bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 A UPF1 variant drives conditional remodeling of nonsense-mediated mRNA decay
2
3 Sarah E. Fritz1, Soumya Ranganathan1, and J. Robert Hogg1*
4
5 1Biochemistry and Biophysics Center 6 National Heart, Lung, and Blood Institute 7 National Institutes of Health 8 Bethesda, MD 20892 9 USA 10 11 *Correspondence: [email protected] 12
13 Abstract
14 The nonsense-mediated mRNA decay (NMD) pathway monitors translation termination to 15 degrade transcripts with premature stop codons and regulate thousands of human genes. Due 16 to the major role of NMD in RNA quality control and gene expression regulation, it is important 17 to understand how the pathway responds to changing cellular conditions. Here we show that an
18 alternative mammalian-specific isoform of the core NMD factor UPF1, termed UPF1LL, enables
19 condition-dependent remodeling of NMD specificity. UPF1LL associates more stably with
20 potential NMD target mRNAs than the major UPF1SL isoform, expanding the scope of NMD to 21 include many transcripts normally immune to the pathway. Unexpectedly, the enhanced
22 persistence of UPF1LL on mRNAs supports induction of NMD in response to rare translation
23 termination events. Thus, while canonical NMD is abolished by translational repression, UPF1LL 24 activity is enhanced, providing a mechanism to rapidly rewire NMD specificity in response to 25 cellular stress.
1 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 Introduction
2 Nonsense-mediated mRNA decay (NMD) is an evolutionarily conserved mRNA quality-
3 control pathway that degrades transcripts undergoing premature translation termination (Lavysh
4 and Neu-Yilik, 2020; Smith and Baker, 2015). In addition, NMD performs a critical regulatory
5 role by governing the turnover of ~5-10% of the transcriptome, including mRNAs with upstream
6 open reading frames (uORFs), introns downstream of the stop codon, or long 3’ untranslated
7 regions (UTRs) (Kishor et al., 2019). Despite extensive studies of the large impact of NMD on
8 the transcriptome, the mechanisms by which the pathway selects its regulatory and quality
9 control targets are poorly understood.
10 A suite of conserved NMD factors acts in concert with general mRNA binding proteins
11 and RNA decay enzymes to identify and degrade target mRNAs. The RNA helicase UPF1 is a
12 central coordinator of the NMD pathway, as it directly binds mRNA and functions at multiple
13 steps in the selection and degradation of target transcripts (Kim and Maquat, 2019). Additional
14 core NMD factors UPF2 and UPF3 promote UPF1 activity and link UPF1 to the exon junction
15 complex (EJC), which strongly stimulates decay (Chamieh et al., 2008; Le Hir et al., 2000,
16 2000b). In many eukaryotes, NMD execution also depends on the SMG1, 5, 6, and 7 proteins
17 (Causier et al., 2017; Hodgkin et al., 1989; Page et al., 1999; Pulak and Anderson, 1993).
18 Phosphorylation of UPF1 by the SMG1 kinase is a necessary and critical step in promoting the
19 NMD pathway (Kashima et al., 2006), as phosphorylated UPF1 recruits the SMG6
20 endonuclease and/or general decapping and deadenylation enzymes through the SMG5/7
21 heterodimer (Eberle et al., 2009; Huntzinger et al., 2008; Loh et al., 2013).
22 In addition to the functions of specialized NMD proteins, substrate selection and
23 degradation by the NMD pathway requires the translation termination machinery to detect in-
24 frame stop codons (Karousis and Mühlemann, 2018). Although the exact molecular details
25 remain to be elucidated, widely accepted models of NMD state that interactions between core
26 NMD factors and a terminating ribosome are necessary for decay. Because of the strict
2 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 dependence of NMD on translation termination, decay efficiency of canonical NMD targets is
2 expected to be tightly linked to translation efficiency. However, in contrast to this prevailing
3 model, there is evidence that NMD efficiency for some targets is actually enhanced during
4 conditions of impaired translation (Martinez-Nunez et al., 2017). These data warrant a more
5 extensive investigation into the role of translation in shaping target specificity by the NMD
6 pathway, particularly during changing physiological conditions.
7 The ability of UPF1 to bind and hydrolyze ATP is critical to the selection and degradation
8 of potential NMD substrates (Franks et al., 2010; Lee et al., 2015). Specifically, an 11 amino
9 acid regulatory loop in domain 1B of the helicase core functions to regulate the affinity of UPF1
10 for RNA in the presence of ATP (Chakrabarti et al., 2011; Cheng et al., 2007; Gowravaram et
11 al., 2018). When UPF1 binds and hydrolyzes ATP, this regulatory loop protrudes into the RNA
12 binding channel, causing UPF1 to release RNA. Intriguingly, mammals undergo an alternative
13 splicing event to express two UPF1 isoforms that differ only in length of the regulatory loop
14 (Figure 1A). Almost all NMD studies to date have focused on the more abundant UPF1 “short
15 loop” isoform (designated herein UPF1SL), which contains the 11 amino acid regulatory loop that
16 weakens the affinity of UPF1 for RNA in the presence of ATP. Alternative 5’ splice site usage in
17 exon 7 of UPF1 generates a second UPF1 isoform that extends the regulatory loop to 22 amino
18 acids. This naturally occurring UPF1 “long loop” isoform (designated herein UPF1LL), which
19 represents ~15-25% of total UPF1 mRNA in diverse cell and tissue types (Figure 1B), has
20 increased catalytic activity and a higher affinity for RNA in the presence of ATP than the UPF1SL
21 isoform (Gowravaram et al., 2018). However, whether the differential biochemical properties of
22 the UPF1 isoforms affect NMD specificity in cells is unknown.
23 Here, we show that the UPF1LL isoform gives the mammalian NMD pathway the latent
24 ability to remodel NMD target specificity in response to changing physiological conditions. We
25 show that changes to translation efficiency or UPF1 isoform expression conditionally alter which
3 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 mRNAs are selected and degraded by the NMD pathway, greatly expanding the scope of NMD
2 in mammalian gene expression control.
3
4 Results
5
6 UPF1LL contributes to NMD under normal cellular conditions
7 As the endogenous UPF1LL-encoding splice variant accounts for only ~15-25% of total
8 UPF1 mRNA in most human and mouse cells (Figure 1B) (Gowravaram et al., 2018), we first
9 sought to assess the functional significance of basal UPF1LL activity in regulating gene
10 expression. RNA-seq of human HEK-293 cells depleted of the UPF1LL isoform using a short
11 interfering RNA (siRNA) that specifically targeted UPF1LL mRNA levels without perturbing
12 UPF1SL expression (Figure 1C and Figure 1-figure supplement 1A) uncovered a population
13 (~1500 transcripts) that increased in abundance by at least 1.4-fold (Supplementary file 1). To
14 test whether these changes in mRNA abundance were due to activity of UPF1LL in the NMD
15 pathway, we compared our UPF1LL-knockdown RNA-seq dataset with a previously published
16 catalog of high-confidence UPF1 and/or SMG6/SMG7 targets (Colombo et al., 2017).
17 Transcripts up-regulated by UPF1LL knockdown comprised a small but significant subpopulation
18 of mRNAs identified as UPF1 and SMG6/SMG7 substrates, indicating that UPF1LL does
19 contribute to the decay of NMD targets (Figure 1C). Analysis of canonical NMD substrates
20 revealed that knockdown of endogenous UPF1LL did not systematically affect the abundance of
21 mRNAs predicted to undergo decay stimulated by the EJC, as only a small number of such
22 transcripts exhibited enhanced expression upon UPF1LL knockdown (Figure 1-figure
23 supplement 1B). Quantitative PCR of select transcripts, including well validated NMD targets
24 (Lareau et al., 2007; Ni et al., 2007), corroborated the transcriptome-wide RNA-seq results
25 (Figure 1-figure supplement 1C).
4 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
UPF1 UPF1 A SL LL Exon 7 Exon 7 Alternative splice isoforms B
0.50 11 aa 22 aa Regulatory loop length ATP ATP LL
1B 1C 0.25 Fraction UPF1 UPF1 SL UPF1 loop LL loop
0.00 Skin Liver Lung Brain Heart Testis Blood Colon Ovary Nerve Breast Uterus Vagina Kidney Spleen Muscle Thyroid Bladder Pituitary Prostate Stomach RecA2 Pancreas Esophagus Cervix Uteri Blood Vessel Adrenal Gland Salivary Gland Small Intestine Fallopian Tube Adipose Tissue RecA1
UPF1 targets C (Colombo ; < 10⁻15) 806
NT UPF1LL 214 siRNA siRNA 1112
UPF1LL 260 UPF1SL UPF1 targets 1190 LL 144 724 SMG6/7 targets (Colombo ; = 9.6x10⁻12)
Schematic representation of alternative 5' splice site usage in exon 7 of mammalian UPF1 that results in two UPF1 protein isoforms. Ribbon diagram of human UPF1SL helicase core overlaid with that of human UPF1LL. The structure of the regulatory loop in domain 1B for each variant is indicated. This figure was generated using Protein Data Bank accessions
2XZP and 6EJ5 (Chakrabarti et al., 2011; Gowravaram et al., 2018). Box plot of the fraction of UPF1LL mRNA expressed in indicated human tissues, as determined from the Genotype-Tissue Expression (GTEx) project. Boxes indicate interquartile ranges, and bars indicate 10-90% ranges. Semiquantitative RT-PCR of UPF1SL or UPF1LL transcript levels following transfection of HEK-293 cells with a non-targeting (NT) siRNA or a siRNA that specifically targets the UPF1LL isoform. Venn diagram of overlapping targets from RNA-seq that showed significant stabilization with UPF1LL (this dataset), total UPF1, or SMG6/7 knockdown (Colombo et al., 2017).
5 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 UPF1LL overexpression causes downregulation of transcripts with long 3’UTRs
2 The observation that specific depletion of UPF1LL affected a select subpopulation of
3 mRNAs indicated that it may perform distinct functions from those conducted by the major
4 UPF1SL protein. To gain insight into how the biological activities of the two UPF1 isoforms differ,
5 we engineered HEK-293 stable cell lines to express CLIP-tagged UPF1LL or UPF1SL, with a
6 GFP-expressing stable line as a control. We elected to express CLIP-tagged UPF1 constructs in
7 order to enable downstream biochemical analysis of differential UPF1 mRNA binding (Kishor et
8 al., 2020).
9 Previous work has suggested that SMG1-mediated UPF1 phosphorylation increases
10 with increasing residence time of UPF1 on mRNPs (Durand et al., 2016). Because UPF1LL has
11 been found to have higher affinity for RNA in the presence of ATP than UPF1SL in vitro
12 (Gowravaram et al., 2018), we hypothesized that it would more stably associate with mRNPs in
13 cells, in turn promoting its phosphorylation. Consistent with this hypothesis, immunoblotting with
14 a phospho-(S/T)Q antibody revealed that CLIP-UPF1LL was more highly phosphorylated than
15 CLIP-UPF1SL (Figure 2A). The two splice variants exhibited a similar capacity to be
16 phosphorylated, as treatment of cells with puromycin increased the phosphorylation of both
17 UPF1 proteins to equivalent levels, perhaps due to reduced displacement by elongating
18 ribosomes (Dang et al., 2009; Zünd et al., 2013).
19 To analyze the target specificity of the UPF1 splice variants, we performed RNA-seq on
20 CLIP-UPF1 and control cells (Supplementary file 2). Previous studies have indicated that
21 overexpression of UPF1SL does not enhance overall NMD activity, potentially due to feedback
22 control by the pathway (Huang et al., 2011). Consistent with these findings, analysis of mRNAs
23 previously observed to be up-regulated upon total UPF1 knockdown (Baird et al., 2018)
24 revealed that ~5 to 10-fold stable overexpression of CLIP-UPF1SL had a negligible impact on the
25 expression of previously identified NMD targets (Figure 2-figure supplement 1A). Expression of
26 CLIP-UPF1LL caused small increases in the levels of some previously identified NMD targets but
6 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
A - puromycin + puromycin B stably overexpress: CLIP-UPF1 SL LL SL LL SL or RNA-seq of total RNA GFP GFP UPF1 UPF1 UPF1 UPF1 CLIP-UPF1 P-CLIP-UPF1 LL P-Endogenous-UPF1 CLIP-UPF1 Endogenous-UPF1 1.0 β-Actin
5 UPF1 = 0.23 4 SL 0.5 UPF1LL 3 3’UTRs < 477 nt (n = 3126) = 0.021 2 3’UTRs 477-1579 nt (n = 3732; > 0.99) Cumulative frequency 1 3’UTRs > 1579 nt (n = 3815; < 1x10-15) Relative level phophorylation 0 0.0 - puromycin + puromycin -1.0 -0.5 0.0 0.5 1.0
UPF1LL vs UPF1SL log2(fold change)
relative mRNA abundance decreased increased
C D canonical NMD targets protected by PTBP1/hnRNP L **** 1 1.0 **** UPF1 **** LL
UPF1SL 0.5
(fold change) 0 2 log 0.0 SL *
vs UPF1 -0.5 -1 * LL *
(relative mRNA abundance) * 2 UPF1 3'UTRs > 1579 nt -1.0 log * * n= 251 1075 1447 1042
SMG1SMG5 PTENFMR1 DCP2CDK6FGF9 CSRP1 eIF5A2 OSBPL8 PTBP1/hnRNP L motif density within the 3'UTR TBL1XR1FAM171B
SRSF2SRSF3 (PTC+)SRSF6 (PTC+) (PTC+)
. Western blots of CLIP-UPF1SL and CLIP-
UPF1LL phosphorylation status at steady state (- puromycin) or following incubation with puromycin (+ puromycin). Graph below represents the relative phosphorylation level of each CLIP-UPF1 loop variant, as determined by dividing the phosphorylation signal for each protein (P-CLIP-UPF1) by that of its corresponding total amount (CLIP-UPF1) and normalizing to CLIP-UPF1SL at steady state. Error bars indicate SD (n = 3). Statistical significance was determined by unpaired Student’s t-test. All immunoblots shown were generated from the same membrane. CDF plot of relative mRNA abundance as determined from RNA-seq following CLIP-UPF1LL vs CLIP-UPF1SL overexpression. mRNAs were binned according to 3’UTR lengths (short, medium, long). Statistical significance was determined by K-W test, with Dunn's correction for multiple comparisons. Box plot of relative mRNA abundance as determined from RNA-seq following CLIP-UPF1LL vs CLIP-UPF1SL overexpression. mRNAs with long 3’UTRs (> 1579 nt) were binned by PTBP1 and/or hnRNP L motif density within the 3'UTR. Statistical significance was determined by K-W test, with Dunn's correction for multiple comparisons (**** < 1x10-12). Boxes indicate interquartile ranges, and bars indicate Tukey whiskers. qPCR analysis of indicated transcripts from CLIP-UPF1 overexpression RNA-seq experiments. Relative fold changes are in reference to the GFP expressing control line. Significance of CLIP-UPF1LL overexpression was compared to CLIP-UPF1SL. Asterisk (*) indicates < 0.05, as determined by unpaired Student’s t-test. Error bars indicate SD (n = 3).
