R

B(

Sa

b

a

ARRAA

KRPmNNSUSUUUUSSSS

C

h1

Seminars in Cell & Developmental Biology 75 (2018) 78–87

Contents lists available at ScienceDirect

Seminars in Cell & Developmental Biology

j ourna l ho me pa g e: www.elsev ier .com/ locate /semcdb

eview

eyond quality control: The role of nonsense-mediated mRNA decayNMD) in regulating gene expression

ofia Nasif a, Lara Contua,b, Oliver Mühlemanna,∗

Department of Chemistry and Biochemistry, University of Bern, CH-3012 Bern, SwitzerlandGraduate School for Cellular and Biomedical Sciences, University of Bern, CH-3012 Bern, Switzerland

r t i c l e i n f o

rticle history:eceived 3 July 2017eceived in revised form 25 August 2017ccepted 28 August 2017vailable online 1 September 2017

eywords:NA turnoverosttranscriptional gene regulationRNA surveillanceMDeuronal differentiationpermatogenesisnfolded protein responsetress responsePF1

a b s t r a c t

Nonsense-mediated mRNA decay (NMD) has traditionally been described as a quality control systemthat rids cells of aberrant mRNAs with crippled protein coding potential. However, transcriptome-wideprofiling of NMD deficient cells identified a plethora of seemingly intact mRNAs coding for functionalproteins as NMD targets. This led to the view that NMD constitutes an additional post-transcriptionallayer of gene expression control involved in the regulation of many different biological pathways. Here,we review our current knowledge about the role of NMD in embryonic development and tissue-specificcell differentiation. We further summarize how NMD contributes to balancing of the integrated stressresponse and to cellular homeostasis of splicing regulators and NMD factors through auto-regulatoryfeedback loops. In addition, we discuss recent evidence that suggests a role for NMD as an innate immuneresponse against several viruses. Altogether, NMD appears to play an important role in a broad spectrumof biological pathways, many of which still remain to be discovered.

© 2017 Elsevier Ltd. All rights reserved.

PF2PF3APF3BMG1MG5MG6

MG7

ontents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 792. NMD factors are essential for normal embryonic development and viability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 793. Roles for NMD factors in tissue-specific differentiation programs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

3.1. Tight regulation of NMD activity is required for normal spermatogenesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 803.2. NMD factors are required for neuronal development and physiology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

4. NMD is involved in the integrated stress response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 825. RNA viruses are targeted by or able to evade the host NMD machinery. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .846. Alternative splicing coupled with NMD (AS-NMD) modulates post-transcriptional gene expression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 847. NMD factors are autoregulated by the NMD pathway . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 858. NMD factor mutations in human diseases and the therapeutic potential of modulating NMD activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

9. Concluding remark. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

∗ Corresponding author.E-mail address: oliver.muehlemann@dcb.unibe.ch (O. Mühlemann).

ttps://doi.org/10.1016/j.semcdb.2017.08.053084-9521/© 2017 Elsevier Ltd. All rights reserved.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .85

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

evelo

1

frpNC[mf

bnhasiormastfsauccNttdtIpsiThiga[mlt

f[meoSSrrpdpciaip

S. Nasif et al. / Seminars in Cell & D

. Introduction

In eukaryotes, nonsense mutations truncate the open readingrame (ORF) of genes prematurely and have long been known toeduce the half-life of the respective mRNA [1,2]. This widespreadhenomenon has been termed nonsense-mediated mRNA decay orMD for short [3]. Genetic screens in Saccharomyces cerevisiae andaenorhabditis elegans identified the first proteins required for NMD4,5] and since then homologous proteins have been identified in

ost other eukaryotic organisms. More recently, additional NMDactors were identified in C. elegans [6–8] and human cells [9].

Despite a wealth of biochemical data documenting interactionsetween the different NMD factors, the exact molecular mecha-ism of NMD is not yet understood. There is an emerging consensus,owever, that NMD results from improper translation terminationt stop codons occurring in places of the mRNP that lack neces-ary termination stimulating signals and/or contain terminationnhibitors. It is likely that in mammalian cells many regulatorsf translation termination exist, however functional data is cur-ently available for only a few. The most well understood is theajor cytoplasmic poly(A) binding protein (PABPC1), which acts

s a potent NMD suppressor when bound in the vicinity of thetop codon [10,11]. Whether “improper” termination means thathose translation events that trigger NMD are mechanistically dif-erent from those that leave the mRNA intact, or whether it is onlylower termination that gives enough time for the assembly ofn NMD complex and subsequent mRNA degradation, is currentlynresolved. The observation that UGA stop codons that normallyode for selenocysteine elicit NMD under limiting selenium con-entrations [12,13] supports the current model postulating thatMD results from prolonged ribosome stalling at stop codons due

o the absence of termination promoting signals. Slow or stalledranslation termination allows the core NMD factor UPF1–an ATP-ependent RNA helicase − and the SMG1 complex to interact viahe release factors eRF1 and eRF3 with the stalled ribosome [14].nteraction of UPF2 and possibly UPF3B with the SMG1-UPF1 com-lex activates the SMG1 kinase, resulting in the phosphorylation ofeveral serine-glutamine (SQ) and threonine-glutamine TQ motifsn the N and C-terminal part of UPF1 that flank its helicase domain.he endonuclease SMG6 [15,16] and the heterodimer SMG5/SMG7ave been found to interact with phosphorylated UPF1 [17] and to

nduce the degradation of the targeted mRNA. Recent evidence sug-ests that in human cells NMD is predominantly initiated through

SMG6-mediated endonucleolytic cleavage near the stop codon18,19] and as a backup system through SMG7-mediated recruit-

ent of the CCR4/NOT complex, which deadenylates the mRNA,eading to decapping followed by 5′-to-3′ exonucleolytic degrada-ion by XRN1 [20].

Interestingly, it seems that not all proteins known to be requiredor NMD in mammalian cells have homologues in other organisms21] (Table 1), indicating that differences in the regulation and

aybe even in the molecular mechanism of NMD among differ-nt species are very likely to exist. For example, no homologuesf the serine/threonine protein kinase SMG1 and its regulatorsMG8 and SMG9 appear to be present in S. cerevisiae. Given thatMG1-mediated phosphorylation of the central NMD factor UPF1epresents a crucial step in NMD in a majority of organisms, itemains unclear whether S. cerevisiae has developed an alternativehosphorylation-independent way of activating NMD or whether aifferent yet unknown kinase fulfills this function. That a few phos-horylation sites present in mammalian UPF1 are conserved in S.erevisiae Upf1 would argue for the latter possibility [22]. SMG1

s also absent from Arabidopsis thaliana, while SMG1 homologuesre found in all other examined plants [23], raising the intrigu-ng question of how A. thaliana could afford to lose SMG1 duringlant evolution without simultaneously incapacitating NMD. It

pmental Biology 75 (2018) 78–87 79

is not only the SMG1 complex that is missing in some species;homologues of the mammalian SMG5/SMG7 heterodimer and ofSMG6 are also absent in some eukaryotes. For example, Drosophilamelanogaster lacks a SMG7 homologue and NMD appears to beinduced exclusively by SMG6-mediated endonucleolytic cleavageof the mRNA in flies. No homologues of SMG5, SMG6 and SMG7have been unambiguously identified in S. cerevisiae, although thereis some evidence that the protein Ebs1 might be the orthologue ofSMG7 [24]. Overall it looks as if NMD comes in different flavours indifferent organisms, with only UPF1, UPF2 and UPF3 forming theevolutionarily conserved core.

Since the focus of this review is on the biological processes inwhich NMD plays a critical role in shaping the respective geneexpression program rather than on the mechanistic aspects of NMD,we decided to provide only a brief sketch of the current mech-anistic NMD model in mammalian cells. Several recent reviewsdescribe in much greater detail the current working models andopen mechanistic questions, and readers interested in the mech-anism of NMD are referred to these publications [25–27]. Amongthese many open questions, one of the most pressing is to betterunderstand the translation termination conditions that result inNMD and how they differ from the conditions required for propertermination. It was discovered early on that the presence of anexon–exon junction located >50 nucleotides downstream of thestop codon was a hallmark of mammalian mRNAs subject to NMD[28]. Rather than being strictly required for NMD, however, it waslater found that these exon–exon junctions function as enhancersof NMD efficacy (i.e. leading to a more pronounced destabilizationof the mRNA) and that long 3′ UTRs lacking exon–exon junctionscan also trigger NMD [29,10,30]. Similar to long 3′ UTRs, shortupstream ORFs (uORFs) and ORF-interrupting premature termi-nation codons (PTCs) are also well known NMD-inducing featuresthat result in ribosomes terminating translation distantly from thepoly(A) tail. Transcriptome profiling revealed that NMD modulatesthe cellular abundance of many mRNAs that appear to code for fulllength functional proteins [31–35], indicating that beyond its welldocumented quality control function, NMD also contributes to reg-ulating gene expression and therewith is likely to play an importantrole in the regulation of specific biological pathways. In the follow-ing, we review our current knowledge about the different biologicalfunctions of NMD.

