Target Discrimination in Nonsense-Mediated mRNA Decay Requires Upf1 ATPase Activity

Abstract

RNA quality-control pathways get rid of faulty RNAs and therefore must be able to discriminate these RNAs from those that are normal. Here we present evidence that the adenosine triphosphatase (ATPase) cycle of the SF1 helicase Upf1 is required for mRNA discrimination during nonsense-mediated decay (NMD). Mutations affecting the Upf1 ATPase cycle disrupt the mRNA selectivity of Upf1, leading to indiscriminate accumulation of NMD complexes on both NMD target and non-target mRNAs. In addition, two modulators of NMD-translation and termination codon-proximal poly(A) binding protein-depend on the ATPase activity of Upf1 to limit Upf1-non-target association. Preferential ATPase-dependent dissociation of Upf1 from non-target mRNAs in vitro suggests that selective release of Upf1 contributes to the ATPase dependence of Upf1 target discrimination. Given the prevalence of helicases in RNA regulation, ATP hydrolysis may be a widely used activity in target RNA discrimination.

Figure 1. Upf1 ATPase cycle mutants are defective in selective NMD target association

(A) NMD target (PTC) and non-target (NTC) mRNA reporter pairs used in RNA-IP assays based on human β-globin, β-globin with an insertion from GAPDH (β-GAP), or GPx1. The NMD-inducing termination codon is denoted in bold. (B) Northerns of β-globin and control mRNAs in input (0.5%) samples or coprecipitated with Flag-Upf1 using anti-Flag antibody (α-Flag IP). Flag-Upf1 recovered in IPs is shown alongside a two-fold titration of input on anti-Flag Westerns below Northerns. The graph on the right represents recovery of β-globin mRNAs (NTC or PTC) with Flag-Upf1 over input normalized to recovery of the internal control (lanes 3,4), after subtraction of background in negative control IPs (lanes 1,2). Data are represented as mean +/− SEM for five biological replicates. (C) Similar to panel B for RNA-IPs of β-GAP mRNAs. (D) Similar to panel B, comparing IPs with Flag-Upf1 WT, D637A/E638A (DEAA), or K498A (KA). Western of recovered Flag-Upf1 variants is shown below Northerns. (E) Graph representing the ratio of normalized IP recovery for NMD target (PTC) to non-target (NTC) mRNAs with the indicated Upf1 variants. A value of one, denoted by the dotted line, reflects an absence of discrimination between target and non-target. Data are represented as mean +/− SEM for three to four biological repeats. (F) Similar to panel C but without normalization, comparing percent IP recovery of β-GAP mRNAs and the internal control by Flag-Upf1 WT, F192E (FE), DEAA or DEAA/FE mutants. Asterisks denote P-values: *≤0.05, **≤0.001 (paired student’s t-test, two-tailed). See also Figure S1

Figure 2. Transcriptome-wide loss in mRNA selectivity for Upf1 ATPase cycle mutants

(A-C) Scatter plots of reads per kilobase transcript per million mapped reads (RPKM) from RNA-seq of input samples versus IPs for Flag-Upf1 WT (A), DEAA (B) and KA (C). Genes with IP/input ratios for WT Upf1 of greater than 2.05 (cut-off based on 5% false discovery rate (FDR) established using cells expressing Flag epitope only, Figure S2) are shown in red (Upf1-enriched), while genes with log₂ (IP/input) between −0.5 and +0.5 are shown in blue (non-enriched). All remaining genes are shown in grey. (D) Cumulative fraction of Upf1-enriched and non-enriched genes with IP enrichment represented as log₂ (IP RPKM/input lysate RPKM) for Flag-Upf1 WT, KA, and DEAA, along with Flag only. Difference between WT Upf1 (Upf1-enriched) curve compared to all other curves was statistically significant (p-value <0.05 for all comparisons, KS-test). See also Figure S2 and Table S1

Figure 3. ATP hydrolysis-deficient Upf1 accumulates as a phosphoprotein in complexes with Smg proteins on both target and non-target mRNA

(A) Northerns of β-GAP and control reporter mRNAs in inputs (0.3%) or RNA-IPs with αphospho-S1116 Upf1 (α-p-Upf1) or α-Upf1 antibodies from okadiac acid-treated cells expressing Flag-Upf1 WT or DEAA. Control RNA-IPs performed with α-phospho-Upf1 from lysates treated with or without λ protein phosphatase are shown on the right. Westerns of Flag-Upf1 recovered in IPs alongside a two-fold titration of Flag-Upf1 WT input are shown below Northerns. (B) Graphs representing mean IP recovery of β-GAP mRNAs normalized by recovery of the internal control +/− SEM for triplicate biological repeats of α-phospho-Upf1 RNA-IPs and duplicates of α-Upf1 RNA-IPs. (C) Northerns of β-GAP and control mRNAs in inputs (0.6%) or coprecipitated with Flag-Smg5, -Smg6, or -Smg7 in cells coexpressing Myc-Upf1 WT or DEAA. Westerns of Flag-Smg proteins recovered in IPs, along with copurifying Myc-Upf1 are shown below Northerns. (D) Similar to panel C for RNA-IPs with Myc-Upf1 WT or DEAA, with 2% inputs loaded. Western of Myc-Upf1 recovered in IPs alongside a two-fold titration of Myc-Upf1 WT input is shown below Northerns. (E) Quantifications, similar to those in panel B, for RNA-IPs shown in panels C, D with SEM for triplicate (Smg5, Smg7, Upf1) or duplicate (Smg6) biological repeats. Asterisks denote P-values: *≤0.05, **≤0.01 (paired student’s t-test, two-tailed). See also Figure S3