7 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 did not cause systematic changes in the abundance of putative EJC-stimulated NMD targets,
2 although some mRNAs in this class were modestly perturbed (Figure 2-figure supplement 1B
3 and 1C). These findings support that overexpression of CLIP-UPF1LL does not substantially
4 impair cellular NMD functions but may cause some alteration of NMD target specificity.
5 Having determined that overexpression of UPF1LL had only minor effects on the
6 abundance of established NMD targets, we next asked whether UPF1LL overexpression affected
7 the abundance of mRNAs not normally regulated by NMD. Transcripts with long 3’UTRs
8 represent a large population of potential NMD targets (Hurt et al., 2013; Yepiskoposyan et al.,
9 2011), only some of which are degraded by the pathway under normal conditions (Toma et al.,
10 2015). Indeed, subdivision of the transcriptome according to 3’UTR lengths (first tertile: < 477
11 nt; second tertile: 477-1578 nt; third tertile: > 1579 nt) revealed differential effects of the UPF1
12 isoforms on mRNAs containing long 3’UTRs. Specifically, mRNAs with the longest 3’UTRs were
13 more likely to be reduced in abundance by UPF1LL overexpression than mRNAs with shorter
14 3’UTRs (Figure 2B). These results demonstrate that UPF1LL has an expanded reach that
15 includes a population of long 3’UTR-containing transcripts that are not normally affected by the
16 pathway.
17
18 UPF1LL downregulates mRNAs normally protected from NMD
19 We have identified hundreds to thousands of mammalian long 3’UTRs that evade UPF1
20 recognition and decay by association with the protective RNA-binding proteins (RBPs)
21 polypyrimidine tract binding protein 1 (PTBP1) and heterogeneous nuclear ribonucleoprotein L
22 (hnRNP L) (Ge et al., 2016; Kishor et al., 2018, 2020). The observation that UPF1LL has an
23 expanded reach to target long 3’UTRs not normally affected by NMD led us to ask whether
24 mRNAs protected by PTBP1 and/or hnRNP L undergo enhanced downregulation with UPF1LL
25 overexpression. To explore this possibility, we further subdivided the population of long 3’UTR-
26 containing mRNAs according to the density of PTBP1 and hnRNP L binding sites in their
8 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 3’UTRs. We previously showed that transcripts containing a low density of PTBP1 and/or
2 hnRNP L binding sites were more likely to be bound and degraded by UPF1 than transcripts
3 containing a high density of binding sites for the protective proteins (Kishor et al., 2018).
4 Remarkably, mRNAs with a moderate or high density of 3’UTR PTBP1 and/or hnRNP L binding
5 sites were more likely to be down-regulated by UPF1LL overexpression than those with a low
6 density of binding sites (Figure 2C). Quantitative PCR of select transcripts confirmed these
7 transcriptome-wide results of RNA-seq (Figure 2D), including CSRP1, which we have previously
8 established as a long 3’UTR-containing mRNA that undergoes decay upon hnRNP L
9 knockdown or mutation of hnRNP L binding sites in its 3’UTR (Kishor et al., 2018).
10 To investigate whether the observed changes in mRNA abundance with UPF1LL
11 overexpression were due to decreased stability, we used REMBRANDTS software to infer
12 changes in mRNA stability based upon differences in the relative abundance of exonic and
13 intronic reads from each gene (Alkallas et al., 2017). The effects of UPF1SL overexpression on
14 mRNA abundance were not correlated with changes to mRNA stability (Spearman’s ⍴ = -0.45; P
15 = 10-5; Figure 2-figure supplement 1D), corroborating the previous observation that moderate
16 UPF1SL overexpression does not enhance NMD activity (Huang et al., 2011). In contrast, the
17 effects of UPF1LL overexpression on mRNA abundance and changes in stability were highly
18 correlated (Spearman’s ⍴ = 0.65; P < 10-15; Figure 2-figure supplement 1E), with transcripts
19 having long 3’UTRs and a high density of PTBP1 and/or hnRNP L binding motifs showing
20 preferential destabilization with UPF1LL overexpression (Figure 2-figure supplement 1F and 1G).
21 These data support the conclusion that the UPF1LL isoform is able to overcome the protective
22 proteins to promote decay of mRNAs normally shielded from NMD.
9 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
A ssRNA for hybridization UPF1 loading region
PTBP1 binding site
+ UPF1∆CH + UPF1∆CH + PTBP1
PTBP1
ATP ATP PTBP1 binding site UPF1SL UPF1SL + ATP and quench strand vs.
PTBP1 ADP P
ATP
UPF1SL PTBP1 binding site UPF1LL
ATP
UPF1SL
B No protein control C No protein control
UPF1SL∆CH + PTBP1 UPF1LL∆CH + PTBP1
UPF1SL∆CH UPF1LL∆CH
30000 30000 (A.U.)
20000 20000
10000 10000 Relative Fluorescence Relative Fluorescence (A.U.)
0 500 1000 0 500 1000 Time (seconds) Time (seconds)
Scheme of the fluorescence-based
unwinding assay to monitor UPF1 translocation in real time (Fritz et al., 2020). UPF1SLΔCH translocation along an RNA substrate harboring a high-affinity PTBP1 binding site in the absence and presence ofPTBP1.
Results as in (B) but with UPF1LLΔCH. Results of four technical replicates are shown for each dataset and represent at least three independent experiments. Shaded region indicates SD.
10 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 UPF1LL is less sensitive to PTBP1-mediated inhibition of translocation
2 We have proposed that the protective RBPs PTBP1 and hnRNP L promote UPF1
3 dissociation from potential NMD substrates prior to decay induction (Fritz et al., 2020). In
4 support of this model, deletion of the regulatory loop from the helicase core of recombinant
5 UPF1SL rendered UPF1 less sensitive to PTBP1 inhibition in vitro (Fritz et al., 2020). Because
6 UPF1 variants with a truncated regulatory loop and the physiological UPF1LL isoform containing
7 the extended 22 amino acid loop both exhibit a greater affinity for RNA in the presence of ATP
8 than the PTBP1-sensitive UPF1SL isoform (Gowravaram et al., 2018), we hypothesized that
9 UPF1LL could overcome the translocation inhibition by PTBP1.
10 We recently established a real-time assay to monitor UPF1 helicase activity (Fritz et al.,
11 2020). In this assay, UPF1 translocation and duplex unwinding causes a fluorescently labeled
12 oligonucleotide to be displaced from the assay substrate (Figure 3A, left). An excess of
13 complementary oligonucleotide labeled with a dark quencher is provided, causing a decrease in
14 fluorescence with increased UPF1 unwinding. Inhibition of UPF1 translocation results in
15 sustained fluorescence over time, allowing for the determination of inhibitory effects on UPF1
16 unwinding activity. We leveraged this system to compare UPF1LL versus UPF1SL translocation
17 on a duplexed RNA substrate harboring a high affinity PTBP1 binding site (Figure 3A, right)
18 (Fritz et al., 2020). For these experiments, we compared the activity of UPF1 proteins containing
19 the helicase core but lacking the autoinhibitory N-terminal cysteine-histidine domain (UPF1ΔCH)
20 (Figure 3-figure supplement 1A and 1B) (Chakrabarti et al., 2011; Fiorini et al., 2012; Fritz et al.,
21 2020). The ability of UPF1SLΔCH to unwind the duplexed RNA was impaired by the presence of
22 PTBP1 (Figure 3B), as previously observed (Fritz et al., 2020). In contrast, inhibition of
23 UPF1LLΔCH translocation by PTBP1 was reduced by extension of the regulatory loop (Figure
24 3C). These results indicate that UPF1LL can overcome the translocation inhibition by PTBP1,
25 reinforcing the conclusion that PTBP1-mediated UPF1 dissociation depends on the steric clash
26 between the UPF1 regulatory loop and RNA.
11 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 UPF1LL is able to bind NMD-resistant transcripts
2 The resistance of UPF1LL to inhibition by PTBP1 in vitro suggested that it may have
3 altered RNA-binding specificity in cells. We therefore used the CLIP-UPF1 stable lines to ask
4 whether the distribution of UPF1LL across the transcriptome differs from that of UPF1SL. CLIP-
5 UPF1 complexes were isolated from whole cell extracts by streptavidin affinity purification using
6 a CLIP-biotin substrate (Figure 4A and Figure 4-figure supplement 1A), with GFP-expressing
7 cell lines as a negative control for interaction specificity (Kishor et al., 2020). Bound mRNAs
8 were then extracted and used for sequencing library preparation. Because recovery of RNA
9 from GFP samples was at least 100-fold lower than from CLIP-UPF1 affinity purifications, only
10 UPF1 samples were analyzed by RNA-seq. UPF1 occupancy was assessed by normalizing the
11 abundance of transcripts in RIP-seq samples to their abundance in total RNA-seq (UPF1 RIP-
12 seq efficiency; Supplementary file 2). The two UPF1 isoforms recovered thousands of mRNAs
13 with similar efficiency, and both strongly enriched mRNAs up-regulated by total UPF1 depletion
14 (Figure 4-figure supplement 1B), consistent with previous observations of preferential UPF1
15 association with NMD targets (Hurt et al., 2013; Kishor et al., 2018; Zünd et al., 2013). The
16 efficiency of mRNA co-purification with both UPF1SL and UPF1LL increased with 3’UTR length
17 (Figure 4B), reinforcing previous observations that UPF1 binding correlates with 3’UTR length
18 (Baker and Hogg, 2017; Hogg and Goff, 2010; Hurt et al., 2013; Kurosaki et al., 2014).
19 Distinctly, the preferential downregulation of long 3’UTR-containing mRNAs by UPF1LL was
20 accompanied by a greater propensity to associate with the longest class of 3’UTRs than UPF1SL
21 (Figure 4B). Indeed, mRNAs down-regulated upon UPF1LL overexpression were highly enriched
22 in UPF1LL relative to UPF1SL RIP-seq (Figure 4C; compare solid and dotted blue lines), and we
23 therefore designated mRNAs with UPF1LL:UPF1SL RIP-seq efficiency greater than 1.5 as
24 putative direct UPF1LL substrates for further analysis.
25 Based on the resistance of UPF1LL to PTBP1 in the in vitro unwinding assays (Figure 3),
26 we hypothesized that UPF1LL would additionally occupy mRNAs with a high density of protective
12 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
A Streptavidin antcBead Magnetic mRNA mRNA CLIP + UPF1 Biotin UPF1 CLIP Biotin Benzylcytosine Covalent biotinylation in extracts Affinity purification of biotinylated CLIP-UPF1
B C 1.0 1.0
3’UTRs < 477 nt (n = 4082) mRNAs down in UPF1LL overexpression (n = 1247) UPF1LL ( = 0.0065) UPF1 ( < 1x10-15) UPF1 ( = 1.4x10-11) 0.5 UPF1SL 0.5 LL SL 3’UTRs 477-1578 nt (n = 4081) mRNAs up in UPF1LL overexpression (n = 1208) UPF1 ( = 7.7x10-8) LL -15 UPF1LL ( < 1x10 ) UPF1SL ( = 0.25) UPF1SL
3’UTRs > 1579 nt (n = 4098) Cumulative frequency Remaining mRNAs (n = 9806) Cumulative frequency UPF1 ( < 1x10-15) LL UPF1LL UPF1SL UPF1SL 0.0 0.0 0 10 20 0 10 20 30 UPF1 RIP-seq efficiency (IP/Input) UPF1 RIP-seq efficiency (IP/Input)
enrichment in enrichment in RIP RIP decreased increased decreased increased
D E short 3'UTRs medium 3'UTRs long 3'UTRs <477 nt 477-1578 nt >1579 nt canonical NMD targets protected by PTBP1/hnRNP L **** 200 **** UPF1 2 **** LL * **** * *** **** UPF1 * **** * SL 150 * RIP-seq efficiency 100 * * SL 1 * * * * *
vs UPF1 50 * LL UPF1 RIP efficiency (IP/Input)
UPF1 0 0 n= 1449 555 388 734 950 1066 816 900 1139 1963 2335 1930
PTENFMR1 DCP2CDK6FGF9 PTBP1/ SMG5CSRP1 eIF5A2 hnRNP L TBL1XR1FAM171BOSBPL8 motif density within 3'UTRs SRSF2SRSF3 (PTC+)SRSF6 (PTC+) (PTC+)
Scheme for the CLIP-UPF1 affinity purification (RIP) assay.