2. NMD factors are essential for normal embryonicdevelopment and viability

NMD factors have an essential role in mammalian embryoge-nesis, since the knockout of almost every one of these genes isembryonic lethal. For instance, UPF1 knockout (KO) mice are onlyviable in the pre-implantation period (up to embryonic day 3.5[E3.5]), and attempts to maintain the UPF1−/− pre-implantationblastocysts in culture have failed due to a strong induction of apo-ptosis [36]. Similarly, UPF2 KO mice die in utero between E3.5 andE9.5 [37]. Efforts to generate SMG1 KO mice also resulted in devel-opmental arrest at E8.5, with SMG1−/− embryos showing profounddevelopmental defects [38]. Alike UPF1−/− mice, SMG6 KO micealso die around the implantation period, due to impaired differ-entiation of SMG6−/− cells [39]. Very recently, embryonic lethalphenotypes were also reported for UPF3A and SMG9 deficient mice.UPF3A KO mice die between E4.5 and E8.5 [40]. In the case of SMG9,E15.5 homozygous embryos with a severe hypomorph SMG9 alleleshowed a range of variable and incompletely penetrant pheno-

types including hemorrhage, and exencephaly [41]. To date, theonly reported constitutive KO of an NMD factor that is viable inmice is UPF3B [42]. Even though UPF3B KO mice have no overtphenotypic defects, UPF3B mutations have been found in humans

80 S. Nasif et al. / Seminars in Cell & Developmental Biology 75 (2018) 78–87

Table 1Core NMD factors. Human (Homo sapiens, Hs) factors and their homologues in Saccharomyces cerevisiae (Sc), Caenorhabditis elegans (Ce), Drosophila melanogaster (Dm) andplants. *SMG1 homologues are present in all examined plants except Arabidopsis thaliana. **Dispensable for NMD in S. cerevisiae.Protein (Hs)

Protein (Hs) Functional description HomologuesRefs.

Sc Ce Dm Plants

UPF1 • Central NMD factor Upf1 SMG-2 UPF1 UPF1 [114]• ATP-dependent RNA helicase [115]• Binding platform for decay machinery [116]

[17]

UPF2 • Regulates the activation and helicase activity of UPF1 Upf2 SMG-3 UPF2 UPF2 [117]• Scaffold that brings together UPF1 and UPF3 [118]

UPF3A • Acts as an NMD repressor by antagonizing the formation of a complex between UPF2 and EJC

Upf3 SMG-4 UPF3 UPF3[40]• Can sometimes act as a weak NMD factor

• Up when UPF3B is low [118]

UPF3B • Brings together UPF1, UPF2 and the EJC

SMG1 • S/T kinase that phosphorylates UPF1 none SMG-1 SMG1 SMG1* [119][23]

SMG5 • Forms a heterodimer with SMG7 and is involved in exonucleolytic degradation of transcripts none SMG-5 SMG5 none [20]• Involved in UPF1 dephosphorylation [120]

SMG6 • Ribonuclease required for endonucleolytic degradation of transcripts none SMG-6 SMG6 none [15][16][19][18]

lytic d

wta

cihrtcptTbaglttmtd

isanubSsmtvaiNi

SMG7 • Forms a heterodimer with SMG5 and is involved in exonucleo

ith a wide range of intellectual disabilities [43,44–47], implyinghat UPF3B might be required for normal neuronal developmentnd/or function (see section 3.2).

There is evidence that some NMD factors function in additionalellular pathways apart from NMD. For example, UPF1 is involvedn telomere maintenance, cell cycle progression and degradation ofistone mRNAs (reviewed in [48]) and SMG1 and SMG6 were alsoeported to play a role in telomere maintenance [49]. Even thoughheir loss results in embryonic lethality, it still remains to be elu-idated if this can be attributed to inhibition of the NMD pathwayer se, or if, alternatively, all these KOs converge into similar pheno-ypes as a consequence of the inactivation of their non-NMD roles.here is evidence suggesting that NMD inactivation is responsi-le for the observed developmental defects. While SMG6 KO micere inviable, SMG6 KO mouse embryonic stem cells (mESC) can beenerated and cultured in vitro [39]. SMG6 deficient mESCs can pro-iferate normally and are indistinguishable from control mESCs, buthey fail to differentiate into all three germ layers (Fig. 1). Attemptso rescue SMG6−/− mESCs with either NMD-deficient or telomere

aintenance-deficient mutant forms of SMG6 showed that onlyhe expression of NMD-proficient versions of SMG6 relieved theifferentiation block of SMG6−/− mESCs [39].

Since NMD regulates the levels of many endogenous transcripts,t can be envisioned that it is the upregulation of one or severalpecific transcripts that renders the KO of NMD factors inviable,nd not the massive accumulation of faulty transcripts, as origi-ally thought. In the case of the differentiation blockade in micepon SMG6 KO, it was demonstrated that c-Myc mRNA is targetedy NMD and that elevated c-Myc is responsible for keeping theMG6−/− mESC from differentiating in vitro [39]. More evidence toupport this notion came from a suppressor screen in Drosophilaelanogaster seeking to restore viability of a NMD mutant, where

hey found that deletion of the Gadd45 coding region restored fulliability to Upf225G mutants (a partially viable, hypomorphic Upf2

llele), and partially suppressed the complete lethality observedn Upf1 and Upf2 null mutants [50]. The authors could show thatMD deficient flies have increased levels of cell death and this

s suppressed by Gadd45 elimination. Similarly, Gadd45b knock-

egradation of transcripts Ebs1** SMG-7 none SMG7 [20][24]

down was shown to partially rescue the reduction in cell viabilityobserved following Upf1 knockdown in mammalian cells. Theseresults suggest that the upregulation of Gadd45 is an importantcontributor – although probably not the only one – to lethality uponNMD inactivation [50].

Additional evidence that NMD can influence very early devel-opmental decisions comes from a study in human embryonic stemcells (hESCs). Lou and colleagues showed that UPF1 depletionaltered the expression of genes involved in several major signalingcascades and suggested that NMD promotes the differentiation ofhESCs into mesoderm and inhibits their differentiation into defini-tive endoderm by regulating the balance between TGF-� and BMPsignaling (Fig. 1) [51].

3. Roles for NMD factors in tissue-specific differentiationprograms

Conditional knockout (cKO) strategies were used to overcomethe early embryonic lethality and study the roles of NMD factorsat later stages during development, in a tissue-specific manner.Induced deletion of UPF2 in the hematopoietic system resulted incomplete loss of hematopoietic stem and progenitor cell popula-tions and death of the cKO mice (Fig. 1). However, UPF2 deletion inthe myeloid lineage had no phenotypic consequences, suggestingthat UPF2 has a vital role in proliferating, but not terminally differ-entiated, cells [37]. On the other hand, conditional ablation of UPF2from the developing liver is incompatible with postnatal life due tofailure of terminal differentiation. Furthermore, UPF2 is requiredfor adult normal liver homeostasis and liver regeneration [52].

3.1. Tight regulation of NMD activity is required for normalspermatogenesis

Several lines of evidence support a role for NMD in germ cell

development and male fertility (Fig. 2). Conditional ablation ofUPF2 from embryonic Sertoli cells (SC) – somatic cells that nourishsperm cells through spermatogenesis – in mice leads to testicularatrophy and male sterility due to depletion of SC and germ cells

S. Nasif et al. / Seminars in Cell & Developmental Biology 75 (2018) 78–87 81

Fig. 1. Roles of NMD factors in cellular differentiation. Diagram depicting cellular differentiation pathways, from pluripotent stem cells to terminally differentiated cell types.For simplicity, only some cell types are shown. The self-renewal capacity of stem cells is indicated by a circular arrow. Manipulation of the levels of NMD factors has beens with

i knockn

(tws(opUic(ootcltdUacoaltawpb

hown to affect some differentiation steps and this is depicted in the scheme, alongs represented in red, whereas stimulation is shown in green. (KO) knock out; (KD)eural stem cells.

