may include or subsequently recruit DHX34, an RNA helicase that may aid in interaction of UPF1 with its activators UPF2 and UPF3B (or its paralog UPF3A). Formation of the SURF complex leads to UPF1 phosphorylation (represented by yellow triangles labeled “P”) by SMG1. B, the EJC-dependent NMD requires the presence of an EJC downstream of the terminating ribosome that enhances NMD by recruiting UPF2 and UPF3B (or UPF3A) to mRNA. Following assembly of the SURF complex at the terminating ribosome and UPF1 interaction with its activators UPF2 and UPF3B (or UPF3A), which can be aided by DHX34, UPF1 gets phosphorylated by SMG1. C, after initial phosphorylation, UPF1 is phosphorylated more extensively. It is unknown how long the ribosome and release factors remain associated with the phosphorylated UPF1 (indicated by faded shapes). The extensive phosphorylation of UPF1 leads to the recruitment of SMG6 and SMG5/7 heterodimer, which initiate mRNA degradation via multiple mechanisms including SMG6-catalyzed mRNA endonucleolytic cleavage close to termination codons. EJC, exon junction complex; NMD, nonsense-mediated mRNA decay.

      3′ UTR-bound proteins

      Another major reason for a translation termination event to be identified as abnormal is the presence of UPF1 activating factors in 3' UTRs. The most prominent example is the exon junction complex (EJC), which is deposited upstream of mRNA exon–exon junctions during mRNA splicing (
      • Gehring N.H.
      • Lamprinaki S.
      • Hentze M.W.
      • Kulozik A.E.
      The hierarchy of exon-junction complex assembly by the spliceosome explains key features of mammalian nonsense-mediated mRNA decay.
      ,
      • Reichert V.L.
      • Le Hir H.
      • Jurica M.S.
      • Moore M.J.
      5’ exon interactions within the human spliceosome establish a framework for exon junction complex structure and assembly.
      ,
      • Woodward L.A.
      • Mabin J.W.
      • Gangras P.
      • Singh G.
      The exon junction complex: A lifelong guardian of mRNA fate.
      ) and is disassembled by the ribosome during translation (
      • Woodward L.A.
      • Mabin J.W.
      • Gangras P.
      • Singh G.
      The exon junction complex: A lifelong guardian of mRNA fate.
      • Dostie J.
      • Dreyfuss G.
      Translation is required to remove Y14 from mRNAs in the cytoplasm.
      ). Generally, NTCs occur in the last exon, so all EJCs are removed before translation terminates. However, if a stop codon occurs sufficiently upstream of the last exon–exon junction such that at least one EJC remains bound in the 3′ UTR, abnormal termination and NMD ensues (Fig. 2B) (
      • Yi Z.
      • Sanjeev M.
      • Singh G.
      The branched nature of the nonsense-mediated mRNA decay pathway.
      ). The presence of at least one exon–exon junction, and hence the EJC, downstream of a stop codon is one of the strongest predictors of NMD susceptibility for human mRNAs (
      • Lindeboom R.G.H.
      • Supek F.
      • Lehner B.
      The rules and impact of nonsense-mediated mRNA decay in human cancers.
      ). Conversely, if translation terminates either in the last exon or close to the last exon–exon junction in the penultimate exon, the mRNA evades NMD possibly because of removal of the EJC from the last exon junction (
      • Lindeboom R.G.H.
      • Supek F.
      • Lehner B.
      The rules and impact of nonsense-mediated mRNA decay in human cancers.
      ). Current NMD models suggest that terminating ribosome-bound UPF1 senses downstream EJCs through bridging interactions mediated by UPF2 and UPF3 proteins (
      • Ivanov P.V.
      • Gehring N.H.
      • Kunz J.B.
      • Hentze M.W.
      • Kulozik A.E.
      Interactions between UPF1, eRFs, PABP and the exon junction complex suggest an integrated model for mammalian NMD pathways.
      ,
      • López-Perrote A.
      • Castaño R.
      • Melero R.
      • Zamarro T.
      • Kurosawa H.
      • Ohnishi T.
      • et al.
      Human nonsense-mediated mRNA decay factor UPF2 interacts directly with eRF3 and the SURF complex.
      ,
      • Gehring N.H.
      • Kunz J.B.
      • Neu-Yilik G.
      • Breit S.
      • Viegas M.H.
      • Hentze M.W.
      • et al.
      Exon-junction complex components specify distinct routes of nonsense-mediated mRNA decay with differential cofactor requirements.
      ,
      • Gehring N.H.
      • Neu-Yilik G.
      • Schell T.
      • Hentze M.W.
      • Kulozik A.E.
      Y14 and hUpf3b form an NMD-activating complex.
      ). However, recent evidence that human UPF3 paralogs UPF3A and UPF3B can activate NMD even when they lack EJC binding ability suggests that interaction between EJC and UPF3 proteins is not necessary for premature termination (
      • Yi Z.
      • Arvola R.M.
      • Myers S.
      • Dilsavor C.N.
      • Alhasan R.A.
      • Carter B.N.
      • et al.
      Mammalian UPF3A and UPF3B can activate nonsense-mediated mRNA decay independently of their exon junction complex binding.
      ,
      • Wallmeroth D.
      • Lackmann J.-W.
      • Kueckelmann S.
      • Altm€ Uller J.
      • Dieterich C.
      • Boehm V.
      • et al.
      Human UPF3A and UPF3B enable fault-tolerant activation of nonsense-mediated mRNA decay.
      ). It is possible that another mode of communication exists between the terminating ribosome and the downstream EJC (
      • Yi Z.
      • Arvola R.M.
      • Myers S.
      • Dilsavor C.N.
      • Alhasan R.A.
      • Carter B.N.
      • et al.
      Mammalian UPF3A and UPF3B can activate nonsense-mediated mRNA decay independently of their exon junction complex binding.
      ,
      • Wallmeroth D.
      • Lackmann J.-W.
      • Kueckelmann S.
      • Altm€ Uller J.
      • Dieterich C.
      • Boehm V.
      • et al.
      Human UPF3A and UPF3B enable fault-tolerant activation of nonsense-mediated mRNA decay.
      ). One possible function of 3' UTR-bound EJCs could be to recruit UPF2 and UPF3 proteins to 3′ UTRs to increase their likelihood of UPF1 activation (Fig. 2B) (
      • Gehring N.H.
      • Lamprinaki S.
      • Hentze M.W.
      • Kulozik A.E.
      The hierarchy of exon-junction complex assembly by the spliceosome explains key features of mammalian nonsense-mediated mRNA decay.
      ,
      • Gehring N.H.
      • Kunz J.B.
      • Neu-Yilik G.
      • Breit S.
      • Viegas M.H.
      • Hentze M.W.
      • et al.
      Exon-junction complex components specify distinct routes of nonsense-mediated mRNA decay with differential cofactor requirements.
      ,
      • Gehring N.H.
      • Neu-Yilik G.
      • Schell T.
      • Hentze M.W.
      • Kulozik A.E.
      Y14 and hUpf3b form an NMD-activating complex.
      ,
      • Gangras P.
      • Gallagher T.L.
      • Patton R.D.
      • Yi Z.
      • Parthun M.A.
      • Tietz K.T.
      • et al.
      Zebrafish rbm8a and magoh mutants reveal EJC developmental functions and new 3′UTR intron-containing NMD targets.
      ,
      • Mabin J.W.
      • Woodward L.A.
      • Patton R.D.
      • Yi Z.
      • Jia M.
      • Wysocki V.H.
      • et al.
      The exon junction complex undergoes a compositional switch that alters mRNP structure and nonsense-mediated mRNA decay activity.
      ). Importantly, some genes naturally contain introns in their 3' UTRs, which causes normal termination events on their mRNAs to resemble abnormal termination because of the presence of a downstream EJC (
      • Giorgi C.
      • Yeo G.W.
      • Stone M.E.
      • Katz D.B.
      • Burge C.
      • Turrigiano G.
      • et al.
      The EJC factor eIF4AIII modulates synaptic strength and neuronal protein expression.
      ,
      • Bicknell A.A.
      • Cenik C.
      • Chua H.N.
      • Roth F.P.
      • Moore M.J.
      Introns in UTRs: why we should stop ignoring them.
      ). Thus, the presence of an EJC in the 3′ UTR is a potent trigger for abnormal termination and NMD, but how EJC enhances these processes remains far from clearly understood.
      In addition to the EJC, the serine–arginine (SR) proteins, which are also recruited to RNA during its biogenesis in the nucleus (
      • Slišković I.
      • Slišković S.
      • Eich H.
      • Müller-Mcnicoll M.
      Exploring the multifunctionality of SR proteins.
      ), can also enhance NMD (
      • Zhang Z.
      • Krainer A.R.
      Involvement of SR proteins in mRNA surveillance.
      ,
      • Sato H.
      • Hosoda N.
      • Maquat L.E.
      Efficiency of the pioneer round of translation affects the cellular site of nonsense-mediated mRNA decay.
      ) possibly via their binding to 3' UTRs. SRSF1, a prototypical SR protein, has been suggested to interact directly with UPF1 and activate NMD when present in 3′ UTRs (
      • Aznarez I.
      • Nomakuchi T.T.
      • Tetenbaum-Novatt J.
      • Rahman M.A.
      • Fregoso O.
      • Rees H.
      • et al.
      Mechanism of nonsense-mediated mRNA decay stimulation by splicing factor SRSF1.
      ). Another SR protein, SRSF2, can promote NMD by enhancing EJC deposition/binding on RNA (
      • Rahman M.A.
      • Lin K.-T.T.
      • Bradley R.K.
      • Abdel-Wahab O.
      • Krainer A.R.
      Recurrent SRSF2 mutations in MDS affect both splicing and NMD.
      ). It is possible that other SR proteins including SRSF1 can also act in NMD via the EJC. All 12 canonical SR proteins and many SR-like proteins are physically associated with the EJCs purified from human cells with SR protein binding sites enriched in the vicinity of the EJC (
      • Singh G.
      • Kucukural A.
      • Cenik C.
      • Leszyk J.D.
      • Shaffer S.A.
      • Weng Z.
      • et al.
      The cellular EJC interactome reveals higher-order mRNP structure and an EJC-SR protein Nexus.
      ,
      • Saulière J.
      • Murigneux V.
      • Wang Z.
      • Marquenet E.
      • Barbosa I.
      • Le Tonquèze O.
      • et al.
      CLIP-seq of eIF4AIII reveals transcriptome-wide mapping of the human exon junction complex.
      ).
      Regardless of the signal that induces NMD, once termination is determined to be abnormal, subsequent steps in the NMD pathway are set in motion that are summarized in Figure 2. For a more detailed discussion of the modes of NMD initiation as well as the resulting mRNA degradation mechanisms, we refer the readers to excellent recent reviews (
      • Yi Z.
      • Sanjeev M.
      • Singh G.
      The branched nature of the nonsense-mediated mRNA decay pathway.
      ,
      • Celik A.
      • Kervestin S.
      • Jacobson A.
      Nmd: at the crossroads between translation termination and ribosome recycling.
      ,
      • Raimondeau E.
      • Bufton J.C.
      • Schaffitzel C.
      New insights into the interplay between the translation machinery and nonsense-mediated mRNA decay factors.
      ).

      Interactions between UPF proteins and termination machinery

      Following the recruitment of NMD factors by one of the NMD-inducing signals, the core NMD factors engage with the terminating ribosome and associated release factors, and possibly even with ribosome recycling machinery. The precise mechanistic details of how NMD factors influence and alter events during termination to initiate NMD remain obscure. However, some details of the interactions between NMD and termination/recycling machinery have emerged, mainly from work in yeast and mammalian cell culture models. Notably, there are several discrepancies in the observed interactions and roles of the UPF factors during translation termination. Such differences could stem from different model systems and experimental approaches employed or from our incomplete understanding of the players and interactions between NMD and translation termination factors.