CDF plot of recovered mRNAs as determined from RIP-seq efficiency in CLIP-UPF1LL or CLIP-UPF1SL affinity
purifications. mRNAs were binned according to 3'UTR length. Significance of mRNA co-purification withCLIP-UPF1LL
was compared to that of CLIP-UPF1SL, and was determined by K-S test. CDF plot as in (B), with mRNAs binned
by sensitivity (either up or down-regulated) to UPF1LL overexpression. Significance of indicated mRNA co-purification
with CLIP-UPF1LL or CLIP-UPF1SL is relative to those unaffected by UPF1LL overexpression (remaining mRNAs), and was determined by K-W test with Dunn’s correction for multiple comparisons. Box plot of recovered mRNAs as
determined from RIP-seq efficiency in CLIP-UPF1LL vs CLIP-UPF1SL affinity purifications. mRNAs were binned according to 3’UTR length (short, medium, long) and then further subdivided by PTBP1 and/or hnRNP L motif density within the 3'UTR. Statistical significance was determined by K-W test, with Dunn's correction for multiple comparisons (* < 0.03, *** < 0.0002, **** < 2x10-11). Boxes indicate interquartile ranges, and bars indicate Tukey whiskers. qPCR analysis of indicated transcripts from UPF1 RIP-seq experiments. Relative fold enrichment was determined
by dividing the recovered mRNA by its corresponding input amount. Significance of the CLIP-UPF1LL RIP was
compared to that of CLIP-UPF1SL. Asterisk (*) indicates < 0.05, as determined by unpaired Student’s t-test. Error bars indicate SD (n = 3).
13 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 protein binding sites. Consistent with this idea, UPF1LL and UPF1SL exhibited almost identical
2 recovery of mRNAs with long 3’UTRs containing a low density of PTBP1 and hnRNP L motifs,
3 while mRNAs with moderate or high protective protein motif density were preferentially
4 recovered by UPF1LL (Figure 4D). Quantitative PCR of select transcripts confirmed these
5 transcriptome-wide results of RIP-seq (Figure 4E). These data support that UPF1LL can
6 circumvent the protective effects of PTBP1 and/or hnRNP L by remaining bound to otherwise
7 protected mRNAs.
8
9 Coordinated downregulation of UPF1LL targets during ER stress and ISR induction
10 We next investigated whether the population of mRNAs bound and regulated by UPF1LL
11 tended to encode proteins with particular properties or functions that might indicate
12 physiological conditions in which UPF1LL influences gene expression. Based on previous
13 observations that genes in functionally related pathways are often coordinately regulated at the
14 posttranscriptional level (Keene, 2007), we performed a gene ontology (GO) enrichment
15 analysis (Eden et al., 2009) to identify commonalities among mRNAs preferentially bound by
16 UPF1LL in the RIP-seq experiments. This analysis revealed a high degree of enrichment for
17 proteins that rely on the endoplasmic reticulum (ER) for biogenesis, including a large number of
18 integral membrane proteins (204 of 686 UPF1LL-enriched genes assigned to GO terms; Figure
19 5A; Supplementary file 3). We used a previous survey of ER-localized translation (Jan et al.,
20 2014) to corroborate the results of the GO analysis, finding that many UPF1LL targets were
21 indeed found to be preferentially translated at the ER (Figure 5-figure supplement 1A).
22 The ER is a major site of protein synthesis and folding, and thus serves as a nexus for
23 signaling to prevent proteotoxic stress. In response to the accumulation of unfolded proteins,
24 cells activate the integrated stress response (ISR), which restores homeostasis by repressing
25 translation and inducing expression of a battery of stress response genes (Costa-Mattioli and
26 Walter, 2020). A major effect of ISR-mediated translational repression is corresponding
14 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
Sensitivity to UPF1LL A B UPF1 RIP-seq efficiency overexpression Cellular component **** n.s. **** GO-term 1 enrichment organelle membrane
part part (fold change) 2 > 10⁻³ plasma intrinsic membrane component of 0 part membrane 10⁻³ to 10⁻⁵
intracellular intrinsic component integral component of 10⁻⁵ to 10⁻⁷ organelle of plasma -1 part membrane membrane n= 8759 167 696 9298 325
10⁻⁷ to 10⁻⁹ Tunicamycin vs DMSO log endoplasmic integral component reticulum of plasma SL LL < 10⁻⁹ part membrane Other
endoplasmic overexpression reticulum LL membrane Equally recovered Enriched withEnriched UPF1 with UPF1 Downregulated with UPF1
C D thapsigargin 3 1.0 Down in thapsigargin and rescued UPF1LL siRNA -15 by siUPF1LL (n = 517; < 1x10 ) * NT siRNA Up in thapsigargin 2 (n = 2397; = 0.003) * Other (n = 10787) 1 * * * * * * 0.5 0 * * *
(relative mRNA abundance) * * 2 *
-1 Cumulative frequency log
-2 0.0 -1 0 1 SMG1SMG5 LDLR GJB2 NXT2 eIF5A2 GCNT1EFNB3 TIMP2 TMEM19 GPR161VPS37D TMEM165 METTL7A UPF1LLsiRNA vs NT siRNA log2(fold change) TNFRSF10D SRSF2SRSF3 (+PTC) (+PTC) relative mRNA abundance canonical NMD targets UPF1LL targets decreased increased
Gene ontology analysis of UPF1LL target mRNAs. Box plot of relative mRNA abundance as determined from RNA-seq following treatment of HEK-293 cells with 1 µM tunicamycin for 6hr
(GSE97384). mRNAs were binned by RIP-seq efficiency in CLIP-UPF1LL or CLIP-UPF1SL affinity
purifications (left) or sensitivity to CLIP-UPF1LL overexpression (right). Statistical significance was determined by K-W test, with Dunn's correction for multiple comparisons (left; **** < 1x10-15), or K-S test (right; **** < 1x10-6). Boxes indicate interquartile ranges, and bars indicate Tukey whiskers. qPCR analysis of indicated transcripts following treatment of HEK-293 cells with 1 µM thapsigargin for 6hr.
Relative fold changes are in reference to vehicle-treated, NT siRNA. Significance of UPF1LL knockdown was compared to the NT siRNA control in thapsigargin treatment. Asterisk (*) indicates < 0.05, as determined by unpaired Student’s t-test. Error bars indicate SD (n = 3). CDF plot of relative mRNA
abundance as determined from RNA-seq following UPF1LL knockdown in HEK-293 cells. mRNAs were
binned by sensitivity to thapsigargin and rescue by UPF1LL knockdown. Statistical significance was determined by K-W test, with Dunn's correction for multiple comparisons.
15 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 stabilization of canonical NMD targets, including several mRNAs encoding factors integral to the
2 activation and resolution of the stress response (Goetz and Wilkinson, 2017). Because of our
3 finding that UPF1LL preferentially binds and downregulates mRNAs enriched for membrane and
4 ER-associated gene products, we asked whether UPF1LL activity is regulated in response to ER
5 stress and induction of the ISR. In sharp contrast to previously characterized NMD targets,
6 analysis of publicly-available RNA-seq data of HEK-293 cells treated with the ER stress-
7 inducing agent tunicamycin (Park et al., 2017) revealed that mRNAs preferentially bound and
8 down-regulated by UPF1LL were systematically less abundant during ER stress (Figure 5B). In
9 comparison, transcripts preferentially bound by UPF1SL, and more likely to represent canonical
10 NMD targets, were up-regulated. This effect was not limited to a specific cell type, as UPF1LL
11 targets were similarly down-regulated in HeLa cells treated with tunicamycin (Figure 5-figure
12 supplement 1B) (Boswell et al., 2017). Mouse NIH3T3 cells treated with the ER stress-inducing
13 agent thapsigargin (Namkoong et al., 2018) also exhibited a coordinated decrease in the
14 abundance of UPF1LL targets (Figure 5-figure supplement 1C), indicating a conserved response
15 across mammals in response to distinct ISR-inducing conditions.
16 We next asked whether the downregulation of specific mRNAs during ISR induction
17 occurs via UPF1LL. To this end, we transfected HEK-293 cells with non-targeting (NT) or
18 UPF1LL-specific siRNAs and then treated cells with thapsigargin (Figure 5-figure supplement
19 1D). Changes in gene expression caused by UPF1LL depletion were not due to differential ISR
20 induction, as previously established stress response genes were comparably up-regulated in
21 response to thapsigargin treatment following NT and UPF1LL-specific knockdown (Figure 5-
22 figure supplement 1E) (Ashburner et al., 2000; The Gene Ontology Consortium, 2019).
23 Strikingly, we identified a population of mRNAs (~500 transcripts) that decreased in abundance
24 with thapsigargin treatment and were rescued upon UPF1LL knockdown (Supplementary file 4).
25 Quantitative PCR of select transcripts confirmed these transcriptome-wide results of RNA-seq
26 (Figure 5C). Many transcripts down-regulated in thapsigargin and rescued by UPF1LL
16 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 knockdown were also significantly up-regulated with basal depletion of the UPF1LL isoform
2 (Figure 5D). In contrast, mRNAs that were up-regulated in thapsigargin were only modestly
3 perturbed by UPF1LL knockdown under basal conditions. Thapsigargin treatment alone did not
4 alter the relative levels of UPF1LL and UPF1SL mRNAs (Figure 5-figure supplement 1F),
5 indicating that UPF1LL activity in ER stress was likely due to activity of the existing population of
6 protein rather than a consequence of altered UPF1 splicing. Taken together, these data suggest
7 that the ISR enhances UPF1LL-dependent NMD, while suppressing decay of canonical UPF1SL
8 target mRNAs.
9
10 UPF1LL activity is enhanced by translational repression
11 We have identified that endogenous UPF1LL regulates the abundance of a population of
12 transcripts that undergo enhanced downregulation during ER stress and induction of the ISR.
13 We next sought to determine the underlying mechanism controlling UPF1LL activity under
14 cellular stress conditions. Activation of the ISR induces phosphorylation of eIF2ɑ (Figure 5-
15 figure supplement 1G), driving global downregulation of protein synthesis due to impaired eIF2-
16 GTP-Met-tRNAi ternary complex recycling and reduced delivery of the initiator Met-tRNAi to
17 translating ribosomes (Baird and Wek, 2012; Wek, 2018; Young and Wek, 2016). We therefore
18 asked whether translational repression would induce UPF1LL activity outside of the context of
19 the ISR. To explore this possibility, we first assessed the response of UPF1LL target transcripts
20 in previously published RNA-seq data from HEK-293 cells treated with the translation elongation
21 inhibitor emetine (Martinez-Nunez et al., 2017). We observed that mRNAs preferentially bound
22 and down-regulated by CLIP-UPF1LL were overwhelmingly decreased in abundance with
23 emetine treatment (Figure 6A and Figure 6-figure supplement 1A). Notably, a similar effect was
24 evident upon impairment of translation initiation, as we observed decreased expression of direct
25 CLIP-UPF1LL target mRNAs with decreased translation initiation upon hippuristanol treatment
26 (Figure 6B and Figure 6-figure supplement 1B) in a publicly available RNA-seq dataset
17 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
18 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
Volcano plot of relative mRNA abundance as determined from RNA-seq following treatment of HEK-293 cells with 50 ng/mL emetine for 4hr (GSE89774). mRNAs were binned by sensitivity to CLIP-UPF1LL overexpression. Horizontal line indicates the significance threshold ≤ 0.05. Volcano plot as in (A), following treatment of MCF7 cells with 150 nM hippuristanol for 1hr (GSE134888). CDF plot of relative mRNA abundance as determined from RNA-seq following UPF1LL-specific knockdown in HEK-293 cells and treatment with 50 µg/mL puromycin for 4hr. mRNAs were binned by sensitivity to puromycin in the absence of UPF1LL knockdown. Statistical significance was determined by K-S test. qPCR analysis of indicated transcripts following treatment of HEK-293 cells with 50 µg/mL puromycin for 4hr. Relative fold changes are in reference to vehicle-treated, NT siRNA. Significance of UPF1LL knockdown was compared to the NT siRNA control in puromycin treatment. Asterisk (*) indicates < 0.05, as determined by unpaired Student’s t-test. Error bars indicate SD (n = 3). CDF plot of UPF1LL
RIP-seq efficiency. mRNAs were binned by sensitivity to puromycin and rescue byUPF1LL knockdown. Statistical significance was determined by K-W test, with Dunn's correction for multiple comparisons. qPCR analysis of indicated transcripts following treatment of HEK-293 cells with indicated concentrations of puromycin for 4hr. Relative fold changes are in reference to vehicle-treated control. Significance of puromycin treatment on relative transcript abundance was compared to the vehicle-treated control. Asterisk (*) indicates < 0.05, as determined by two-way ANOVA. Error bars indicate SD (n = 3).
19 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 (Waldron et al., 2019). Extending our analysis to independent datasets of emetine (Zhang et al.,
2 2019) and hippuristanol (Karginov and Hannon, 2013), as well as RNA-seq data generated with
3 the elongation inhibitor anisomycin (Hollerer et al., 2016), confirmed that the downregulation of
4 UPF1LL target transcripts was not specific to a cell type or mechanism of translation inhibition
5 (Figure 6-figure supplement 1C, 1D, and 1E).