GC) [53]. RNA-seq analysis of total testis revealed that UPF2 dele-ion causes a transcriptome-wide dysregulation of gene expression,ith approximately 30% of the upregulated genes containing a

top codon >50 nucleotides upstream of an exon–exon junctionFig. 2B) [53]. More recently, Bao and colleagues studied the rolef UPF2 in GC. They generated two different UPF2 cKO mice:rospermatogonia-specific UPF2 cKO and spermatocyte-specificPF2 cKO. In both cases, they found that adult UPF2 cKO males were

nfertile, exhibited a drastic reduction in testis size and that UPF2KO seminiferous tubules were almost completely devoid of GCFig. 2C) [54]. These two cKO studies show that UPF2 is essential notnly for the first wave of spermatogenesis during testicular devel-pment but also for the subsequent spermatogenic cycles in adultestes [54]. Surprisingly, transcriptome profiling showed that UPF2KO spermatocytes and round spermatids had a selective upregu-ation of transcripts with long 3′UTRs, whereas the classical NMDargets (those that harbor an exon–exon junction >50 nucleotidesownstream of the termination codon) were not affected [54].PF1 and UPF2 are highly expressed in post-meiotic germ cellsnd they are localized in germ cell-specific perinuclear granulesalled chromatoid bodies (CBs) [55]. While studying the functionf Tudor domain-containing protein 6 (TDRD6) in CBs, Fanourgakisnd colleagues found that in the absence of TDRD6, UPF1 fails toocalize to CBs and can no longer interact with UPF2. Resemblinghe phenotype observed for UPF2 cKOs, TDRD6−/− spermatocytes

nd round spermatids show a specific accumulation of transcriptsith long 3′UTRs (Fig. 2E) [55]. Taken together, these studies sup-ort the notion that in GC, transcripts with long 3′UTRs are targetedy NMD in a UPF2 and CB-dependent manner, whereas degra-

the species or cell type in which it was described. Inhibition of differentiation steps down; (Mm) Mus musculus; (hESC) human embryonic stem cells; (mNSC) mouse

dation of transcripts harboring an EJC in the 3′UTR is UPF2 andCB-independent. In mice, the testis is the only adult tissue withhigh UPF3A protein expression and recent evidence suggests thatUPF3A can work as an NMD repressor that sequesters UPF2 fromNMD substrates [40]. UPF3A cKO in meiotic GC resulted in micewith reduced sperm counts characterized by loss of later-stageGC. Consistent with UPF3A acting as a NMD repressor, many NMDsubstrate RNAs were downregulated in UPF3A cKO spermatocytes,when compared to controls (Fig. 2D) [40]. Altogether, these stud-ies highlight the importance of a tight control of NMD efficiency tosustain normal spermatogenesis.

3.2. NMD factors are required for neuronal development andphysiology

There is some indication that downregulation of NMD activityis required for neural differentiation. Lou and colleagues found thatthe levels of many NMD factors, and consequently NMD efficiency,are reduced during in vitro neural differentiation of mouse neuronalstem cells (mNSCs) and human neural progenitor cells (hNPCs) [56].Forced maintenance of UPF1 expression levels in differentiatingP19 cells blocked retinoic acid (RA)-induced neuronal differentia-tion, whereas UPF1 knockdown triggered neuronal differentiationin the absence of RA (Fig. 1) [56]. Analysis of UPF1-depleted P19 cellsrevealed that some proliferation inhibitors and pro-neural factors

are direct NMD targets, providing a molecular basis for the roleof UPF1 in sustaining the stem cell state and antagonizing neu-ral differentiation [56]. miR-128 is a brain-specific miRNA whoseexpression is dramatically increased during neural differentiation

82 S. Nasif et al. / Seminars in Cell & Developmental Biology 75 (2018) 78–87

Fig. 2. Tight regulation of NMD activity is required for normal spermatogenesis. (A) Scheme representing a cross section of a seminiferous tubule with one Sertoli cell (SC)and several germ cells (GC) in their differentiation stages from spermatogonia to spermatozoa. In round spermatids, UPF1 is localized in chromatoid bodies (CB). (LC) Leydigcells. Panels B-E show schematic representations of a cross section of a seminiferous tubule in different knock out (KO) mice. (B) Sertoli cell-specific UPF2 KO (SC-UPF2 KO)results in testicular atrophy with depletion of SC and GC. Analyses of total testis showed upregulation of RNAs with one exon–exon junction >50 nucleotides downstreamof the termination codon (dEJC) [53]. (C) Early stage germ cell-specific UPF2 KO (GC-UPF2 KO) results in depletion of GC and testicular atrophy. Purified germ cells showi c UPFd strucs d roun

b[idtcreaiX

pgiratcsCcrnoe

ncreased levels of transcripts with long 3′ UTR [54]. (D) Early stage germ cell-specifiecreased levels of transcripts with dEJC [40]. (E) TDRD6 KO results in distorted CBpermatids is observed and transcripts with long 3′ UTRs are upregulated in purifie

oth in vivo and in vitro and UPF1 is a direct target of miR-12857]. miR-128 and Upf1 RNA levels are inversely correlated dur-ng neuron maturation and differentiation in vitro and during brainevelopment in vivo [56,57]. Lou and colleagues found evidencehat in addition to miR-128 repressing UPF1 levels, NMD activityan indirectly downregulate the levels of miR-128. This mutuallyepressive circuit is predicted to lock the cells in either an undiffer-ntiated (with high NMD activity) or differentiated (with low NMDctivity) state (Fig. 3A) [56]. Remarkably, the UPF1/miR-128 circuits conserved and controls cell proliferation and differentiation inenopus laevis [56].

NMD has further been reported to influence commissural axonalathfinding in the spinal cord [58]. Normal commissural axonuidance depends on the expression of two alternatively splicedsoforms of the ROBO3 receptor, ROBO3.1 and ROBO3.2. The latteretains an intron that contains a stop codon, and is predicted to ben NMD target. Axonal growth cones are initially attracted towardshe midline floor plate, in a ROBO3.1-dependent manner. Afterrossing the midline, axons are repelled from the floor plate and thiswitch is accompanied by a sharp induction of ROBO3.2 expression.olak and colleagues found that ROBO3.2 mRNA is present in pre-rossing commissural axons in a translation repressed state that

enders it NMD-resistant. After crossing the midline, floor plate sig-als induce local translation of ROBO3.2 and allow NMD to ensuen the transcript. Neural growth cones of commissural neurons arenriched in UPF1, UPF2 and SMG1 proteins. When UPF2 was selec-

3A KO (GC-UPF3A KO) results in depletion of late stage GC. Purified germ cells showture and mislocalization of UPF1 to the perinuclear space. Depletion of elongatedd spermatids [55].

tively ablated from spinal commissural neurons, increased levelsof ROBO3.2 mRNA and protein were observed in post-crossingaxons, and this lead to abnormal axon guidance, characterized byincreased repulsion from the floor plate [58]. In addition, other tran-scripts of the Robo family were shown to be upregulated upon NMDimpairment, further supporting a role for NMD in axon guidance[57,59].

4. NMD is involved in the integrated stress response

Several stress-associated mRNAs are targeted by NMD inunstressed cells [18,32,35]. Interestingly, many of them escapeNMD degradation upon ER stress, amino acid starvation andhypoxia [60–64]. One such transcript is the one encoding the acti-vating transcription factor 4 (ATF4). ATF4 is the best characterizedeffector of the integrated stress response (ISR), an adaptive signal-ing pathway present in eukaryotes that is activated in response todiverse stress stimuli. Activation of ISR leads to phosphorylationof eIF2� and to a global attenuation of cap-dependent translationwith a concomitant increase in translation of selected stress-relatedmRNAs [65]. In unstressed cells, ATF4 mRNA is subjected to NMDdegradation because it contains several uORFs in its 5′ UTR. How-

ever, upon eIF2� phosphorylation, the main ORF in ATF4 mRNAis translated and ATF4 protein accumulates [66,67]. ATF4 increaseresults in the activation of a gene expression program that pro-motes cellular recovery.