      Evidence from yeast

      In Saccharomyces cerevisiae, all three UPF proteins have been shown to interact with ribosomes (
      • Atkin A.L.
      • Schenkman L.R.
      • Eastham M.
      • Dahlseid J.N.
      • Lelivelt M.J.
      • Culbertson M.R.
      Relationship between yeast polyribosomes and Upf proteins required for nonsense mRNA decay.
      ,
      • Ganesan R.
      • Leszyk J.
      • Jacobson A.
      Selective profiling of ribosomes associated with yeast Upf proteins.
      ,
      • Atkin A.L.
      • Altamura N.
      • Leeds P.
      • Culbertson M.R.
      The majority of yeast UPF1 co-localizes with polyribosomes in the cytoplasm.
      ,
      • Ganesan R.
      • Mangkalaphiban K.
      • Baker R.E.
      • He F.
      • Jacobson A.
      Progressive polysomal association of Upf1, Upf2, and Upf3 as ribosomes traverse mRNA coding regions.
      ). This interaction happens on polysomes (
      • Atkin A.L.
      • Schenkman L.R.
      • Eastham M.
      • Dahlseid J.N.
      • Lelivelt M.J.
      • Culbertson M.R.
      Relationship between yeast polyribosomes and Upf proteins required for nonsense mRNA decay.
      ,
      • Atkin A.L.
      • Altamura N.
      • Leeds P.
      • Culbertson M.R.
      The majority of yeast UPF1 co-localizes with polyribosomes in the cytoplasm.
      ,
      • Ganesan R.
      • Mangkalaphiban K.
      • Baker R.E.
      • He F.
      • Jacobson A.
      Progressive polysomal association of Upf1, Upf2, and Upf3 as ribosomes traverse mRNA coding regions.
      ), where the mRNA degradation phase of NMD is also reported to occur (
      • Hu W.
      • Petzold C.
      • Coller J.
      • Baker K.E.
      Nonsense-mediated mRNA decapping occurs on polyribosomes in Saccharomyces cerevisiae.
      ,
      • Maderazo A.B.
      • Belk J.P.
      • He F.
      • Jacobson A.
      Nonsense-containing mRNAs that accumulate in the absence of a functional nonsense-mediated mRNA decay pathway are destabilized rapidly upon its restitution.
      ). Such interactions can occur because of the direct binding of UPF factors to ribosomes or because of UPF interactions with release/recycling factors. Selective profiling of footprints of ribosomes bound to each of the three yeast UPF proteins shows that detection of UPF–ribosome complexes progressively increases toward the 3′ end of coding sequences (
      • Ganesan R.
      • Mangkalaphiban K.
      • Baker R.E.
      • He F.
      • Jacobson A.
      Progressive polysomal association of Upf1, Upf2, and Upf3 as ribosomes traverse mRNA coding regions.
      ). These UPF factor–ribosome complexes are slightly more enriched on NMD-targeted transcripts as compared with nontargets. This finding suggests that unlike in human cells where translocating ribosomes strip UPF1 from RNA (
      • Hogg J.R.
      • Goff S.P.
      Upf1 senses 3’UTR length to potentiate mRNA decay.
      ), translating ribosomes in yeast cells do not remove UPF1 from mRNAs. Alternatively, it is possible that the UPF1–ribosome complex formation is unaffected by the dissociation of mRNA-bound UPF1 by translocating ribosomes. Nonetheless, this new study suggests that UPF proteins can engage with ribosomes before they reach annotated stop codons (
      • Ganesan R.
      • Mangkalaphiban K.
      • Baker R.E.
      • He F.
      • Jacobson A.
      Progressive polysomal association of Upf1, Upf2, and Upf3 as ribosomes traverse mRNA coding regions.
      ). While it is not known if these are elongating or terminating ribosomes, these observations suggest that the NMD machinery may continuously monitor the termination status of ribosomes as they progress along a coding sequence.
      How UPF proteins interact with ribosomes is best understood in the case of UPF1. In an in vitro system, recombinant UPF1 can directly bind to 80S ribosomes by binding to the L1 stalk of the 25S rRNA (Fig. 3) (
      • Schuller A.P.
      • Zinshteyn B.
      • Enam S.U.
      • Green R.
      Directed hydroxyl radical probing reveals Upf1 binding to the 80S ribosomal E site rRNA at the L1 stalk.
      ). This binding positions UPF1 at the ribosomal E site of tRNA-bound ribosomes. Interestingly, UPF1 has also been shown to interact with the 40S ribosomal subunit via a direct interaction with Rps26 (Fig. 3) (
      • Min E.E.
      • Roy B.
      • Amrani N.
      • He F.
      • Jacobson A.
      Yeast Upf1 CH domain interacts with Rps26 of the 40S ribosomal subunit.
      ). While the contribution of UPF1–ribosome interactions to its function in NMD is yet to be completely understood, the UPF1–40S ribosome interaction may aid in ribosome recycling following termination at PTCs, as loss of UPF1 impairs this process (
      • Ghosh S.
      • Ganesan R.
      • Amrani N.
      • Jacobson A.
      Translational competence of ribosomes released from a premature termination codon is modulated by NMD factors.
      ). Notably, the role of UPF1 in ribosome recycling at PTCs is also dependent on its ATP hydrolysis activity and UPF2 binding ability (
      • Serdar L.D.
      • Whiteside D.J.L.
      • Baker K.E.
      ATP hydrolysis by UPF1 is required for efficient translation termination at premature stop codons.
      ).
      Figure thumbnail gr3
      Figure 3Interactions between UPF proteins and termination machinery. Left, a terminating ribosome. Right, a recycling ribosome. The UPF proteins interact with the ribosome and termination factors in both a direct (black lines) and indirect (purple lines) fashion. All three UPF proteins can interact with the 80S ribosome and the 40S subunit although the specific contacts within these assemblies are known only for UPF1. All three UPF proteins are also reported to interact with termination factors. The numbers on each arrow represent possible functions of each interaction listed in the legend below. References that provide evidence for the interactions between UPF proteins and termination machinery are included in parentheticals.
      Early studies in yeast revealed a genetic interaction between UPF factors and translation termination. In yeast cells lacking UPF proteins, termination at PTCs was inefficient such that readthrough of translation was observed at a much higher rate (
      • Maderazo A.B.
      • He F.
      • Mangus D.A.
      • Jacobson A.
      Upf1p control of nonsense mRNA translation is regulated by Nmd2p and Upf3p.
      ,
      • Wang W.
      • Czaplinski K.
      • Rao Y.
      • Peltz S.W.
      The role of Upf proteins in modulating the translation read-through of nonsense-containing transcripts.
      ). Such an effect was most pronounced for UPF1 as compared with UPF2 and UPF3, indicating a more direct role of UPF1 in termination. The role of UPF2 and UPF3 in nonsense suppression was suggested to be more of regulatory nature and independent of their function in inducing rapid decay of PTC-containing mRNAs. Importantly, inhibition of mRNA decay by compromising mRNA decapping or 5′-3′ exonucleolytic decay did not result in increased nonsense suppression (
      • Celik A.
      • Kervestin S.
      • Jacobson A.
      Nmd: at the crossroads between translation termination and ribosome recycling.
      ). Another study using bicistronic reporters for monitoring readthrough at nonsense codons did not find a role for UPF proteins in this process, suggesting that UPF function in nonsense suppression could be context dependent (
      • Harger J.W.
      • Dinman J.D.
      Evidence against a direct role for the Upf proteins in frameshifting or nonsense codon readthrough.
      ).
      The role for UPF proteins in termination is further supported by physical interactions between all three yeast UPF proteins and eRF1 and eRF3 in cell extracts as well as when produced recombinantly in vitro (Fig. 3) (
      • Wang W.
      • Czaplinski K.
      • Rao Y.
      • Peltz S.W.
      The role of Upf proteins in modulating the translation read-through of nonsense-containing transcripts.
      ,
      • Czaplinski K.
      • Ruiz-Echevarria M.J.
      • Paushkin S.V.
      • Han X.
      • Weng Y.
      • Perlick H.A.
      • et al.
      The surveillance complex interacts with the translation release factors to enhance termination and degrade aberrant mRNAs.
      ). UPF1 can bind simultaneously to eRF3, UPF2, and UPF3, whereas UPF2 and UPF3 compete with eRF1 to bind to eRF3 (
      • Wang W.
      • Czaplinski K.
      • Rao Y.
      • Peltz S.W.
      The role of Upf proteins in modulating the translation read-through of nonsense-containing transcripts.
      ). Furthermore, UPF1 RNA-binding activity antagonizes its interaction with eRF3 (
      • Czaplinski K.
      • Ruiz-Echevarria M.J.
      • Paushkin S.V.
      • Han X.
      • Weng Y.
      • Perlick H.A.
      • et al.
      The surveillance complex interacts with the translation release factors to enhance termination and degrade aberrant mRNAs.
      ). Interestingly, UPF1 function in nonsense suppression and mRNA decay could be separated by mutations impairing its various biochemical activities. A UPF1 mutant that could bind ATP but not hydrolyze it was still able to function in nonsense suppression but was not able to induce rapid mRNA decay (
      • Weng Y.
      • Czaplinski K.
      • Peltz S.W.
      Genetic and biochemical characterization of mutations in the ATPase and helicase regions of the Upf1 protein.
      ). On the other hand, a UPF1 mutant with reduced UPF2 binding could not support nonsense suppression even though it was able to destabilize PTC-containing mRNAs (
      • Weng Y.
      • Czaplinski K.
      • Peltz S.W.
      Identification and characterization of mutations in the UPF1 gene that affect nonsense suppression and the formation of the Upf protein complex but not mRNA turnover.
      ). Taken together, these data from early work in yeast hinted that NMD-inducing termination is comprised of a series of steps that involve interplay between UPF factors and the terminating ribosome, but molecular details of this process remain largely a mystery.