6 Having established a relationship between translational repression and enhanced
7 UPF1LL activity, we next sought to directly test the role of UPF1LL in modulating mRNA
8 abundance in response to impaired translation. To this end, we transfected HEK-293 cells with
9 non-targeting (NT) or UPF1LL-specific siRNAs and then treated cells with the elongation inhibitor
10 puromycin (50 μg/mL; Figure 6-figure supplement 1F). In RNA-seq analyses, we identified a
11 population of mRNAs (~700 transcripts) that decreased in abundance by at least 1.4-fold with
12 puromycin treatment and were rescued upon UPF1LL knockdown (Figure 6C and
13 Supplementary file 4). Quantitative PCR of select transcripts confirmed these transcriptome-
14 wide results (Figure 6D). Inferred mRNA stability changes using REMBRANDTS software
15 indicated that the observed changes in mRNA abundance upon puromycin treatment and
16 UPF1LL knockdown were due to changes in mRNA stability (Figure 6-figure supplement 1G). In
17 support of this conclusion, UPF1LL target transcripts decayed with apparent first-order kinetics
18 upon puromycin treatment, with significant loss of the most highly affected transcripts in as little
19 as 15 minutes (Figure 6-figure supplement 1H). Similar to conditions of ER stress, puromycin
20 treatment alone did not alter UPF1 splicing (Figure 6-figure supplement 1I), indicating that
21 enhanced UPF1LL activity during conditions of impaired translation efficiency was likely due to
22 the existing population of UPF1LL protein.
23 We next evaluated whether our UPF1LL overexpression RNA-seq and RIP-seq studies
24 were predictive of the response of UPF1LL target transcripts to impaired translation efficiency.
25 Indeed, mRNAs that were selectively down-regulated with UPF1LL overexpression showed a
26 significant decrease in abundance with puromycin treatment, an effect that was likely due to
20 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 enhanced decay according to REMBRANDTS analysis (Figure 6-figure supplement 1J).
2 Correspondingly, mRNAs that decreased in abundance with puromycin treatment in a UPF1LL-
3 dependent manner were enriched in the UPF1LL RIP-seq (Figure 6E). These results indicate
4 that UPF1LL substrates are more efficiently targeted during conditions of impaired translation
5 due to both translation repression and the intrinsic capacity of UPF1LL to stably associate with
6 these transcripts.
7 Because of the well-established requirement for translation in NMD, we hypothesized
8 that the UPF1LL-dependent decay we observed under conditions of translational repression was
9 due to rare translation termination events. To test this hypothesis, we treated cells with a
10 titration of puromycin from 25 µg/mL to 400 µg/mL. If UPF1LL is capable of responding to rare
11 translation termination events, its activity should be promoted at low concentrations of
12 puromycin that permit a low level of termination events to persist, but inhibited by high
13 concentrations of puromycin that more efficiently block translation. In line with these
14 expectations, we observed a dose-dependent response, in which downregulation of
15 representative UPF1LL target transcripts FMR1, eIF5A2, ERBB2, and CDK16 was most efficient
16 at lower puromycin concentrations (Figure 6F). Treatment with high concentrations of puromycin
17 did not have a significant effect on the levels of these UPF1LL target mRNAs, consistent with a
18 requirement for translation termination events. Corroborating these findings, we observed
19 globally more efficient downregulation of puromycin-sensitive UPF1LL targets with 25 µg/mL
20 puromycin than 100 µg/mL puromycin in RNA-seq (Figure 6-figure supplement 1K). Based on
21 these results, we conclude that translation termination is strictly required for all NMD events but
22 that changes in translation efficiency can drive downregulation of a novel class of substrates by
23 the UPF1LL isoform.
21 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 Discussion
2 The idea that the mammalian NMD pathway consists of multiple branches with distinct
3 factor requirements and substrate specificities has been proposed by several groups, but the
4 underlying mechanisms and regulatory roles of NMD branching are poorly understood (Chan et
5 al., 2007; Huang et al., 2011; Yi et al., 2020). Our identification of differential activities of UPF1SL
6 and UPF1LL is an unforeseen example of NMD pathway branching, which can be controlled at
7 the cellular level by changes in the abundance of protective RBPs, translation, or UPF1 splicing.
8 When cells have high translational capacity, both isoforms contribute to NMD, but the more
9 abundant UPF1SL isoform predominates. Contrary to expectations, we find that the NMD
10 pathway does not cease to function under conditions of acute translational repression. Instead,
11 the pathway switches from UPF1SL-dependent targeting of frequent translation termination
12 events to UPF1LL-dependent targeting of rare termination events. In this way, UPF1LL enables
13 cellular NMD specificity to be dramatically remodeled in response to changing cellular
14 conditions, including induction of the ISR. Because altered translation efficiency and induction of
15 cellular stress pathways are pervasive in cancer, genetic disease, and infection, the expanded
16 scope of UPF1LL-dependent decay has far-reaching implications for the physiological roles of
17 the mammalian NMD pathway in health and disease.
18 We propose a new model by which the alternative UPF1LL isoform expands the scope of
19 NMD under conditions of translation inhibition (Figure 7). The differential RNA-binding
20 properties of UPF1SL and UPF1LL in the presence of ATP and protective proteins provide a
21 molecular mechanism by which NMD specificity is tuned in response to cellular conditions (Fritz
22 et al., 2020; Gowravaram et al., 2018). We show that UPF1LL activity can be promoted either by
23 overexpression of the UPF1LL protein or by translational repression. When UPF1LL expression is
24 enhanced, it is able to overcome inhibition by the protective RBPs PTBP1 and hnRNP L, which
25 normally induce dissociation of UPF1SL from potential NMD target mRNAs. Under conditions of
26 translational repression, canonical NMD targets predominantly regulated by UPF1SL are
22 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
23 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 stabilized, while UPF1LL targets are more efficiently degraded. Together, our data inform a
2 model in which the increased residence time of UPF1LL on RNAs enables it to respond to rare
3 translation termination events, while the faster dissociation of UPF1SL from potential NMD
4 targets renders it unable to complete decay when translation is repressed. The rate-limiting
5 step(s) downstream of UPF1 binding to potential NMD targets remain to be clearly defined, but
6 an attractive possibility is that persistent binding of UPF1LL to mRNAs promotes its
7 phosphorylation by SMG1, readying UPF1LL to scaffold assembly of productive decay
8 complexes (Durand et al., 2016).
9 The majority of studies of mammalian UPF1 functions have been conducted in
10 transformed, rapidly dividing, highly metabolically active cells under conditions that promote
11 efficient translation. These experimental conditions have enabled the discovery that UPF1 and
12 NMD as a whole have a major impact on transcriptome expression, affecting 5-10% of all
13 human genes. However, our work strongly suggests that these studies have been blind to major
14 physiological functions of UPF1. Notably, we find that UPF1LL is able to regulate several
15 proteins of central importance in cancer and other diseases, including fragile X mental
16 retardation 1 (FMR1), the low-density lipoprotein receptor (LDLR), and oncogenes PTEN,
17 EIF5A2, and ERBB2, among others (Dockendorff and Labrador, 2019; Goldstein and Brown,
18 2009; Harbeck et al., 2019; Mathews and Hershey, 2015; Song et al., 2012). Along with EIF5A2
19 and FMR1, several additional translation-dependent UPF1LL targets are themselves important
20 regulators of translation and mRNA decay, including the signal recognition particle receptor
21 SRPR, the major mRNA decapping enzyme DCP2, and the miRNA-processing endonuclease
22 DICER1 (Akopian et al., 2013; Michlewski and Cáceres, 2019; Mugridge et al., 2018).
23 The ability of UPF1LL to target specific mRNAs in response to translational repression
24 positions NMD to function as a mechanism to reset the transcriptome upon cellular stress.
25 Translational repression via eIF2α phosphorylation is a central feature of responses to diverse
26 cellular insults, including nutrient deprivation, oncogene activation, viral infection, hypoxia, and
24 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 accumulation of unfolded proteins (Pakos-Zebrucka et al., 2016). We show that induction of ER
2 stress with thapsigargin induces UPF1LL-dependent downregulation of hundreds of transcripts, a
3 population particularly enriched for mRNAs encoding proteins trafficked through the ER. The
4 extent to which UPF1LL complements or collaborates with IRE1-mediated mRNA decay or the
5 recently identified ER-associated NMD factor NBAS will require further study (Hollien and
6 Weissman, 2006; Longman et al., 2020), but our data suggest that UPF1LL may help to relieve
7 proteotoxic stress by reducing the abundance of transcripts encoding proteins with complex
8 biosynthetic needs.
9 Finally, we note that, even in the absence of stress, most cells and tissues in vivo likely
10 have lower basal rates of translation than those attained in exponentially growing transformed
11 cell lines. Based on our findings, understanding how NMD shapes gene expression in diverse
12 tissue types in health and disease will require not just analysis of NMD targets characterized in
13 transformed cells, but also transcripts that are conditionally targeted in response to changing
14 translational states.
25 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 Materials and methods
2 UPF1 isoform nomenclature
3 UPF1SL refers to the protein encoded by Ensembl transcript ID ENST00000262803.9 and
4 UPF1LL by Ensembl transcript ID ENST00000599848.5. These isoforms are also referred to as
5 UPF1 isoform 2 and UPF1 isoform 1, respectively (Gowravaram et al., 2018; Nicholson et al.,
6 2014).
7
8 In vitro helicase assays
9 Unwinding assays were performed as previously described in detail (Fritz et al., 2020). Briefly
10 75 nM of the pre-assembled RNA duplex substrate (described below) was combined with 1x
11 unwinding reaction buffer (10 mM MES pH 6.0, 50 mM KOAc, 0.1 mM EDTA), 2 mM MgOAc,
12 and 1 unit RNasin in a well of a half-area, black, flat-bottom 96-well plate (Corning 3993).
13 PTBP1 (80 nM) was then added, the sample mixed, and incubated at room temperature for 10
14 minutes. Then, UPF1 (80 nM) was added, the sample mixed, and incubated at room
15 temperature for 10 minutes. Finally, BHQ1 quencher (0.56 µM final;
16 GTGTGCGTACAACTAGCT / 3BHQ_1) and 2 mM ATP was added to initiate the unwinding
17 reaction. An InfiniteR F200 Pro microplate reader and associated i-controlTM 1.9 software
18 (Tecan) was used to monitor Alexa Fluor 488 fluorescence every 10 seconds for 20 minutes at
19 37°C. Measured fluorescence intensities were normalized to the zero timepoint for each
20 condition to obtain relative fluorescence values. Four technical replicates for each condition
21 were obtained at the same time.
22 To generate the RNA duplex, a DNA oligo template (5’
23 TAATACGACTCACTATAGGGACACAAAACAAAAGACAAAAACACAAAACAAAAGACAAAAAC
24 ACAAAACAAAAGACAAAAAGCCTCTCCTTCTCTCTGCTTCTCTCTCGCTGTGTGCGTACAAC
25 TAGCT 3’) was PCR-amplified and then in vitro transcribed using the MEGAshortscriptTM T7
26 Transcription kit (Invitrogen). An 11:7 ratio of the helicase RNA substrate to 5’ Alex Fluor 488
26 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 fluorescent oligo strand (Alexa Fluor 488/ AGCTAGTTGTACGCACAC) was incubated with 2
2 mM MgOAC and 1x unwinding reaction buffer at 95°C for 3 minutes and 30 seconds and then
3 slowly cooled to 30°C. All DNA oligos were obtained from Integrated DNA Technologies. The 5’
4 fluorescent oligo probe and 3’ BHQ1 quencher were acquired as RNase-free, HPLC-purified.
5 Cloning, expression, and purification of recombinant UPF1LLΔCH, UPF1SLΔCH, and
6 PTBP1 were conducted as previously described (Fritz et al., 2020; Gowravaram et al., 2018). In
7 this study, UPF1LLΔCH was purified in the same manner as UPF1SLΔCH (Fritz et al., 2020).
8
9 Mammalian cell lines and generation of CLIP-UPF1 expression lines
10 HEK-293 cells used in the endogenous UPF1LL knockdown and RNA-seq experiments were
11 received from ATCC (CRL-3216) and maintained at 37°C and 5% CO2 in DMEM with 10% FBS
TM TM 12 (Gibco) and 1% pen/strep. Stable integration of CLIP-UPF1SL into human Flp-In T-REx -293
13 cells (Invitrogen) and subsequent maintenance of this stable line was previously described
14 (Kishor et al., 2020). CLIP-UPF1LL expression lines were generated and maintained in an
15 identical manner.