S. Nasif et al. / Seminars in Cell & Developmental Biology 75 (2018) 78–87 83

Fig. 3. NMD is part of regulatory feedback loops. (A) UPF1 and miR-128 form a mutually repressive circuit that locks the cells in either the stem cell state or in a differentiatedstate. In neural stem cells, NMD activity is high and it indirectly downregulates miR-128 to lock the cells in the undifferentiated state (dashed arrow). Upon a neurogenicsignal, miR-128 is upregulated and it promotes neural differentiation by directly downregulating UPF1. MiR-128 direct repression of UPF1 locks the cells in the neuronalstate [56,57]. (B) NMD and the unfolded protein response (UPR) regulate each other to allow a proper ER stress response. In unstressed cells, NMD degrades transcriptsencoding components of the UPR, and prevents UPR activation upon innocuous stimuli. Upon strong ER stress the UPR is activated and, in turn, inhibits NMD to allow a robuststress-response. NMD-mediated inhibition of UPR is important during the termination phase, to protect the cells from stress-induced apoptosis [60]. (C) Chromatin target ofP . Excei ers tht

NioocwmaictwtifiEsiUiaeatItt

RMT1 (Chtop) levels are autoregulated in a homeostatic feedback loop by AS-NMDntron 2 retention and gives rise to a premature termination codon (PTC). This rendhe levels of functional Chtop [100].

By downregulating stress-responsive genes in unstressed cells,MD activity may prevent the triggering of a stress response upon

nnocuous stimuli. However, following exposure to strong stress-rs, specific mRNAs can evade NMD and are upregulated as partf the stress response. In support of this view, a bi-directionalontrol between NMD and the unfolded protein response (UPR)as recently reported [60]. The UPR is an adaptive intracellularechanism that is activated when misfolded or unfolded proteins

ccumulate in the lumen of the ER. Activation of the UPR resultsn changes in the transcriptional and translational landscape of theell, aimed at restoring ER homeostasis. When this is accomplished,he UPR is inactivated, to avoid programmed cell death that other-ise follows prolonged ER stress [68]. Karam and colleagues found

hat many mRNAs encoding UPR components are targeted by NMD,ncluding the conserved ER stress transducer IRE1�. Using loss-of-unction studies, the authors showed that NMD-deficiency resultsn exacerbated UPR activation upon normally innocuous levels ofR stress, suggesting that NMD activity increases the threshold oftress required to trigger the UPR [60]. In addition, NMD activ-ty was found to be important during the termination phase ofPR, to attenuate the UPR and protect the cells from ER stress-

nduced apoptosis. Not only can NMD repress the UPR, the UPR canlso inhibit NMD by promoting eIF2� phosphorylation [60]. Thisstablishes a bi-directional circuit in which NMD represses UPRctivation after harmless ER stress while, upon strong ER stress,

he UPR represses NMD to allow a robust stress response (Fig. 3B).n support of this feedback loop, Sieber and colleagues uncoveredhat NMD inhibition induces the upregulation de novo protein syn-hesis of downstream targets of the PERK and IRE1 branches of the

ss levels of Chtop result in binding of Chtop to its own pre-mRNA, which promotese Chtop alternatively spliced mRNA isoform susceptible to NMD, thereby reducing

UPR [69], whereas Li and colleagues found that thapsigargin treat-ment, which induces ER stress, inhibits NMD via activation of thePERK branch of the UPR [61].

Autophagy is a surveillance mechanism that rids the cells ofcytoplasmic deleterious proteins and also serves as a source ofamino acids during metabolic stress. Many cellular stressors, likeamino acid deprivation and hypoxia, inhibit NMD and activateautophagy, and an inverse correlation between these two pathwayswas reported [70]. Inhibition of NMD activates autophagy, in a par-tially ATF4-dependent manner, and increases intracellular aminoacid levels, while NMD hyperactivation blunts stress-inducedautophagy [70].

Furthermore, NMD is also involved in the mitigation of oxida-tive stress [63]. SLC7A11 is a subunit of the xCT cysteine/glutamateamino acid transport system, and its mRNA is a NMD targetwhose expression is upregulated in response to cellular stress[63]. The authors showed that NMD inhibition by stress-inducedeIF2� phosphorylation leads to stabilization of SLC7A11, increasein intracellular glutathione levels and establishment of an adaptiveresponse to oxidative stress.

It was recently reported that NMD inhibition is also part of thepro-apoptotic response to stress [71,72]. UPF1 and UPF2 are cleavedby caspases upon apoptotic insults, and the respective truncatedproteins have dominant negative effects. Their expression leads toupregulation of NMD targets and to induction of apoptosis [71,72].

This suggests that after an apoptotic insult, NMD inhibition andcaspase-induced UPF fragments reinforce the apoptotic response.Finally, it should be noted that some NMD factors, for example

8 evelo

St

5N

ipilradchiwlwsVSo(UTg

tdVAtfPaTL(aRcmtftfsc

aGPacutoAR1itt

4 S. Nasif et al. / Seminars in Cell & D

MG1 and SMG7, play roles in the DNA damage response that seemo be independent of NMD [73,74].

. RNA viruses are targeted by or able to evade the hostMD machinery

The multicistronic nature of viral genomes introduces featuresnto their transcripts that could trigger the host NMD pathway. Inarticular, the presence of uORFs results in stop codons in positions

n which they would be expected to activate NMD. Similarly, trans-ation of the first ORF of the genome of a plus-strand RNA virusesembles the situation of translating an mRNA with a long 3′ UTR,nother feature that may elicit NMD. Indeed, a number of studiesemonstrating the degradation of viral transcripts by NMD haveome to light, suggesting that NMD could represent part of theost cell innate immune response. Such studies have been done

n plants, insects, as well as mammalian cell systems. In plants, itas shown that NMD targets two subgenomic RNAs − both with

ong 3′ UTRs − of potato virus X (PVX) and that restriction by NMDas dependent on the length of the 3′ UTR [75]. In addition, NMD

uppression in tobacco plants led to an increase in Turnip Crinkleirus (TCV) genomic RNA [75]. The NMD factors, UPF1, SMG5 andMG7 were shown to be involved in restricting viral replicationf two alphaviruses, Semliki Forest Virus (SFV) and Sindbis VirusSINV), in mammalian cells [76]. Surprisingly, shortening of the 3′

TR of SFV still rendered the viral genome sensitive to NMD [76].he mechanism by which NMD recognizes and degrades the SFVenome therefore remains unclear.

Since viral genomes present targets for NMD, it is unsurprisinghat viruses have evolved mechanisms to protect them from degra-ation by the host cell NMD machinery. For example, Hepatitis Cirus (HCV) was shown to inhibit NMD in hepatoma cell lines [77].lthough the complete mechanism remains unclear, NMD inhibi-

ion by HCV infection was thought to involve the direct interactionound between the HCV core protein and the EJC recycling factorYM [77]. Mechanisms to evade NMD have also been reported for

number of retroviruses, including Human T-lymphotropic Virusype 1 (HTLV-1), Rous Sarcoma Virus (RSV), and Moloney Murineeukemia Virus (MoMLV), [78–82]. HTLV-1 encodes viral proteinsTAX and REX) that were demonstrated to inhibit NMD by bindingnd sequestering components of the pathway [78,83]. Resistance ofSV to NMD was found to be conferred by a 400 nucleotide stretch,ontaining 11 CU-rich clusters, termed the RSV RNA stability ele-ent (RSE) [79,80]. The RSE, located immediately downstream of

he stop codon of the first ORF (gag), was found to be a binding siteor PTBP1 [81]. Ge and colleagues showed that PTBP1 binding tohe RSE prevents recruitment of UPF1, thereby inhibiting NMD, andurther demonstrated that insertion of PTBP1 binding sites down-tream of the stop codon was sufficient to rescue NMD-sensitiveellular mRNAs from degradation [81].

Proper viral assembly of MoMLV, as for all retroviruses, requires balance between the production of polyproteins (encoded by theag gene) and viral reverse transcriptase (RT) (encoded by a Gag-ol fusion protein). In the MoMLV genome, the Gag and Pol genesre encoded in the same reading frame, separated by a UAG stopodon [82]. To allow controlled expression of the Pol gene, MoMLVses strategies to suppress the stop codon between the two genes,hereby promoting read-through. One such strategy is the presencef an RNA pseudoknot downstream of the UAG stop codon [84,85].

recent study that determined the crystal structure of MoMLVT in complex with full length mouse eukaryotic release factor

(eRF1) followed by functional assays, revealed that the bind-ng of MoMLV RT to eRF1 competitively inhibits binding of eRF3,hereby suppressing translation termination and promoting read-hrough [82]. Using a dual-fluorescent reporter system, the authors

pmental Biology 75 (2018) 78–87

demonstrated that the RNase H domain of MoMLV RT enhancedread-through above RNA pseudoknot-directed read-through [82].In addition, mRNA decay assays revealed that the RNase H domainof MoMLV RT was also involved in mRNA stability, suggesting thatit has a protective role against NMD. The authors noted that sinceUPF1 interacts with eRF1 and eRF3 to trigger NMD [11,86], dis-ruption of the interaction of UPF1 with this complex, by MoMLVRT binding, could be the mechanism by which the viral mRNA isprotected from the NMD machinery [82].

