widespread exon junction complex footprints in the rna ... · or yeast ever shorter telomeres1;...

LARGE-SCALE BIOLOGY ARTICLE

Widespread Exon Junction Complex Footprints in the RNADegradome Mark mRNA Degradation before SteadyState Translation[OPEN]

Wen-Chi Lee,1 Bo-Han Hou,1 Cheng-Yu Hou,1,2 Shu-Ming Tsao, Ping Kao,3 and Ho-Ming Chen4

Agricultural Biotechnology Research Center, Academia Sinica, Taipei 11529, Taiwan

ORCID IDs: 0000-0001-6717-2542 (W.-C.L.); 0000-0001-5133-3073 (B.-H.H.); 0000-0003-4195-9165 (C.-Y.H.); 0000-0001-9614-3652 (S.-M.T.); 0000-0003-0464-3778 (P.K.); 0000-0001-5979-0121 (H.-M.C.)

Exon junction complexes (EJCs) are deposited on mRNAs during splicing and displaced by ribosomes during the pioneerround of translation. Nonsense-mediatedmRNA decay (NMD) degrades EJC-bound mRNA, but the lack of suitable methodologyhas prevented the identification of other degradation pathways. Here, we show that the RNA degradomes of Arabidopsis(Arabidopsis thaliana), rice (Oryza sativa), worm (Caenorhabditis elegans), and human (Homo sapiens) cells exhibit anenrichment of 59 monophosphate (59P) ends of degradation intermediates that map to the canonical EJC region. Inhibition of59 to 39 exoribonuclease activity and overexpression of an EJC disassembly factor in Arabidopsis reduced the accumulation ofthese 59P ends, supporting the notion that they are in vivo EJC footprints. Hundreds of Arabidopsis NMD targets possess evidentEJC footprints, validating their degradation during the pioneer round of translation. In addition to premature termination codons,plant microRNAs can also direct the degradation of EJC-bound mRNAs. However, the production of EJC footprints from NMDbut not microRNA targets requires the NMD factor SUPPRESSORWITH MORPHOLOGICAL EFFECT ON GENITALIA PROTEIN7.Together, our results demonstrating in vivo EJC footprinting in Arabidopsis unravel the composition of the RNA degradome andprovide a new avenue for studying NMD and other mechanisms targeting EJC-bound mRNAs for degradation before steadystate translation.

INTRODUCTION

Manyeukaryoticpre-mRNAsundergosplicing, inwhich interveningsequences (introns) are excised and the exon junction complex(EJC) is deposited in a region 20 to 24 nucleotides upstreamof theexon-exon junction (LeHiretal., 2000).During thepioneer roundoftranslation, EJCs residing upstream of the termination codon aredisassembled and removed through the ribosome-associatedprotein Partner of Y14 and Mago (PYM; Gehring et al., 2009).For most eukaryotic mRNAs, the termination codon is present inthe last exon. Hence, after the pioneer round of translation, thesemRNAs no longer carry any EJCs and subsequently enter steadystate cycles of translation,whichcontribute to thebulk productionof proteins. However, if a termination codon situated more than50 to 55 nucleotides upstream of an exon-exon junction is

encountered by a ribosome during the pioneer round of trans-lation, it is generally recognized as a premature termination codon(PTC) and promotes nonsense-mediated mRNA decay (NMD;Nagy and Maquat, 1998).Although many studies have demonstrated that the EJCs

downstreamof a termination codonare crucial for NMD,NMDcanalso occur in an EJC-independent manner. In the absence ofEJCs, a long 39 untranslated region (UTR) sequence is able topromote NMD of mRNAs harboring a natural termination codon(Behm-Ansmant et al., 2007). Budding yeast (Saccharomycescerevisiae) lacks the genes encoding core EJC componentsbut possesses the NMD pathway (Conti and Izaurralde, 2005;Bannerman et al., 2018). Additionally, in the wormCaenorhabditiselegans and the fly Drosophila melanogaster, PTC definition isindependent of exon-exon boundaries and EJCs are dispensablefor NMD (Gatfield et al., 2003; Longman et al., 2007). While si-lencing of the EJC core components Mago and Y14 or over-expression of PYM inhibits NMD triggered by 39 UTR-locatedintrons in plants (Kerényi et al., 2008; Nyikó et al., 2013), thenumber of studies of plant EJCs is limited and little is known aboutthe binding sites of plant EJCs.Translation termination at a PTC promotes the assembly of the

UPFRAMESHIFT1 (UPF1) surveillance complex that activesNMD(Hug et al., 2016). In animal cells, PTC-containing mRNAs aredestabilized through endonucleolytic cleavage near the PTCcatalyzed by SMG6/EST1A (a member of the gene family named

1 These authors contributed equally to this work.2 Current address: Sofiva Genomics, Taipei 10043, Taiwan.3 Current address: Gregor Mendel Institute of Molecular Plant Biology,1030 Vienna, Austria.4 Address correspondence to [email protected] author responsible for distribution of materials integral to the findingspresented in this article in accordance with the policy described in theInstructions for Authors (www.plantcell.org) is: Ho-Ming Chen ([email protected]).[OPEN]Articles can be viewed without a subscription.www.plantcell.org/cgi/doi/10.1105/tpc.19.00666

The Plant Cell, Vol. 32: 904–922, April 2020, www.plantcell.org ã 2020 ASPB.

Dow

nloaded from https://academ

ic.oup.com/plcell/article/32/4/904/6115765 by guest on 24 June 2021

https://orcid.org/0000-0001-6717-2542https://orcid.org/0000-0001-5133-3073https://orcid.org/0000-0003-4195-9165https://orcid.org/0000-0001-9614-3652https://orcid.org/0000-0003-0464-3778https://orcid.org/0000-0001-5979-0121http://orcid.org/0000-0001-6717-2542http://orcid.org/0000-0001-5133-3073http://orcid.org/0000-0003-4195-9165http://orcid.org/0000-0001-9614-3652http://orcid.org/0000-0001-9614-3652http://orcid.org/0000-0003-0464-3778http://orcid.org/0000-0001-5979-0121http://crossmark.crossref.org/dialog/?doi=10.1105/tpc.19.00666&domain=pdf&date_stamp=2020-03-26mailto:[email protected]://www.plantcell.orgmailto:[email protected]:[email protected]://www.plantcell.org/cgi/doi/10.1105/tpc.19.00666http://www.plantcell.org

afterC.elegansSuppressorwithMorphological effectonGenitaliaor yeast Ever Shorter Telomeres1; Gatfield and Izaurralde, 2004;Huntzinger et al., 2008; Eberle et al., 2009) or through dead-enylationordecapping (ChenandShyu,2003;Lejeuneetal., 2003;Couttet and Grange, 2004). Although the NMD factors UPF1,UPF2, and SMG7 are conserved in animals and plants, plantsappear to lack SMG6 orthologs and presumably remove PTC-containing mRNAs through exonucleolytic pathways (Kerényiet al., 2008). Previously, identification or validation ofNMD targetsoften relied on the detection of changes in gene expression inNMDmutants or NMD factor-depleted cells because the specificdegradation products of NMD targets were poorly defined (Heet al., 2003; Kalyna et al., 2012; Drechsel et al., 2013; Colomboet al., 2017; Lloydet al., 2018). Nevertheless, relying onchanges ingene expression could be problematic because impairment of theNMD pathway can alter the expression of many genes that areindirectly regulated by this pathway.

High-throughput approaches for profiling 59 monophosphateends of RNA degradation intermediates (hereafter referred to as59P ends), such as Degradome sequencing (Degradome-Seq;Addo-Quaye et al., 2008), Parallel Analysis of RNA Ends (PARE;German et al., 2008), Genome-wide Mapping of Uncapped andCleaved Transcripts (GMUCT; Gregory et al., 2008), and 59Psequencing (Pelechanoetal., 2015), provideamore reliableway todetect the direct targets of RNA degradation pathways. Theseapproaches have been applied for transcriptome-wide identifi-cation of endonucleolytic cleavage events directed by microRNA(miRNA) indiverseplant speciesandcatalyzedbySMG6 inhuman(Homo sapiens) cells as well as the substrates of exoribonu-clease4 (XRN4),whichcatalyzes59 to39RNAdecay inArabidopsis(Arabidopsis thaliana). For plant miRNA targets, 59P ends pre-dominantly correspond to the middle of miRNA complementarysites (Addo-Quaye et al., 2008; German et al., 2008; Li et al., 2010;Pantaleo et al., 2010; Shamimuzzaman andVodkin, 2012; Li et al.,2013a). Arabidopsis miRNA-directed cleavage products over-accumulate in the Arabidopsis xrn4mutant (German et al., 2008).Global profiling of human SMG6-dependent 59P ends by PARErevealed a sequence motif enriched at the SMG6 cleavage site(Schmidt et al., 2015). Notably, depletion of SMG6 increased thenumber of 59P ends located at themRNA cap site for many SMG6substrates, suggesting competition between endonucleolyticcleavage and decapping in the human NMD pathway. BecauseSMG6 appears to be absent in plants, plant NMD targets arepresumably degraded through XRNs and the exosome after de-cappinganddeadenylation. Intriguingly,Arabidopsisxrn4notonlyoveraccumulates intermediates with 59P ends located at the capsites of NMD targets but also intermediates with 59P ends locatedin the coding region (CDS) or 39 UTR of some targets (Nagarajanet al., 2019), suggesting that endonucleolytic cleavagemight alsoaccount for the turnover of plant NMD targets.

Endonucleolytic cleavage is not the only mechanism that canresult in predominant accumulation of 59P ends at specific sites.Blocking of 59 to 39RNAdecay by RNAbinding proteins likely alsoyields 59P ends clearly delineated by the 59 edge of the RNAbindingprotein (Houet al., 2014).Consistentwith this speculation,previous analyses of yeast and plant 59P end data indicated thata portion of 59P ends are ribosome-protected mRNA fragmentsgenerated through cotranslational RNA decay (Pelechano et al.,

2015; Hou et al., 2016; Yu et al., 2016). The in vivo ribosomefootprints in the RNA degradome thus provide an alternative toribosome profiling as way to infer ribosome dynamics. Stress-inducedpauses in translation at specificcodonsandgene regionsassociatedwith collided ribosomeshavebeenuncovered throughanalyses of RNA degradome data. As many eukaryotic mRNAsare posttranscriptionally processed by splicing, EJCs might alsomake a noteworthy contribution to the pool of RNA degradationintermediates if they are able to hinder XRN-mediated RNA decayand leave footprints during in vivo mRNA degradation as ribo-somesdo. If this is the case, EJC footprints in theRNAdegradomecan serve as signatures ofmRNAdegradation before steady statetranslation, as EJCs residing upstream of the termination codonare presumably removed after the pioneer round of translation.Theoretically, only mRNA degradation occurring before or duringthe pioneer round of translation, such as NMD, can lead to theaccumulation of EJC footprints in CDSs.Here, we provide multiple lines of evidence supporting the

notion that the RNA degradome contains EJC footprints. Weobserved EJC footprints in some miRNA targets in addition toNMD targets, demonstrating that RNA degradome data can beapplied to the study of mRNA degradation before steady statetranslation.