      Evidence from mammalian cells

      In mammalian cells, like in yeast, UPF proteins are also detected on polysomes (
      • Sinha N.K.
      • Ordureau A.
      • Best K.M.
      • Saba J.A.
      • Zinshteyn B.
      • Sundaramoorthy E.
      • et al.
      EDF1 coordinates cellular responses to ribosome collisions.
      ,
      • Juszkiewicz S.
      • Slodkowicz G.
      • Lin Z.
      • Freire-Pritchett P.
      • Peak-Chew S.Y.
      • Hegde R.S.
      Ribosome collisions trigger cis-acting feedback inhibition of translation initiation.
      ). Furthermore, human UPF1, UPF2, and UPF3B can bind to the 80S ribosome, individually and as a complex (Fig. 3) (
      • Neu-Yilik G.
      • Raimondeau E.
      • Eliseev B.
      • Yeramala L.
      • Amthor B.
      • Deniaud A.
      • et al.
      Dual function of UPF3B in early and late translation termination.
      ). However, it remains to be seen if the UPF1 interaction with the ribosome is direct, through 25S rRNA or Rps26 as reported in yeast (
      • Min E.E.
      • Roy B.
      • Amrani N.
      • He F.
      • Jacobson A.
      Yeast Upf1 CH domain interacts with Rps26 of the 40S ribosomal subunit.
      ). Regarding UPF interactions with release factors in mammalian/human cells, there are conflicting data in the literature. Early studies that focused only on UPF1 showed that it can interact with both eRF1 and eRF3, binding to the GTPase domain of eRF3 (
      • Ivanov P.V.
      • Gehring N.H.
      • Kunz J.B.
      • Hentze M.W.
      • Kulozik A.E.
      Interactions between UPF1, eRFs, PABP and the exon junction complex suggest an integrated model for mammalian NMD pathways.
      ,
      • Kashima I.
      • Yamashita A.
      • Izumi N.
      • Kataoka N.
      • Morishita R.
      • Hoshino S.
      • et al.
      Binding of a novel SMG-1-Upf1-eRF1-eRF3 complex (SURF) to the exon junction complex triggers Upf1 phosphorylation and nonsense-mediated mRNA decay.
      ,
      • Czaplinski K.
      • Ruiz-Echevarria M.J.
      • Paushkin S.V.
      • Han X.
      • Weng Y.
      • Perlick H.A.
      • et al.
      The surveillance complex interacts with the translation release factors to enhance termination and degrade aberrant mRNAs.
      ). Similarly, it was found that UPF2 can also interact with eRF3 to further activate UPF1 within the SURF complex (
      • López-Perrote A.
      • Castaño R.
      • Melero R.
      • Zamarro T.
      • Kurosawa H.
      • Ohnishi T.
      • et al.
      Human nonsense-mediated mRNA decay factor UPF2 interacts directly with eRF3 and the SURF complex.
      ). The UPF2–eRF3 interaction occurs in opposition to the UPF2 interaction with UPF3B, though the UPF2–UPF3B interaction is stronger (
      • López-Perrote A.
      • Castaño R.
      • Melero R.
      • Zamarro T.
      • Kurosawa H.
      • Ohnishi T.
      • et al.
      Human nonsense-mediated mRNA decay factor UPF2 interacts directly with eRF3 and the SURF complex.
      ).
      A more recent investigation compared interactions of the three human UPF proteins with termination factors. Both UPF1 and UPF3B were found to interact strongly with release factors in human cell extracts, whereas UPF2 binding to release factors was weaker (Fig. 3) (
      • Neu-Yilik G.
      • Raimondeau E.
      • Eliseev B.
      • Yeramala L.
      • Amthor B.
      • Deniaud A.
      • et al.
      Dual function of UPF3B in early and late translation termination.
      ). However, only UPF3B was found to influence translation termination in a reconstituted system in vitro. Depending on the stage of translation termination, UPF3B was suggested to have two opposing functions (
      • Neu-Yilik G.
      • Raimondeau E.
      • Eliseev B.
      • Yeramala L.
      • Amthor B.
      • Deniaud A.
      • et al.
      Dual function of UPF3B in early and late translation termination.
      ). Before stop codon recognition, particularly if release factors are limiting, UPF3B can prevent translation termination (
      • Neu-Yilik G.
      • Raimondeau E.
      • Eliseev B.
      • Yeramala L.
      • Amthor B.
      • Deniaud A.
      • et al.
      Dual function of UPF3B in early and late translation termination.
      ). After termination, UPF3B can promote ribosome recycling (
      • Neu-Yilik G.
      • Raimondeau E.
      • Eliseev B.
      • Yeramala L.
      • Amthor B.
      • Deniaud A.
      • et al.
      Dual function of UPF3B in early and late translation termination.
      ). Both of these UPF3B functions are antagonized by UPF2 (
      • Neu-Yilik G.
      • Raimondeau E.
      • Eliseev B.
      • Yeramala L.
      • Amthor B.
      • Deniaud A.
      • et al.
      Dual function of UPF3B in early and late translation termination.
      ), suggesting that function of UPF3B during termination is distinct from its role in NMD activation together with UPF2. The dual roles of UPF3B in translation termination were attributed to the direct interactions between UPF3B and release factors, although the stronger eRF3–UPF3B interaction may stabilize the interaction between UPF3B and eRF1, allowing the three proteins to form a complex (
      • Neu-Yilik G.
      • Raimondeau E.
      • Eliseev B.
      • Yeramala L.
      • Amthor B.
      • Deniaud A.
      • et al.
      Dual function of UPF3B in early and late translation termination.
      ). It was suggested that before stop codon recognition, the interaction between UPF3B and eRF3 leads to sequestration of the release factors, delaying translation termination (
      • Neu-Yilik G.
      • Raimondeau E.
      • Eliseev B.
      • Yeramala L.
      • Amthor B.
      • Deniaud A.
      • et al.
      Dual function of UPF3B in early and late translation termination.
      ). Subsequently, after the first step of translation termination occurs, the interaction between UPF3B and release factors is proposed to promote ribosome recycling (
      • Neu-Yilik G.
      • Raimondeau E.
      • Eliseev B.
      • Yeramala L.
      • Amthor B.
      • Deniaud A.
      • et al.
      Dual function of UPF3B in early and late translation termination.
      ). Whether these UPF3B functions in translation termination also hold true in living cells remains to be seen.
      Notably, vertebrate genomes also encode a UPF3B paralog, UPF3A, which is ubiquitously expressed but enriched in select tissues such as the testis and brain (
      • Serin G.
      • Gersappe A.
      • Black J.D.
      • Aronoff R.
      • Maquat L.E.
      Identification and characterization of human orthologues to Saccharomyces cerevisiae Upf2 protein and Upf3 protein (Caenorhabditis elegans SMG-4).
      ). The available evidence suggests that UPF3A has weak NMD activation capability but can also antagonize the ability of UPF3B to activate NMD in certain contexts (
      • Yi Z.
      • Arvola R.M.
      • Myers S.
      • Dilsavor C.N.
      • Alhasan R.A.
      • Carter B.N.
      • et al.
      Mammalian UPF3A and UPF3B can activate nonsense-mediated mRNA decay independently of their exon junction complex binding.
      ,
      • Wallmeroth D.
      • Lackmann J.-W.
      • Kueckelmann S.
      • Altm€ Uller J.
      • Dieterich C.
      • Boehm V.
      • et al.
      Human UPF3A and UPF3B enable fault-tolerant activation of nonsense-mediated mRNA decay.
      ,
      • Kunz J.B.
      • Neu-Yilik G.
      • Hentze M.W.
      • Kulozik A.E.
      • Gehring N.H.
      Functions of hUpf3a and hUpf3b in nonsense-mediated mRNA decay and translation.
      ,
      • Chan W.K.
      • Bhalla A.D.
      • Le Hir H.
      • Nguyen L.S.
      • Huang L.
      • Gécz J.
      • et al.
      A UPF3-mediated regulatory switch that maintains RNA surveillance.
      ,
      • Shum E.Y.
      • Jones S.H.
      • Shao A.
      • Dumdie J.
      • Krause M.D.
      • Chan W.K.
      • et al.
      The antagonistic gene paralogs Upf3a and Upf3b govern nonsense-mediated RNA decay.
      ). Interestingly, the NMD activation potential of UPF3A becomes comparable to that of UPF3B when a poorly characterized region of the protein that lies within the UPF2- and EJC-binding domains is replaced with an equivalent region of UPF3B (
      • Yi Z.
      • Arvola R.M.
      • Myers S.
      • Dilsavor C.N.
      • Alhasan R.A.
      • Carter B.N.
      • et al.
      Mammalian UPF3A and UPF3B can activate nonsense-mediated mRNA decay independently of their exon junction complex binding.
      ). It is noteworthy that this “mid-domain” of UPF3B was shown to directly interact with eRF3 (
      • Neu-Yilik G.
      • Raimondeau E.
      • Eliseev B.
      • Yeramala L.
      • Amthor B.
      • Deniaud A.
      • et al.
      Dual function of UPF3B in early and late translation termination.
      ) and bears point mutations linked to human neurodevelopmental disorders (
      • Alrahbeni T.
      • Sartor F.
      • Anderson J.
      • Miedzybrodzka Z.
      • McCaig C.
      • Müller B.
      Full UPF3B function is critical for neuronal differentiation of neural stem cells.
      ). How UPF3A contributes toward overall “UPF3” function during termination remains an important question.
      Most of the research into human NMD has focused on one specific isoform of UPF1. Recently, another UPF1 isoform possessing a longer regulatory loop and greater affinity for RNA was described (
      • Fritz S.E.
      • Ranganathan S.
      • Wang C.D.
      • Hogg J.R.
      An alternative UPF1 isoform drives conditional remodeling of nonsense-mediated mRNA decay.
      ). This isoform regulates a different subset of NMD targets than the more widely studied UPF1 isoform (
      • Fritz S.E.
      • Ranganathan S.
      • Wang C.D.
      • Hogg J.R.
      An alternative UPF1 isoform drives conditional remodeling of nonsense-mediated mRNA decay.
      ). It is unknown if this long loop UPF1 isoform interacts with the termination machinery in the same manner as the short loop isoform. It is also possible that the increased RNA affinity of the UPF1 long loop isoform may influence some of the events that inhibit traditional NMD, as discussed later in this review.

      Is NMD-inducing termination slower than normal termination?