Intriguingly, a number of reports have shown that HIV, in con-trast to the above-mentioned studies, requires UPF1 for infectivityand nuclear export of viral RNA [87–89]. The role of UPF1 in HIVinfection was shown to be independent of UPF2, indicating anNMD-independent role for UPF1 in promoting HIV infectivity [87].

6. Alternative splicing coupled with NMD (AS-NMD)modulates post-transcriptional gene expression

Mammalian cells employ extensive alternative pre-mRNA splic-ing to generate multiple mRNA isoforms produced from a singlegene. Differentially spliced mRNA isoforms may give rise tochanges in protein structure and function, thereby contributing toorganism- or tissue-specific proteome diversity. However, differen-tial splicing may also generate transcript isoforms with PTCs, whichare recognized and degraded by NMD.

Many RNA binding proteins, including splicing regulators them-selves, use AS-NMD as a mechanism to regulate their ownabundance in a homeostatic feedback loop. This was shown formost SR proteins in a number of independent studies [18,90,91].Additionally, many hnRNPs were also suggested to be autoregu-lated by AS-NMD via homeostatic feedback loops [91]. For example,Rossbach and colleagues demonstrated that inclusion of exon 6a inhuman hnRNP L introduces a PTC, rendering the isoform sensitiveto NMD [92]. Binding of hnRNP L to an intronic region imme-diately upstream of exon 6a promotes inclusion of exon 6a. Anexcess of hnRNP L therefore triggers NMD of its own mRNA [92].Polypyrimidine tract binding protein (PTB), an important regulatorof alternative splicing, presents another example of autoregulationby AS-NMD. High levels of PTB promotes exon 11 skipping, whichresults in an NMD-sensitive PTB isoform [93].

PTB proteins also regulate alternative splicing events of othermRNA transcripts, some of which produce isoforms susceptibleto NMD [94,95]. Such is the case for HPS1, a subunit of a gua-nine nucleotide exchange factor (GEF) essential for biogenesis oflysosome-related organelles. The strong correlation of expressionlevels of HPS1 and PTB protein 1 (PTBP1) across mammalian tis-sues is indicative of this regulatory circuit [94]. Similarly, PTBP1levels regulate alternative splicing of PSD-95, a postsynaptic scaf-fold protein important in glutamatergic synapses in neurons [95].High levels of expression of PTBP1 and PTBP2 in non-neural cellsresult in exon 18 skipping, causing a frameshift and hence a PTCin exon 19, thereby rendering this transcript isoform sensitive toNMD. This results in low levels of abundance of functional PSD-95in non-neural cells [95]. The proportion of exon 18 inclusion wasshown to increase upon differentiation of hESCs into hNPCs andthen into neurons, suggesting that PSD-95 exon 18 inclusion, andhence functional PSD-95 expression, is neural-specific in humancells [95]. Several other studies have also reported a role for AS-NMD in regulating cellular differentiation programs [96–98].

More recently, proteins involved in chromatin modificationwere also identified as being regulated by AS-NMD. This was

revealed through deep RNA-sequencing analysis of highly con-served AS-NMD exons between human and mouse brain cortex[99]. In line with this, a recent study revealed that expressionof Chromatin target of PRMT1 (Chtop) is regulated by AS-NMD

evelo

vCoPimSNatf

7

pNfSft[wfmtsa[

8t

UalmliTscf[rrCaLsplRlsti

g([rma

S. Nasif et al. / Seminars in Cell & D

ia an autoregulatory homeostatic feedback loop (Fig. 3C) [100].htop was shown to bind its own mRNA to control retentionf intron 2 during splicing. Retention of intron 2 results in aTC, which renders Chtop mRNA susceptible to NMD [100]. It ismportant to note that AS-NMD has also been reported in non-

ammalian systems, including plants [101], zebrafish [102] and. cerevisiae [103]. As functional and mechanistic evidence for AS-MD continues to emerge, it is becoming clearer that AS-NMD is

biologically relevant post-transcriptional mechanism that fine-unes gene expression by modulating the overall abundance ofunctional protein coding isoforms.

. NMD factors are autoregulated by the NMD pathway

Independent studies by Huang and colleagues [42] and Yepisko-osyan and colleagues [34] revealed that suppression of NMD byMD factor knockdowns resulted in a significant increase in NMD

actor mRNA levels of UPF1, UPF2, UPF3B, SMG1, SMG5, SMG6 andMG7. An autoregulatory feedback loop for the regulation of NMDactors by NMD was thus identified. All seven NMD factor mRNAshat were identified possess long 3‘ UTRs, and some possess uORFs34,42]. The 3′ UTRs of UPF1, SMG5, SMG7 and the uORF of SMG5ere shown to induce degradation of reporter constructs, there-

ore identifying them as the NMD-triggering features [30,34,42].RNAs encoding NMD factors have also been identified as NMD

argets in other organisms, including SMG7 and UPF3 in Arabidop-is thaliana and SMG5 in Drosophila melanogaster, suggesting theutoregulatory NMD mechanism may be evolutionarily conserved104–106].

. NMD factor mutations in human diseases and theherapeutic potential of modulating NMD activity

As mentioned above, loss-of-function mutations in the humanPF3B gene are associated with diverse forms of intellectual dis-bility (ID) [43–47]. Molecular analysis of lymphoblastoid cellines (LCLs) from ID patients showed that UPF3B loss-of-function

utations cause a strong reduction in UPF3B mRNA and proteinevels, with a concomitant increase in UPF3A protein levels thatnversely correlates to the severity of the patients’ phenotype [107].ranscriptome analysis showed that approximately 5% of the tran-criptome was affected in UPF3B mutant patients compared toontrol individuals and GO term analysis showed that some dif-erentially regulated transcripts are related to neuronal processes107]. In addition, UPF3B ablation in primary hippocampal neu-ons or expression of disease-related UPF3B missense variants inat neuronal stem cells resulted in altered neurite growth [59,108].opy number variants of UPF3A, SMG6 and UPF2 are significantlyssociated with neurodevelopmental disorders [107]. Moreover,CLs from ID patients with heterozygous deletion of UPF2 genehow a transcriptome profile that resembles that of UPF3B mutantatients, suggesting NMD deregulation as a common molecu-

ar mechanism for these neuro-developmental disorders [109].ecently, a multiple congenital anomaly syndrome was found to be

inked to loss-of-function mutations in SMG9, and global expres-ion analysis showed that LCLs from affected individuals have aranscriptome profile that is significantly different to that of controlndividuals [41].

Two recent studies report that somatic mutations in the UPF1ene are frequently found in pancreatic adenosquamous carcinomaASC) [110] and in inflammatory myofibroblatic tumors (IMTs)

111]. In both cases, the mutations reduce UPF1 protein levels andesult in impaired NMD in the tumor tissue relative to adjacent nor-al tissue. In IMTs, it was proposed that downregulation of NMD

ctivity results in increased levels of chemokines that would in turn

pmental Biology 75 (2018) 78–87 85

promote the immune cell infiltration that is characteristic in IMTs[111].

NMD can have either a protective or a harmful role in diseasepathology. In cases where the PTC-containing transcript gives riseto a truncated protein with dominant negative activities, degrada-tion of the faulty transcript by NMD supposes a beneficial role forthe cell. There are however cases, in which expression of a truncatedprotein would help to ameliorate the phenotypic consequences ofa disease-causing mutation and, in these cases, bypassing NMDwould serve as a therapeutic approach. There have been severalattempts aimed at using read-through therapies alone or in com-bination with NMD inhibition to treat genetic disorders (reviewedin [112]). Yet, since NMD can shape the transcriptome in differ-ent cell types, general inhibition of NMD may have unforeseeableside effects. An alternative would be to inhibit NMD in a gene-specific way, as was recently reported [113]. This study showed thatantisense oligonucleotides (ASO) that block EJC deposition in themRNA downstream of a PTC can be used to inhibit NMD in a gene-specific manner. Moreover, combination of ASO treatment withread-through promoting compounds restored the expression of afull-length protein from a PTC containing transcript [113]. Althoughthe therapeutic value of gene-specific NMD inhibition still remainsto be tested, it emerges as an elegant and highly specific way totreat diseases caused by nonsense mutations.

