Transcriptome-wide identification of NMD-targeted human mRNAs reveals extensive redundancy between SMG6- and SMG7-mediated degradation pathways

Abstract

Besides degrading aberrant mRNAs that harbor a premature translation termination codon (PTC), nonsense-mediated mRNA decay (NMD) also targets many seemingly "normal" mRNAs that encode for full-length proteins. To identify a bona fide set of such endogenous NMD targets in human cells, we applied a meta-analysis approach in which we combined transcriptome profiling of knockdowns and rescues of the three NMD factors UPF1, SMG6, and SMG7. We provide evidence that this combinatorial approach identifies NMD-targeted transcripts more reliably than previous attempts that focused on inactivation of single NMD factors. Our data revealed that SMG6 and SMG7 act on essentially the same transcripts, indicating extensive redundancy between the endo- and exonucleolytic decay routes. Besides mRNAs, we also identified as NMD targets many long noncoding RNAs as well as miRNA and snoRNA host genes. The NMD target feature with the most predictive value is an intron in the 3' UTR, followed by the presence of upstream open reading frames (uORFs) and long 3' UTRs. Furthermore, the 3' UTRs of NMD-targeted transcripts tend to have an increased GC content and to be phylogenetically less conserved when compared to 3' UTRs of NMD insensitive transcripts.

Keywords: RNA turnover; SMG6; SMG7; UPF1; bioinformatics analysis; mRNA-seq; nonsense-mediated mRNA decay; post-transcriptional gene regulation.

FIGURE 1.

Monitoring of UPF1, SMG6, and SMG7 knockdown (KD), double knockdown (dKD), and rescue experiments compared to a control knockdown (Ctrl). (A) Lysates corresponding to 2 × 10⁵ cell equivalents of HeLa cells transiently transfected with the indicated knockdown and rescue constructs were analyzed by Western blotting. After electrophoretic separation of the proteins on 10% SDS–PAGE and transfer to nitrocellulose membranes, membrane sections were incubated with antibodies against UPF1, SMG6, SMG7, and CPSF73, the latter serving as loading control. The anti-SMG7 antibody gives a double band of which only the upper band (arrow) corresponds to SMG7. (B) Relative mRNA levels of known endogenous NMD-targeted mRNAs (GAS5, RP9P, SMG5, ATF4), normalized to β-actin mRNA (ACTB), were determined for all conditions 72-h post-transfection by RT-qPCR. Mean values and standard deviations of three independent experiments are shown, with the samples in the control knockdown (Ctrl) set to 1.0.

FIGURE 2.

High overlap of putative NMD targets identified in the different conditions. (A) Bar plot displaying the results of a k-means clustering procedure performed on the log₂FC measured for each gene in the indicated conditions. The number of clusters was set to two. The y-axis shows the average log₂FC of all the genes present in cluster 1 (comprising 40,000 genes) and cluster 2 (4000 genes). KD refers to the log₂FC (KD/Ctrl) and rescue to the log₂FC (rescue/KD), respectively. The double knockdown of SMG6 and SMG7 (dKD) was rescued either with SMG6 (dKD SMG6 rescue) or with SMG7 (dKD SMG7 rescue). (B) Histogram of genes with positive KD-rescue log₂FC, which is expected from NMD targets. The top 1000 targets in each of our conditions were determined. Each of these sets corresponds to a cluster of columns in the plot. The y-axis shows the fraction of these targets that have a positive log₂FC in the other conditions. The log₂FC used for this analysis is the average between the log₂FC (KD/Ctrl) and the inverse of the log₂FC (rescue/KD) (see Materials and Methods). (C) Bar plots showing the RNA levels of the three factors under study upon each single KD.

FIGURE 3.

SMG6 and SMG7 dKD and individual rescues reveal the highly redundant activity of these two NMD factors. (A) Scatter plot comparing SMG6 and SMG7 rescues from the dKD. The picture shows the log₂FC of the analysis of SMG6 rescue versus dKD (x-axis) and SMG7 rescue versus dKD (y-axis). Colored in red are genes significantly down-regulated in either of the two conditions. The histograms on the y-axis (on the right side) and x-axis (on the top) show the SMG7 log₂FC distribution of all the significant down-regulated targets in SMG6 rescue and vice versa, respectively. (B) Simulation to estimate the variation between SMG6 and SMG7 results. Based on the negative binomial parameters computed from our data, new counts data sets were simulated. Additional variation was added to provide an accurate estimate of the individual difference between dKD_SMG6 and dKD_SMG7 rescues. The picture compares the correlation scores (y-axis) found in the simulations at different levels of variation (x-axis) with the observed one in our data set (black horizontal line).

FIGURE 4.

NMD targets transcripts classified as noncoding, small-RNA host RNAs, and products of pervasive transcription. (A) Pie chart illustrating the top 1000 NMD targets categorized according to their biotype. Seventy-eight percent of these NMD targets code for protein, 9% are pseudogenes, 6% lincRNAs, and 4% antisense transcripts. (B) The number of possible ORFs in noncoding RNAs correlates with the likelihood of undergoing NMD. The expressed noncoding genes are partitioned in different bins depending on how many theoretical ORFs can be predicted on their sequence. The y-axis reports the percentage of NMD targets (top 1000) of all the genes in each bin (e.g., 8% of genes with three ORFs are NMD targets). (C) Top NMD targets are enriched in snoRNA and miRNA host genes. Each bar shows the percentage of genes that are among the NMD targets (top 1000) in every class of genes. (D) Transcripts initiating upstream of the canonical transcription start site (TSS) are partially cleared by NMD. The number of reads upstream of every annotated TSS has been computed in the indicated conditions. This quantity was divided by the total counts of every gene and every condition was analyzed comparing KDs to Ctrl and rescues to KDs, as in the normal analysis (see Materials and Methods). A box plot showing the log₂FC of these quantities is displayed.

FIGURE 5.

NMD targets are enriched in known and novel characteristic features. (A) Bar plot showing the enrichment of mRNAs with introns in the 3′ UTR and uORFs among NMD targets. All mRNAs with introns in the 3′ UTR have been removed from the analysis of uORFs. (B) Box plots comparing 3′ UTR features between NMD targets and a matched control group. All mRNAs with 3′ UTR introns or uORFs have been removed from the analysis. This resulted in a total of 660 significant isoforms. The control set is always a set of mRNAs that have the same expression levels of the NMD mRNAs in the Ctrl condition.