RESULTS

59P Ends Are Enriched in the Canonical EJC Region

If EJCs are able to block 59 to 39decay of RNAand leave footprintsin the RNA degradome as ribosomes do, it is expected that en-richment of 59Pends in the canonical EJC regionwill be observed.To test this possibility, we analyzed the previously reported 59Penddata of five evolutionarily distant species. Consistentwith thisprediction, metagene analyses of both Arabidopsis PARE data(German et al., 2008) and rice (Oryza sativa) Degradome-Seq data(Li et al., 2010) revealed that the relative occurrence of 59P endswas higher in a region 25 to 30 nucleotides upstream of the 39 endof the exon (Figure 1A). As these twodata sets only contain the 59Pendsofpolyadenylateddegradation fragments,wealsoexaminedArabidopsis PARE data for nonpolyadenylated degradation frag-ments generated by Nagarajan et al. (2019). Interestingly, 59P endenrichment was also detected in the same region of non-polyadenylated degradation fragments but to a lesser degreecompared with that in polyadenylated degradation fragments(Supplemental Figure 1A). Analysis of the Arabidopsis 59P endsobtainedusing theGMUCTmethod (Willmannetal., 2014) gaveanenrichment pattern in this region similar to that observed for thePARE data (German et al., 2008; Supplemental Figure 1B). Sim-ilarly, a 59P peak spanning the same region was also detectedwhen analyzing the GMUCT data for human HEK293 cells(Willmann et al., 2014; Figure 1A). Analysis of Degradome-Seqdata forworm (Park et al., 2013), inwhichEJCs are not required fortriggering NMD (Longman et al., 2007), also revealed an enrich-ment of 59P ends in a similar region with a 1 nucleotide offset(Figure 1A). However, analysis of the PAREdata for budding yeast(Harigaya and Parker, 2012), which lacks EJCs and has only a fewintron-containing genes (Bannerman et al., 2018), revealed no

Degradation of EJC-Bound mRNAs 905

Dow



http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1

enrichment of 59P ends in this region (Figure 1A). The commonpatterns observed in the 59P end data sets of four evolutionarilydistant species indicate that this phenomenon is truly conservedacross species possessing EJCs. Given that the canonicaldeposition site of human EJCs is centered at a position 24

nucleotides upstream of the exon-exon junction (Saulière et al.,2012;Singhetal., 2012)andthatdepositionofanEJCresults in theprotection of an 8-nucleotide fragment from complete RNasedigestion (Le Hir et al., 2000; Figure 1A, top), the 59P peaksspanning the 225 to 230 region correspond to the 59 edges ofEJCs deposited in the canonical region. Therefore, these 59Ppeaks are very likely to be EJC-protected 59 termini of RNAdegradation intermediates naturally produced in cells.In addition to the analysis of 59P end distribution in a 50-

nucleotide region upstream of the exon-exon junction, we alsoinvestigated the most abundant 59P peaks (maximum 59P peaks)occurring in the canonical EJC region (positions 226 to 230)within transcripts of intron-containing genes. Analysis of pre-viously published 59P end data revealed that 13.1 (2411 genes),9.4 (1650 genes), 8.2 (710 genes), and 8.1% (1142 genes) of themaximum59Ppeakswithin transcriptsofArabidopsis, rice, humancells, and worm, respectively, were located in the canonical EJCregion (Figure 1B). These values were significantly higher thanexpected (P < 102100, x2 test), as the total length of the canonicalEJC regions represents less than 2% of the total length of splicedtranscripts. By contrast, the percentages of maximum 59P peaksfalling in the nearby 5-nucleotide regions in Arabidopsis, rice,and human cells were mostly lower than 2%, which is close tothe expected value for a 5-nucleotide region calculated by di-viding the total length of each 5-nucleotide region by that ofspliced transcripts (Figure 1B). Hence, EJC footprints appear toaccount for a noticeable portion of major mRNA degradationintermediates.

Exoribonucleases Are Involved in in Vivo EJC Footprinting

To investigate whether EJC footprints in the RNA degradome aregenerated through 59 to 39 decay, as was previously reported forribosome footprints (Pelechano et al., 2015; Yu et al., 2016), weprofiled 59P ends in the Arabidopsis wild type and two mutantsdefective in 59 to 39 RNA decay, xrn4-6 and fiery1-6 (fry1-6), andcompared the abundances of 59Pendsmapping to canonical EJCregions. Although 59P ends remained enriched in the canonicalEJC region in the twomutants (Figure 2A), xrn4-6 had significantlyfewer 59P ends in the canonical EJC region than the wild type ina global analysis of Arabidopsis exons (P < 0.001, Kruskal-Wallistest followed by a pairwiseWilcoxon test with false discovery rate[FDR] correction; Figure 2B). The decrease in the number of 59Pends in the canonical EJC region was more evident in fry1-6 (P <0.001, Kruskal-Wallis test followed by a pairwise Wilcoxon testwith FDR correction), wherein the activities of XRN2, XRN3, andXRN4 are likely all suppressed (Gy et al., 2007). Notably, the 59Pends enriched around positions 220 to 215 are likely also frag-ments protected byRNAbinding proteins, because the number of59P ends mapping to this region also appeared to be reduced inxrn4-6 and fry1-6. By contrast, the abundance of 59Pends aroundthe transcription start site (TSS) of genes harboring exons waslargely increased in xrn4-6 and slightly but significantly elevated infry1-6 compared with the wild type (P < 0.001, Kruskal-Wallis testfollowed by a pairwise Wilcoxon test with FDR correction; Fig-ure 2C). This result strongly suggests that EJCs can stericallyhinderXRN-mediated59 to39decay, resulting in theEJCfootprintsobserved in the RNA degradome. Moreover, fry1-6 exhibited

Figure1. TheCanonical EJCRegion IsaConservedHotSpot for59PEnds.

(A) Distribution of the relative frequency of 59P end occurrences in a 50-nucleotide region upstream of the exon-exon junction. PARE data forArabidopsis wild-type inflorescences (German et al., 2008) and yeast(Harigaya and Parker 2012), Degradome-Seq data for rice wild-typeseedlings (Li et al., 2010) and adult stage worms (Park et al., 2013), andGMUCT data for HEK293 T cells (Willmann et al., 2014) were used in themetageneanalyses. The illustration at the topshows thecanonical positionof an EJC and the size of a fragment protected by an EJC. The numbers ofexons $ 50 nucleotides in size and with a 59P end count $ 1 that wereincluded in the analysis are shown in parentheses.(B) Distribution of the maximum 59P peaks within the transcripts of intron-containing genes in a 50-nucleotide region upstream of the exon-exonjunction. The numbers of intron-containing genes that had the maximum59P peak with a count$ 3 and were included in the analyses are shown inparentheses.

906 The Plant Cell

Dow



a more pronounced reduction of EJC footprints than xrn4-6,implying that, in addition to the cytoplasm-localized XRN4(Kastenmayer and Green, 2000), nuclear XRNs (XRN2 and XRN3)orotherproteinswithexoribonucleaseactivitymayalsocontributeto in vivo EJC footprinting.

Prominent EJC Footprints Are Prevalent in NMD Targets

Given that ribosomal arrest by a PTC can trigger decapping,followed by 59 to 39RNA decay, the EJCs deposited on the exonsdownstream of a PTC were predicted to hinder XRN-mediated

Figure 2. XRNs Are Involved in the Production of 59P Ends Corresponding to the Canonical EJC Region in Arabidopsis.

(A)Distribution of 59Pends in a 50-nucleotide region upstreamof the exon-exon junction inwild-type, xrn4-6, and fry1-6 seedlings. Thedistributions for twoindependent biological replicates (R1 andR2) are shown. The numbers of exons$ 50 nucleotides in size with a 59P end count$ 1 that were included in theanalysis are shown in parentheses.(B) and (C)Comparison of 59P end abundance in the canonical EJC region (B) and at the TSS (C) of the genes harboring the selected exons in Arabidopsiswild-type, xrn4-6, and fry1-6 seedlings. The numbers of EJC regions (count > 10 TP40M) and TSSs (count > 5 TP40M) included in the analysis are shown inparentheses. Different letters above theboxplots indicate significant differencesbetweengroups (P

degradation and lead to the accumulation of degradation inter-mediates with 59P ends in the EJC region. Indeed, three Arabi-dopsis genes, TFIIIA, LPEAT2, and RS2Z33, which are known toproduce PTC-containing mRNAs due to alternative splicing (AS;Yoine et al., 2006; Gloggnitzer et al., 2014), displayed prominent59P peaks in the EJC regions of the exons 39 to the PTC (Figure 3;Supplemental Figure 2). Notably, in the wild type, the maximum59Ppeaks in these three geneswere all located in the secondexon39enddownstreamof thePTC, although themaximum59Ppeak inLPEAT2 was not in the canonical EJC region but at a position 37nucleotides upstream of the exon-exon junction. Moreover, theabundances of these maximum peaks, including the two in thecanonical EJC region, were reduced in both xrn4-6 and fry1-6compared with the wild type, but to a greater extent in fry1-6.However, a comparable or higher number of 59P ends corre-sponding to the putative TSS and regions outside the canonicalEJC site were observed in xrn4-6 and fry1-6 than in the wild type,which excluded the possibility that a change in gene expressionaccounted for the reduction of EJC footprints in the mutants.

Aprevious analysis ofASevents suppressed in adoublemutantof the NMD factors UPF1 and UPF3 of Arabidopsis revealeda large number of genes producing splicing variants with classicalNMD features (Drechsel et al., 2013).We thusused this gene list totest theassociationbetween theNMD-suppressedASevents andthe occurrence of maximum 59P peaks in the canonical EJC re-gion.Using a threshold applied in theprevious report (upregulatedin the double mutant of lba1 upf3-1 with FDR < 0.1; lba1 is alsoknown as upf1-1; Drechsel et al., 2013), 1082 genes with NMD-suppressed AS events were identified. Among them, 37.2% (402genes) and 32.8% (355 genes) possessed amaximum59Ppeak inthe EJC region in replicates 1 and 2 of the Arabidopsis 10-d-oldwild-type seedlingPAREdatawegenerated, respectively (Figures4Aand4B).Bycontrast, the total incidenceofmaximum59Ppeaksin the EJC region for all intron-containing genes in theArabidopsisgenome was 15.3% (3379 genes) and 13.8% (3064 genes) inreplicates 1 and 2, respectively, which is significantly lower thanthe incidence of genes with NMD-suppressed AS events (P <102100, binomial test). This result implies that EJC footprints arethe major and common degradation products of putative NMDtargets. The predominance of EJC footprints in many putativeNMD targets also validated their degradation during the pioneerround of translation. Because it has been shown thatmany intron-retention events are not sensitive to NMD (Kalyna et al., 2012;Drechsel et al., 2013), we also compared the incidence of maxi-mum 59P peaks in the EJC region for four types of AS eventssuppressedbyNMD. Interestingly, a significantlyhigher incidenceof maximum 59P peaks in the EJC region was observed for al-ternative 59 or 39 splice site events than for intron-retention events(P < 1025, binomial test), although these events were all sup-pressed byNMD (Figure 4B). However, for 2306 intron-containinggenes harboring a maximum 59P peak in the EJC region inboth replicates of the PARE data we generated, only 293 genes(12.7%) had NMD-suppressed AS events (Figure 4A). A largenumber of maximum 59P peaks in the EJC region did not overlapwith NMD-suppressed AS events reported previously (Drechselet al., 2013), implying that either the list was not complete or thereare additional pathways involved in the degradation of EJC-bound RNA.