9. Concluding remark

As described in this review, NMD exerts important biologicalfunctions that reach beyond mRNA surveillance. Although we arestill in the beginning of unravelling the different regulatory cir-cuits in which NMD is implicated, the existing evidence suggeststhat NMD affects gene expression in a diverse range of biologicalcontexts. The unifying principle in all these cases appears to be theoccurrence of stop codons in mRNP environments that are not opti-mal for fast enough and/or correct translation termination, leadingto the degradation of the respective mRNA. This “survival of thefittest mRNA” bears reminiscence to a recurring and fundamentalconcept in biology.

Acknowledgments

The research in the lab of OM is supported by the NCCR RNA &Disease funded by the Swiss National Science Foundation (SNSF),by the SNSF grant31003A-162986, and by the canton of Bern.

References

[1] R. Losson, F. Lacroute, Interference of nonsense mutations with eukaryoticmessenger RNA stability, Proc. Natl. Acad. Sci. U. S. A. 76 (10) (1979)5134–5137.

[2] L.E. Maquat, et al., Unstable beta-globin mRNA in mRNA-deficient beta othalassemia, Cell 27 (3 Pt. 2) (1981) 543–553.

[3] S.W. Peltz, et al., mRNA destabilization triggered by premature translationaltermination depends on at least three cis-acting sequence elements and onetrans-acting factor, Genes Dev. 7 (9) (1993) 1737–1754.

[4] P. Leeds, et al., The product of the yeast UPF1 gene is required for rapidturnover of mRNAs containing a premature translational termination codon,Genes Dev. 5 (12A) (1991) 2303–2314.

[5] R. Pulak, P. Anderson, mRNA surveillance by the Caenorhabditis elegans smggenes, Genes Dev. 7 (10) (1993) 1885–1897.

[6] A. Yamashita, et al., SMG-8 and SMG-9, two novel subunits of the SMG-1complex, regulate remodeling of the mRNA surveillance complex duringnonsense-mediated mRNA decay, Genes Dev. 23 (9) (2009) 1091–1105.

[7] D. Longman, et al., Mechanistic insights and identification of two novelfactors in the C elegans NMD pathway, Genes Dev. 21 (9) (2007) 1075–1085.

[8] A. Casadio, et al., Identification and characterization of novel factors that act

in the nonsense-mediated mRNA decay pathway in nematodes, flies andmammals, EMBO Rep. 16 (January (1)) (2015) 71–78.

[9] A. Alexandrov, et al., Fluorescence amplification method for forward geneticdiscovery of factors in human mRNA degradation, Mol. Cell 65 (1) (2017)191–201.

8 evelo

6 S. Nasif et al. / Seminars in Cell & D

[10] A.B. Eberle, et al., Posttranscriptional gene regulation by spatialrearrangement of the 3’ untranslated region, PLoS Biol. 6 (4) (2008) e92.

[11] P.V. Ivanov, et al., Interactions between UPF1, eRFs, PABP and the exonjunction complex suggest an integrated model for mammalian NMDpathways, EMBO J. 27 (5) (2008) 736–747.

[12] P.M. Moriarty, et al., Selenium deficiency reduces the abundance of mRNAfor Se-dependent glutathione peroxidase 1 by a UGA-dependent mechanismlikely to be nonsense codon-mediated decay of cytoplasmic mRNA, Mol.Cell. Biol. 18 (5) (1998) 2932–2939.

[13] A. Zupanic, et al., Modeling and gene knockdown to assess the contributionof nonsense-mediated decay, premature termination, and selenocysteineinsertion to the selenoprotein hierarchy, RNA 22 (7) (2016) 1076–1084.

[14] I. Kashima, et al., Binding of a novel SMG-1-Upf1-eRF1-eRF3 complex (SURF)to the exon junction complex triggers Upf1 phosphorylation andnonsense-mediated mRNA decay, Genes Dev. 20 (3) (2006) 355–367.

[15] A.B. Eberle, et al, SMG6 promotes endonucleolytic cleavage of nonsensemRNA in human cells, Nat. Struct. Mol. Biol. 16 (1) (2009) 49–55.

[16] E. Huntzinger, et al., SMG6 is the catalytic endonuclease that cleaves mRNAscontaining nonsense codons in metazoan, RNA 14 (12) (2008) 2609–2617.

[17] Y. Okada-Katsuhata, et al., N- and C-terminal Upf1 phosphorylations createbinding platforms for SMG-6 and SMG-5:SMG-7 during NMD, Nucleic AcidsRes. 40 (3) (2012) 1251–1266.

[18] S. Lykke-Andersen, et al., Human nonsense-mediated RNA decay initiateswidely by endonucleolysis and targets snoRNA host genes, Genes Dev. 28(22) (2014) 2498–2517.

[19] V. Boehm, et al., 3′UTR length and messenger ribonucleoproteincomposition determine endocleavage efficiencies at termination codons,Cell Rep. 9 (2) (2014) 555–568.

[20] B. Loh, et al., The SMG5-SMG7 heterodimer directly recruits the CCR4-NOTdeadenylase complex to mRNAs containing nonsense codons via interactionwith POP2, Genes Dev. 27 (19) (2013) 2125–2138.

[21] O. Muhlemann, et al., Recognition and elimination of nonsense mRNA,Biochim. Biophys. Acta 1779 (2008) 538–549.

[22] C. Lasalde, et al., Identification and functional analysis of novelphosphorylation sites in the RNA surveillance protein Upf1, Nucleic AcidsRes. 42 (3) (2014) 1916–1929.

[23] J.P. Lloyd, B. Davies, SMG1 is an ancient nonsense-mediated mRNA decayeffector, Plant J. 76 (5) (2013) 800–810.

[24] B. Luke, et al., Saccharomyces cerevisiae Ebs1 p is a putative ortholog ofhuman Smg7 and promotes nonsense-mediated mRNA decay, Nucleic AcidsRes. 35 (22) (2007) 7688–7697.

[25] E.D. Karousis, et al., Nonsense-mediated mRNA decay: novel mechanisticinsights and biological impact, Wiley Interdiscip. Rev. RNA 7 (5) (2016)661–682.

[26] F. He, A. Jacobson, Nonsense-mediated mRNA decay: degradation ofdefective transcripts is only part of the story, Annu. Rev. Genet. 49 (2015)339–366.

[27] T. Fatscher, et al., Mechanism, factors, and physiological role ofnonsense-mediated mRNA decay, Cell. Mol. Life Sci. 72 (23) (2015)4523–4544.

[28] E. Nagy, L.E. Maquat, A rule for termination-codon position withinintron-containing genes: when nonsense affects RNA abundance, TrendsBiochem. Sci. 23 (6) (1998) 198–199.

[29] M. Buhler, et al., EJC-independent degradation of nonsenseimmunoglobulin-mu mRNA depends on 3’ UTR length, Nat. Struct. Mol. Biol.13 (5) (2006) 462–464.

[30] G. Singh, et al., A competition between stimulators and antagonists of Upfcomplex recruitment governs human nonsense-mediated mRNA decay,PLoS Biol. 6 (4) (2008) e111.

[31] J.T. Mendell, et al., Nonsense surveillance regulates expression of diverseclasses of mammalian transcripts and mutes genomic noise, Nat. Genet. 36(10) (2004) 1073–1078.

[32] H. Tani, et al., Identification of hundreds of novel UPF1 target transcripts bydirect determination of whole transcriptome stability, RNA Biol. 9 (11)(2012) 1370–1379.

[33] J. Wittmann, et al., hUPF2 silencing identifies physiologic substrates ofmammalian nonsense-mediated mRNA decay, Mol. Cell. Biol. 26 (4) (2006)1272–1287.

[34] H. Yepiskoposyan, et al., Autoregulation of the nonsense-mediated mRNAdecay pathway in human cells, RNA 17 (12) (2011) 2108–2118.

[35] M. Colombo, et al., Transcriptome-wide identification of NMD-targetedhuman mRNAs reveals extensive redundancy between SMG6- andSMG7-mediated degradation pathways, RNA 23 (2) (2017) 189–201.

[36] S.M. Medghalchi, et al., Rent1, a trans-effector of nonsense-mediated mRNAdecay, is essential for mammalian embryonic viability, Hum. Mol. Genet. 10(2) (2001) 99–105.