      An important question in the NMD field is if the translation termination reaction itself is different on NMD-inducing stops as compared with the normal stops. Studies that have compared termination at these two types of stop codons have arrived at conflicting conclusions. Estimation of ribosome occupancy/dwell time in yeast extracts showed that, as compared with normal stops, translation termination on NMD-inducing stop codons yields distinct toeprints suggesting that ribosomes persist longer on NMD stops, perhaps because of inefficient/slow termination (
      • Amrani N.
      • Ganesan R.
      • Kervestin S.
      • Mangus D.A.
      • Ghosh S.
      • Jacobson A.
      A faux 3′-UTR promotes aberrant termination and triggers nonsense-mediated mRNA decay.
      ). Such ribosomes were also capable of backtracking and reinitiating translation at nearby upstream start codons. This “aberrant” behavior of terminating ribosomes was abolished in extracts lacking UPF1 or when either normal 3′ UTR or PABP was positioned in the immediate vicinity of the stop codons (
      • Amrani N.
      • Ganesan R.
      • Kervestin S.
      • Mangus D.A.
      • Ghosh S.
      • Jacobson A.
      A faux 3′-UTR promotes aberrant termination and triggers nonsense-mediated mRNA decay.
      ). Similar ribosome toeprints, diagnostic of slow ribosome release from NMD-inducing stop codons, were also observed on human β-globin mRNA in rabbit reticulocyte lysates (
      • Peixeiro I.
      • Inácio Â.
      • Barbosa C.
      • Silva A.L.
      • Liebhaber S.A.
      • Romão L.
      Interaction of PABPC1 with the translation initiation complex is critical to the NMD resistance of AUG-proximal nonsense mutations.
      ). These earlier observations argued that NMD-inducing termination reactions are slow and aberrant and hence different from normal termination. However, a more recent study in human cells did not detect differences in ribosome occupancy at NMD-inducing and normal stop codons either in in vitro translation extracts or using ribosome profiling to probe ribosome occupancy in vivo (
      • Karousis E.D.
      • Gurzeler L.A.
      • Annibaldis G.
      • Dreos R.
      • Mühlemann O.
      Human NMD ensues independently of stable ribosome stalling.
      ). While the differences in ribosome occupancy at NMD-inducing stops in human extracts and rabbit reticulocyte lysates were attributed to an mRNA cleavage activity giving rise to ribosome toeprints in reticulocyte lysates (
      • Karousis E.D.
      • Gurzeler L.A.
      • Annibaldis G.
      • Dreos R.
      • Mühlemann O.
      Human NMD ensues independently of stable ribosome stalling.
      ), discrepancies between the yeast and human studies remain unexplained. Thus, similarities or differences between the termination reaction at normal versus NMD-inducing stop codons remain unresolved.
      One consequence of prolonged ribosome residence on stop codons because of inefficient or slow termination (or ribosome recycling) will be collisions between the terminating (or recycling ribosome) and the trailing ribosome. Ribosome collisions have been well documented to occur when an elongating ribosome runs into another stalled elongating ribosome (
      • Doma M.K.
      • Parker R.
      Endonucleolytic cleavage of eukaryotic mRNAs with stalls in translation elongation.
      ,
      • Juszkiewicz S.
      • Chandrasekaran V.
      • Lin Z.
      • Kraatz S.
      • Ramakrishnan V.
      • Hegde R.S.
      ZNF598 is a quality control sensor of collided ribosomes.
      ,
      • Letzring D.P.
      • Dean K.M.
      • Grayhack E.J.
      Control of translation efficiency in yeast by codon-anticodon interactions.
      ). There is some evidence suggesting that an elongating ribosome can also collide with a terminating ribosome, most frequently if termination is prolonged or inefficient. In mammalian cells expressing the dominant negative form of the release factor eRF1AAQ, which arrests translation termination, the ribosome collision sensor ZNF598 is found to associate with the collided ribosomes at stop codons (
      • Juszkiewicz S.
      • Chandrasekaran V.
      • Lin Z.
      • Kraatz S.
      • Ramakrishnan V.
      • Hegde R.S.
      ZNF598 is a quality control sensor of collided ribosomes.
      ,
      • Garzia A.
      • Jafarnejad S.M.
      • Meyer C.
      • Chapat C.
      • Gogakos T.
      • Morozov P.
      • et al.
      The E3 ubiquitin ligase and RNA-binding protein ZNF598 orchestrates ribosome quality control of premature polyadenylated mRNAs.
      ). Furthermore, in both yeast and human cells, the footprints of collided ribosomes, termed a disome, are enriched where the A-site of the leading ribosome is located at a stop codon (
      • Zhao T.
      • Chen Y.M.
      • Li Y.
      • Wang J.
      • Chen S.
      • Gao N.
      • et al.
      Disome-seq reveals widespread ribosome collisions that promote cotranslational protein folding.
      • Han P.
      • Shichino Y.
      • Schneider-Poetsch T.
      • Mito M.
      • Hashimoto S.
      • Udagawa T.
      • et al.
      Genome-wide survey of ribosome collision.
      ). It remains unknown if ribosome collisions occur more frequently at the NMD-inducing stop codons as compared with normal stop codons, as would be expected if NMD-inducing termination was slower. What is now known, however, is that the ribosome collision–sensing proteins ZNF598 and EDF1 are dispensable for NMD as their depletion in mammalian cells did not affect the decay of NMD reporters or endogenous NMD targets (
      • Zinshteyn B.
      • Sinha N.K.
      • Enam S.U.
      • Koleske B.
      • Green R.
      Translational repression of NMD targets by GIGYF2 and EIF4E2.
      ,
      • De S.
      • Mühlemann O.
      A comprehensive coverage insurance for cells: Revealing links between ribosome collisions, stress responses and mRNA surveillance.
      ). It is possible that collisions at stop codons are sensed by yet another mechanism, or such collisions are inconsequential for NMD. On the other hand, NMD factors may be important to ensure efficient translation termination at NMD-inducing stop codons, and their depletion can potentially cause termination arrest leading to ribosome collisions. This has been observed in human cells on an NMD reporter following NMD arrest via pharmacological inhibition of the UPF1 kinase, SMG1 (
      • Zinshteyn B.
      • Sinha N.K.
      • Enam S.U.
      • Koleske B.
      • Green R.
      Translational repression of NMD targets by GIGYF2 and EIF4E2.
      ). The role of ribosome collisions in NMD and the contributions of NMD factors in responding to ribosome collisions remain important avenues for future investigations.
      In summary, despite the progress outlined in the preceeding paragraphs, the picture of how UPF proteins engage with translation termination machinery during NMD and how the NMD-inducing termination reaction proceeds remains incomplete in any system. To fill the remaining gaps, we need to understand if the various interactions of UPF factors with the ribosome, termination factors, and with mRNA and mRNA-binding proteins occur in a particular ordered fashion or if these events occur in a more stochastic manner to initiate NMD. A deeper understanding of the kinetics of normal versus NMD-inducing termination reaction is also needed along with the appreciation of how these events are influenced by the NMD factors. Some of these key outstanding questions can be resolved through the application of new tools and technologies that have recently illuminated the biophysical underpinnings and provided an atomic-level view of the basic steps of normal translation termination (
      • Brown A.
      • Shao S.
      • Murray J.
      • Hegde R.S.
      • Ramakrishnan V.
      Structural basis for stop codon recognition in eukaryotes.
      ,
      • Matheisl S.
      • Berninghausen O.
      • Becker T.
      • Beckmann R.
      Structure of a human translation termination complex.
      ,
      • Lawson M.R.
      • Lessen L.N.
      • Wang J.
      • Prabhakar A.
      • Corsepius N.C.
      • Green R.
      • et al.
      Mechanisms that ensure speed and fidelity in eukaryotic translation termination.
      ). In addition, a single-molecule microscopy–based direct visualization of NMD induction at an individual mRNA in live cells is now possible (
      • Hoek T.A.
      • Khuperkar D.
      • Lindeboom R.G.H.
      • Sonneveld S.
      • Verhagen B.M.P.
      • Boersma S.
      • et al.
      Single-molecule imaging Uncovers rules governing nonsense-mediated mRNA decay.
      ). This work has revealed that not every termination event at a PTC leads to NMD and has opened the door to further dissect NMD mechanisms inside cells.

      Alternative outcomes during elongation and termination that can lead to NMD activation or escape

      As the ribosome transitions through the various states of each elongation cycle, occasional unexpected outcomes can lead to recoding of the message (
      • Dever T.E.
      • Dinman J.D.
      • Green R.
      Translation elongation and recoding in eukaryotes.
      ). For example, an elongating ribosome can shift into a different reading frame, which will not only change the sequence of the encoded polypeptide but also change the termination context and thereby alter mRNA stability via NMD. Similarly, a terminating ribosome can take alternative routes that may result in continued translation. For example, during the termination reaction, eRF1 may be outcompeted by a noncognate aminoacyl-tRNA for recognition of the stop codon, causing the ribosome to readthrough the stop codon and continue translation. If such an event occurs at a PTC, a consequence can be a full-length polypeptide and failure to induce NMD. As discussed later, these variations in outcomes during translation are dictated by local mRNA sequence and structural features. In addition, certain locations within an ORF (e.g., close to the start codon) are more prone to alternative outcomes that override termination and cause an escape from NMD (
      • Zhang J.
      • Maquat L.E.
      Evidence that translation reinitiation abrogates nonsense-mediated mRNA decay in mammalian cells.
      ,
      • Inácios Â.
      • Silva A.L.
      • Pinto J.
      • Ji X.
      • Morgado A.
      • Almeida F.
      • et al.
      Nonsense mutations in close proximity to the initiation codon fail to trigger full nonsense-mediated mRNA decay.
      ). In the following sections, we discuss several examples of events during translation that unexpectedly cause or counteract NMD that have emerged from decades of work on mechanisms of translation and NMD, particularly of mRNAs bearing nonsense mutations that cause disease. Importantly, following ribosome frameshifting or readthrough, the eventual NMD outcome will be dictated by the location of the final stop codon and its impact on the NMD-inducing signals, the 3' UTR-bound NMD activators, and/or the 3′ UTR length.

      Ribosome frameshifting during elongation can lead to NMD

      Ribosome frameshifting occurs when a translating ribosome moves out of its normal reading frame on mRNA and continues translation in a new reading frame. Such a shift in ribosome reading frame can occur because of an active mechanism, referred to as programmed ribosome frameshifting (PRF), which causes ribosomes to “slip” upstream or downstream, mostly by one nucleotide (
      • Dinman J.D.
      Mechanisms and implications of programmed translational frameshifting.
      ). Ribosomes can also move into a new reading frame because of a change in the protein-coding sequence by a length that is not a multiple of 3, either because of insertion/deletion of nucleotides (frameshift mutations) or because of AS. Regardless of the cause of ribosome frameshifting, all such events have the potential to result in an NMD-inducing stop codon (
      • Celik A.
      • Baker R.
      • He F.
      • Jacobson A.
      High-resolution profiling of NMD targets in yeast reveals translational fidelity as a basis for substrate selection.
      ).

      Ribosome frameshifting caused by viral sequences

      PRF can occur because of certain special mRNA sequence features. The most prominent and widespread signal for PRF consists of mRNA structural elements such as a pseudoknot or a stem–loop that are connected by a spacer element to an upstream “slippery” sequence (
      • Dinman J.D.
      Mechanisms and implications of programmed translational frameshifting.
      ). Such a tripartite PRF signal is common in genomes of many RNA viruses (
      • Korniy N.
      • Samatova E.
      • Anokhina M.M.
      • Peske F.
      • Rodnina M.V.
      Mechanisms and biomedical implications of -1 programmed ribosome frameshifting on viral and bacterial mRNAs.
      ,
      • Chang K.-C.
      • Wen J.-D.
      Programmed -1 ribosomal frameshifting from the perspective of the conformational dynamics of mRNA and ribosomes.
      ). Having multiple ORFs in the same transcript staggered in different reading frames allows viruses to compact more genetic information into a smaller genome (
      • Chang K.-C.
      • Wen J.-D.
      Programmed -1 ribosomal frameshifting from the perspective of the conformational dynamics of mRNA and ribosomes.
      ). While PRF is often used to produce multiple products from one transcript, it can also be used as a regulatory mechanism, as a frameshift in the reading frame would introduce a PTC and induce NMD (
      • Hammell A.B.
      • Taylor R.C.
      • Peltz S.W.
      • Dinman J.D.
      Identification of putative programmed -1 ribosomal frameshift signals in large DNA databases.
      ,
      • Plant E.P.
      • Wang P.
      • Jacobs J.L.
      • Dinman J.D.
      A programmed -1 ribosomal frameshift signal can function as a cis-acting mRNA destabilizing element.
      ). The degree of transcript instability caused via NMD by programmed ribosomal frameshifting is a function of the strength of the frameshifting signal and mRNA secondary structure to slow the translating ribosomes (
      • Belew A.T.
      • Advani V.M.
      • Dinman J.D.
      Endogenous ribosomal frameshift signals operate as mRNA destabilizing elements through at least two molecular pathways in yeast.
      ).
      As a −1PRF will destabilize viral transcripts via NMD, many viruses have developed mechanisms to protect their transcripts from NMD while producing proteins via frameshifting. Such is the case of human-T-cell leukemia virus type-1 (HTLV-1), the virus responsible for adult T-cell leukemia (
      • Nakano K.
      • Ando T.
      • Yamagishi M.
      • Yokoyama K.
      • Ishida T.
      • Ohsugi T.
      • et al.
      Viral interference with host mRNA surveillance, the nonsense-mediated mRNA decay (NMD) pathway, through a new function of HTLV-1 Rex: implications for retroviral replication.
      ). Reporter RNAs based on HTLV-1 genome are sensitive to NMD as a result of PRF (
      • Nakano K.
      • Ando T.
      • Yamagishi M.
      • Yokoyama K.
      • Ishida T.
      • Ohsugi T.
      • et al.
      Viral interference with host mRNA surveillance, the nonsense-mediated mRNA decay (NMD) pathway, through a new function of HTLV-1 Rex: implications for retroviral replication.
      ). Furthermore, the HTLV-1-encoded Rex protein blocks NMD of these reporters, suggesting a mechanism for HTLV-1 to reduce the NMD sensitivity of its genomic RNA (
      • Nakano K.
      • Ando T.
      • Yamagishi M.
      • Yokoyama K.
      • Ishida T.
      • Ohsugi T.
      • et al.
      Viral interference with host mRNA surveillance, the nonsense-mediated mRNA decay (NMD) pathway, through a new function of HTLV-1 Rex: implications for retroviral replication.
      ). The use of PRF and mechanisms to avoid NMD as a result are not unique to viruses that infect humans. The plant infecting turnip crinkle virus employs a different method to avoid NMD (
      • May J.P.
      • Yuan X.
      • Sawicki E.
      • Simon A.E.
      RNA virus evasion of nonsense-mediated decay.
      ). Downstream of the termination codon in the turnip crinkle virus genomic RNA is an unstructured region of RNA that reduces NMD efficiency (
      • May J.P.
      • Yuan X.
      • Sawicki E.
      • Simon A.E.
      RNA virus evasion of nonsense-mediated decay.
      ). Mechanisms that viruses use to avoid ribosomal frameshifting–triggered NMD are a potential therapeutic target for treating diseases caused by viruses.