16
17 Western for phosphorylation status of UPF1 variants
18 HEK-293 FLP-In CLIP-UPF1 stable cell lines were seeded at a density of 1x106 cells in 10 cm
19 plates and then treated with 200 ng/mL doxycycline hyclate (Sigma) for 48 hours to induce
20 CLIP-UPF1 expression. On the day of harvest, cells were treated with or without 100 µg/mL
21 puromycin (Sigma) for 2 hours. Whole cell lysate was then generated by the freeze/thaw
22 method as previously described (Fritz et al., 2018; Hogg and Collins, 2007a, 2007b; Hogg and
23 Goff, 2010). A total of 10 µg protein was run on a NuPageTM 4-12% Bis-Tris Protein Gel
24 (Invitrogen) using MOPS buffer according to manufacturer instructions and subsequently
25 transferred to a nitrocellulose membrane according to the NuPageTM manufacturer's protocol
26 (Invitrogen). Membranes were incubated with a blocking buffer for fluorescent Western blotting
27 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 (Rockland) for 1 hour at room temperature and then incubated overnight at 4°C with the
2 indicated primary antibody. Primary antibodies used were: anti-Phospho-(Ser/Thr) ATM/ATR
3 (rabbit polyclonal, Cell Signaling, #2851, 1:1000), anti-RENT1/UPF1 (goat polyclonal, Bethyl,
4 A300-038A, 1:1000), or anti-β-actin (mouse monoclonal, Cell Signaling, #3700, 1:1000).
5 Membranes were subsequently washed three times with 1xTBS supplemented with 0.1%
6 Tween-20 and then incubated with the appropriate secondary antibody for 1 hour at room
7 temperature. Secondary antibodies used were: anti-rabbit IgG (H&L) Antibody DyLightTM 680
8 Conjugated (Rockland, 1:10,000), anti-goat IgG (H&L) Antibody DyLightTM 800 Conjugated
9 (Rockland, 1:10,000), or anti-mouse IgG (H&L) Antibody DyLightTM 680 Conjugated (Rockland,
10 1:10,000). Membranes were washed three times with 1xTBS supplemented with 0.1% Tween-
11 20 and then two times with 1xTBS. Western blot images were obtained on an Amersham
12 Typhoon imaging system (GE Healthcare Life Sciences) and quantified using ImageStudio
13 software (LI-COR Biosciences).
14
15 CLIP-UPF1 overexpression RIP-seq and RNA-seq
16 CLIP-UPF1SL overexpression RIP-seq and RNA-seq sample preparation were previously
17 reported (Kishor et al., 2020); CLIP-UPF1LL datasets were generated in parallel. Briefly, CLIP-
18 UPF1 stable cell lines or a GFP-expressing control line were seeded in 6 x 15 cm plates and
19 then treated with 200 ng/mL doxycycline hyclate (Sigma) for 48 hours to induce CLIP-UPF1
20 expression. Cells were harvested 48 hours post-induction at 80-85% confluency and whole cell
21 lysate generated by the freeze/thaw method as previously described (Fritz et al., 2018; Hogg
22 and Collins, 2007a, 2007b; Hogg and Goff, 2010). Equilibrated cell extracts, reserving 1/10th for
23 downstream input analysis, were then combined with 10 µM CLIP-Biotin (New England Biolabs)
24 and rotated (end-over-end) for 1 hour at 4°C to allow CLIP-UPF1 to react with the CLIP-Biotin
25 substrate and yield covalently biotinylated protein. Unbound CLIP-Biotin was subsequently
26 removed by passing the samples through Zeba™ Spin Desalting Columns, 40K MWCO, 2 mL
28 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 (Thermo Scientific) according to manufacturer instructions. Buffer exchange was performed with
2 HLB-150 supplemented with 1 mM DTT and 0.1% NP-40. Pre-washed DynabeadsTM MyOneTM
3 Streptavidin T1 (Invitrogen) were then added and the samples rotated (end-over-end) for 1 hour
4 at 4°C to immobilize the biotin-bound CLIP-UPF1 complexes. The samples were then washed
5 three times with 500 µL of HLB-150 supplemented with 1 mM DTT and 0.1% NP-40 and one
6 one-hundredth reserved for downstream Western blot analysis of CLIP-UPF1 pull-down
7 efficiency. The remainder was combined with TRIzolTM (Invitrogen) and RNA was isolated
8 according to manufacturer instructions. DNase-treatment was subsequently performed using
9 RQ1 RNase-Free DNase (Promega) and RNA isolated by acid phenol chloroform extraction
10 according to standard protocol. This resulted in approximately 1 µg of total RNA that was then
11 subjected to high-throughput sequencing. In parallel, the reserved input lysate (approximately 3
12 µg of total RNA) was also sent for high-throughput sequencing. A total of three biological
13 replicates from each condition were processed. Sequencing libraries were prepared from input
14 and bound RNA using the Illumina TruSeq Stranded Total RNA Human kit and sequenced on
15 an Illumina HiSeq 3000 instrument.
16 To validate select transcripts by RT-qPCR, an equivalent volume of reserved RNA (3 µL)
17 was used as a template for cDNA synthesis using the Maxima First Strand cDNA synthesis kit
18 for RT-qPCR (Thermo Scientific) according to manufacturer instructions. The resulting cDNA
19 was diluted with nuclease-free water and subsequently analyzed by qPCR using iTaq Universal
20 SYBR Green Supermix (BioRad) on a Roche LightCycler 96 instrument (Roche). Sequences for
21 gene-specific primers used for amplification are listed in supplementary file 5. For input
22 samples, relative fold change was determined by calculating 2-ΔΔCT values using GAPDH for
23 normalization. For pull-down samples, relative fold enrichment was determined by dividing the
24 Cq value of the pull-down by its corresponding input, multiplying by 100 and then normalizing to
25 the relative recovery of the SMG1 transcript.
29 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 For assessment of CLIP-UPF1 expression and pull-down efficiency by Western blot,
2 reserved input and IP samples were run on a NuPageTM 4-12% Bis-Tris Protein Gel (Invitrogen)
3 using MOPS buffer according to manufacturer instructions and subsequently transferred to a
4 nitrocellulose membrane according to the NuPageTM manufacturer's protocol (Invitrogen).
5 Membranes were incubated with a blocking buffer for fluorescent Western blotting (Rockland)
6 for 1 hour at room temperature and then incubated overnight at 4°C with the indicated primary
7 antibody. Primary antibodies used were: anti-RENT1 (goat polyclonal, Bethyl, A300-038A,
8 1:1000) and anti-β-actin (mouse monoclonal, Cell Signaling, #3700, 1:1000). Membranes were
9 subsequently washed three times with 1xTBS supplemented with 0.1% Tween-20 and then
10 incubated with the appropriate secondary antibody for 1 hour at room temperature. Secondary
11 antibodies used were: anti-goat IgG (H&L) Antibody DyLightTM 680 Conjugated (Rockland, 605-
12 744-002, 1:10,000) and anti-mouse IgG (H&L) Antibody DyLightTM 680 Conjugated (Rockland,
13 610-744-124, 1:10,000). Membranes were washed three times with 1xTBS supplemented with
14 0.1% Tween-20 and then two times with 1xTBS. Western blot images were obtained on an
15 Amersham Typhoon imaging system (GE Healthcare Life Sciences) and quantified using
16 ImageStudio software (LI-COR Biosciences).
17
18 Endogenous UPF1LL knockdown with puromycin or thapsigargin treatment and RNA-seq
5 19 3x10 HEK-293 cells were reverse transfected with 40 nM of a non-targeting or UPF1LL-specific
20 (Forward sequence: 5’ GGUAAUGAGGAUUUAGUCA 3’; Reverse sequence: 5’
21 UGACUAAAUCCUCAUUACC 3’) siRNA using Lipofectamine RNAiMAX according to
22 manufacturer instructions. Forty-eight hours post-siRNA transfection, cells were replated at a
23 density of 5x105 cells per well of a 6-well plate. This step was critical to achieve 70-75%
24 confluency on the day of drug treatment and cell harvest. The next day, cells were treated with
25 vehicle control, 25, 50, or 100 µg/mL puromycin (Sigma) for 4 hours, or 1 µM thapsigargin for 6
26 hours. A total of three replicates were generated for each condition. Total RNA was then
30 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 isolated using the RNeasy Plus Mini Kit (QIAGEN). Sequencing libraries were prepared from 2
2 µg total RNA using the Illumina TruSeq Stranded Total RNA Human kit and sequenced on an
3 NovaSeq 6000 instrument. For RT-qPCR validation of changes in target gene expression, 500
4 ng of total RNA was used as input for cDNA synthesis using the Maxima First Strand cDNA
5 synthesis kit for RT-qPCR (Thermo Scientific) according to manufacturer instructions. The
6 resulting cDNA was diluted with nuclease-free water and subsequently analyzed by qPCR using
7 iTaq Universal SYBR Green Supermix (BioRad) on a Roche LightCycler 96 instrument (Roche).
8 Sequences for gene-specific primers used for amplification are listed in supplementary file 5.
9 Relative fold changes were determined by calculating 2-ΔΔCT values using GAPDH for
10 normalization.
11
12
13 Western for phospho-eIF2ɑ
14 HEK-293 cells treated with 1 µM thapsigargin were lysed in 1X Passive Lysis Buffer (Promega)
15 supplemented with HaltTM Protease and Phosphatase Inhibitor Cocktail (ThermoScientific)
16 according to manufacturer instructions. A total of 5 µg protein was run on a NuPageTM 4-12%
17 Bis-Tris Protein Gel (Invitrogen) using MOPS buffer according to manufacturer instructions and
18 subsequently transferred to a nitrocellulose membrane according to the NuPageTM
19 manufacturer's protocol (Invitrogen). Detection of phospho and total eIF2ɑ was performed as
20 previously described (Young-Baird et al., 2020).
21
22 Semiquantitative PCR to detect UPF1 isoform ratios
23 Generated cDNA (1/40th) from RNA-seq samples was used as input for PCR amplification with
24 Phusion High-Fidelity DNA polymerase (New England Biolabs) and UPF1 specific primers that
25 flank the regulatory loop sequence (Forward: 5’ AACAAGCTGGAGGAGCTGTGGA 3’; Reverse:
26 5’ ACTTCCACACAAAATCCACCTGGAAGTT 3’). The PCR cycling conditions used were: an
31 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 initial denaturation at 98°C for 30 seconds and then 22 cycles of 98°C for 10 seconds, 63°C for
2 30 seconds, 72°C for 15 seconds. PCR products were then run on a 8% NovexTM TBE gel
3 (Invitrogen) according to manufacturer instructions and subsequently stained with SYBR® Gold
4 Nucleic Acid Stain (Invitrogen). Images were obtained on an Amersham Typhoon imaging
5 system (GE Healthcare Life Sciences) and quantified using ImageStudio software (LI-COR
6 Biosciences).
7
8 Computational analyses
9 Raw fastq reads from CLIP-UPF1 overexpression RNA-seq and RIP-seq data were trimmed
10 with Cutadapt using the following parameters: --times 2 -e 0 -O 5 --quality-cutoff 6 -m 18 -a
11 AGATCGGAAGAGCACACGTCTGAACTCCAGTCAC -A
12 AGATCGGAAGAGCGTCGTGTAGGGAAAGAGTGTAGATCTCGGTGGTCGCCGTATCATT
13 -b AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA -b
14 TTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTT (Martin, 2011). Trimmed
15 reads were quantified against a custom HEK-293 transcriptome index, prepared as described
16 (Kishor et al., 2018), using Kallisto software with parameters --bias -b 1000 -t 16 --single --rf-
17 stranded -l 200 -s 20 (Bray et al., 2016). RIP-seq enrichment values were obtained by dividing
18 TPM values from IP samples by TPM values from input samples. Differential gene expression
19 analysis was performed using RUVSeq and edgeR (Risso et al., 2014; Robinson et al., 2009).
20 Normalization and batch correction was performed with the RUVg function, based on transcripts
21 with invariant expression among all samples. The edgeR TMM method was used to obtain
22 normalized differential expression values and to calculate FDRs. The most abundant transcript
23 isoform from each gene was used for further analysis. PTBP1 and hnRNP L binding motif
24 position-specific scoring matrices were downloaded from the RBPmap database and used for
25 motif finding in 3’UTRs derived from the custom HEK-293 transcriptome with HOMER, as
26 described (Heinz et al., 2010; Kishor et al., 2018; Paz et al., 2014). IsoformSwitchAnalyzeR was
32 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 used to annotate PTCs in the custom HEK-293 transcriptome, as described (Kishor et al., 2018;
2 Vitting-Seerup and Sandelin, 2019), and NMD-sensitive transcripts were classified on the basis
3 of a log2 fold change of 0.5 or greater in previous RNA-seq studies of total UPF1 knockdown in
4 HEK-293 cells (GSE105436) (Baird et al., 2018). For analysis of RNA-seq data from HEK-293
5 cells treated with non-targeting or anti-UPF1LL siRNAs, raw fastq reads from the NovaSeq 6000
6 platform were trimmed with fastp (Chen et al., 2018), with the parameters --
7 detect_adapter_for_pe and --trim_poly_g. Trimmed reads were aligned with HISAT2 to the
8 hg19/GRCh37 genome and transcriptome index provided by the authors (Kim et al., 2019).
9 Reads mapping to Ensembl GRCh37 release 75 gene annotations were quantified with
10 featureCounts (Liao et al., 2014), and differential gene expression was analyzed using
11 limma/voom, as implemented by the Degust server (Powell, 2015; Ritchie et al., 2015).
12 Alternative splicing was analyzed using rMATS 4.0.1 (Shen et al., 2014), and Sashimi plots
13 were generated using Integrative Genomics Viewer 2.8.2 (Thorvaldsdóttir et al., 2013).
14
15 GTEx data analysis
16 GTEx data were downloaded from the GTEx Portal on 11/11/2020. To determine the relative
17 representation of UPF1LL and UPF1SL mRNA isoforms, transcript TPM values for transcript
18 ENST00000599848.5 (UPF1LL) were divided by the total TPM values derived from transcripts
19 ENST00000599848.5 (UPF1LL) and ENST00000262803.9 (UPF1SL). GTEx samples were
20 assigned to the indicated tissue types using the sample attributes provided in GTEx Analysis v8.