AS events in CYP38, PPD6, and STR14 were suppressed byNMD (Drechsel et al., 2013) but were not annotated in The Ara-bidopsis Information Resource (TAIR) database. The maximum59Ppeaks in these threegeneswere located in theEJC region, andthese peaks had a 59P count$ 1000 tags per 40 million (TP40M;Figure 4C; Supplemental Figure 3). Although the 59P end profilesof these three genes highly resembled those of validated plantmiRNA targets reported previously (Addo-Quaye et al., 2008;German et al., 2008), the maximum 59P peaks were dampened inxrn4-6 and fry1-6. This indicates that these extremely abundant59P peaks do not correspond to 59P ends derived from endo-nucleolytic cleavagebut instead theprotected fragments of EJCs.Notably, these peaks were located in the first or third exon 39 enddownstream of the predicted PTCs caused by AS. This againsupports thenotion that thedominantEJC footprints in theexon39ends downstream of predicted PTCs could be evidence of NMDtriggered by those PTCs.

PTC-Triggered RNA Decay Leads to the Production ofEJC Footprints

To verify that a PTC conditions the production of EJC foot-prints, we took advantage of existing nonsense mutants ofPHYB and PHO2 (Reed et al., 1993; Delhaize and Randall,1995), in which a PTC was introduced by ethyl methanesul-fonatemutagenesis. ThePTCs in the nonsensemutants shoulddirect mRNA decay through the NMD pathway and are pre-dicted to trigger the production of EJC footprints downstreamof the PTC. We compared degradation intermediates of PHYBand PHO2 mRNAs in the wild type and the nonsense mutantsphyB-9 and pho2 using a modified RNA ligase-mediated (RLM)59RACE protocol. In both nonsensemutants, the 59 ends of themost abundant amplification products of the mRNA degra-dation fragments were mainly mapped to positions 27 to 28nucleotides upstream of the exon-exon junction downstreamof the PTCs (Figures 5A and 5B). However, these degradationintermediates were not prominent in the wild type, although themiR159-guided 39 cleavage product of MYB65, the positivecontrol for modified RLM 59 RACE, was detected in both thewild type and the nonsense mutants. The modified RLM 59RACE results for phyB-9 and pho2mimic the PARE data of thethree known NMD targets with PTCs introduced by AS (Fig-ure 3), and together these results support the conclusion thatPTCs elicit the accumulation of EJC footprints in downstreamexons. Moreover, PTCs caused by AS or point mutationstended to result in the generation of 59P ends mainly in the firstor second EJC region downstream of PTCs (Figures 3, 4C, 5A,and 5B), suggesting that EJC footprints may serve as an im-mediate readout of PTC-mediated NMD irrespective of thecause of the PTC.We further validated the contribution of XRN4 to EJC footprint

production from nonsense mRNAs by generating the phyB-9xrn4-6 and pho2 xrn4-6 double mutants. As expected, a lowernumber of putative EJC footprints was observed in the doublemutants than in the nonsense mutants (Figure 5C). This resultagain confirms the role of XRN4 in the turnover of PTC-containingmRNAs and the generation of EJC footprints.

908 The Plant Cell

Dow




Analysis of EJC Footprints Reveals NMD Targets

Using the common features of EJC footprints in the NMD targetsshown in Figures 3, 4, and 5, we globally validated potentialtargets of NMD triggered by AS or introns in the 39UTR that were

annotated in the Arabidopsis and rice genome databases. Forprotein-coding transcripts with AS events, if the maximum 59Ppeak was located in the first or second canonical EJC regiondownstream of an AS site, they were identified as putative tar-getsofNMD triggeredbyAS,which frequently results in aPTC. In

Figure 4. Genes Producing NMD-Suppressed AS Variants More Frequently Possess a Maximum 59P Peak in the EJC Region.

(A) Venn diagram showing the overlap between genes producing NMD-suppressed AS variants (NMD1 AS) and genes possessing a maximum 59P peak(M59P) in the EJC region. The NMD-suppressed AS variants were extracted from the lba1 upf3-1 data reported by Drechsel et al. (2013). Two replicates (R1and R2) of PARE data for 10-d-old (10d) seedlings generated by this study were used for the identification of genes possessing a maximum 59P peak witha count $ 3 located in the EJC region.(B)Relative frequency of genes possessing amaximum59Ppeak in the EJC region. Frequencies are shown for genes containing introns (intron1 gene) andthose producing transcripts derived from the following AS events: NMD1 AS, intron retention, exon skipping, and alternative 59 and 39 splice sites (alt. 59ssandalt. 39ss). Thenumberof genes ineachcategorypossessingamaximum59Ppeak in theEJCregion isshownabove thebar.Asterisks indicate significantdifferences between categories based on the binomial test (*, P < 1025; **, P < 10215; ***, P < 102100).(C)Positional distribution of 59Pends (fromPAREdata generatedby this study) in threeArabidopsis genes producingNMD-suppressedASvariants.Withinthe59Pendplots, redpeaks indicatepeaks in thecanonicalEJC regionandblackarrows indicate themaximumpeaks in10-d-oldwild-typeseedlings.Withinthe genemodels, light gray boxes indicate UTRs, dark gray boxes indicate CDSs, thin lines indicate introns, red asterisks indicate PTCs, and orange boxesindicate regions altered by NMD-suppressed AS events according to Drechsel et al. (2013). Only the 59P ends residing in exons are displayed in the plots.Data for replicate 2 are shown in Supplemental Figure 3.


Dow



http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1

addition, transcripts possessing an intron-containing 39 UTR werealso identified as putative NMD targets if the maximum 59P peakwas located in the first or second canonical EJC region of the exon39 end at least 50 nucleotides downstream of the annotatedtermination codon. Analysis of the PARE data for Arabidopsiswild-type seedlings and flowers that we generated and the pre-viouslypublishedDegradome-Seqdata for ricewild-typeseedlings(Li et al., 2010) revealed 254 Arabidopsis transcripts and 168 ricetranscripts as putative NMD targets (Figure 6A; SupplementalData Sets 1 and 2). Remarkably, 19.3 and 44% of the putativeNMD targets identified from the analysis of Arabidopsis and rice59P end data, respectively, have an intron-containing 39 UTR(Figure 6A). Compared with the large number of NMD-sensitiveAS events reported by Drechsel et al. (2013), the number ofArabidopsis NMD targets we identified is small, because manyAS events reported by Drechsel et al. (2013), including AS ofCYP38, PPD6, and STR14, were not annotated in TAIR 10.However, 224 of the targets we identified in Arabidopsis did not

show evidence of NMD-suppressed AS in the previous report(upregulated in the double mutant of lba1 upf3-1with FDR < 0.1;Supplemental Data Set 1; Drechsel et al., 2013). Some of thenovel NMD targets, such as Arabidopsis ARR8 and KCBP andrice transcripts encoding an SPA4-like protein (LOC_Os01g52640)and a transposon protein (LOC_Os04g03884), had high numbersof EJC footprints in the 39 UTR (Figures 6B and 6C). For the Ara-bidopsis NMD targets we obtained, the analysis of gene Ontology(GO) terms revealed the highest enrichment for transcripts involvedin “regulation of alternative mRNA splicing, via spliceosome” (22.27-fold enrichment, P < 0.001, Fisher’s exact test with Bonferronicorrection; Supplemental Table 1). This enriched GO term is con-sistent with the previous finding that in Arabidopsis, NMD is animportant mechanism regulating the splicing of genes encodingSer/Argproteins,whichplaykeyroles insplicing (PalusaandReddy,2010). Other significantly enriched GO terms unrelated to splicingincluded “organic cyclic compound metabolic process,” “nitrogencompoundmetabolic process,” and “chloroplast” (P < 0.05, Fisher’sexact testwithBonferroni correction). For the riceNMD targetsweidentified, however, no enriched GO terms were recovered. Thedifference in GO enrichment analysis results between the twospecies might be due to differences in the 59P end data sets weanalyzed, the frequencyof introns in the39UTR,or theannotationsof splicing variants in the two genomes. It may also suggest thatthe role of the NMD pathway in gene regulation could be distinctamong plant species.

Dysfunction of SMG7 Reduces the Number of EJCFootprints in Putative NMD Targets

To verify that EJC footprints in NMD targets are degradationproducts generated by the NMD pathway, we profiled 59P ends inthe Arabidopsis wild type and mutants of key NMD factors andcompared the abundances of EJC footprints between them. InArabidopsis, the crucial roles of UPF1 and SMG7 in the NMDpathway have been demonstrated (Arciga-Reyes et al., 2006;Riehs-Kearnan et al., 2012). Given that null mutants of UPF1 areseedling lethal (Arciga-Reyes et al., 2006), smg7-1, a severe butviable mutant of SMG7 with a T-DNA insertion in the CDS, waschosen for the comparison (Riehs et al., 2008). However, ho-mozygous smg7-1plants are sterile andmorphologically differentfromwild-type plants when grown at 22°C due to an autoimmuneresponse (Riehs et al., 2008; Riehs-Kearnan et al., 2012). To re-duce the possibility that differences might be caused by the au-toimmune response, we selected smg7-1 homozygous plants bygenotyping and grew them at 28°C, under which the autoimmuneresponse is repressed and wild-type and smg7-1 plants moreclosely resemble each other (Gloggnitzer et al., 2014). We thencollected the inflorescences of wild-type and smg7-1 plants andisolated RNA for PARE library construction. Similar to the wildtype,smg7-1displayedanenrichmentof59Pends in thecanonicalEJC region (Figure 7A).However, in smg7-1, the abundanceof 59Pendswassignificantly lower in thecanonical EJC regionbuthigherat the TSS compared with the wild type (P < 0.01, Wilcoxon rank-sum test; Figures 7B and 7C). Furthermore, themedian of the log2fold change in maximum EJC footprint abundance betweensmg7-1 and the wild type was significantly lower for transcriptsexperiencing NMD-suppressed AS events than for other transcripts

Figure 5. PTCs Trigger the Formation of EJC Footprints in DownstreamExons.