      Ribosomal frameshifting signals in eukaryotic mRNAs

      Ribosomal frameshifting linked to NMD has also been co-opted as a gene regulatory method in eukaryotes. In S. cerevisiae many genes encoding proteins associated with telomeres have programmed frameshifting signals (
      • Belew A.T.
      • Advani V.M.
      • Dinman J.D.
      Endogenous ribosomal frameshift signals operate as mRNA destabilizing elements through at least two molecular pathways in yeast.
      ). The programmed frameshift is used to maintain the proper stochiometric ratio between chromosomes and telomerases by degrading transcripts via NMD when more telomerase is not needed (
      • Belew A.T.
      • Advani V.M.
      • Dinman J.D.
      Endogenous ribosomal frameshift signals operate as mRNA destabilizing elements through at least two molecular pathways in yeast.
      ). When the UPF proteins are knocked out in yeast, their telomeres are shortened, which in turn shortens cellular life span (
      • Enomoto S.
      • Glowczewski L.
      • Lew-Smith J.
      • Berman J.G.
      Telomere cap components influence the rate of senescence in telomerase-deficient yeast cells.
      ). Previously, human CCR5 mRNA was shown to undergo programmed ribosomal frameshift that introduces a PTC and subjects the mRNA to NMD (
      • Belew A.T.
      • Meskauskas A.
      • Musalgaonkar S.
      • Advani V.M.
      • Sulima S.O.
      • Kasprzak W.K.
      • et al.
      Ribosomal frameshifting in the CCR5 mRNA is regulated by miRNAs and the NMD pathway.
      ). However, a recent re-evaluation of the frameshift reporters from the original study have disputed the occurrence of programmed ribosomal frameshifting in CCR5 mRNA (
      • Khan Y.A.
      • Loughran G.
      • Steckelberg A.-L.
      • Brown K.
      • Kiniry S.J.
      • Stewart H.
      • et al.
      Evaluating ribosomal frameshifting in CCR5 mRNA decoding.
      ). Therefore, even though some examples exist of mammalian genes containing retrovirus-derived elements that induce -1PRF (
      • Wills N.M.
      • Moore B.
      • Hammer A.
      • Gesteland R.F.
      • Atkins J.F.
      A functional -1 ribosomal frameshift signal in the human paraneoplastic Ma3 gene.
      ,
      • Clark M.B.
      • Jänicke M.
      • Gottesbühren U.
      • Kleffmann T.
      • Legge M.
      • Poole E.S.
      • et al.
      Mammalian gene PEG10 expresses two reading frames by high efficiency -1 frameshifting in embryonic-associated tissues.
      ,
      • Manktelow E.
      • Shigemoto K.
      • Brierley I.
      Characterization of the frameshift signal of Edr, a mammalian example of programmed -1 ribosomal frameshifting.
      ), but as of yet, there is no evidence of -1PRF signals in mammalian genes that are not derived from retroelements.
      In addition to PRF, other sequence features in coding sequences can increase the frequency of frameshifting by the elongating ribosome, thereby potentially inducing NMD. One such feature is poly(A) tracts within the mRNA coding region that result from repetitive lysine codons. While translating poly(A) tracts, ribosomes “slide” on these sequences, shift into a new reading frame, and can lead to premature termination followed by NMD, a phenomenon that has been observed both in yeast and human cells (
      • Arthur L.L.
      • Pavlovic-Djuranovic S.
      • Koutmou K.S.
      • Green R.
      • Szczesny P.
      • Djuranovic S.
      Translational control by lysine-encoding A-rich sequences.
      ,
      • Koutmou K.S.
      • Schuller A.P.
      • Brunelle J.L.
      • Radhakrishnan A.
      • Djuranovic S.
      • Green R.
      Ribosomes slide on lysine-encoding homopolymeric A stretches.
      ). Notably, the ribosome sliding behavior and resulting mRNA suppression by NMD is the strongest on uninterrupted A stretches and is weaker on polylysine-encoding stretches that include AAG as lysine codons. The connection between poor elongation and ribosome frameshifting to induce NMD apparently also extends to coding sequences where suboptimal codons are more frequent. An investigation of global NMD targets in yeast found that mRNAs that are otherwise normal looking but have a preference for nonoptimal codons are targeted to NMD (
      • Celik A.
      • Baker R.
      • He F.
      • Jacobson A.
      High-resolution profiling of NMD targets in yeast reveals translational fidelity as a basis for substrate selection.
      ). A higher rate of out-of-frame translation observed on these mRNAs suggests that translation of nonoptimal codons is likely to induce ribosome frameshifting, thereby causing premature termination and NMD. It remains to be seen whether similar mRNA features can subject mRNAs to NMD in organisms other than S. cerevisiae.
      When frameshift mutations force a ribosome into a new reading frame, NMD can also ensue when translation terminates in this new reading frame, provided such termination occurs in a context that deems it to be abnormal. Importantly, the basic NMD-inducing contexts discussed previously (e.g., EJC in 3′ UTR) are further governed by features such as the sequence surrounding PTCs, and the location of PTCs in relation to the ORF, which are discussed in more detail later. Nevertheless, whether frameshifting mutations trigger or escape NMD has an important bearing on gene expression from the mutated allele, which can provide new therapeutic avenues (
      • Litchfield K.
      • Reading J.L.
      • Lim E.L.
      • Xu H.
      • Liu P.
      • Al-Bakir M.
      • et al.
      Escape from nonsense-mediated decay associates with anti-tumor immunogenicity.
      ).

      Stop codon readthrough can mute NMD

      When a ribosome reaches a stop codon, it may not always terminate translation. Instead, a noncognate aminoacyl-tRNA can outcompete eRF1 for recognition of stop codon in the A-site, thereby leading to continuation of translation into the downstream region. Such stop codon readthrough can occur naturally at any stop codon, albeit at an extremely low rate. However, certain sequence contexts can make ribosome readthrough more likely. As readthrough at stop codon yields an elongating rather than a terminating ribosome, leaky translation at stop codons antagonizes NMD (
      • Keeling K.M.
      • Lanier J.
      • Du M.
      • Salas-Marco J.
      • Gao L.
      • Kaenjak-Angeletti A.
      • et al.
      Leaky termination at premature stop codons antagonizes nonsense-mediated mRNA decay in S. cerevisiae.
      ). If ribosome translocation continues for extended distances past stop codons, this can further desensitize the mRNA for NMD by stripping off NMD activating factors such as UPF1 and EJC from downstream regions (Fig. 4). In S. cerevisiae, which does not rely on the EJC for NMD, a meager stop codon readthrough rate of 0.5%, or one out of every 200 ribosomes, can reduce NMD efficiency (
      • Keeling K.M.
      • Lanier J.
      • Du M.
      • Salas-Marco J.
      • Gao L.
      • Kaenjak-Angeletti A.
      • et al.
      Leaky termination at premature stop codons antagonizes nonsense-mediated mRNA decay in S. cerevisiae.
      ). Similarly, in mammalian cells, a readthrough rate of less than 1% is sufficient to displace the EJCs and/or UPF1 to inhibit NMD (
      • Hogg J.R.
      • Goff S.P.
      Upf1 senses 3’UTR length to potentiate mRNA decay.
      ). The low amount of stop codon readthrough required to reduce NMD efficiency is among the many lines of evidence that supports NMD to occur during any round of translation (
      • Hogg J.R.
      • Goff S.P.
      Upf1 senses 3’UTR length to potentiate mRNA decay.
      ,
      • Keeling K.M.
      • Lanier J.
      • Du M.
      • Salas-Marco J.
      • Gao L.
      • Kaenjak-Angeletti A.
      • et al.
      Leaky termination at premature stop codons antagonizes nonsense-mediated mRNA decay in S. cerevisiae.
      ). If NMD was only triggered during the pioneer round of translation, a much higher rate of stop codon readthrough would be required to stabilize mRNA with PTCs.
      Figure thumbnail gr4
      Figure 4Stop codon readthrough can prevent NMD. Top, factors that promote stop codon readthrough: (1) the stop codon triplet used, where the UGA codon is the most permissive stop codon, (2) the identity of the nucleotide at +4 position where stop codon readthrough is promoted the most by a C, and (3) an AU-rich region downstream of the stop codon. Bottom, factors that inhibit stop codon readthrough: These are typically found adjacent to the normal termination codons and include a GC-rich region immediately downstream of the stop codon and the presence of poly-A binding protein 1 (PABP1) in spatial proximity of the termination codon. NMD, nonsense-mediated mRNA decay.