[37] J. Weischenfeldt, et al., NMD is essential for hematopoietic stem andprogenitor cells and for eliminating by-products of programmed DNArearrangements, Genes Dev. 22 (10) (2008) 1381–1396.

[38] D.R. McIlwain, et al., Smg1 is required for embryogenesis and regulatesdiverse genes via alternative splicing coupled to nonsense-mediated mRNA

decay, Proc. Natl. Acad. Sci. U. S. A. 107 (27) (2010) 12186–12191.

[39] T. Li, et al., Smg6/Est1 licenses embryonic stem cell differentiation vianonsense-mediated mRNA decay, EMBO J. 34 (12) (2015) 1630–1647.

[40] E.Y. Shum, et al., The antagonistic gene paralogs Upf3 a and Upf3 b governnonsense-mediated RNA decay, Cell 165 (2) (2016) 382–395.

pmental Biology 75 (2018) 78–87

[41] R. Shaheen, et al., Mutations in SMG9, encoding an essential component ofnonsense-mediated decay machinery, cause a multiple congenital anomalysyndrome in humans and mice, Am. J. Hum. Genet. 98 (4) (2016) 643–652.

[42] L. Huang, et al., RNA homeostasis governed by cell type-specific andbranched feedback loops acting on NMD, Mol. Cell 43 (6) (2011) 950–961.

[43] F. Laumonnier, et al., Mutations of the UPF3B gene, which encodes a proteinwidely expressed in neurons, are associated with nonspecific mentalretardation with or without autism, Mol. Psychiatry 15 (7) (2010) 767–776.

[44] P.S. Tarpey, et al., Mutations in UPF3 B, a member of the nonsense-mediatedmRNA decay complex, cause syndromic and nonsyndromic mentalretardation, Nat. Genet. 39 (9) (2007) 1127–1133.

[45] S.A. Lynch, et al., Broadening the phenotype associated with mutations inUPF3 B: two further cases with renal dysplasia and variable developmentaldelay, Eur. J. Med. Genet. 55 (8–9) (2012) 476–479.

[46] A.M. Addington, et al., A novel frameshift mutation in UPF3B identified inbrothers affected with childhood onset schizophrenia and autism spectrumdisorders, Mol. Psychiatry 16 (3) (2010) 238–239.

[47] X. Xu, et al., Exome sequencing identifies UPF3B as the causative gene for aChinese non-syndrome mental retardation pedigree, Clin. Genet. 83 (6)(2013) 560–564.

[48] N. Imamachi, et al., Up-frameshift protein 1 (UPF1): multitalentedentertainer in RNA decay, Drug Discov. Ther. 6 (2) (2012) 55–61.

[49] C.M. Azzalin, et al., Telomeric repeat containing RNA and RNA surveillancefactors at mammalian chromosome ends, Science 318 (5851) (2007)798–801.

[50] J.O. Nelson, et al., Degradation of Gadd45 mRNA by nonsense-mediateddecay is essential for viability, Elife 5 (2016).

[51] C.-H. Lou, et al., Nonsense-mediated RNA decay influences humanembryonic stem cell fate, Stem Cell Rep. 6 (6) (2016) 844–857.

[52] L.A. Thoren, et al., UPF2 is a critical regulator of liver development, functionand regeneration, PLoS One 5 (7) (2010), e11650.

[53] J. Bao, et al., UPF2, a nonsense-mediated mRNA decay factor, is required forprepubertal Sertoli cell development and male fertility by ensuring fidelityof the transcriptome, Development 142 (2) (2015) 352–362.

[54] J. Bao, et al., UPF2-dependent nonsense-Mediated mRNA decay pathway isessential for spermatogenesis by selectively eliminating longer 3’UTRtranscripts, PLoS Genet. 12 (5) (2016) e1005863.

[55] G. Fanourgakis, et al., Chromatoid body protein TDRD6 supports long 3’ UTRtriggered nonsense mediated mRNA decay, PLoS Genet. 12 (5) (2016),e1005857.

[56] C.H. Lou, et al., Posttranscriptional control of the stem cell and neurogenicprograms by the nonsense-mediated RNA decay pathway, Cell Rep. 6 (4)(2014) 748–764.

[57] I.G. Bruno, et al., Identification of a microRNA that activates gene expressionby repressing nonsense-mediated RNA decay, Mol. Cell 42 (4) (2011)500–510.

[58] D. Colak, et al., Regulation of axon guidance by compartmentalizednonsense-mediated mRNA decay, Cell 153 (6) (2013) 1252–1265.

[59] L.A. Jolly, et al., The UPF3B gene, implicated in intellectual disability, autism,ADHD and childhood onset schizophrenia regulates neural progenitor cellbehaviour and neuronal outgrowth, Hum. Mol. Genet. 22 (23) (2013)4673–4687.

[60] R. Karam, et al., The unfolded protein response is shaped by the NMDpathway, EMBO Rep. 16 (5) (2015) 599–609.

[61] Z. Li, et al., Inhibition of nonsense-mediated RNA decay by ER stress, RNA 23(3) (2017) 378–394.

[62] L.B. Gardner, Hypoxic inhibition of nonsense-mediated RNA decay regulatesgene expression and the integrated stress response, Mol. Cell. Biol. 28 (11)(2008) 3729–3741.

[63] L. Martin, L.B. Gardner, Stress-induced inhibition of nonsense-mediated RNAdecay regulates intracellular cystine transport and intracellular glutathionethrough regulation of the cystine/glutamate exchanger SLC7 A11, Oncogene34 (32) (2015) 4211–4218.

[64] D. Wang, et al., Inhibition of nonsense-mediated RNA decay by the tumormicroenvironment promotes tumorigenesis, Mol. Cell. Biol. 31 (17) (2011)3670–3680.

[65] K. Pakos-Zebrucka, et al., The integrated stress response, EMBO Rep. 17 (10)(2016) 1374–1395.

[66] K.M. Vattem, R.C. Wek, Reinitiation involving upstream ORFs regulates ATF4mRNA translation in mammalian cells, Proc. Natl. Acad. Sci. U. S. A. 101 (31)(2004) 11269–11274.

[67] C.P. Chan, et al., Internal ribosome entry site-mediated translationalregulation of ATF4 splice variant in mammalian unfolded protein response,Biochim. Biophys. Acta 1833 (10) (2013) 2165–2175.

68 K.A. Moore, J. Hollien, The unfolded protein response in secretory cellfunction, Annu. Rev. Genet. 46 (2012) 165–183.

[69] J. Sieber, et al., Proteomic analysis reveals branch-specific regulation of theunfolded protein response by nonsense-mediated mRNA decay, Mol. Cell.Proteomics 15 (5) (2016) 1584–1597.

[70] J. Wengrod, et al., Inhibition of nonsense-mediated RNA decay activatesautophagy, Mol. Cell. Biol. 33 (11) (2013) 2128–2135.

[71] J. Jia, et al., Caspases shutdown nonsense-mediated mRNA decay duringapoptosis, Cell Death Differ. 22 (November (11)) (2015) 1754–1763.

[72] M.W. Popp, L.E. Maquat, Attenuation of nonsense-mediated mRNA decayfacilitates the response to chemotherapeutics, Nat Commun. 6 (2015) 6632.

evelo

surveillance complex and is involved in the regulation ofnonsense-mediated mRNA decay, Genes Dev. 15 (17) (2001) 2215–2228.

S. Nasif et al. / Seminars in Cell & D

[73] K.M. Brumbaugh, et al., The mRNA surveillance protein hSMG-1 functions ingenotoxic stress response pathways in mammalian cells, Mol. Cell 14 (5)(2004) 585–598.

[74] H. Luo, et al., SMG7 is a critical regulator of p53 stability and function inDNA damage stress response, Cell Disco. 2 (2016) 15042.

[75] D. Garcia, et al., Nonsense-mediated decay serves as a general viralrestriction mechanism in plants, Cell Host Microbe 16 (3) (2014) 391–402.

[76] G. Balistreri, et al., The host nonsense-mediated mRNA decay pathwayrestricts Mammalian RNA virus replication, Cell Host Microbe 16 (3) (2014)403–411.

[77] H.R. Ramage, et al., A combined proteomics/genomics approach linkshepatitis C virus infection with nonsense-mediated mRNA decay, Mol. Cell57 (2) (2015) 329–340.

[78] K. Nakano, et al., Viral interference with host mRNA surveillance, thenonsense-mediated mRNA decay (NMD) pathway, through a new functionof HTLV-1 Rex: implications for retroviral replication, Microbes Infect. 15(6–7) (2013) 491–505.