(A) and (B) Modified RLM 59 RACE assays of RNA degradation inter-mediatesgenerated fromwild-type andnonsensemutantmRNAsofPHYB(A) and PHO2 (B). Arrowheads indicate the RACE products excised andcloned forSangersequencing (leftpanels). Thepositionaldistributionof the59Pends of the cloned RACEproducts fromnonsensemutants is shown inthe right panels. Within the gene models, light gray boxes indicate UTRs,dark gray boxes indicate CDSs, thin lines indicate introns, asterisks in-dicate PTCs, and half arrows indicate gene-specific primers used formodified RLM 59 RACE. The 39 remnant of MYB65 derived from miR159-guided cleavagewas used as a positive control formodifiedRLM59RACE.(C)Comparison of EJC footprints from nonsensemutant mRNAs betweenphyB-9 and phyB-9 xrn4-6 and between pho2 and pho2 xrn4-6 usingmodified RLM 59 RACE assays. Arrowheads indicate the RACE productssequenced in (A) and (B).

910 The Plant Cell

Dow



http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1

(P < 1024, Kolmogorov-Smirnov two-sided test; Figure 7D),supporting the notion that SMG7 plays a crucial role in the pro-duction of EJC footprints from NMD targets. Consistent with theanalysis of PARE data (Figures 3 and 4C), the analysis of modifiedRLM59RACEdata also confirmed the involvement of SMG7 in theproduction of EJC footprints for threeNMD targets in both flowersand 14-d-old seedlings (Supplemental Figure 4). The reduction inthe number of EJC footprints in smg7-1 seedlings supports thenotion that the effect was due to the NMD function of SMG7 in-stead of its unique function inmeiosis (Riehs et al., 2008). Like the59Ppeaks in theEJC regionsof sixNMDtargetsshown inFigures3and 4C, themaximum 59P peak in LPEAT2, which was positionedoutside the canonical EJC region, was also dampened in smg7-1(Figure 3; Supplemental Figure 2), indicating that it might also bea degradation product of the SMG7-dependent NMD pathway.Moreover, for threepreviously reportedNMD-sensitive transcriptsthat had xrn4-enhanced 59P peaks in the CDS or 39 UTR(Nagarajan et al., 2019), xrn4-enhanced 59P peaks in eRF1-1(Eukaryotic Release Factor1-1, NMD factor and target) were de-tected in the flowers of the wild type but were absent in those ofsmg7-1 (Supplemental Figure 5). This result further confirmsthe notion that the xrn4-enhanced 59P peaks in eRF1-1 arealso intermediates of degradation initiated by the NMD pathway(Nagarajan et al., 2019).

Some miRNA Targets Possess EJC Footprints

Surprisingly, 59P peaks corresponding to putative EJC footprintswere also observed downstream ofmiRNA-guided cleavage sites

in the targets of Arabidopsis miR159 (MYB33), miR160 (ARF10),and miR396 (GRF1; Figure 8A; Supplemental Figure 6). Likewise,rice miR159, miR160, and miR396 targets also harbored EJCfootprints in the exon ends downstream of themiRNA target sites(Figure 8B). The accumulation of putative EJC footprints in thethreeArabidopsismiRNAtargetswasreduced inxrn4-6andnearlyabolished in fry1-6; however, the amplitude of miRNA-guidedcleavage peaks was increased in both xrn4-6 and fry1-6 com-paredwith thewild type.Wealsoconfirmedthedecreasednumberof EJC footprints and the increased number of miRNA-guidedcleavage 39 remnants for these three miRNA targets in xrn4-6and fry1-6 by performing modified RLM 59 RACE assays(Supplemental Figure 7). We also assayed XRN2 and XRN3function in the production of EJC footprints from these threemiRNA targets. The levels of 39 remnants of miRNA-guidedcleavage and EJC footprints were comparable between thewild type, xrn2-1, and xrn3-3 (Supplemental Figure 7). We thusfurther examined the level of EJC footprints in the XRN triplemutant xrn2-1 xrn3-3 xrn4-6. Like xrn4-6, the triple mutant alsoaccumulated a higher number of miRNA-guided cleavage 39remnants but fewer EJC footprints compared with the wild type.However, the accumulation of these two types of degradationfragments in the triple mutant was neither different from that inxrn4-6 nor comparable to that in fry1-6. This result does notcompletely rule out a role for XRN3 in EJC footprinting becausexrn3-3 is a knockdown mutant with a T-DNA insertion in thepromoter region (Gy et al., 2007). These results confirm that XRN4degrades the39cleavage remnantsofplantmiRNA targets (Souretet al., 2004) and suggest that in addition to XRN4, other enzymes

Figure 6. Analysis of EJC Footprints Reveals Targets of NMD Triggered by AS and Intron-Containing 39 UTRs.

(A) Fractions of putative NMD targets where NMD is triggered by AS or an intron-containing 39 UTR (Intron1 39 UTR). The numbers of total putative NMDtargets recovered from the analysis are in parentheses.(B) and (C) Positional distribution of 59P ends for targets of NMD triggered by intron-containing 39 UTRs in Arabidopsis (B) and rice (C). 59P ends inArabidopsis were identified using the Arabidopsis PARE data (replicate 1) generated by this study, and those in rice were identified using the Degradome-Seq data for rice wild-type seedlings reported by Li et al. (2010). Within the 59P end plots, red peaks indicate peaks in the canonical EJC region and blackarrows indicate themaximumpeaks.Within thegenemodels, light gray boxes indicateUTRs, dark gray boxes indicateCDSs, and thin lines indicate introns.Only the 59P ends residing in exons are displayed in the plots.


Dow



http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1

with 59 to 39 exonuclease activity also contribute to the turnover ofmiRNA cleavage remnants and EJC footprinting.

Notably, not all of the 59P peaks evident at miRNA target siteswere associated with prominent EJC footprints in downstreamexon ends. For instance, while a high number of 59P ends cor-responding to the miR398 target site of Arabidopsis CSD1 andCSD2 were observed, no or poor 59P peaks appeared in thecanonical EJC regions 39 to the miR398 target site (Figure 9A;Supplemental Figure8).More intriguingly, themiR398 target site isclose to the canonical EJC region in CSD1 and overlaps with thecanonical EJC region in CSD2. Hence, EJCs might mask themiR398 target site on the CSD1 and CSD2mRNAs that have not

been translated and prevent miRNA regulation before the pioneerround of translation. Consistent with this result, rice miR398-regulated CSD1 and two miR444-regulated MADS genes, whichpossess a miRNA target site in or near the canonical EJC region,also lacked evident EJC footprints (Figure 9B). In addition to themiR398 and miR444 targets, a miR408 target encoding planta-cyanin (ARPN) in Arabidopsis also possessed a prominent 59Ppeakat themiRNA-guidedcleavage sitebut poor peaks in theEJCregion. However, unlike the miR398 and miR444 targets, themiRNA target site on ARPN is located at a position ;200 nu-cleotides upstream of an exon-exon junction that is unlikely to bemaskedbyEJCs. Taken together, these results suggest that there

Figure 7. EJC Footprints in NMD Targets but Not miRNA Targets Are Less Abundant in smg7-1.

(A)Distribution of 59Pend counts in a 50-nucleotide region upstream of the exon-exon junction. Counts are shown for two replicates (R1 andR2) each fromwild-type and smg7-1 flowers. The numbers of analyzed exons that are $ 50 nucleotides in size and with a 59P end count $ 1 TP40M are shown inparentheses.(B)and (C)Comparisonof 59Pendcounts in thecanonical EJC region (B)andat theTSS (C)of thegenesharboring the selectedexonsbetween thewild typeand smg7-1. Asterisks indicate significant differences between groups (P < 0.01,Wilcoxon rank-sum test). Box boundaries andwhiskers indicate quartilesand the range of 59P end abundance, respectively. The numbers of analyzed EJC regions (count > 10 TP40M) or TSSs (count > 5 TP40M) are shown inparentheses.(D)Scatter anddensityplotsdepicting log2 fold change (smg7-1/wild type) and log2 average59Pendcounts for themaximumEJCpeaks ingenes. Eachdatapoint represents a maximum EJC peak with an average 59P end count $ 10. EJC peaks for genes producing NMD-suppressed AS variants (NMD1 AS),miRNA targets, andother genes (other) are shown. Arabidopsis genesproducingNMD-suppressedASvariants andgenes that are validatedmiRNA targetswere extracted from data sets reported by Drechsel et al. (2013) and Zheng et al. (2012), respectively. The number of analyzed genes in each category isshown in parentheses. The density plots to the right cover all data points in each category, and straight lines indicate the median of each distribution.Asterisks indicate significant differences in the comparisons of the NMD1 AS group and the miRNA group with the group of other genes (Kolmogorov-Smirnov two-sided test). The colors and locations of asterisks indicate the associated category and the direction of the shift in the distribution.

912 The Plant Cell

Dow




might bemultiple factors controlling the interaction between plantmiRNAs and EJC-bound targets.

Arabidopsis miRNAs Direct EJC-Bound mRNA Degradation

Unlike NMD targets, Arabidopsis miRNA targets did not exhibitchanges in the number of EJC footprints in smg7-1 (Figure 8A;Supplemental Figure 6). Indeed, a global analysis of maximumEJC footprint abundance between smg7-1 and the wild type re-vealed a median log2 fold change value close to 0 for miRNAtargets.The log2 foldchangeofmiRNAtargets issignificantlyhigherthan that of other transcripts (P < 0.01, Kolmogorov-Smirnov two-sided test; Figure 7D). This trend is opposite to that observed forNMD-suppressedASeventsandsuggests thatmostEJCfootprintsinmiRNA targetsare likelyproduced throughanNMD-independentpathway.