      Effect of stop codon context

      Early genetic studies identified that various stop codon contexts can influence termination efficiency and hence ribosome readthrough rate (
      • Dabrowski M.
      • Bukowy-Bieryllo Z.
      • Zietkiewicz E.
      Translational readthrough potential of natural termination codons in eucaryotes – the impact of RNA sequence.
      ). More recent genome-wide analyses of translational readthrough in both yeast and human cells have confirmed these previous observations (
      • Wangen J.R.
      • Green R.
      Stop codon context influences genome-wide stimulation of termination codon readthrough by aminoglycosides.
      ,
      • Mangkalaphiban K.
      • He F.
      • Ganesan R.
      • Wu C.
      • Baker R.
      • Jacobson A.
      Transcriptome-wide investigation of stop codon readthrough in Saccharomyces cerevisiae.
      ). The UGA stop codon is more permissive to readthrough than the other two stop codons (Fig. 4). The sequence surrounding the stop codon, both upstream (
      • Wangen J.R.
      • Green R.
      Stop codon context influences genome-wide stimulation of termination codon readthrough by aminoglycosides.
      ,
      • Mottagui-Tabar S.
      • Tuite M.F.
      • Isaksson L.A.
      The influence of 5’ codon context on translation termination in Saccharomyces cerevisiae.
      ) and downstream (
      • Brown C.M.
      • Stockwell P.A.
      • Trotman C.N.A.
      • Tate W.P.
      Sequence analysis suggests that tetra-nucleotides signal the termination of protein synthesis in eukaryotes.
      ,
      • Keeling K.M.
      • Lanier J.
      • Du M.
      • Salas-Marco J.
      • Gao L.
      • Kaenjak-Angeletti A.
      • et al.
      Leaky termination at premature stop codons antagonizes nonsense-mediated mRNA decay in S. cerevisiae.
      ,
      • Wangen J.R.
      • Green R.
      Stop codon context influences genome-wide stimulation of termination codon readthrough by aminoglycosides.
      ,
      • Tate W.P.
      • Poole E.S.
      • Dalphin M.E.
      • Major L.L.
      • Crawford D.J.G.
      • Mannering S.A.
      The translational stop signal: codon with a context, or extended factor recognition element?.
      ), can dramatically influence the rate of readthrough. The sequences immediately downstream of the stop codon have the strongest effect on readthrough rate. For instance, a cytosine immediately after the stop codon (at the +4 position) results in an increased readthrough rate across all eukaryotes tested (Fig. 4) (
      • Wangen J.R.
      • Green R.
      Stop codon context influences genome-wide stimulation of termination codon readthrough by aminoglycosides.
      ,
      • Mangkalaphiban K.
      • He F.
      • Ganesan R.
      • Wu C.
      • Baker R.
      • Jacobson A.
      Transcriptome-wide investigation of stop codon readthrough in Saccharomyces cerevisiae.
      ). As eRF1 also recognizes the +4 nt during translation termination leading to mRNA compaction (
      • Brown A.
      • Shao S.
      • Murray J.
      • Hegde R.S.
      • Ramakrishnan V.
      Structural basis for stop codon recognition in eukaryotes.
      ,
      • Brown C.M.
      • Stockwell P.A.
      • Trotman C.N.A.
      • Tate W.P.
      Sequence analysis suggests that tetra-nucleotides signal the termination of protein synthesis in eukaryotes.
      ), it is conceivable that a cytosine at this position is least effective in stabilizing such a conformation, which could explain why UGAC is the most readthrough permissive stop codon context (
      • Wangen J.R.
      • Green R.
      Stop codon context influences genome-wide stimulation of termination codon readthrough by aminoglycosides.
      ). An AU-rich sequence 15 nt upstream and downstream of stop codons also increases the rate of stop codon readthrough (
      • Wangen J.R.
      • Green R.
      Stop codon context influences genome-wide stimulation of termination codon readthrough by aminoglycosides.
      ). The +4 nt and AU-rich regions downstream of the stop codon promote readthrough by preventing eRF1 from cleaving the peptidyl-tRNA bond (Fig. 4) (
      • Lawson M.R.
      • Lessen L.N.
      • Wang J.
      • Prabhakar A.
      • Corsepius N.C.
      • Green R.
      • et al.
      Mechanisms that ensure speed and fidelity in eukaryotic translation termination.
      ). An earlier study suggested that potential base pairing between the readthrough promoting sequence and 18S rRNA may be responsible for altering termination efficiency (
      • Namy O.
      • Hatin I.
      • Rousset J.P.
      Impact of the six nucleotides downstream of the stop codon on translation termination.
      ). The penultimate codon can also affect readthrough frequency in S. cerevisiae (
      • Wangen J.R.
      • Green R.
      Stop codon context influences genome-wide stimulation of termination codon readthrough by aminoglycosides.
      ,
      • Mottagui-Tabar S.
      • Tuite M.F.
      • Isaksson L.A.
      The influence of 5’ codon context on translation termination in Saccharomyces cerevisiae.
      ); the AUA, ACA, and GAC codons can enhance stop codon readthrough, whereas CGU, GCU, and UUA codons suppress it (
      • Mangkalaphiban K.
      • He F.
      • Ganesan R.
      • Wu C.
      • Baker R.
      • Jacobson A.
      Transcriptome-wide investigation of stop codon readthrough in Saccharomyces cerevisiae.
      ). There also exist sequence contexts that prevent readthrough and promote termination. For example, high GC content in regions immediately downstream of NTCs may have evolved to help prevent readthrough (Fig. 4) (
      • Wangen J.R.
      • Green R.
      Stop codon context influences genome-wide stimulation of termination codon readthrough by aminoglycosides.
      ). When a ribosome terminates at a PTC introduced by a mutation or an error in the middle of a transcript, it is unlikely that such a GC-rich region is present immediately downstream, thus making PTCs more prone than NTCs to ribosome readthrough.
      The effect of readthrough promoting sequence contexts on NMD is yet to be systematically evaluated, but the available anecdotal evidence supports the profound influence of increased readthrough on NMD efficiency (
      • Lindeboom R.G.H.
      • Supek F.
      • Lehner B.
      The rules and impact of nonsense-mediated mRNA decay in human cancers.
      ,
      • Zhu X.
      • Zhang H.
      • Mendell J.T.
      Ribosome recycling by ABCE1 links Lysosomal function and iron homeostasis to 3’ UTR-directed regulation and nonsense-mediated decay.
      ). In certain mammalian genes, a UGA stop codon followed by CUAG sequence promotes stop codon readthrough (
      • Loughran G.
      • Chou M.Y.
      • Ivanov I.P.
      • Jungreis I.
      • Kellis M.
      • Kiran A.M.
      • et al.
      Evidence of efficient stop codon readthrough in four mammalian genes.
      ). When these readthrough-inducing stop codon contexts from the OPRL1 and AQP4 genes are introduced into an NMD reporter, it dramatically decreases reporter NMD efficiency in mammalian cells (
      • Zinshteyn B.
      • Sinha N.K.
      • Enam S.U.
      • Koleske B.
      • Green R.
      Translational repression of NMD targets by GIGYF2 and EIF4E2.
      ). Importantly, nonsense mutations in the F9 gene, which cause hemophilia B because of inactivation of coagulation factor IX, lead to appreciable readthrough and only a mild phenotype when the nonsense mutation is in a readthrough permissible UUGAC context (
      • Pinotti M.
      • Caruso P.
      • Canella A.
      • Campioni M.
      • Tagariello G.
      • Castaman G.
      • et al.
      Ribosome readthrough accounts for secreted full-length factor IX in hemophilia B patients with nonsense mutations.
      ). Similarly, a nonsense mutation in the identical UUGAC context in LAMA3, which encodes laminin subunit a-3, causes increased readthrough, substantially increased full-length polypeptide expression, and less severe phenotype in a patient with junctional epidermolysis bullosa (
      • Pacho F.
      • Zambruno G.
      • Calabresi V.
      • Kiritsi D.
      • Schneider H.
      Efficiency of translation termination in humans is highly dependent upon nucleotides in the neighbourhood of a (premature) termination codon.
      ). A systematic analysis of potentially NMD-resistant nonsense variants also identified two other readthrough-prone human genes although the sequence context of these nonsense variants was not described (
      • Jagannathan S.
      • Bradley R.K.
      Translational plasticity facilitates the accumulation of nonsense genetic variants in the human population.
      ).
      A specialized stop codon readthrough mechanism is observed in the case of mRNAs encoding selenoproteins where “UGA” stop codons can be recoded as selenocysteine, a nonstandard amino acid (
      • Moriarty P.M.
      • Reddy C.C.
      • Maquat L.E.
      Selenium deficiency reduces the abundance of mRNA for Se-dependent glutathione peroxidase 1 by a UGA-dependent mechanism likely to be nonsense codon-mediated decay of cytoplasmic mRNA.
      ). In this case, the selenocysteine recoding and readthrough is dictated by a cis-acting structured selenocysteine insertion sequence located in 3′ UTRs (
      • Moriarty P.M.
      • Reddy C.C.
      • Maquat L.E.
      Selenium deficiency reduces the abundance of mRNA for Se-dependent glutathione peroxidase 1 by a UGA-dependent mechanism likely to be nonsense codon-mediated decay of cytoplasmic mRNA.
      ). In selenium-deficient conditions, translation terminates at the UGA “selenocysteine” codon and mRNA is degraded via NMD (
      • Moriarty P.M.
      • Reddy C.C.
      • Maquat L.E.
      Selenium deficiency reduces the abundance of mRNA for Se-dependent glutathione peroxidase 1 by a UGA-dependent mechanism likely to be nonsense codon-mediated decay of cytoplasmic mRNA.
      ). Under selenium-replete conditions (in certain organs such as the liver), UGA is recoded as selenocysteine leading to readthrough and mRNA escape from NMD (
      • Moriarty P.M.
      • Reddy C.C.
      • Maquat L.E.
      Selenium deficiency reduces the abundance of mRNA for Se-dependent glutathione peroxidase 1 by a UGA-dependent mechanism likely to be nonsense codon-mediated decay of cytoplasmic mRNA.
      ).
      Specialized sequences that promote stop codon readthrough are also often employed by viruses to maximize the coding potential of their meager genetic resources (
      • Firth A.E.
      • Brierley I.
      Non-canonical translation in RNA viruses.
      ,
      • Hogg J.R.
      Viral evasion and manipulation of host RNA quality control pathways.
      ). Viral RNAs often use such a strategy to encode multiple polypeptides from the same viral mRNA. An example is the Moloney murine leukemia virus that uses translational recoding of stop codon to either produce Gag polyprotein or a Gag–Pol fusion polypeptide (
      • Firth A.E.
      • Brierley I.
      Non-canonical translation in RNA viruses.
      ,
      • Hogg J.R.
      Viral evasion and manipulation of host RNA quality control pathways.
      ). The viral reverse transcriptase, which is part of the Gag polyprotein, interacts with the C-terminal domain of eRF1 to prevent eRF1–eRF3 interaction and causes stop codon readthrough (
      • Firth A.E.
      • Brierley I.
      Non-canonical translation in RNA viruses.
      ,
      • Hogg J.R.
      Viral evasion and manipulation of host RNA quality control pathways.
      ). Other viral sequences that promote ribosome readthrough form stable RNA structures (
      • Hogg J.R.
      Viral evasion and manipulation of host RNA quality control pathways.
      ). Although the readthrough promoting viral sequences can lower NMD efficiency when inserted in NMD reporters (
      • Baker S.L.
      • Hogg J.R.
      A system for coordinated analysis of translational readthrough and nonsense-mediated mRNA decay.
      ), it is not clear if viruses use such strategies to specifically suppress NMD to promote their propagation. Still, such viral control elements offer an attractive tool to investigate the mechanisms of ribosome readthrough and its effect on NMD.