21
22 Data availability
23 CLIP-UPF1 overexpression RIP-seq and RNA-seq data are available from the NCBI GEO
24 database with accession number GSE134059. The endogenous UPF1LL knockdown RNA-seq
25 data, including puromycin or thapsigargin treatment, are available with accession number
33 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 GSE162699. All publicly available datasets used in this study are directly identified in the text
2 and figure legends.
3
4 Acknowledgements
5 We thank Sutapa Chakrabarti for the UPF1LLΔCH recombinant expression plasmid and
6 members of the Hogg lab and Nicholas R. Guydosh for critical reading of the manuscript. We
7 are grateful to Sandy Mattijssen and Richard J. Maraia for helpful discussion and to Sara K.
8 Young-Baird for thapsigargin reagents, helpful discussion, and assistance with phospho-eIF2ɑ
9 immunoblotting. High-throughput sequencing was conducted by Yan Luo and Poching Liu in the
10 NHLBI DNA Sequencing and Genomics Core. The Genotype-Tissue Expression (GTEx) Project
11 was supported by the Common Fund of the Office of the Director of the National Institutes of
12 Health, and by NCI, NHGRI, NHLBI, NIDA, NIMH, and NINDS. This work was supported by the
13 Intramural Research Program, National Institutes of Health, National Heart, Lung, and Blood
14 Institute and utilized the computational resources of the NIH HPC Biowulf cluster
15 (http://hpc.nih.gov).
16
17 Author contributions
18 S.E.F. and J.R.H. conceived and designed the study, analyzed and interpreted the data, and
19 wrote the manuscript. S.E.F. acquired the UPF1 RIP-seq and RNA-seq data and performed
20 associated cellular studies. S.R. acquired the unwinding data. All authors read and approved
21 the manuscript.
22
23 Competing interests
24 The authors declare no competing interests.
34 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 References
2 Akopian, D., Shen, K., Zhang, X., and Shan, S.-O. (2013). Signal recognition particle: an 3 essential protein-targeting machine. Annu. Rev. Biochem. 82, 693–721.
4 Alkallas, R., Fish, L., Goodarzi, H., and Najafabadi, H.S. (2017). Inference of RNA decay rate 5 from transcriptional profiling highlights the regulatory programs of Alzheimer’s disease. Nat. 6 Commun. 8, 909.
7 Ashburner, M., Ball, C.A., Blake, J.A., Botstein, D., Butler, H., Cherry, J.M., Davis, A.P., 8 Dolinski, K., Dwight, S.S., Eppig, J.T., et al. (2000). Gene ontology: tool for the unification of 9 biology. The Gene Ontology Consortium. Nat. Genet. 25, 25–29.
10 Baird, T.D., and Wek, R.C. (2012). Eukaryotic initiation factor 2 phosphorylation and 11 translational control in metabolism. Adv. Nutr. 3, 307–321.
12 Baird, T.D., Cheng, K.C.-C., Chen, Y.-C., Buehler, E., Martin, S.E., Inglese, J., and Hogg, J.R. 13 (2018). ICE1 promotes the link between splicing and nonsense-mediated mRNA decay. Elife 7, 14 e33178.
15 Baker, S.L., and Hogg, J.R. (2017). A system for coordinated analysis of translational 16 readthrough and nonsense-mediated mRNA decay. PLoS One 12, e0173980.
17 Boswell, S.A., Snavely, A., Landry, H.M., Churchman, L.S., Gray, J.M., and Springer, M. (2017). 18 Total RNA-seq to identify pharmacological effects on specific stages of mRNA synthesis. Nat. 19 Chem. Biol. 13, 501–507.
20 Bray, N.L., Pimentel, H., Melsted, P., and Pachter, L. (2016). Near-optimal probabilistic RNA- 21 seq quantification. Nat. Biotechnol. 34, 525–527.
22 Causier, B., Li, Z., De Smet, R., Lloyd, J.P.B., Van de Peer, Y., and Davies, B. (2017). 23 Conservation of Nonsense-Mediated mRNA Decay Complex Components Throughout 24 Eukaryotic Evolution. Sci. Rep. 7, 16692.
25 Chakrabarti, S., Jayachandran, U., Bonneau, F., Fiorini, F., Basquin, C., Domcke, S., Le Hir, H., 26 and Conti, E. (2011). Molecular mechanisms for the RNA-dependent ATPase activity of Upf1 27 and its regulation by Upf2. Mol. Cell 41, 693–703.
28 Chamieh, H., Ballut, L., Bonneau, F., and Le Hir, H. (2008). NMD factors UPF2 and UPF3 29 bridge UPF1 to the exon junction complex and stimulate its RNA helicase activity. Nat. Struct. 30 Mol. Biol. 15, 85–93.
31 Chan, W.K., Huang, L., Gudikote, J.P., and Chang, Y.F. (2007). An alternative branch of the 32 nonsense-mediated decay pathway. EMBO J.
33 Chen, S., Zhou, Y., Chen, Y., and Gu, J. (2018). fastp: an ultra-fast all-in-one FASTQ 34 preprocessor. Bioinformatics 34, i884–i890.
35 Cheng, Z., Muhlrad, D., Lim, M.K., Parker, R., and Song, H. (2007). Structural and functional 36 insights into the human Upf1 helicase core. EMBO J. 26, 253–264.
37 Colombo, M., Karousis, E.D., Bourquin, J., Bruggmann, R., and Mühlemann, O. (2017).
35 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 Transcriptome-wide identification of NMD-targeted human mRNAs reveals extensive 2 redundancy between SMG6- and SMG7-mediated degradation pathways. RNA 23, 189–201.
3 Costa-Mattioli, M., and Walter, P. (2020). The integrated stress response: From mechanism to 4 disease. Science 368.
5 Dang, Y., Low, W.-K., Xu, J., Gehring, N.H., Dietz, H.C., Romo, D., and Liu, J.O. (2009). 6 Inhibition of nonsense-mediated mRNA decay by the natural product pateamine A through 7 eukaryotic initiation factor 4AIII. J. Biol. Chem. 284, 23613–23621.
8 Dockendorff, T.C., and Labrador, M. (2019). The Fragile X Protein and Genome Function. Mol. 9 Neurobiol. 56, 711–721.
10 Durand, S., Franks, T.M., and Lykke-Andersen, J. (2016). Hyperphosphorylation amplifies UPF1 11 activity to resolve stalls in nonsense-mediated mRNA decay. Nat. Commun. 7, 12434.
12 Eberle, A.B., Lykke-Andersen, S., Mühlemann, O., and Jensen, T.H. (2009). SMG6 promotes 13 endonucleolytic cleavage of nonsense mRNA in human cells. Nat. Struct. Mol. Biol. 16, 49–55.
14 Eden, E., Navon, R., Steinfeld, I., Lipson, D., and Yakhini, Z. (2009). GOrilla: a tool for discovery 15 and visualization of enriched GO terms in ranked gene lists. BMC Bioinformatics 10, 48.
16 Fiorini, F., Bonneau, F., and Le Hir, H. (2012). Biochemical characterization of the RNA helicase 17 UPF1 involved in nonsense-mediated mRNA decay. Methods Enzymol. 511, 255–274.
18 Franks, T.M., Singh, G., and Lykke-Andersen, J. (2010). Upf1 ATPase-dependent mRNP 19 disassembly is required for completion of nonsense- mediated mRNA decay. Cell 143, 938– 20 950.
21 Fritz, S.E., Haque, N., and Hogg, J.R. (2018). Highly efficient in vitro translation of authentic 22 affinity-purified messenger ribonucleoprotein complexes. RNA 24, 982–989.
23 Fritz, S.E., Ranganathan, S., Wang, C.D., and Hogg, J.R. (2020). The RNA-binding protein 24 PTBP1 promotes ATPase-dependent dissociation of the RNA helicase UPF1 to protect 25 transcripts from nonsense-mediated mRNA decay. J. Biol. Chem. 295, 11613–11625.
26 Ge, Z., Quek, B.L., Beemon, K.L., and Robert Hogg, J. (2016). Polypyrimidine tract binding 27 protein 1 protects mRNAs from recognition by the nonsense-mediated mRNA decay pathway. 28 eLife Sciences 5, e11155.
29 Goetz, A.E., and Wilkinson, M. (2017). Stress and the nonsense-mediated RNA decay pathway. 30 Cell. Mol. Life Sci. 74, 3509–3531.
31 Goldstein, J.L., and Brown, M.S. (2009). The LDL receptor. Arterioscler. Thromb. Vasc. Biol. 29, 32 431–438.
33 Gowravaram, M., Bonneau, F., Kanaan, J., Maciej, V.D., Fiorini, F., Raj, S., Croquette, V., Le 34 Hir, H., and Chakrabarti, S. (2018). A conserved structural element in the RNA helicase UPF1 35 regulates its catalytic activity in an isoform-specific manner. Nucleic Acids Res. 46, 2648–2659.
36 Harbeck, N., Penault-Llorca, F., Cortes, J., Gnant, M., Houssami, N., Poortmans, P., Ruddy, K., 37 Tsang, J., and Cardoso, F. (2019). Breast cancer. Nat Rev Dis Primers 5, 66.
36 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 Heinz, S., Benner, C., Spann, N., Bertolino, E., Lin, Y.C., Laslo, P., Cheng, J.X., Murre, C., 2 Singh, H., and Glass, C.K. (2010). Simple combinations of lineage-determining transcription 3 factors prime cis-regulatory elements required for macrophage and B cell identities. Mol. Cell 4 38, 576–589.
5 Hodgkin, J., Papp, A., Pulak, R., Ambros, V., and Anderson, P. (1989). A new kind of 6 informational suppression in the nematode Caenorhabditis elegans. Genetics 123, 301–313.
7 Hogg, J.R., and Collins, K. (2007a). RNA-based affinity purification reveals 7SK RNPs with 8 distinct composition and regulation. RNA 13, 868–880.
9 Hogg, J.R., and Collins, K. (2007b). Human Y5 RNA specializes a Ro ribonucleoprotein for 5S 10 ribosomal RNA quality control. Genes Dev. 21, 3067–3072.
11 Hogg, J.R., and Goff, S.P. (2010). Upf1 senses 3′ UTR length to potentiate mRNA decay. Cell 12 143, 379–389.
13 Hollerer, I., Curk, T., Haase, B., Benes, V., Hauer, C., Neu-Yilik, G., Bhuvanagiri, M., Hentze, 14 M.W., and Kulozik, A.E. (2016). The differential expression of alternatively polyadenylated 15 transcripts is a common stress-induced response mechanism that modulates mammalian 16 mRNA expression in a quantitative and qualitative fashion. RNA 22, 1441–1453.
17 Hollien, J., and Weissman, J.S. (2006). Decay of endoplasmic reticulum-localized mRNAs 18 during the unfolded protein response. Science 313, 104–107.
19 Huang, L., Lou, C.-H., Chan, W., Shum, E.Y., Shao, A., Stone, E., Karam, R., Song, H.-W., and 20 Wilkinson, M.F. (2011). RNA homeostasis governed by cell type-specific and branched 21 feedback loops acting on NMD. Mol. Cell 43, 950–961.
22 Huntzinger, E., Kashima, I., Fauser, M., Saulière, J., and Izaurralde, E. (2008). SMG6 is the 23 catalytic endonuclease that cleaves mRNAs containing nonsense codons in metazoan. RNA 14, 24 2609–2617.
25 Hurt, J.A., Robertson, A.D., and Burge, C.B. (2013). Global analyses of UPF1 binding and 26 function reveal expanded scope of nonsense-mediated mRNA decay. Genome Res. 23, 1636– 27 1650.
28 Jan, C.H., Williams, C.C., and Weissman, J.S. (2014). Principles of ER cotranslational 29 translocation revealed by proximity-specific ribosome profiling. Science 346.
30 Karginov, F.V., and Hannon, G.J. (2013). Remodeling of Ago2-mRNA interactions upon cellular 31 stress reflects miRNA complementarity and correlates with altered translation rates. Genes & 32 Development 27, 1624–1632.
33 Karousis, E.D., and Mühlemann, O. (2018). Nonsense-Mediated mRNA Decay Begins Where 34 Translation Ends. Cold Spring Harb. Perspect. Biol.
35 Kashima, I., Yamashita, A., Izumi, N., Kataoka, N., Morishita, R., Hoshino, S., Ohno, M., 36 Dreyfuss, G., and Ohno, S. (2006). Binding of a novel SMG-1-Upf1-eRF1-eRF3 complex 37 (SURF) to the exon junction complex triggers Upf1 phosphorylation and nonsense-mediated 38 mRNA decay. Genes Dev. 20, 355–367.
39 Keene, J.D. (2007). RNA regulons: coordination of post-transcriptional events. Nat. Rev. Genet.
37 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 8, 533–543.
2 Kim, Y.K., and Maquat, L.E. (2019). UPFront and center in RNA decay: UPF1 in nonsense- 3 mediated mRNA decay and beyond. RNA 25, 407–422.
4 Kim, D., Paggi, J.M., Park, C., Bennett, C., and Salzberg, S.L. (2019). Graph-based genome 5 alignment and genotyping with HISAT2 and HISAT-genotype. Nat. Biotechnol. 37, 907–915.
6 Kishor, A., Ge, Z., and Hogg, J.R. (2018). hnRNP L-dependent protection of normal mRNAs 7 from NMD subverts quality control in B cell lymphoma. EMBO J.