Given that EJC footprints were evident downstream of miRNAtarget sites but rare in the upstream regions,wehypothesized thatplant miRNAs are able to target EJC-bound mRNAs for degra-dation and trigger the production of EJC footprints. To test thishypothesis, we first assayed the role of AGO1,whichmediates thefunction ofmost plantmiRNAs, in theproduction of EJC footprints

from the threemiRNA targets shown inFigure 8A in thenull ago1-3mutant (Arribas-Hernández et al., 2016). In ago1-3, there werefewer miRNA-guided cleavage remnants and EJC footprintscorresponding to themiR160 targetARF10 and themiR396 targetGRF1 (Figure 10A). Interestingly, for the miR159 target MYB33,which exhibited no change in mRNA level in ago1-3 (Arribas-Hernández et al., 2016), the levels of miRNA-guided cleavageremnants and EJC footprints in ago1-3were comparable to thosein the wild type. Also, the mutation of AGO1 did not affect EJCfootprint accumulation in the two NMD targets CYP38 andLPEAT2. Hence, these results confirm the link between thefunction of AGO1 and the production of EJC footprints in somemiRNA targets.To further validate the ability of plant miRNAs to direct EJC

footprint accumulation, we created an artificial miRNA target byfusing the 59 UTR of Arabidopsis NITROGEN LIMITATION AD-APTATION (NLA), which contains a target site of miR827, to anintron-containing GUS gene and introduced this construct(I-GUS827) into Arabidopsis (Figure 10B). As miR827 is barelydetectable when phosphate levels are sufficient but is highlyinduced upon phosphate starvation (Figure 10C; Hsieh et al.,2009), we thus were able to test the ability of miR827 to induce

Figure 8. Some Arabidopsis and Rice miRNA Targets Possess Dominant EJC Footprints Downstream of miRNA-Guided Cleavage Sites.

Positional distribution is shown for 59Pendsof selectedArabidopsis (A) and rice (B)miRNA targetswith prominent EJC footprints basedon theArabidopsisPARE data generated in this study and the rice Degradome-Seq data generated by Li et al. (2010). Within the 59P end plots, red peaks indicate peaks in thecanonical EJC region and red arrowheads indicate peaks corresponding to the miRNA-guided cleavage site. Within the gene models, light gray boxesindicate UTRs, dark gray boxes indicate CDSs, and thin lines indicate introns. Only 59P ends residing in exons are displayed in the plots. Data for replicate2 of Arabidopsis are shown in Supplemental Figure 6.


Dow




EJC footprint production by comparing RNA degradation in-termediates produced from the I-GUS827 construct underphosphate-sufficient and -starvation conditions using modifiedRLM 59 RACE. As expected, distinct amplification productsof the RNA degradation intermediates with 59P ends pre-dominantly mapping to positions 26 to 27 nucleotides upstreamof the exon-exon junction were detected in the plants grownunder phosphate-starvation conditions but not in those grownwith sufficient phosphate (Figures 10D and 10E). In contrast tothe I-GUS827 construct, the control construct containing anintron-free GUS gene fused with the NLA 59 UTR (GUS827) didnot result in the accumulation of RNA degradation fragmentstruncated at the same site under phosphate-starvation con-ditions (Figure 10D). Nevertheless, the number of RNA degra-dation fragments with the 59 end determined by miR827-guidedcleavage was comparable between plants expressing the twoconstructs. This result validates the ability of plant miRNAs totarget the mRNAs associated with EJCs for degradation andinitiate the accumulation of transcripts with 59P ends in the EJCregion.

Unexpectedly, AS was observed in the I-GUS827 construct wecreated. In the AS form, the dominant sequence matching cloned59P ends was shifted farther upstream along with the alternative59 splice site compared with that in the normal splice form(Figure 10E). However, the distancebetween the predominant 59Ppeak and the exon-exon junction remained 27 nucleotides in theAS form. As the deposition of EJC on mRNA is dependent onsplicing, the shift of the 59P peak in the AS form reinforces ourhypothesis that the 59P ends matching a region 26 to 30 nu-cleotides upstream of the exon-exon junction mostly representin vivo EJC footprints.

Overexpression of an EJC Disassembly Factor EliminatesEJC Footprints

To provide direct evidence that the 59P ends in the EJC region areEJC-protected fragments, we cloned an EJC disassembly factor,PYM, from Arabidopsis and transiently expressed a PYM over-expression construct in Arabidopsis protoplasts to disassemble

Figure 9. Some Arabidopsis and Rice miRNA Targets Possess Poor EJC Footprints Downstream of miRNA-Guided Cleavage Sites.

Positional distribution is shown for 59Pendsof selectedArabidopsis (A)and rice (B)miRNA targetswithpoor EJC footprints basedon theArabidopsisPAREdatagenerated in thisstudyand the riceDegradome-SeqdatageneratedbyLietal. (2010).Within the59Pendplots, redpeaks indicatepeaks in thecanonicalEJC region and red arrowheads indicate peaks corresponding to themiRNA-guided cleavage site.Within the genemodels, light gray boxes indicate UTRs,dark gray boxes indicate CDSs, and thin lines indicate introns. Only 59P ends residing in exons are displayed in the plots. Data for replicate 2 of Arabidopsisare shown in Supplemental Figure 8.

914 The Plant Cell

Dow




EJCs (Figure 11). We predicted that disassembly of EJCs by PYMoverexpression would promote the degradation of existing EJC-protected degradation intermediates and inhibit the production ofnew EJC footprints. For NMD targets, the production of inter-mediates with 59P ends matching the canonical EJC regions ofCYP38 and PPD6 was abolished when PYM was overexpressed(Figure 11). Degradation fragments corresponding to the maxi-mum 59P peak of LPEAT2, which was positioned outside thecanonical EJC region and SMG7-dependent (Figure 3), were alsoabsent in PYM-overexpressing protoplasts. This result indicatesthat this maximum 59P peak is an EJC footprint and supports thenotion that EJCs are also deposited on noncanonical sites inArabidopsis. Inaddition to investigating theSMG7-dependent59Ppeaks in NMD targets (Figures 3 and 4C), we also examined theeffect of PYM overexpression on the SMG7-independent 59Ppeaks in the EJC region of miRNA targets shown in Figure 8A.Compared with the number of degradation intermediates in

control protoplasts, the number of 39 cleavage remnants directedbymiRNAs in the threemiRNA targets (MYB33,ARF10, andGRF1)was increased, whereas degradation intermediates with 59P endscorresponding to the EJC region were barely detectable in thePYM-overexpressing protoplasts (Figure 11). Taken together,these results strongly indicate that the 59P ends mapping to thecanonical EJC regions of these transcripts are in vivo EJC foot-prints irrespective of their dependency on SMG7.

DISCUSSION

The RNA degradome is composed of assorted RNA degradationintermediates derived from diverse RNA degradation pathways.While endonucleolytic cleavage results in the production ofcleavage remnants with well-defined 59P ends, growing evidencesupports the idea that the hindrance of XRN-mediated 59 to 39decay by RNA binding proteins such as Pumilio/fem-3 mRNA

Figure 10. AGO1 Function, Splicing, and miRNA Expression Determine EJC Footprint Production in miRNA Targets.

(A)RLM59RACEassaysofdegradation intermediatesgenerated from threemiRNA targetsand twoNMDtargets in thewild typeandago1-3.Openandsolidarrowheads indicatemiRNA-guidedcleavagesitesandEJCbindingsites, respectively, ingenemodels and thecorrespondingRACEproductsseparatedbygel electrophoresis. The arrow indicates themaximum59peak in theLPEAT2genemodel and in thecorrespondingRACEproduct onanelectrophoretic gel.Within thegenemodels, light gray boxes indicateUTRs, dark gray boxes indicateCDSs, thin lines indicate introns, orangeboxes indicate regions altered byAS, asterisks indicatePTCs, andhalf arrows indicate gene-specificprimers. ThesteadystatemRNA level ofArabidopsisACT8wasused to control the inputamount of total RNA and reverse transcription among samples. An equal amount of tobacco acid pyrophosphatase-treated total RNA fromHeLa cells wasspiked into each sample, and the TSS of human b-actin was used to control for differences in ligation efficiency among samples.(B) Constructs expressing miR827 artificial targets: GUS with an intron (I-GUS827) and without an intron (GUS827). 35S, Cauliflower mosaic virus 35Spromoter. Boxes indicate exons, thin lines indicate introns, and half arrows indicate gene-specific primers used for modified RLM 59 RACE.(C) RNA gel blot analysis of miR827 in Arabidopsis transgenic lines harboring miR827 artificial targets. Transgenic seedlings grown under phosphate-sufficient (1P)orphosphate-deficient (2P)conditionswereused for theassays.U6wasusedasa loadingcontrol.Numbers to the rightof theblot showsizesin nucleotide.(D) Modified RLM 59 RACE assays of RNA degradation intermediates generated from miR827 artificial targets in Arabidopsis transgenic lines. The openarrowhead indicates the expected RACE product for the 39 cleavage remnant of the artificial targets directed bymiR827, and the solid arrowhead indicatesthe RACE product excised and cloned for Sanger sequencing. The miR159-guided 39 cleavage remnant ofMYB65 was used as a positive control for themodified RLM 59 RACE assay.(E) Positional distribution of the 59P ends of cloned RACE products from I-GUS827 transgenic plants grown under phosphate-deficient conditions.


Dow



binding factors and ribosomescanalsopreciselydelineate the59Pends of degradation products (Hou et al., 2014, 2016; Pelechanoet al., 2015; Yu et al., 2016). Here, we demonstrate that, like ri-bosomes, EJCs also have the capacity to stop 59 to 39 decay andleave precise footprints in the RNA degradome (Figure 12). Be-cause EJCs are deposited on RNA during splicing and are dis-placed by ribosomes during the pioneer round of translation(Figures 12Aand12B), EJC footprints can serve asmarkers for thedegradation of mRNA before steady state translation. If a splicedmRNA is degraded after the pioneer round of translation, 59 to 39mRNA decay should not yield EJC footprints upstream of thetermination codon while cotranslational mRNA decay duringsteadystate translationmayyield ribosome footprints (Figure12B;Pelechano et al., 2015; Hou et al., 2016; Yu et al., 2016). By

contrast, NMD targets are mainly degraded during the pioneerround of translation andmay produce EJC footprints downstreamof PTCs (Figure 12C). Additionally, some plant miRNA targetsseem to be attacked by miRNAs before translation, which mayresult in EJC footprints located 39 to the miRNA target sites(Figure 12D). As some EJC footprints appear not to be linked toNMD or miRNA pathways (Figures 4A and 7D), other post-transcriptional gene regulationmechanismsmay also account forEJC-boundmRNAdegradationbeforeorduring thepioneer roundof translation. In summary, our study demonstrates that the RNAdegradome contains EJC footprints, which account for a sizableportion of major RNA degradation intermediates. The analysis ofEJC footprints in the RNA degradome confirms the degradationof plant NMD targets during the pioneer round of translation andreveals EJC-bound mRNAs as targets of some plant miRNAs.