Figure 4. Translation prevents Upf1 accumulation on non-target mRNA in a Upf1 ATPase-dependent manner

(A) Northerns for β-GAP and control reporter mRNAs in inputs (0.5%) or coprecipitated with Flag-Upf1. Schematic of the β-GAP mRNAs used in RNA-IPs is shown below Northerns. HP denotes a stable RNA hairpin in the 5′ UTR that blocks translation. Graphs on the right represent mean IP recovery over input of β-GAP mRNA normalized by recovery of the internal control after subtraction of background from negative control IPs, +/− SEM for triplicate biological repeats. (B) Similar to panel A for β-globin RNA-IPs, except performed in the presence of Smg6 depletion to prevent Smg6-mediated cleavage of the β-globin PTC mRNA (see Figure S1). Asterisks denote P-values: *≤0.1, **≤0.05, *** ≤0.01 (paired student’s t-test, two-tailed). See also Figure S4

Figure 5. PTC-proximal poly(A) binding protein prevents Upf1 accumulation on mRNA in a Upf1 ATPase-dependent manner

(A) Northerns for β-GAP and control reporter mRNAs in inputs (0.3%) or coprecipitated with Flag-Upf1 WT or DEAA from cells coexpressing Myc-tagged proteins as indicated. MS2 denotes fusion with MS2 coat protein. Schematic of the β-GAP PTC-6xMS2 mRNA used is shown below Northerns. Graphs on the right represent mean IP recovery of reporter mRNA normalized by recovery of the internal control after subtraction of background from negative control IPs, +/− SEM for triplicate biological repeats. (B) Similar to panel A for β-GAP PTC reporter mRNAs containing 30 adenosines (A30) or a 30-nucleotide degenerate sequence (N30) instead of MS2 coat protein binding sites (schematic shown below Northerns), with 0.5% inputs loaded on Northerns and IPs performed in the absence of MS2 fusion proteins. Asterisks denote P-values: *≤0.1, **≤0.05, *** ≤0.01 (paired student’s t-test, two-tailed). See also Figure S5

Figure 6. Upf1 ATPase mutants accumulate in 3’UTRs with enhanced binding downstream of termination codons and near 3′ ends

(A) Mean read density for Upf1 WT and mutant CLIP assays across the metagene, normalized to the total number of reads per gene. (B) Cumulative fraction of genes with 3’UTR read abundance represented as a fraction of total reads in the gene, normalized to nucleotide length, shown for all mRNAs on the left, and for WT enriched and non-WT enriched mRNAs, as defined in Figure 2, on the right. Differences between WT and mutant curves were statistically significant (p-value < 0.001; KS statistic 0.30 for both DEAA and KA compared to WT for non-WT Enriched and 0.20 and 0.19 for DEAA and KA, respectively, compared to WT for WT Enriched) (C) Mean read densities, shown as percentages of total reads mapped in the region depicted per mRNA, around the first nucleotide of the stop codon, shown on the left, and the 3′ end of annotated transcripts, shown on the right. Solid lines represent regions where differences were found to be significant with a P-value <0.05 (Bonferroni corrected). See also Figure S6 and Table S2

Figure 7. Upf1 ATP hydrolysis is required for faster release of Upf1 from non-target over target mRNA

(A) Northerns for β-globin and control reporter mRNAs in unbound fractions or coprecipitated with Flag-Upf1 WT from cell lysates treated with MgCl₂/ATP for the number of minutes indicated above lanes. Graph under Northerns shows the ratio of βglobin mRNA recovery in IPs to the internal control β-GAP PTC mRNA after subtraction of background from negative control Flag IPs as a function of MgCl₂/ATP-treatment duration, normalized to values at time 0. Data are represented as the normalized mean ratio +/− SEM for two to three biological replicates. P-value calculations were restricted to time points with triplicate measurements. (B) Northerns of β-globin and control mRNAs in unbound fractions or coprecipitated with Flag-Upf1 from untreated lysates (-) or lysates treated with MgCl₂/ATP or MgCl₂/AMP PNP for 360 minutes. Graphs under Northerns represent mean ratios +/− SEM for triplicate biological repeats calculated as in panel A, normalized to values for untreated lysate samples. (C) Similar to panel B comparing release of β-globin +/− HP mRNAs used in Figure 4B. Graphs under Northerns are from triplicate biological repeats normalized to values for AMP-PNP-treated samples. Asterisks denote P-values: *≤0.1, **≤0.05, ***≤0.01 (paired student t-test, two-tailed). See also Figure S7 (D) Model depicting the ATPase-dependent mRNA discrimination step by Upf1 preceding NMD complex assembly. A second mRNA-selective commitment step might occur prior to mRNA degradation. See text for detail.