      Influence of trans-acting factors

      In addition to the sequence context surrounding the stop codon, disruption of the factors that promote translation termination can enhance readthrough. In different model systems, limited levels of release factors lead to reduced termination efficiency and increased ribosome readthrough (
      • Stansfield I.
      • Eurwilaichitr L.
      • Akhmaloka
      • Tuite M.F.
      Depletion in the levels of the release factor eRF1 causes a reduction in the efficiency of translation termination in yeast.
      ,
      • Carnes J.
      • Jacobson M.
      • Leinwand L.
      • Yarus M.
      Stop codon suppression via inhibition of eRF1 expression.
      ,
      • Baradaran-Heravi A.
      • Balgi A.D.
      • Hosseini-Farahabadi S.
      • Choi K.
      • Has C.
      • Roberge M.
      Effect of small molecule eRF3 degraders on premature termination codon readthrough.
      ). Under these conditions, readthrough can be even greater with a longer 3' UTR and hence lesser PABP1 interaction with the terminating ribosome (
      • Wu C.
      • Roy B.
      • He F.
      • Yan K.
      • Jacobson A.
      Poly(A)-binding protein regulates the efficiency of translation termination.
      ,
      • Mangkalaphiban K.
      • He F.
      • Ganesan R.
      • Wu C.
      • Baker R.
      • Jacobson A.
      Transcriptome-wide investigation of stop codon readthrough in Saccharomyces cerevisiae.
      ). This outcome is different from that under normal conditions when a reduced PABP1-terminating ribosome interaction would promote NMD and possibly reflects a different function of PABP1 during inefficient termination. Loss of other factors that enhance translation termination, such as Dbp5/Gle1 and their activator IP6, can also increase readthrough rates (
      • Alcázar-Román A.R.
      • Bolger T.A.
      • Wente S.R.
      Control of mRNA export and translation termination by inositol hexakisphosphate requires specific interaction with Gle1.
      ,
      • Gross T.
      • Siepmann A.
      • Sturm D.
      • Windgassen M.
      • Scarcelli J.J.
      • Seedorf M.
      • et al.
      The DEAD-Box RNA helicase Dbp5 functions in translation termination.
      ,
      • Beißel C.
      • Grosse S.
      • Krebber H.
      Dbp5/DDX19 between translational readthrough and nonsense mediated decay.
      ). Disruption of ribosome recycling, which will result in extended residence times of the terminated ribosome on a transcript, thereby increasing the chances of stop codon recognition by a noncognate aminoacyl-tRNA, also enhances ribosome readthrough. A reduction in levels of factors involved in ribosome splitting/60S subunit recycling, such as ABCE1 depletion in human cells and HCR1 disruption in yeast cells, leads to increased ribosome readthrough, and hence the removal of NMD activators from the 3′ UTRs and reduced NMD efficiency (
      • Beznosková P.
      • Cuchalová L.
      • Wagner S.
      • Shoemaker C.J.
      • Gunišová S.
      • von der Haar T.
      • et al.
      Translation initiation factors eIF3 and HCR1 control translation termination and stop codon read-through in yeast cells.
      ,
      • Zhu X.
      • Zhang H.
      • Mendell J.T.
      Ribosome recycling by ABCE1 links Lysosomal function and iron homeostasis to 3’ UTR-directed regulation and nonsense-mediated decay.
      ,
      • D’Orazio K.N.
      • Lessen L.N.
      • Veltri A.J.
      • Neiman Z.
      • Pacheco M.
      • Loll-Krippleber R.
      • et al.
      Genetic screens identify connections between ribosome recycling and nonsense mediated decay.
      ,
      • Annibaldis G.
      • Domanski M.
      • Dreos R.
      • Contu L.
      • Carl S.
      • Klaÿ N.
      • et al.
      Readthrough of stop codons under limiting ABCE1 concentration involves frameshifting and inhibits nonsense-mediated mRNA decay.
      ). Similar outcomes are reported for the loss of 40S subunit recycling factors TMA20 and TMA22 in yeast cells (
      • Wangen J.R.
      • Green R.
      Stop codon context influences genome-wide stimulation of termination codon readthrough by aminoglycosides.
      ,
      • D’Orazio K.N.
      • Lessen L.N.
      • Veltri A.J.
      • Neiman Z.
      • Pacheco M.
      • Loll-Krippleber R.
      • et al.
      Genetic screens identify connections between ribosome recycling and nonsense mediated decay.
      ).
      Disruption of NMD machinery also influences the amount of readthrough observed at PTCs. This observation supports a role for UPF factors in enhancing the efficiency of termination and/or ribosome recycling during premature termination. UPF1 knockout in S. cerevisiae increases stop codon readthrough rate by approximately twofold (
      • Keeling K.M.
      • Lanier J.
      • Du M.
      • Salas-Marco J.
      • Gao L.
      • Kaenjak-Angeletti A.
      • et al.
      Leaky termination at premature stop codons antagonizes nonsense-mediated mRNA decay in S. cerevisiae.
      ). Moreover, under these conditions, the difference in readthrough rates between readthrough permissive and standard stop codon contexts is much smaller than in wildtype, indicating an overall increase in readthrough (
      • Keeling K.M.
      • Lanier J.
      • Du M.
      • Salas-Marco J.
      • Gao L.
      • Kaenjak-Angeletti A.
      • et al.
      Leaky termination at premature stop codons antagonizes nonsense-mediated mRNA decay in S. cerevisiae.
      ). Thus, the relationship between UPF1 and stop codon readthrough is a reciprocal one. Just as ribosomes failing to terminate can displace UPF1, the presence of UPF1 can suppress stop codon readthrough (
      • Keeling K.M.
      • Lanier J.
      • Du M.
      • Salas-Marco J.
      • Gao L.
      • Kaenjak-Angeletti A.
      • et al.
      Leaky termination at premature stop codons antagonizes nonsense-mediated mRNA decay in S. cerevisiae.
      ). Importantly, the increased readthrough upon NMD inhibition in yeast can also be attributed to an indirect mechanism. Depletion of UPF factors in yeast stabilizes the PTC containing ALR1 mRNA that encodes a magnesium transporter. The elevation of ALR1 levels in both NMD-deficient and NMD-proficient cells causes an increase in magnesium uptake, which leads to an increase in translational readthrough (
      • Johansson M.J.O.
      • Jacobson A.
      Nonsense-mediated mRNA decay maintains translational fidelity by limiting magnesium uptake.
      ). Thus, the direct impact of NMD factors on outcomes of termination reaction in yeast and eukaryotes in general remains an important open question.

      Therapeutic potential of ribosome readthrough

      Readthrough of nonsense codons has long been recognized as a potential therapeutic strategy for diseases where a PTC leads to NMD and expression of a truncated protein. In such cases, promoting stop codon readthrough would allow for the production of a full-length protein that may retain wildtype function and ameliorate disease (
      • Jagannathan S.
      • Bradley R.K.
      Translational plasticity facilitates the accumulation of nonsense genetic variants in the human population.
      ). A variety of small chemicals have been identified that can promote readthrough at stop codons. We refer readers to an excellent recent review for an in-depth discussion of such readthrough-promoting drugs, their mode of action, and various challenges in their clinical use (
      • Dabrowski M.
      • Bukowy-Bieryllo Z.
      • Zietkiewicz E.
      Advances in therapeutic use of a drug-stimulated translational readthrough of premature termination codons.
      ). We want to highlight two key points here regarding these readthrough-promoting drugs. First, just like natural ribosomal readthrough, the drug-induced readthrough is likely to be influenced by stop codon sequence context, as was recently shown for the aminoglycoside class of antibiotics (
      • Wangen J.R.
      • Green R.
      Stop codon context influences genome-wide stimulation of termination codon readthrough by aminoglycosides.
      ). Notably, aminoglycosides are the most widely investigated readthrough-promoting drugs that hold therapeutic potential in a wide variety of contexts, such as epilepsy (
      • Huang X.
      • Tian M.
      • Hernandez C.C.
      • Hu N.
      • MacDonald R.L.
      The GABRG2 nonsense mutation, Q40X, associated with Dravet syndrome activated NMD and generated a truncated subunit that was partially rescued by aminoglycoside-induced stop codon read-through.
      ), Duchenne muscular dystrophy (
      • Malik V.
      • Rodino-Klapac L.R.
      • Viollet L.
      • Wall C.
      • King W.
      • Al-Dahhak R.
      • et al.
      Gentamicin-induced readthrough of stop codons in Duchenne muscular dystrophy.
      ), and cancer (
      • Bordeira-Carriço R.
      • Pêgo A.P.
      • Santos M.
      • Oliveira C.
      Cancer syndromes and therapy by stop-codon readthrough.
      ). Second, another readthrough-promoting chemical, PTC124 (also known as ataluren) that is already in clinical use for Duchenne muscular dystrophy treatment, was shown to specifically promote readthrough at PTCs but not NTCs (
      • Welch E.M.
      • Barton E.R.
      • Zhuo J.
      • Tomizawa Y.
      • Friesen W.J.
      • Trifillis P.
      • et al.
      PTC124 targets genetic disorders caused by nonsense mutations.
      ). This observation also provides an argument in favor of potential mechanistic differences between normal and premature termination.

      Translation reinitiation following termination also silences NMD

      Similar to stop codon readthrough is ribosome reinitiation, where a terminated ribosome restarts translation from a nearby AUG codon and into the 3′ UTR (
      • Ghosh S.
      • Ganesan R.
      • Amrani N.
      • Jacobson A.
      Translational competence of ribosomes released from a premature termination codon is modulated by NMD factors.
      ). Estimates from S. cerevisiae suggest that approximately 1 in 40 ribosomes can reinitiate at a downstream start codon after terminating at a nonsense codon (
      • Ghosh S.
      • Ganesan R.
      • Amrani N.
      • Jacobson A.
      Translational competence of ribosomes released from a premature termination codon is modulated by NMD factors.
      ). The impact of ribosome reinitiation is similar to readthrough as reinitiated ribosomes would translocate through sequences 3′ to PTCs, thereby dislodging NMD-activating proteins from these regions and making mRNAs resistant to NMD (Fig. 5). If reinitiation downstream of a PTC occurs from an in-frame start codon, it will produce an N-terminally truncated protein, which may or may not be beneficial (or even harmful) to the cell depending on the location of PTC. However, if the ribosome reinitiates in another frame, it will result in a novel polypeptide with little resemblance to the expected protein. It is also notable that translation reinitiation that causes NMD escape can also occur at noncanonical start codons downstream of a PTC although this happens at a lower rate (
      • Jagannathan S.
      • Bradley R.K.
      Translational plasticity facilitates the accumulation of nonsense genetic variants in the human population.
      ,
      • Neu-Yilik G.
      • Amthor B.
      • Gehring N.H.
      • Bahri S.
      • Paidassi H.
      • Hentze M.W.
      • et al.
      Mechanism of escape from nonsense-mediated mRNA decay of human β-globin transcripts with nonsense mutations in the first exon.
      ).
      Figure thumbnail gr5
      Figure 5Ribosome reinitiation after a PTC can prevent NMD. A, when the ribosome terminates at a PTC, it is recognized by downstream UPF1, with the aid of the EJC, and the transcript undergoes NMD. See for more details. B, when ribosome recycling occurs poorly at a premature termination codon, for example, because of limiting recycling factors like ABCE1, this can lead to the ribosome reinitiation at a nearby downstream start codon. Reinitiation causes displacement of NMD activating UPF and EJC proteins, escape from NMD and production of an N-terminally truncated or a unique polypeptide. EJC, exon junction complex; NMD, nonsense-mediated mRNA decay; PTC, premature termination codon.