Location and Abundance of EJC Footprints in the PlantRNA Degradome

Although cross-linked immunoprecipitation followed by next-generation sequencing has not been adopted to profile EJCbinding regions on a genome-wide scale in Arabidopsis and rice,the identification of a conserved hot spot for 59P ends located ina region 25 to 30 nucleotides upstream of the exon-exon junction(Figure 1) implies that the deposition of EJCs on plant and animalRNAs presumably follows similar rules. Two transcriptome-widestudies of human EJC binding regions using cross-linked im-munoprecipitation followed by next-generation sequencing re-vealed that only about half of the EJC binding sitesmapped to thecanonical position (Saulière et al., 2012; Singh et al., 2012). Thebinding of EJCs to noncanonical sites likely also occurs in plantsand accounts for the 59P peaks positioned outside the canonicalEJC regions. For instance, the maximum 59P peak in ArabidopsisLPEAT2 appeared to be the footprint of an EJC deposited ata noncanonical site, as it was located immediately downstreamofthe PTC, and the peak was dampened in smg7-1 and largelyreduced when the EJC disassembly factor PYM was overex-pressed (Figures 3 and 11). If half of EJCs are deposited atnoncanonical sites in plants as in humans, the contribution of EJCfootprints to maximum 59P peaks located within transcripts islikely to have been underestimated, as we only considered themaximum59Ppeakspositioned in thecanonicalEJCregion.Giventhat EJC footprints in the RNA degradome can be prominent,abundant, and prevalent, as we demonstrated in this study, in-vestigations of inference by endonucleolytic cleavage directed bymiRNAs or othermechanisms usingRNAdegradome data shouldbe undertaken cautiously. Dependency on XRN activity couldserve as a key to differentiate between protected fragments de-rived from RNA binding proteins and endonucleolytic cleavageproducts.The number of 59P ends mapping to EJC regions varied dra-

matically within PTC-containing mRNAs and miRNA targets(Figures 3 and 8). Intriguingly, the maximum EJC-protected 59Ppeaks tended to reside in thefirst or secondexonenddownstreamof a PTC or an miRNA-directed cleavage site. This result impliesthat in both cases 59 to 39 decay ismore frequently blocked by thefirst EJC met by the XRN. However, in some cases, the exonsfarther downstream also had 59P endsmapping predominantly to

Figure 11. Overexpression of PYMAbolishes EJC Footprints in NMD andmiRNA Targets.

ArabidopsisPYMwasclonedand transientlyoverexpressed inArabidopsisprotoplasts. The steady state mRNA levels of PYM and a housekeepinggene, ACT8, in control (2) and PYM-transfected (1) protoplasts weremeasured using the cDNA from the modified RLM 59 RACE assays. Theamounts of RNA degradation intermediates from three NMD targets andthree miRNA targets in control and PYM-overexpressing protoplasts werecompared using modified RLM 59 RACE assays. An equal amount of to-bacco acid pyrophosphatase-treated total RNA from HeLa cells wasspiked into each sample, and the TSS of human b-actin was used to inferligation efficiency. Open and solid arrowheads indicate themiRNA-guidedcleavage sites and EJC binding sites, respectively, in genemodels and thecorresponding RACE products separated by gel electrophoresis. Thearrow indicates the maximum 59 peak in the LPEAT2 gene model and thecorresponding RACE product on an electrophoretic gel. The identity of allmarked RACE products was confirmed by Sanger sequencing. Within thegene models, light gray boxes indicate UTRs, dark gray boxes indicateCDSs, thin lines indicate introns, orange boxes indicate regions altered byAS, asterisks indicate PTCs, and half arrows indicate gene-specificprimers.

916 The Plant Cell

Dow



the EJC region, although to a lesser degree, implying that some-times XRN-mediated decay might be able to overcome the hin-drance of EJCs. Notably, in PTC-containing transcripts such asArabidopsis LPEAT2 and RS2Z33, some exons possessed poor59P peaks in the canonical EJC region while both the 59 and 39neighboring exons had prominent EJC peaks (Figure 3). Since noother evident 59P peaks were present in these exons, which lackedcanonical EJC footprints, it appears that EJCs are less frequentlydeposited on these exons than their neighboring exons. Thispossibility is supported by the previous finding that the loading ofhuman EJCs varies among exons even within the same transcript(Saulière et al., 2012; Singh et al., 2012). Differential EJC loadingprobably also accounts for the poor accumulation of 59P ends insomeEJC regionsdownstreamofaPTCormiRNA target sitewithinthe same transcripts in plants.

Application of EJC Footprints to the Study of NMD

Previously, the identificationor thevalidationof endogenousNMDsubstrates required NMD mutants or treatment with NMD inhib-itors such as cycloheximide, which is not feasible or problematicin some species or tissues. Moreover, the upregulation of someassayed RNAs is an indirect consequence of NMD pathway

impairment. Our study shows that EJC footprints can be a directreadout of NMD; these footprints will be useful in the validation ofpotential NMD substrates and inference of NMD activity. Thepresence of a predominant 59P peak in the EJC region located atthe first or second exon end downstream of the predicted PTC isstrong evidence of NMD-mediated degradation. Using 59P enddata for wild-type species alone, we have validated hundreds ofNMD targets in Arabidopsis and rice, where NMD is potentiallytriggered by AS or intron-containing 39UTRs (Supplemental DataSets 1 and 2). As RNA degradome data have been generated inmany plant species formiRNA target identification, the analysis ofpublicly available RNA degradome data sets may advance ourunderstanding of gene regulation mediated by the NMD pathwayindifferent species.Nevertheless, this applicationmightbe limitedto NMD targets with at least one intron after the PTC; NMDeventstriggered by long intron-free 39 UTRs are unlikely to be confirmedby this approach. Furthermore, RNA degradome data may allowassessment of the impact of stress on the transcriptome throughtheNMDpathway. Inplants, abiotic stressconditions suchashightemperature, drought, and salt stress have been shown to largelyalter the accumulation of splicing variants, which mostly havefeatures associated with NMD (Chang et al., 2014; Feng et al.,2015; Thatcher et al., 2016; Jiang et al., 2017). However, whether

Figure 12. Model for EJC-Bound mRNA Degradation before and during Steady State Translation.

(A)During splicing, EJCs are deposited onmRNAs at aposition so that thedistancebetween the 59 edgeof theEJCand the exon-exon junction is;26 to 30nucleotides.(B) After the pioneer round of translation, all EJCs upstream of the termination codon are removed from the mRNAs. The XRN-mediated 59 to 39 co-translational mRNA decay occurring during steady state translation yields ribosome footprints but not EJC footprints in the CDS.(C)During thepioneer roundof translationofPTC-containingmRNAs, ribosomesdisplace theEJCsupstreamof thePTCwhile triggeringNMDwhenstallingat the PTC. The XRN-mediated 59 to 39 decay of PTC-containing mRNAs produces footprints of the EJCs positioned downstream of the PTC.(D) Before the pioneer round of translation, endonucleolytic cleavage directed by miRNAs results in EJC-bound cleavage remnants, which are furthertrimmed by XRNs and become EJC footprints.


Dow




NMD activity is regulated by these stresses has not been ad-dressed. Measuring EJC footprints from well-characterized NMDtargets provides a straightforward approach to investigate theregulation of NMD activity by stress.

Targeting of EJC-Bound mRNAs by Plant miRNAs

Here, we showed that the cleavage events guided by some, butnot all, Arabidopsis and ricemiRNAswerenotably associatedwiththeproductionofEJC footprints (Figure8;Supplemental Figure6).If plant miRNAs only direct cleavage of the mRNAs that havefinished thepioneer roundof translation, themRNAdecay initiatedby plant miRNAs would not lead to the accumulation of EJCfootprints in the CDS. This result therefore suggests that at leasta fraction of some plant miRNAs such as miR159, miR160, andmiR396 target EJC-bound mRNAs that have not completed thepioneer roundof translation (Figure12D). Althoughprominent EJCpeaks are only evident downstreamof themiRNA-guided cleavagesite in some targets (Figure 8), EJCs upstream of the miRNA targetsite likely remain associated with the mRNA during miRNA-targetinteraction (Figure 12D). After cleavage initiatedbymiRNAs, if theseupstream EJCs can stop 39 to 59 RNA decay by the exosome, theymay also leave footprints in the RNAdegradome. However, 39 to 59RNA decay of nonpolyadenylated 59 cleavage remnants will gen-erate degradation intermediates with a 39 end delineated by the 39edge of the EJC. Hence, even if the EJC footprints upstream ofthe miRNA-guided cleavage sites are present in the RNA de-gradome, they cannot be captured by PARE or similar methods,which profile the 59 ends but not 39 ends of degradation frag-ments with a polyadenylated tail.

While several previous studies havedemonstrated that AGO1 isassociated with polysomes (Lanet et al., 2009; Li et al., 2013b,2016), the miRNA-mediated regulation of EJC-bound mRNAsappearsnot to requireprior translation. Thereare twoscenarios forthe targeting of EJC-bound mRNAs by plant miRNAs. Since theloading ofmature plantmiRNAs to AGO1 can occur in the nucleus(Bologna et al., 2018), it is possible that some miRNA-inducedsilencing complexes can function before being exported from thenucleus, which is rich in EJC-bound mRNAs. Alternatively, cy-toplasmic miRNA-induced silencing complexes might be able torecognize their targets, which remain associated with EJCs, anddirect cleavage before translation. Considering these two sce-narios, themiRNA targetswith a lower efficiency of nuclear exportor translation during the pioneer round may have a higher chanceof being cleaved before translation and accumulating EJC foot-prints.Theanalysisof the59PendsofmiRNAtargets innuclearandcytoplasmic fractions will help to answer this question. By con-trast, for miRNAs such as miR398 and miR444, which pair withtarget sites close to the EJC region and direct cleavage withouttriggering EJC footprint production (Figure 9), prior translation oftargeted mRNAs is likely to be essential for their regulation. OthermiRNA targets that possess target sites outside the EJC regionbut lack prominent EJC footprints, such as Arabidopsis ARPN(Figure 9), may have few EJCs deposited on the transcripts. Al-ternatively, their cleavage remnants might be mainly degradedthrough the 39 to 59RNA decay pathway, which is not able to yieldEJC-protected 59 RNA ends.

Interplay between the miRNA and NMD Pathways

In human cells, it was demonstrated that miRNAs are able to beloaded on EJC-boundmRNAs, thereby repressing the translationand decay of some NMD targets (Choe et al., 2010). As ourstudy also demonstrated that some plant miRNAs can targetEJC-bound mRNAs, this raises the possibility that some planttranscripts are targeted by both RNA degradation pathways. In-terestingly, a previous analysis of miRNA and NMD targets inArabidopsis showed that there was little overlap between the twogroups and that these groups were enriched in distinct biologicalprocesses (Zhang et al., 2013). In spite of these findings, interplaybetween these two RNA degradation pathways might occur inother plant species. In Solanum and Fabaceae species, 22-nucleotide miRNAs repress the expression of a large number ofnucleotide binding site leucine-rich repeat (NBS-LRR) resistancegenes (Zhai et al., 2011; Shivaprasad et al., 2012). Interestingly,normal transcripts of some NBS-LRR genes, which harbor longor intron-containing 39 UTRs, are NMD targets in Arabidopsis(Gloggnitzer et al., 2014). If NMDalso regulatesNBS-LRRgenes inSolanum or Fabaceae species, the miRNA and NMD pathwaysmight be interchangeable or cooperate to inhibit the plant immunesystem.As total RNA is generally used for genome-wide profiling of 59P

ends, the existence of EJC footprints in the RNA degradome al-lowsus to infer thatmRNAdegradationoccursbeforesteadystatetranslation in the cytoplasm and nucleus by known and unknownmechanisms. Our work leads to new possibilities for obtainingagreater understandingofmRNAdegradationbefore steadystatetranslation, which is currently underappreciated and has manyunknowns.