8 Kishor, A., Fritz, S.E., and Robert Hogg, J. (2019). Nonsense-mediated mRNA decay: The 9 challenge of telling right from wrong in a complex transcriptome. Wiley Interdisciplinary 10 Reviews: RNA 10.
11 Kishor, A., Fritz, S.E., Haque, N., Ge, Z., Tunc, I., Yang, W., Zhu, J., and Hogg, J.R. (2020). 12 Activation and inhibition of nonsense-mediated mRNA decay control the abundance of 13 alternative polyadenylation products. Nucleic Acids Res.
14 Kurosaki, T., Li, W., Hoque, M., Popp, M.W.-L., Ermolenko, D.N., Tian, B., and Maquat, L.E. 15 (2014). A post-translational regulatory switch on UPF1 controls targeted mRNA degradation. 16 Genes Dev. 28, 1900–1916.
17 Lareau, L.F., Inada, M., Green, R.E., Wengrod, J.C., and Brenner, S.E. (2007). Unproductive 18 splicing of SR genes associated with highly conserved and ultraconserved DNA elements. 19 Nature 446, 926–929.
20 Lavysh, D., and Neu-Yilik, G. (2020). UPF1-Mediated RNA Decay-Danse Macabre in a Cloud. 21 Biomolecules 10, 999.
22 Lee, S.R., Pratt, G.A., Martinez, F.J., Yeo, G.W., and Lykke-Andersen, J. (2015). Target 23 Discrimination in Nonsense-Mediated mRNA Decay Requires Upf1 ATPase Activity. Mol. Cell 24 59, 413–425.
25 Le Hir, H., Moore, M.J., and Maquat, L.E.M. (2000). Pre-mRNA splicing alters mRNP 26 composition: evidence for stable association of proteins at exon-exon junctions. Genes Dev. 14, 27 1098–1108.
28 Le Hir, H., Izaurralde, E., Maquat, L.E., and Moore, M.J. (2000). The spliceosome deposits 29 multiple proteins 20-24 nucleotides upstream of mRNA exon-exon junctions. EMBO J. 19, 30 6860–6869.
31 Liao, Y., Smyth, G.K., and Shi, W. (2014). featureCounts: an efficient general purpose program 32 for assigning sequence reads to genomic features. Bioinformatics 30, 923–930.
33 Loh, B., Jonas, S., and Izaurralde, E. (2013). The SMG5-SMG7 heterodimer directly recruits the 34 CCR4-NOT deadenylase complex to mRNAs containing nonsense codons via interaction with 35 POP2. Genes Dev. 27, 2125–2138.
36 Longman, D., Jackson-Jones, K.A., Maslon, M.M., Murphy, L.C., Young, R.S., Stoddart, J.J., 37 Hug, N., Taylor, M.S., Papadopoulos, D.K., and Cáceres, J.F. (2020). Identification of a 38 localized nonsense-mediated decay pathway at the endoplasmic reticulum. Genes Dev.
38 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 Martin, M. (2011). Cutadapt removes adapter sequences from high-throughput sequencing 2 reads. EMBnet.journal 17, 10–12.
3 Martinez-Nunez, R.T., Wallace, A., Coyne, D., Jansson, L., Rush, M., Ennajdaoui, H., Katzman, 4 S., Bailey, J., Deinhardt, K., Sanchez-Elsner, T., et al. (2017). Modulation of nonsense mediated 5 decay by rapamycin. Nucleic Acids Res. 45, 3448–3459.
6 Mathews, M.B., and Hershey, J.W.B. (2015). The translation factor eIF5A and human cancer. 7 Biochim. Biophys. Acta 1849, 836–844.
8 Michlewski, G., and Cáceres, J.F. (2019). Post-transcriptional control of miRNA biogenesis. 9 RNA 25, 1–16.
10 Mugridge, J.S., Coller, J., and Gross, J.D. (2018). Structural and molecular mechanisms for the 11 control of eukaryotic 5’-3' mRNA decay. Nat. Struct. Mol. Biol. 25, 1077–1085.
12 Namkoong, S., Ho, A., Woo, Y.M., Kwak, H., and Lee, J.H. (2018). Systematic Characterization 13 of Stress-Induced RNA Granulation. Mol. Cell 70, 175–187.e8.
14 Ni, J.Z., Grate, L., Donohue, J.P., Preston, C., Nobida, N., O’Brien, G., Shiue, L., Clark, T.A., 15 Blume, J.E., and Ares, M., Jr (2007). Ultraconserved elements are associated with homeostatic 16 control of splicing regulators by alternative splicing and nonsense-mediated decay. Genes Dev. 17 21, 708–718.
18 Nicholson, P., Josi, C., Kurosawa, H., Yamashita, A., and Mühlemann, O. (2014). A novel 19 phosphorylation-independent interaction between SMG6 and UPF1 is essential for human 20 NMD. Nucleic Acids Res. 42, 9217–9235.
21 Page, M.F.P., Carr, B.C., Anders, K.R.A., Grimson, A.G., and Anderson, P.A. (1999). SMG-2 is 22 a phosphorylated protein required for mRNA surveillance in Caenorhabditis elegans and related 23 to Upf1p of yeast. Mol. Cell. Biol. 19, 5943–5951.
24 Pakos-Zebrucka, K., Koryga, I., Mnich, K., Ljujic, M., Samali, A., and Gorman, A.M. (2016). The 25 integrated stress response. EMBO Rep. 17, 1374–1395.
26 Park, Y., Reyna-Neyra, A., Philippe, L., and Thoreen, C.C. (2017). mTORC1 Balances Cellular 27 Amino Acid Supply with Demand for Protein Synthesis through Post-transcriptional Control of 28 ATF4. Cell Reports 19, 1083–1090.
29 Paz, I., Kosti, I., Ares, M., Jr, Cline, M., and Mandel-Gutfreund, Y. (2014). RBPmap: a web 30 server for mapping binding sites of RNA-binding proteins. Nucleic Acids Res. 42, W361–W367.
31 Powell, D. (2015). Degust: Visualize, explore and appreciate RNA-seq differential gene- 32 expression data. In COMBINE RNA-Seq Workshop,.
33 Pulak, R., and Anderson, P. (1993). mRNA surveillance by the Caenorhabditis elegans smg 34 genes. Genes Dev. 7, 1885–1897.
35 Risso, D., Ngai, J., Speed, T.P., and Dudoit, S. (2014). Normalization of RNA-seq data using 36 factor analysis of control genes or samples. Nat. Biotechnol. 32, 896–902.
37 Ritchie, M.E., Phipson, B., Wu, D., Hu, Y., Law, C.W., Shi, W., and Smyth, G.K. (2015). limma 38 powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic
39 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 Acids Res. 43, e47.
2 Robinson, M.D., McCarthy, D.J., and Smyth, G.K. (2009). edgeR : a Bioconductor package for 3 differential expression analysis of digital gene expression data. Bioinformatics 26, 139–140.
4 Shen, S., Park, J.W., Lu, Z.-X., Lin, L., Henry, M.D., Wu, Y.N., Zhou, Q., and Xing, Y. (2014). 5 rMATS: robust and flexible detection of differential alternative splicing from replicate RNA-Seq 6 data. Proc. Natl. Acad. Sci. U. S. A. 111, E5593–E5601.
7 Smith, J.E., and Baker, K.E. (2015). Nonsense-mediated RNA decay - a switch and dial for 8 regulating gene expression. Bioessays 37, 612–623.
9 Song, M.S., Salmena, L., and Pandolfi, P.P. (2012). The functions and regulation of the PTEN 10 tumour suppressor. Nat. Rev. Mol. Cell Biol. 13, 283–296.
11 The Gene Ontology Consortium (2019). The Gene Ontology Resource: 20 years and still GOing 12 strong. Nucleic Acids Res. 47, D330–D338.
13 Thorvaldsdóttir, H., Robinson, J.T., and Mesirov, J.P. (2013). Integrative Genomics Viewer 14 (IGV): high-performance genomics data visualization and exploration. Brief. Bioinform. 14, 178– 15 192.
16 Toma, K.G., Rebbapragada, I., Durand, S., and Lykke-Andersen, J. (2015). Identification of 17 elements in human long 3’ UTRs that inhibit nonsense-mediated decay. RNA 21, 887–897.
18 Vitting-Seerup, K., and Sandelin, A. (2019). IsoformSwitchAnalyzeR: Analysis of changes in 19 genome-wide patterns of alternative splicing and its functional consequences. Bioinformatics 20 btz247.
21 Waldron, J.A., Tack, D.C., Ritchey, L.E., Gillen, S.L., Wilczynska, A., Turro, E., Bevilacqua, 22 P.C., Assmann, S.M., Bushell, M., and Le Quesne, J. (2019). mRNA structural elements 23 immediately upstream of the start codon dictate dependence upon eIF4A helicase activity. 24 Genome Biol. 20, 300.
25 Wek, R.C. (2018). Role of eIF2α Kinases in Translational Control and Adaptation to Cellular 26 Stress. Cold Spring Harb. Perspect. Biol. 10.
27 Yepiskoposyan, H., Aeschimann, F., Nilsson, D., Okoniewski, M., and Mühlemann, O. (2011). 28 Autoregulation of the nonsense-mediated mRNA decay pathway in human cells. RNA 17, 2108– 29 2118.
30 Yi, Z., Sanjeev, M., and Singh, G. (2020). The Branched Nature of the Nonsense-Mediated 31 mRNA Decay Pathway. Trends Genet.
32 Young, S.K., and Wek, R.C. (2016). Upstream Open Reading Frames Differentially Regulate 33 Gene-specific Translation in the Integrated Stress Response. J. Biol. Chem. 291, 16927–16935.
34 Young-Baird, S.K., Lourenço, M.B., Elder, M.K., Klann, E., Liebau, S., and Dever, T.E. (2020). 35 Suppression of MEHMO Syndrome Mutation in eIF2 by Small Molecule ISRIB. Mol. Cell 77, 36 875–886.e7.
37 Zhang, J., Harvey, S.E., and Cheng, C. (2019). A high-throughput screen identifies small 38 molecule modulators of alternative splicing by targeting RNA G-quadruplexes. Nucleic Acids
40 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
1 Res. 47, 3667–3679.
2 Zünd, D., Gruber, A.R., Zavolan, M., and Mühlemann, O. (2013). Translation-dependent 3 displacement of UPF1 from coding sequences causes its enrichment in 3’ UTRs. Nat. Struct. 4 Mol. Biol. 20, 936–943.
41 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
A % UPF1LL rMATS 88 mRNA FDR 39 siNT 13.4 ± 1.9 n.a. 4 49 16 46 0 66 46 120 66 59 52 siUPF1LL 0.7 ± 0.7 7.8x10ˉ¹³ 1 3 0 100 46
UPF1SL
UPF1LL B C UPF1LL siRNA 1.0 isoforms without PTCs SMG6/7 knockdown (Colombo ) Validated poison exon * isoforms with PTCs 2 ( = 0.0008) * * * * * * * * * * * * 1 * * * * 0.5 * * * *
0 (relative mRNA abundance) 2 Cumulative frequency log
0.0 -1 -2 -1 0 1 2 SMG1SMG5 SRPR DOK4 GJB2 eIF5A2 CDK16HDAC1STAT2 EEF2KAIFM2 GCNT1EFNB3 ERBB2 UPF1 siRNA vs NT siRNA log (fold change) TMEM56 TMEM19 GPR161VPS37D LL 2 FAM171B METTL7A TNFRSF10D relative mRNA SRSF2SRSF3 (+PTC) (+PTC) abundance canonical UPF1LL targets decreased increased NMD targets
Sashimi plot from
representative RNA-seq samples of siNT and siUPF1LL knockdown cells. Percent spliced in values and FDR were calculated with rMATS software. CDF plot of relative mRNA abundance as determined from RNA-seq following
UPF1LL-specific knockdown in HEK-293 cells. Genes for which expression of an isoform containing a termination codon greater than 50 nt upstream of the final exon junction were analyzed, with mRNA isoforms classified as having or lacking a PTC. Statistical significance was determined by K-S test. qPCR analysis of indicated
transcripts from UPF1LL knockdown RNA-seq experiments. Relative fold changes are in reference to NT siRNA.
Significance of UPF1LL knockdown was compared to the NT siRNA control. Asterisk (*) indicates < 0.05, as determined by unpaired Student’s t-test. Error bars indicate SD (n = 3). Transcripts that increased in abundance at least 1.4-fold in the SMG6/7 knockdown Colombo et al. dataset (2017) are indicated with a red triangle. mRNAs with validated poison exons (Lareau et al., 2007; Ni et al., 2007) are indicated with a red square.