[79] J.E. Weil, K.L. Beemon, A 3’ UTR sequence stabilizes termination codons inthe unspliced RNA of Rous sarcoma virus, RNA 12 (1) (2006) 102–110.

[80] J.B. Withers, K.L. Beemon, The structure and function of the rous sarcomavirus RNA stability element, J. Cell. Biochem. 112 (11) (2011) 3085–3092.

[81] Z. Ge, et al., Polypyrimidine tract binding protein 1 protects mRNAs fromrecognition by the nonsense-mediated mRNA decay pathway, Elife 5 (2016)e11155.

[82] X. Tang, et al., Structural basis of suppression of host translation terminationby Moloney Murine Leukemia Virus, Nat Commu. 7 (2016) 12070.

[83] V. Mocquet, et al., The human T-Lymphotropic virus type 1 tax proteininhibits nonsense-mediated mRNA decay by interacting with INT6/EIF3 eand UPF1, J. Virol. 86 (14) (2012) 7530–7543.

[84] E.B. ten Dam, et al., RNA pseudoknots: translational frameshifting andreadthrough on viral RNAs, Virus Genes 4 (2) (1990) 121–136.

[85] Y.X. Feng, et al., Bipartite signal for read-through suppression in murineleukemia virus mRNA: an eight-nucleotide purine-rich sequenceimmediately downstream of the gag termination codon followed by an RNApseudoknot, J. Virol. 66 (8) (1992) 5127–5132.

[86] K. Czaplinski, et al., The surveillance complex interacts with the translationrelease factors to enhance termination and degrade aberrant mRNAs, GenesDev. 12 (11) (1998) 1665–1677.

[87] L. Ajamian, et al., HIV-1 recruits UPF1 but excludes UPF2 to promotenucleocytoplasmic export of the genomic RNA, Biomolecules 5 (4) (2015)2808–2839.

[88] L. Ajamian, et al., Unexpected roles for UPF1 in HIV-1 RNA metabolism andtranslation, RNA 14 (5) (2008) 914–927.

[89] A.K. Serquina, et al., UPF1 is crucial for the infectivity of humanimmunodeficiency virus type 1 progeny virions, J. Virol. 87 (16) (2013)8853–8861.

[90] L.F. Lareau, et al., Unproductive splicing of SR genes associated with highlyconserved and ultraconserved DNA elements, Nature 446 (7138) (2007)926–929.

[91] J.Z. Ni, et al., Ultraconserved elements are associated with homeostaticcontrol of splicing regulators by alternative splicing and nonsense-mediateddecay, Genes Dev. 21 (6) (2007) 708–718.

[92] O. Rossbach, et al., Auto- and cross-regulation of the hnRNP L proteins byalternative splicing, Mol. Cell. Biol. 29 (6) (2009) 1442–1451.

[93] M.C. Wollerton, et al., Autoregulation of polypyrimidine tract bindingprotein by alternative splicing leading to nonsense-mediated decay, Mol.Cell 13 (1) (2004) 91–100.

[94] F.M. Hamid, E.V. Makeyev, Regulation of mRNA abundance bypolypyrimidine tract-binding protein-controlled alternate 5’ splice sitechoice, PLoS Genet. 10 (11) (2014), e1004771.

[95] S. Zheng, Alternative splicing and nonsense-mediated mRNA decay enforceneural specific gene expression, Int. J. Dev. Neurosci. 55 (December) (2016)102–108.

[96] U. Braunschweig, et al., Widespread intron retention in mammalsfunctionally tunes transcriptomes, Genome Res. 24 (11) (2014) 1774–1786.

pmental Biology 75 (2018) 78–87 87

[97] H. Pimentel, et al., A dynamic alternative splicing program regulates geneexpression during terminal erythropoiesis, Nucleic Acids Res. 42 (6) (2014)4031–4042.

[98] J.J. Wong, et al., Orchestrated intron retention regulates normal granulocytedifferentiation, Cell 154 (3) (2013) 583–595.

[99] Q. Yan, et al., Systematic discovery of regulated and conserved alternativeexons in the mammalian brain reveals NMD modulating chromatinregulators, Proc. Natl. Acad. Sci. U. S. A. 112 (11) (2015) 3445–3450.

[100] K. Izumikawa, et al., Chtop (Chromatin target of Prmt1) auto-regulates itsexpression level via intron retention and nonsense-mediated decay of itsown mRNA, Nucleic Acids Res. 44 (20) (2016) 9847–9859.

[101] G. Drechsel, et al., Nonsense-mediated decay of alternative precursor mRNAsplicing variants is a major determinant of the Arabidopsis steady statetranscriptome, Plant Cell 25 (10) (2013) 3726–3742.

[102] D. Longman, et al., DHX34 and NBAS form part of an autoregulatory NMDcircuit that regulates endogenous RNA targets in human cells, zebrafish andCaenorhabditis elegans, Nucleic Acids Res. 41 (17) (2013) 8319–8331.

[103] T. Kawashima, et al., Widespread use of non-productive alternative splicesites in Saccharomyces cerevisiae, PLoS Genet. 10 (4) (2014), e1004249.

[104] Z. Kerenyi, et al., Inter-kingdom conservation of mechanism ofnonsense-mediated mRNA decay, EMBO J. 27 (11) (2008) 1585–1595.

[105] J. Rehwinkel, et al., Nonsense-mediated mRNA decay factors act in concertto regulate common mRNA targets, RNA 11 (10) (2005) 1530–1544.

[106] H. Saul, et al., The upstream open reading frame of the Arabidopsis AtMHXgene has a strong impact on transcript accumulation through thenonsense-mediated mRNA decay pathway, Plant J. 60 (6) (2009) 1031–1042.

[107] L.S. Nguyen, et al., Transcriptome profiling of UPF3 B/NMD-deficientlymphoblastoid cells from patients with various forms of intellectualdisability, Mol. Psychiatry 17 (11) (2012) 1103–1115.

[108] T. Alrahbeni, et al., Full UPF3B function is critical for neuronal differentiationof neural stem cells, Mol Brain 8 (2015) 33.

[109] L.S. Nguyen, et al., Contribution of copy number variants involvingnonsense-mediated mRNA decay pathway genes to neuro-developmentaldisorders, Hum. Mol. Genet. 22 (9) (2013) 1816–1825.

[110] C. Liu, et al., The UPF1 RNA surveillance gene is commonly mutated inpancreatic adenosquamous carcinoma, Nat. Med. 20 (6) (2014) 596–598.

[111] J. Lu, et al., The nonsense-mediated RNA decay pathway is disrupted ininflammatory myofibroblastic tumors, J. Clin. Invest. 126 (8) (2016)3058–3062.

[112] J.N. Miller, D.A. Pearce, Nonsense-mediated decay in genetic disease: friendor foe? Mutat. Res. Rev. Mutat. Res. 762 (2014) 52–64.

[113] T.T. Nomakuchi, et al., Antisense oligonucleotide-directed inhibition ofnonsense-mediated mRNA decay, Nat. Biotechnol. 34 (2) (2016) 164–166.

[114] Y. Weng, et al., Identification and characterization of mutations in the UPF1gene that affect nonsense suppression and the formation of the Upf proteincomplex but not mRNA turnover, Mol. Cell. Biol. 16 (10) (1996) 5491–5506.

[115] S.E. Applequist, et al., Cloning and characterization of HUPF1, a humanhomolog of the Saccharomyces cerevisiae nonsense mRNA-reducing UPF1protein, Nucleic Acids Res. 25 (4) (1997) 814–821.

[116] T.M. Franks, et al., Upf1 ATPase-dependent mRNP disassembly is requiredfor completion of nonsense- mediated mRNA decay, Cell 143 (6) (2010)938–950.

[117] S. Chakrabarti, et al., Molecular mechanisms for the RNA-dependent ATPaseactivity of Upf1 and its regulation by Upf2, Mol. Cell 41 (6) (2011) 693–703.

[118] H. Chamieh, et al., mulate its RNA helicase activity, Nat. Struct. Mol. Biol. 15(1) (2016) 85–93.

[119] A. Yamashita, et al., Human SMG-1, a novel phosphatidylinositol3-kinase-related protein kinase, associates with components of the mRNA

[120] T. Ohnishi, et al., Phosphorylation of hUPF1 induces formation of mRNAsurveillance complexes containing hSMG-5 and hSMG-7, Mol. Cell 12 (5)(2003) 1187–1200.