METHODS

Plant Materials and Growth Conditions

All wild-type, mutant, and transgenic Arabidopsis (Arabidopsis thaliana)plants used in this study are in the Columbia-0 background. Arabidopsismutants phyB-9 (CS6217), pho2 (CS8508), xrn4-6 (SALK_014209), xrn2-1(SALK_041148), xrn3-3 (SAIL_1172C07), xrn2-1 xrn3-3 xrn4-6, fry1-6(SALK_020882), smg7-1 (SALK_073354), and ago1-3 were describedpreviously byReedet al. (1993),Delhaize andRandall (1995), Bohmert et al.(1998), Gy et al. (2007), Riehs et al. (2008), and Hirsch et al. (2011). Thedouble mutants phyB-9 xrn4-6 and pho2 xrn4-6 were identified in an F2population by genotyping.

All plants except for smg7-1 and its wild-type controls were grown at22°Cwitha16-h-light (110–140mmolm22 s21,PAR)/8-h-darkphotoperiodbefore harvesting for RNAextraction. For smg7-1 and itswild-type control,the seedlings were first grown at 22°C, but after genotyping, smg7-1homozygous individuals and wild-type control plants were transferredto 28°C in order to inhibit autoimmune responses in smg7-1. For PARElibrary construction and modified RLM 59 RACE assays, RNA was ex-tracted from the inflorescences of soil-grown plants or from seedlingsgrownon0.8%(w/v)Bacto-agarplatescontaininghalf-strengthMurashigeand Skoog medium (pH 5.7) and 1% (w/v) Suc. For phosphate-starvationtreatment of wild-type plants and Arabidopsis transgenic lines carryingartificial miRNA targets, seedlings were grown on Pi-sufficient agar platesfor 7 d and then transferred to Pi-deficient or Pi-sufficient agar plates for anadditional 7 d before harvesting for RNA extraction. The Pi-sufficient andPi-deficient agar plates contained 1% (w/v) Bacto-agar, half-strength

918 The Plant Cell

Dow




modifiedHoagland nutrient solution, and 1% (w/v) Suc supplementedwith250 mM and 10 mM KH2PO4, respectively.

PARE Library Construction and Sequencing

For each genotype, ;80 mg of total RNA isolated from two separate bi-ological replicates using PureLink Plant RNA Reagent (Thermo FisherScientific) was used for PARE library construction following the protocoldescribed previously (Zhai et al., 2014). PARE libraries were sequenced onthe Illumina HiSeq 2500 platform.

Preprocessing and Mapping of PARE and GMUCT Reads

Previously published PARE data for Arabidopsis and budding yeast(Saccharomyces cerevisiae), Degradome-Seq data for rice (Oryza sativa)and worm (Caenorhabditis elegans), and GMUCT data for Arabidopsisand human (Homo sapiens) HEK293 cells were downloaded from theGene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/).Accession numbers and details for the publicly available 59P end dataanalyzed are given in Supplemental Data Set 3. For PARE and Degradome-Seq data, trimmed reads of 20 to 21 nucleotides with a quality score$ 30were mapped to the corresponding genome and cDNA sequencesdownloaded from TAIR 10 (https://www.arabidopsis.org/), the MSU RiceGenome Annotation Project (http://rice.plantbiology.msu.edu/; RGAP 7),the Saccharomyces Genome Database (https://www.yeastgenome.org/;S288C), and the WormBase database (https://www.wormbase.org/;WS269) using Bowtie 1.2.1.1 with zero mismatches (Langmead et al.,2009). PAREandDegradome-Seq readswithmore than 10genomic hits orthat mapped to the chloroplast genome, mitochondria genome, rRNAs,tRNAs, small nuclear RNAs, or small nucleolar RNAs were discarded. Forthe human GMUCT data, trimmed reads of 30 to 50 nucleotides witha quality score $ 30 were mapped to the human genome and cDNA se-quences downloaded from the Ensembl database (http://www.ensembl.org/; GRCh38p12) using Bowtie 1.2.1.1 with two mismatches. HumanGMUCT reads with more than 10 genomic hits or that mapped to themitochondria genomes, rRNAs, tRNAs, small nuclear RNAs, or smallnucleolar RNAs were discarded. The processing and mapping of theArabidopsis GMUCT data was similar to that of the human GMUCT data,but the Arabidopsis sequences were used for mapping and filtering. Theabundance of PARE, Degradome-Seq, or GMUCT sequences was firstnormalized to the total remaining reads to TP40Mand then assigned to thegenomic or gene position corresponding to the first nucleotide of thesequence. As many genes have multiple isoforms, only exons with a size$ 50 nucleotides in the longest isoform of a gene were included in thefollowing exon analyses.

Metagene Analysis of 59P Ends

For metagene analysis of 59P ends mapping to the 50-nucleotide regionupstream of the exon-exon junction, the normalized occurrence of 59Pends (Pi) starting at each position was calculated using the formula

Pi 5 ∑n

j51

Ci,j∑49i50Ci,j

where Ci,j is the 59P end count starting at position i of exon j and n is thenumber of analyzed exons with a size$ 50 nucleotides and a 59P end rawcount$ 1. Then, the relative frequency of normalized 59P end occurrencesin the 50-nucleotide region was plotted.

For comparison of 59P endsmapping to the canonical EJC region in thewild typeandmutants,exonspossessinga total59Pendcount#10TP40Min the canonical EJC region (26 to 30 nucleotides upstream of the exon-exon junction) in all of the samples were excluded. The abundance of 59P

endsat theTSSofgenesharboring theexons included in thecomparisonof59P ends in the EJC region was calculated as the total 59P end count in the21-nucleotide region symmetrically flanking the TSS based on the TAIR 10annotation. The TSS regions with a total 59P end count# 5 TP40M in all ofthe samples were excluded from the comparison.

Identification and GO Analysis of NMD Targets with 59P End Data

The wild-type Arabidopsis PARE data generated in this study and theDegradome-Seq data for wild-type rice produced by Li et al. (2010) wereused in the identification of putative NMD targets. These targets wereidentified through the analysis of EJC footprints using a custom R script(https://github.com/LabHMChenABRC/EJC-DegradomeAnalysis). We firstselected protein-coding genes with at least one intron-containing genemodel and amaximal 59P peak with a count$ 10 TP40M in representativegene models. If the position of the maximum 59P peak matched one of thefollowing twocriteria, the transcriptwas identifiedasaputativeNMDtarget.(1) The maximum 59P peak was located in the first or second canonicalEJC region downstream of an AS site based on the existence of a non-representative gene model with an altered translation start or termi-nation site comparedwith that of the representative genemodel. NMDofthis type of target was classified as being triggered by AS. (2) Themaximum 59P peak was located in the first or second canonical EJCregion of the exon 39 end at least 50 nucleotides downstream of theannotated stop codon. NMDof this type of targetwas classified as beingtriggered by an intron-containing 39 UTR. The analysis of enriched GOterms for putative NMD targets was performed using PANTHER version14 (Mi et al., 2019).

Artificial miRNA Target Construction andArabidopsis Transformation

The Cauliflower mosaic virus 35S promoter, a 330-bp fragment of theArabidopsisNLA59UTR,and the intron-freeGUS fromthepBI121vectororthe intron-containing GUS gene from the pBISN1 vector were fused andthen cloned into the pCAMBIA1390 vector for Arabidopsis transformation.Homozygous T3 transgenic lines were selected and used for phosphatetreatments. The PCR primers used for cloning artificial miRNA targets arelisted in Supplemental Table 2.

Modified RLM 59 RACE Assay and Quantification of Steady StatemRNA Levels

The GeneRacer kit (Thermo Fisher Scientific) was used to detect the 59Pendsof specificgenes.TotalRNA(2–3mg) isolatedusing thePureLinkPlantRNA Reagent (Thermo Fisher Scientific) or mirVana miRNA Isolation Kit(Thermo Fisher Scientific) was used as a template in modified RLM 59RACE assays that were performed according to the manual of the Gen-eRacer kit, except that the calf intestinal phosphatase and tobacco acidpyrophosphatase treatmentswere skipped.After RNAwas ligated to the59RNA adapter, an oligo(dT) primer was used to synthesize cDNA, whichserved as the template for PCR analysis with a GeneRacer 59 primer anda gene-specific primer. For the modified RLM 59 RACE of PHO2, nestedPCR was performed with a GeneRacer 59 nested primer and a gene-specific nested primer. PCR products were gel purified and cloned intothe pJET1.2 vector or pCR4-TOPOTA vector (ThermoFisher Scientific) forsequencing. Control HeLa total RNA included in the GeneRacer kit wastreated with tobacco acid pyrophosphatase for a spike-in control. Thesame cDNAwas used for the quantification of steady statemRNA levels ofPYM and ACT8 with pairs of gene-specific primers. The PCR primers arelisted in Supplemental Table 2.


Dow



http://www.ncbi.nlm.nih.gov/geo/http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1https://www.arabidopsis.org/https://github.com/LabHMChenABRC/EJC-DegradomeAnalysishttp://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1

RNA Gel Blot Analysis of Small RNAs

Total RNA (5 mg) isolated using the PureLink Plant RNA Reagent (ThermoFisher Scientific) was used for small RNA gel blot analysis following theprocedures described previously by Lee et al. (2015). The probes used forthe detection of miR827 and U6 are listed in Supplemental Table 2.

PYM Overexpression in Protoplasts

The Arabidopsis PYM coding region amplified from cDNAwas cloned intopJD301 to replace the firefly luciferase sequence using the NEBuilder HiFiDNA Assembly Master Mix according to the manufacturer’s recom-mendations (NewEnglandBiolabs). ThePCRprimersused forPYMcloningare listed in Supplemental Table 2. Protoplast isolation and transfectionwere performed as described previously with minor modifications (Houet al., 2016). Arabidopsis mesophyll protoplasts were isolated from 17-d-old rosette leaves. Together with 5 mg of transfection control plasmids,30 mg of pJD301 or pJD-PYM was transfected into 4 3 104 protoplastsusing polyethylene glycol solution. The transfected protoplasts were in-cubated at 22°C in the dark for 19 to 20 h. Then, 2 3 105 transfectedprotoplasts were pooled and lysed for RNA extraction with the mirVanamiRNA Isolation Kit (Thermo Fisher Scientific).