42 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
A B C 1.0 Other (n = 9975) 1.0 Other (n = 9975) 1.0
Up > 0.5x in total Up > 0.5x in total UPF1 knockdown UPF1 knockdown (n = 678; = 0.01) (n = 678; 2x10-7)
0.5 0.5 0.5
isoforms without PTCs
isoforms with PTCs ( = 0.03) Cumulative frequency Cumulative frequency Cumulative frequency
0.0 0.0 0.0 -1 0 1 -1 0 1 -1.0 -0.5 0.0 0.5 1.0 UPF1 vs UPF1 log (fold change) UPF1SL vs GFP log2(fold change) UPF1LL vs GFP log2(fold change) LL SL 2 relative mRNA relative mRNA relative mRNA abundance abundance abundance decreased increased decreased increased decreased increased
D E 1.0 1.0
0.5 0.5 (Δ stability) (Δ stability) 2 2 0.0 0.0 vs GFP log
vs GFP log -0.5 -0.5 LL SL Spearman's ρ = -0.45 ( = 4x10-5) Spearman's ρ = 0.65 ( < 1x10-15) UPF1
UPF1 -1.0 -1.0 -2 -1 0 1 2 -2 -1 0 1 2 UPF1 vs GFP log (fold change) UPF1SL vs GFP log2 (fold change) LL 2
F G **** 0.5 1.0 **** **** stability) Δ ( 2 0.5 0.0
3’UTRs < 477 nt (n = 2215) vs GFP log LL
Cumulative frequency 3’UTRs 477-1578 nt (n = 2883; = 0.03) UPF1 3’UTRs > 1579 nt (n = 3160; < 1x10-15) 3'UTRs > 1579 nt 0.0 -0.5 -0.5 0.0 0.5 n= 201 880 1205 874
UPF1LL vs GFP log2(Δ stability)
relative mRNA PTBP1/hnRNP L motif density within the 3'UTR stability decreased increased
CDF plot of relative mRNA abundance as determined from RNA-seq following CLIP-UPF1SL overexpression. mRNAs previously determined to be up- regulated upon total UPF1 knockdown (Baird et al., 2018) were analyzed. Statistical significance was determined by K-S test. CDF plot as in (A), following CLIP-UPF1LL overexpression. CDF plot of relative mRNA abundance as determined from RNA-seq following CLIP-UPF1LL vs CLIP-UPF1SL overexpression. Genes for which expression of an isoform containing a termination codon greater than 50 nt upstream of the final exon junction were analyzed, with mRNA isoforms classified as having or lacking a PTC. Statistical significance was determined byK-Stest. Scatterplot of changes in relative mRNA stability as determined by REMBRANDTS analysis of RNA-seq vs relative mRNA abundance as determined by kallisto, comparing CLIP-UPF1SL to GFP overexpression. Scatterplot as in (D), comparing CLIP-UPF1LL to GFP overexpression. CDF plot of relative mRNA stability as determined by REMBRANDTS analysis of RNA-seq following CLIP-UPF1LL or GFP overexpression. mRNAs were binned according to 3’UTR lengths (short, medium, long). Statistical significance was determined by K-W test, with Dunn's correction for multiple comparisons. Box plot of relative mRNA stability as determined by REMBRANDTS analysis of RNA-seq for transcripts having long 3'UTRs (> 1579 nt) following CLIP-UPF1LL or GFP overexpression. mRNAs were binned by PTBP1 and/or hnRNP L motif density within the 3'UTR. Statistical significance was determined by K-W test, with Dunn's correction for multiple comparisons (**** < 7x10-8). Boxes indicate interquartile ranges, and bars indicate Tukey whiskers.
43 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
A B SL LL
Regulatory loop UPF1 UPF1
UPF1 full length CH 1B 1C RecA1 RecA2 80
UPF1SLΔCH
347LPTKDSDMRLM357
UPF1LLΔCH LPTKDS DMRLM
Schematic representation of the domain organization of full-length human UPF1, with the UPF1ΔCH constructs assessed in this study depicted below. The 11 amino acid extension in the regulatory loop of
UPF1LL is indicated in red text. SDS-PAGE of recombinant 6xHIS-CBP-UPF1SLΔCH (MW: 76 kDa)
and 6xHIS-UPF1LLΔCH (MW: 73 kDa) proteins purified from and used in this study.
44 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
A Streptavidin Input Pull down Flow through SL LL SL LL SL LL GFP GFP GFP UPF1 UPF1 UPF1 UPF1 UPF1 UPF1
CLIP-UPF1 Endogenous UPF1
β-actin
B
1.0
mRNAs increased in total 0.5 UPF1 knockdown (n=769)
-25 UPF1LL ( <1x10 ) -25 UPF1SL ( <1x10 ) Remaining mRNAs Cumulative frequency (n=11,442)
UPF1LL
UPF1SL 0.0 0 10 20 UPF1 RIP-seq efficiency (IP/Input)
enrichment in RIP decreased increased
Western blots of CLIP-UPF1SL and CLIP-UPF1LL cell lysates before (input) and after (pull down) biotin- streptavidin affinity purification. Input and flow through lanes represent 5% of total material. CDF plot
of recovered mRNAs as determined from RIP-seq efficiency in CLIP-UPF1LL or CLIP-UPF1SL affinity purifications. mRNAs were binned by sensitivity to total UPF1 knockdown (Baird et al., 2018). Statistical significance was determined by K-S test.
45 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
Sensitivity to UPF1LL A B UPF1 RIP-seq efficiency overexpression 1.0 **** *** **** 2
(fold change) 1 2
0.5 0 High ER translational enrichment (n = 501; < 1x10-15)
-1
Cumulative frequency Low ER translational enrichment (n = 3554) -2 n= 8013 156 596 8471 294
0.0 Tunicamycin vs DMSO log
0 1 2 SL LL Other UPF1LL vs UPF1SL RIP-seq efficiency overexpression enrichment in Equally recovered LL UPF1 RIP Enriched withEnriched UPF1 with UPF1 Downregulated with decreased LL increased UPF1
C Sensitivity to UPF1 D E LL HEK293 UPF1 RIP-seq efficiency overexpression 400 2 **** siNT DMSO *** **** siNT Thapsigargin 300 NT siRNA UPF1LL siRNA siUPF1 DMSO 1 LL (fold change) 2 siUPF1LL Thapsigargin Vehicle Thapsigargin Vehicle Thapsigargin 200 0
100 -1 Total RNA extraction Transcript abundance (CPM) n= 7476 150 543 7903 257 0
Thapsigargin vs DMSO log -2 RNA-seq RT-qPCR
SL LL Other ATF3 DDIT3 DDIT4 ASNS XBP1 CEBPB DNAJB9 DNAJC3 EIF2AK3 overexpression PPP1R15A Equally recovered LL Enriched withEnriched UPF1 with UPF1 Downregulated with ISR target mRNAs UPF1 F G
DMSO Thapsigargin
eIF2α-P % UPF1LL rMATS mRNA FDR DMSO 13.4 ± 1.9 n.a. eIF2α Thapsigargin 15.8 ± 5.1 1 β-actin 1.0 2.5 Relative phosphorylation of eIF2α
CDF plot of recovered mRNAs as determined from RIP-seq efficiency in CLIP-UPF1LL or
CLIP-UPF1SL affinity purifications. mRNAs were binned according tolog2 enrichment for translation at the ER (Jan et al., 2014). Statistical significance was determined by K-S test. Box plot of relative mRNA abundance as determined from RNA-seq following treatment of HeLa cells with 30 µM tunicamycin for 6hr (GSE86857). mRNAs were binned by RIP-seq efficiency in CLIP-UPF1LL or CLIP-UPF1SL affinity purifications (left) or sensitivity to CLIP-UPF1LL overexpression (right). Statistical significance was determined by K-W test, with Dunn's correction for multiple comparisons (left;*** < 0.001; **** < 1x10-14 ), or K-S test (right; **** < 1x10-12). Boxes indicate interquartile ranges, and bars indicate Tukey whiskers. Box plot as in (B), following treatment of NIH3T3 cells with 1 µM thapsigargin for 1.5hr (GSE90869; **** < 1x10-6).
Schematic of the RNA-seq experimental workflow and conditions for UPF1LL knockdown and thapsigargin treatment. Quantification of characterized ISR-target transcript abundance in RNA-seq of the indicated conditions. Errorbars indicate SD (n = 3). Quantification of UPF1LL isoform expression in control and thapsigargin-treated HEK-293 cells from rMATS analyses. Western blot of eIF2α phosphorylation following treatment of HEK-293 cells with 1 µM thapsigargin for 6hr.
46 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
Sensitivity to UPF1LL A B C UPF1 RIP-seq efficiency overexpression 30 Enriched with UPF1 (n = 695) 8 Enriched with UPF1 (n = 700) **** LL LL n.s. ****
Enriched with UPF1SL (n = 170) Enriched with UPF1SL (n = 170)
Equally recovered with UPF1 Equally recovered with UPF1 2 LL 6 LL and UPF1SL (n = 9102) and UPF1SL (n = 8944)
20 (fold change) 2 ) (P) ( 10 10 4 0 -log -log 10 2 -2
0 0 Emetine vs DMSO log n= 9230 175 718 9788 336
-4 -2 0 2 4 -2 -1 0 1 2 SL LL Other
Emetine vs DMSO log2(fold change) Hippuristanol vs DMSO log2(fold change) overexpression relative mRNA relative mRNA Equally recovered LL abundance abundance Enriched withEnriched UPF1 with UPF1 Downregulated with decreased increased decreased increased UPF1
Sensitivity to UPF1LL UPF1 RIP-seq efficiency overexpression HEK293 Sensitivity to UPF1LL D E UPF1 RIP-seq efficiency overexpression F **** *** *** 1.0 **** n.s. **** 1 NT siRNA UPF1LL siRNA 0.5 (fold change) (fold change) 2 2 Vehicle Puromycin Vehicle Puromycin
0 0.0
Total RNA extraction -0.5 -1
n= 9044 171 704 9586 333 n= 8253 151 686 8770 321 RNA-seq RT-qPCR Anisomycin vs Vehicle log -1.0 Hippuristanol vs Vehicle log
SL LL SL LL
Other Other
overexpression overexpression Equally recovered LL Equally recovered LL Enriched withEnriched UPF1 with UPF1 Downregulated with Enriched withEnriched UPF1 with UPF1 Downregulated with UPF1 UPF1 G H puromycin I 1.0 Down in puromycin 1 (n = 929; < 1x10-15) Other (n = 6924) FMR1 % UPF1LL rMATS OSBPL8 mRNA FDR SRPR 0.5 CDK16 DMSO 13.4 ± 1.9 n.a. TNFRSF10D DICER1 Puromycin 16.9 ± 13.2 1 Cumulative frequency LDLR Fraction mRNA remaining TMEM165 0.0 0.1 -1 0 1 0 30 60 90 120
UPF1LL siRNA vs NT siRNA in puromycin log2(Δ stability) Time (minutes)
relative mRNA stability decreased increased
J K
1.0 UPF1 overexpression targets LL Down in puromycin and rescued by siUPF1LL -11 **** (n = 250; = 1x10 ) 2 (n = 700; = 2x10-12) Other (n = 7063) **** Up in puromycin (n = 2560; < 1x10-15) (fold change) 2 Other (n = 10420; = 9x10-6) **** 0.5 0 Cumulative frequency -2
0.0 Puromycin vs Vehicle log -1.0 -0.5 0.0 0.5 1.0 Puromycin vs Vehicle log (Δ stability) 2 25 μg/mL 25 μg/mL 25 μg/mL 100 μg/mL 100 μg/mL 100 μg/mL relative mRNA stability decreased increased
47 bioRxiv preprint doi: https://doi.org/10.1101/2021.01.27.428318; this version posted January 27, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
Volcano plot of relative mRNA abundance as determined from RNA-seq following treatment of HEK-293 cells with 50 ng/mL emetine for 4hr (GSE89774). mRNAs were binned by RIP-seq efficiency in CLIP-UPF1LL or CLIP-UPF1SL affinity purifications. Horizontal line indicates the significance threshold ≤ 0.05. Volcano plot as in (A), following treatment of MCF7 cells with 150 nM hippuristanol for 1hr (GSE134888). Box plot of relative mRNA abundance as determined from RNA-seq following treatment of HMLE-TWIST cells with 10 µM emetine for 24hr (GSE113505). mRNAs were binned by RIP-seq efficiency in CLIP-UPF1LL or CLIP-UPF1SL affinity purifications (left) or sensitivity to CLIP-UPF1LL overexpression (right). Statistical significance was determined by K-W test, with Dunn's correction for multiple comparisons (left; **** < 1x10-15), or K-S test (right; **** < 1x10-7). Boxes indicate interquartile ranges, and bars indicate Tukey whiskers. Box plot as in (C), following treatment of HEK-293 cells with 5 µg/mL anisomycin for 3hr (ERP010553; *** < 0.001, **** < 1x10-11). Box plot as in (C), following treatment of HEK-293 cells with 1 µM hippuristanol for 30min (GSE44378; **** < 1x10-15 (left), **** < 1x10-5 (right)). Schematic of the RNA- seq experimental workflow and conditions for UPF1LL knockdown and puromycin treatment. CDF plot of relative mRNA stability as determined by REMBRANDTS analysis of RNA-seq following UPF1LL knockdown in HEK-293 cells and treatment with 50 µg/mL puromycin for 4hr. mRNAs were binned by sensitivity to puromycin in the absence of UPF1LL knockdown. Statistical significance was determined by K-S test. Fraction of mRNA remaining, as determined by RT-qPCR, for indicated transcripts following treatment of HEK-293 cells with 50 µg/mL puromycin over the course of 2hr. Error bars indicate
SD (n = 3). Quantification of UPF1LL isoform expression in control and puromycin-treated HEK-293 cells from rMATS analyses. CDF plot of relative mRNA stability as determined by REMBRANDTS analysis of RNA-seq following treatment of HEK-293 cells with 50 µg/mL puromycin for 4hr. mRNAs were binned by sensitivity to CLIP-UPF1LL overexpression. Statistical significance was determined by K-S test. Box plot of relative mRNA abundance as determined by RNA-seq following treatment of HEK-293 cells with 25 µg/mL or 100 µg/mL puromycin. mRNAs were binned according to sensitivity to
50 µg/mL puromycin and UPF1LL knockdown. Statistical significance was determined by K-S test. Boxes indicate interquartile ranges, and bars indicate Tukey whiskers.
48