      Factors that influence ribosome reinitiation

      Translation reinitiation that interferes with NMD is observed more frequently when PTCs occur near normal start codons and when an AUG (methionine or Met) codon is also present in close vicinity of such stop codons. These general features are borne out of analysis of several disease-linked nonsense mutations that were found to be NMD refractory and from mutational studies using NMD reporters based on such genes. A triose phosphate isomerase gene–based reporter showed that PTCs within the first 30 nt of the triose phosphate isomerase coding sequence are resistant to downregulation via NMD, and this resistance is dependent on a nearby downstream AUG codon that exists in a favorable translation initiation context (
      • Zhang J.
      • Maquat L.E.
      Evidence that translation reinitiation abrogates nonsense-mediated mRNA decay in mammalian cells.
      ). Similar trends have been observed using reporters based on the β-globin gene wherein nonsense mutations cause β-thalassemia (
      • Chang J.C.
      • Kan Y.W.
      β0 thalassemia, a nonsense mutation in man.
      ). Usually, nonsense mutations in β-globin result in NMD, but those near the 5′ end of the transcript (in exon 1) escape NMD (
      • Inácios Â.
      • Silva A.L.
      • Pinto J.
      • Ji X.
      • Morgado A.
      • Almeida F.
      • et al.
      Nonsense mutations in close proximity to the initiation codon fail to trigger full nonsense-mediated mRNA decay.
      ,
      • Romäo L.
      • Inácio Â.
      • Santos S.
      • Ávila M.
      • Faustino P.
      • Pacheco P.
      • et al.
      Nonsense mutations in human β-globin gene lead to unexpected levels of cytoplasmic mRNA accumulation.
      ). Mutational analysis of β-globin reporters suggests that the NMD resistance of β-globin exon 1 nonsense mutations depends on translation reinitiation at a downstream in-frame AUG (Met55), thereby producing an N-terminally truncated protein (
      • Neu-Yilik G.
      • Amthor B.
      • Gehring N.H.
      • Bahri S.
      • Paidassi H.
      • Hentze M.W.
      • et al.
      Mechanism of escape from nonsense-mediated mRNA decay of human β-globin transcripts with nonsense mutations in the first exon.
      ). Translation reinitiation at downstream AUG codons also underlies NMD resistance of TP53 (
      • Cohen S.
      • Kramarski L.
      • Levi S.
      • Deshe N.
      • Ben David O.
      • Arbely E.
      Nonsense mutation-dependent reinitiation of translation in mammalian cells.
      ) and α-globin (
      • Pereira F.J.C.
      • Teixeira A.
      • Kong J.
      • Barbosa C.
      • Silva A.L.
      • Marques-Ramos A.
      • et al.
      Resistance of mRNAs with AUG-proximal nonsense mutations to nonsense-mediated decay reflects variables of mRNA structure and translational activity.
      ) gene reporters, of multiple genes that encode N-terminally truncated proteins from their nonsense variants (
      • Jagannathan S.
      • Bradley R.K.
      Translational plasticity facilitates the accumulation of nonsense genetic variants in the human population.
      ), and of several disease-causing nonsense alleles that escape NMD (see Refs. (
      • Stump M.R.
      • Gong Q.
      • Zhou Z.
      LQT2 nonsense mutations generate trafficking defective NH2-terminally truncated channels by the reinitiation of translation.
      ,
      • Harries L.W.
      • Bingham C.
      • Bellanne-Chantelot C.
      • Hattersley A.T.
      • Ellard S.
      The position of premature termination codons in the hepatocyte nuclear factor -1 beta gene determines susceptibility to nonsense-mediated decay.
      ,
      • Gurvich O.L.
      • Maiti B.
      • Weiss R.B.
      • Aggarwal G.
      • Howard M.T.
      • Flanigan K.M.
      DMD exon 1 truncating point mutations: amelioration of phenotype by alternative translation initiation in exon 6.
      ) for some examples). In certain diseases, such an escape from NMD can turn out to be beneficial. Nonsense mutations in the ATRX gene are embryonic lethal, but one mutation, R37X (a c.109C>T change in DNA sequence), avoids this fate (
      • Howard M.T.
      • Malik N.
      • Anderson C.B.
      • Voskuil J.L.A.
      • Atkins J.F.
      • Gibbons R.J.
      Attenuation of an amino-terminal premature stop codon mutation in the ATRX gene by an alternative mode of translational initiation.
      ). Reinitiation downstream of the PTC results in enough ATRX transcripts that survive NMD to produce a truncated but partly functional protein at ∼20% of the wildtype amount, which prevents the severe phenotype (
      • Howard M.T.
      • Malik N.
      • Anderson C.B.
      • Voskuil J.L.A.
      • Atkins J.F.
      • Gibbons R.J.
      Attenuation of an amino-terminal premature stop codon mutation in the ATRX gene by an alternative mode of translational initiation.
      ).
      Recent large-scale analyses have further validated that reinitiation downstream of PTCs is one of the key factors for NMD resistance of an mRNA. An analysis of thousands of somatic nonsense mutations in the cancer genome atlas data verified that PTCs within the first 150 nt of coding sequence represent one of the three major classes of NMD evading features (
      • Lindeboom R.G.H.
      • Supek F.
      • Lehner B.
      The rules and impact of nonsense-mediated mRNA decay in human cancers.
      ,
      • Supek F.
      • Lehner B.
      • Lindeboom R.G.H.
      To NMD or not to NMD: nonsense-mediated mRNA decay in cancer and other genetic diseases.
      ,
      • Lindeboom R.G.H.
      • Vermeulen M.
      • Lehner B.
      • Supek F.
      The impact of nonsense-mediated mRNA decay on genetic disease, gene editing and cancer immunotherapy.
      ). Furthermore, while the rapid emergence of engineered nucleases has provided a facile tool to introduce nonsense/frameshift mutations to generate loss-of-function alleles, several studies have noted that such nonsense alleles often escape NMD (
      • Smits A.H.
      • Ziebell F.
      • Joberty G.
      • Zinn N.
      • Mueller W.F.
      • Clauder-Münster S.
      • et al.
      Biological plasticity rescues target activity in CRISPR knock outs.
      ). Reinitiation downstream of PTCs is one of the major reasons for this phenomenon (
      • Lindeboom R.G.H.
      • Vermeulen M.
      • Lehner B.
      • Supek F.
      The impact of nonsense-mediated mRNA decay on genetic disease, gene editing and cancer immunotherapy.
      ,
      • Smits A.H.
      • Ziebell F.
      • Joberty G.
      • Zinn N.
      • Mueller W.F.
      • Clauder-Münster S.
      • et al.
      Biological plasticity rescues target activity in CRISPR knock outs.
      ). Such a limitation can be circumvented by introducing NMD-inducing lesions far away from possible alternative translation initiation sites (
      • Lee J.H.
      • Yu S.
      • Nam T.W.
      • Roh J.-I.
      • Jin Y.
      • Han J.P.
      • et al.
      The position of the target site for engineered nucleases improves the aberrant mRNA clearance in in vivo genome editing.
      ).
      Ribosome reinitiation downstream of a stop codon can also be used to regulate endogenous transcripts via NMD. In a phenomenon observed in the SRSF7 gene, a “split-ORF,” where the normal ORF is split by a PTC and a downstream start codon, is used to regulate SRSF7 levels (
      • Königs V.
      • de Oliveira Freitas Machado C.
      • Arnold B.
      • Blümel N.
      • Solovyeva A.
      • Löbbert S.
      • et al.
      SRSF7 maintains its homeostasis through the expression of Split-ORFs and nuclear body assembly.
      ). Translation reinitiation from the downstream start codon is required to remove the remaining EJCs from the second part of the ORF. Otherwise, mRNA suppression by NMD would not allow accumulation of the polypeptide encoded by the first half of the split-ORF that is required for SRSF7 regulation (
      • Königs V.
      • de Oliveira Freitas Machado C.
      • Arnold B.
      • Blümel N.
      • Solovyeva A.
      • Löbbert S.
      • et al.
      SRSF7 maintains its homeostasis through the expression of Split-ORFs and nuclear body assembly.
      ). Split-ORFs were reported to occur in a large number of human and mouse transcripts and proposed to be an important regulatory mechanism for RNA-binding proteins (
      • Königs V.
      • de Oliveira Freitas Machado C.
      • Arnold B.
      • Blümel N.
      • Solovyeva A.
      • Löbbert S.
      • et al.
      SRSF7 maintains its homeostasis through the expression of Split-ORFs and nuclear body assembly.
      ).

      Possible mechanisms of ribosome reinitiation

      How can a ribosome that has undergone termination reinitiate translation on the same mRNA? Although details of this process remain poorly understood, possible mechanisms have emerged from the investigation of mRNAs that contain uORFs, short translation regulatory ORFs in mRNA 5′UTRs that control ribosome access to the main ORF. uORFs have been well defined in stress response genes in yeast and mammalian cells (
      • Dever T.E.
      • Ivanov I.P.
      • Sachs M.S.
      Conserved upstream open reading frame nascent peptides that control translation.
      ), although their widespread prevalence in human and other genomes has been reported (
      • Dever T.E.
      • Ivanov I.P.
      • Sachs M.S.
      Conserved upstream open reading frame nascent peptides that control translation.
      ). Translation regulation by uORFs is achieved by controlling how frequently a ribosome will reinitiate translation on the start codon of the main/downstream ORF after it terminates on the uORF stop codon. Translation termination at uORF stop codons can also activate NMD as such a termination event is followed by a much longer “3′ UTR” that may even carry one or more EJCs. Investigations into factors that regulate translation reinitiation on the main ORF start codon have revealed that the short length of uORFs and short intercistronic distance between multiple ORFs are key factors that determine reinitiation efficiency (
      • Mohammad M.P.
      • Pondelícková V.M.
      • Zeman J.
      • Gunisová S.
      • Valásek L.S.
      In vivo evidence that eIF3 stays bound to ribosomes elongating and terminating on short upstream ORFs to promote reinitiation.
      ,
      • Wang X.Q.
      • Rothnagel J.A.
      5′-Untranslated regions with multiple upstream AUG codons can support low-level translation via leaky scanning and reinitiation.
      ,
      • Russell P.J.
      • Slivka J.A.
      • Boyle E.P.
      • Burghes A.H.M.
      • Kearse M.G.
      Small uORFs favor translation re-initiation but do not protect mRNAs from nonsense-mediated decay.
      ,
      • Rajkowitsch L.
      • Vilela C.
      • Berthelot K.
      • Ramirez C.V.
      • McCarthy J.E.G.
      Reinitiation and recycling are distinct processes occurring downstream of translation termination in yeast.
      ). Analysis of uORFs in yeast GCN4 gene and human ATF4 gene suggests that shorter distances traversed by ribosomes elongating in uORFs or scanning intercistronic sequences result in retention of certain protein factors such as eIF3B, a subunit of translation factor eIF3 with important roles in translation initiation, termination, and recycling (
      • Mohammad M.P.
      • Pondelícková V.M.
      • Zeman J.
      • Gunisová S.
      • Valásek L.S.
      In vivo evidence that eIF3 stays bound to ribosomes elongating and terminating on short upstream ORFs to promote reinitiation.
      ,
      • Hronová V.
      • Mohammad M.P.
      • Wagner S.
      • Pánek J.
      • Gunišová S.
      • Zeman J.
      • et al.
      Does eIF3 promote reinitiation after translation of short upstream ORFs also in mammalian cells?.
      ). Even core translation initiation factors eIF4E and eIF4G can remain associated with elongating 80S ribosomes for short distances (
      • Bohlen J.
      • Fenzl K.
      • Kramer G.
      • Bukau B.
      • Teleman A.A.
      Selective 40S footprinting reveals cap-Tethered ribosome scanning in human cells.
      ) and can thereby explain how short uORFs can promote reinitiation more efficiently.
      Reinitiation at a downstream start codon occurs in opposition of ribosome recycling (Fig. 5) (
      • Ghosh S.
      • Ganesan R.
      • Amrani N.
      • Jacobson A.
      Translational competence of ribosomes released from a premature termination codon is modulated by NMD factors.
      ). Thus, variations in ribosome recycling may be yet another factor that dictates reinitiation efficiency. For example, when ribosome splitting factor ABCE1 is depleted from yeast cells, ribosomes can reinitiate upstream or downstream of the stop codon, sometimes in a different reading frame (
      • Young D.J.
      • Guydosh N.R.
      • Zhang F.
      • Hinnebusch A.G.
      • Green R.
      Rli1/ABCE1 recycles terminating ribosomes and controls translation reinitiation in 3′UTRs in vivo.
      ). Similarly, knockdown of 40S recycling factors also increases the rate of reinitiation (
      • Pisarev A.V.
      • Skabkin M.A.
      • Pisareva V.P.
      • Skabkina O.V.
      • Rakotondrafara A.M.
      • Hentze M.W.
      • et al.
      The role of ABCE1 in eukaryotic posttermination ribosomal recycling.
      ,
      • Young D.J.
      • Meydan S.
      • Guydosh N.R.
      40S ribosome profiling reveals distinct roles for Tma20/Tma22 (MCT-1/DENR) and Tma64 (eIF2D) in 40S subunit recycling.
      ,
      • Young D.J.
      • Makeeva D.E.
      • Zhang F.
      • Anisimova A.S.
      • Stolboushkina E.A.
      • Ghobakhlou F.
      • et al.
      Tma64/eIF2D, Tma20/MCT-1, and Tma22/DENR recycle post-termination 40S subunits in vivo.
      ,
      • Ghosh S.
      • Ganesan R.
      • Amrani N.
      • Jacobson A.
      Translational competence of ribosomes released from a premature termination codon is modulated by NMD factors.
      ).

      Reinitiation-independent mechanisms of NMD escape?

      Available evidence also suggests that other mechanisms may also dictate if an mRNA with an early stop codon is targeted to NMD. Romão et al. (
      • Silva A.L.
      • Ribeiro P.
      • Inácio A.
      • Liebhaber S.A.
      • Romão L.
      Proximity of the poly(A)-binding protein to a premature termination codon inhibits mammalian nonsense-mediated mRNA decay.