Accession Numbers

The PARE data generated in this study are available in the GEO databaseunder the accession number GSE118215. The previously published PAREdata, Degradome-Seq data, and GMUCT data used in this study areavailable in the GEO database under the accession numbers shown inSupplemental Data Set 3. Sequences of individual genes included in 59Pend analysis using PARE data or modified RLM 59 RACE assays can befound inTAIRor theMSURiceGenomeAnnotationProject databaseunderthe locusnumbers indicated.TheArabidopsismutantsanalyzedandgenescloned can be found in the TAIR database under the following accessionnumbers: AT5G42540 for xrn2-1, AT1G75660 for xrn3-3, AT1G54490 forxrn4-6, AT5G63980 for fry1-6, AT2G18790 for phyB-9, AT2G33770 forpho2, AT5G19400 for smg7-1, AT1G02860 forNLA, AT1G48410 for ago1-3, and AT1G11400 for PYM.

Supplemental Data

Supplemental Figure 1. 59P end analyses of polyadenylated anddeadenylated degradation fragments, using PARE and GMUCT data,all show an enrichment in the canonical EJC region.

Supplemental Figure 2. The EJC regions downstream of PTCs harborSMG7-dependent 59P peaks in Arabidopsis.

Supplemental Figure 3. Genes producing NMD-suppressed ASvariants possess a maximum 59P peak in the EJC region downstreamof the PTC.

Supplemental Figure 4. Dysfunction of SMG7 reduces the number ofEJC footprints in putative NMD targets.

Supplemental Figure 5. The xrn4-enhanced 59P peaks in eRF1-1 aredampened in smg7-1.

Supplemental Figure 6. Some Arabidopsis miRNA targets possessdominant EJC footprints downstream of miRNA-guided cleavage sites.

Supplemental Figure 7. XRN4 contributes to the turnover of 39cleavage remnants and production of EJC footprints in miRNA targets.

Supplemental Figure 8. Some Arabidopsis miRNA targets possesspoor EJC footprints downstream of miRNA-guided cleavage sites.

Supplemental Table 1. GO terms enriched in putative ArabidopsisNMD targets identified from the analysis of EJC footprints in the RNAdegradome.

Supplemental Table 2. Sequences of primers for modified RLM 59RACE, quantification of steady state mRNA, cloning, and RNA gel blotanalysis.

Supplemental Data Set 1. Putative Arabidopsis NMD targets identi-fied from the analysis of EJC footprints in the RNA degradome.

Supplemental Data Set 2. Putative rice NMD targets identified fromthe analysis of EJC footprints in the RNA degradome.

Supplemental Data Set 3. Summary of 59P end data sets used inthis study.

ACKNOWLEDGMENTS

We thank the following staff members and core facilities at AcademiaSinica: Tzyy-Jen Chiou for pho2 and fry1-6 seeds and assistance in thephosphate starvation treatment; Shu-HsingWu for phyB-9 seeds; Erh-MinLai for the pBI121 and pBISN1 vectors; the Genomic Technology Core atthe Institute of Plant and Microbial Biology for PARE library construction;the Plant Tech Core Facility at the Agricultural Biotechnology ResearchCenter for Arabidopsis transformation; Ming-Jung Liu and members ofHo-Ming Chen’s group for critical reading of the manuscript and helpfulsuggestions;Miranda Loney at Academia Sinica andMelissa Lehti-Shiu atMichigan State University for English editing. We also thank Hervé Vau-cheret at INRA for seeds of single and triple mutants of xrn2-1, xrn3-3, andxrn4-6 and Peter Brodersen at the University of Copenhagen for ago1-3seeds. This work was supported by Academia Sinica.

AUTHOR CONTRIBUTIONS

H.-M.C. designed the research; B.-H.H., C.-Y.H., S.-M.T., P.K., and H.-M.C. performed the computational analyses; W.-C.L. carried out theexperiments; B.-H.H., W.-C.L., and H.-M.C. wrote the article; all authorsread and approved the final article.

Received August 28, 2019; revised December 2, 2019; accepted January24, 2020; published January 27, 2020.

REFERENCES

Addo-Quaye, C., Eshoo, T.W., Bartel, D.P., and Axtell, M.J. (2008).Endogenous siRNA and miRNA targets identified by sequencing ofthe Arabidopsis degradome. Curr. Biol. 18: 758–762.

Arciga-Reyes, L., Wootton, L., Kieffer, M., and Davies, B. (2006).UPF1 is required for nonsense-mediated mRNA decay (NMD) andRNAi in Arabidopsis. Plant J. 47: 480–489.

Arribas-Hernández, L., Kielpinski, L.J., and Brodersen, P. (2016).mRNA decay of most Arabidopsis miRNA targets requires sliceractivity of AGO1. Plant Physiol. 171: 2620–2632.

Bannerman, B.P., Kramer, S., Dorrell, R.G., and Carrington, M.(2018). Multispecies reconstructions uncover widespread conser-vation, and lineage-specific elaborations in eukaryotic mRNA me-tabolism. PLoS One 13: e0192633.

Behm-Ansmant, I., Gatfield, D., Rehwinkel, J., Hilgers, V., andIzaurralde, E. (2007). A conserved role for cytoplasmic poly(A)-

920 The Plant Cell

Dow



http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://www.plantcell.org/cgi/content/full/tpc.19.00666/DC1http://dx.doi.org/10.13039/501100001869

binding protein 1 (PABPC1) in nonsense-mediated mRNA decay.EMBO J. 26: 1591–1601.

Bohmert, K., Camus, I., Bellini, C., Bouchez, D., Caboche, M., andBenning, C. (1998). AGO1 defines a novel locus of Arabidopsiscontrolling leaf development. EMBO J. 17: 170–180.

Bologna, N.G., Iselin, R., Abriata, L.A., Sarazin, A., Pumplin, N.,Jay, F., Grentzinger, T., Dal Peraro, M., and Voinnet, O. (2018).Nucleo-cytosolic shuttling of ARGONAUTE1 prompts a revisedmodel of the plant microRNA pathway. Mol. Cell 69: 709–719.e705.

Chang, C.Y., Lin, W.D., and Tu, S.L. (2014). Genome-wide analysis ofheat-sensitive alternative splicing in Physcomitrella patens. PlantPhysiol. 165: 826–840.

Chen, C.Y., and Shyu, A.B. (2003). Rapid deadenylation triggered bya nonsense codon precedes decay of the RNA body in a mammaliancytoplasmic nonsense-mediated decay pathway. Mol. Cell. Biol. 23:4805–4813.

Choe, J., Cho, H., Lee, H.C., and Kim, Y.K. (2010). MicroRNA/Ar-gonaute 2 regulates nonsense-mediated messenger RNA decay.EMBO Rep. 11: 380–386.

Colombo, M., Karousis, E.D., Bourquin, J., Bruggmann, R., andMühlemann, O. (2017). Transcriptome-wide identification of NMD-targeted human mRNAs reveals extensive redundancy between SMG6-and SMG7-mediated degradation pathways. RNA 23: 189–201.

Conti, E., and Izaurralde, E. (2005). Nonsense-mediated mRNA de-cay: Molecular insights and mechanistic variations across species.Curr. Opin. Cell Biol. 17: 316–325.

Couttet, P., and Grange, T. (2004). Premature termination codonsenhance mRNA decapping in human cells. Nucleic Acids Res. 32:488–494.

Delhaize, E., and Randall, P.J. (1995). Characterization of a phos-phate-accumulator mutant of Arabidopsis thaliana. Plant Physiol.107: 207–213.

Drechsel, G., Kahles, A., Kesarwani, A.K., Stauffer, E., Behr, J.,Drewe, P., Rätsch, G., and Wachter, A. (2013). Nonsense-mediated decay of alternative precursor mRNA splicing variants isa major determinant of the Arabidopsis steady state transcriptome.Plant Cell 25: 3726–3742.

Eberle, A.B., Lykke-Andersen, S., Mühlemann, O., and Jensen,T.H. (2009). SMG6 promotes endonucleolytic cleavage of nonsensemRNA in human cells. Nat. Struct. Mol. Biol. 16: 49–55.

Feng, J., Li, J., Gao, Z., Lu, Y., Yu, J., Zheng, Q., Yan, S., Zhang, W.,He, H., Ma, L., and Zhu, Z. (2015). SKIP confers osmotic toleranceduring salt stress by controlling alternative gene splicing in Arabi-dopsis. Mol. Plant 8: 1038–1052.

Gatfield, D., and Izaurralde, E. (2004). Nonsense-mediated mes-senger RNA decay is initiated by endonucleolytic cleavage in Dro-sophila. Nature 429: 575–578.

Gatfield, D., Unterholzner, L., Ciccarelli, F.D., Bork, P., andIzaurralde, E. (2003). Nonsense-mediated mRNA decay in Dro-sophila: At the intersection of the yeast and mammalian pathways.EMBO J. 22: 3960–3970.

Gehring, N.H., Lamprinaki, S., Kulozik, A.E., and Hentze, M.W.(2009). Disassembly of exon junction complexes by PYM. Cell 137:536–548.

German, M.A., et al. (2008). Global identification of microRNA-targetRNA pairs by parallel analysis of RNA ends. Nat. Biotechnol. 26:941–946.

Gloggnitzer, J., Akimcheva, S., Srinivasan, A., Kusenda, B., Riehs,N., Stampfl, H., Bautor, J., Dekrout, B., Jonak, C., Jiménez-Gómez,J.M., Parker, J.E., and Riha, K. (2014). Nonsense-mediated mRNAdecay modulates immune receptor levels to regulate plant antibacterialdefense. Cell Host Microbe 16: 376–390.

Gregory, B.D., O’Malley, R.C., Lister, R., Urich, M.A., Tonti-Filippini, J., Chen, H., Millar, A.H., and Ecker, J.R. (2008). A linkbetween RNA metabolism and silencing affecting Arabidopsis de-velopment. Dev. Cell 14: 854–866.

Gy, I., Gasciolli, V., Lauressergues, D., Morel, J.B., Gombert, J.,Proux, F., Proux, C., Vaucheret, H., and Mallory, A.C. (2007).Arabidopsis FIERY1, XRN2, and XRN3 are endogenous RNA si-lencing suppressors. Plant Cell 19: 3451–3461.

Harigaya, Y., and Parker, R. (2012). Global analysis of mRNA decayintermediates in Saccharomyces cerevisiae. Proc. Natl. Acad. Sci.USA 109: 11764–11769.

He, F., Li, X., Spatrick, P., Casillo, R., Dong, S., and Jacobson, A.(2003). Genome-wide analysis of mRNAs regulated by the nonsense-mediated and 59 to 39 mRNA decay pathways in yeast. Mol. Cell

widespread exon junction complex footprints in the rna ... · or yeast ever shorter telomeres1;...

Documents