Coding regions affect mRNA stability in human cells

Abstract

A new paradigm has emerged that coding regions can regulate mRNA stability in model organisms. Here, due to differences in cognate tRNA abundance, synonymous codons are translated at different speeds, and slow codons then stimulate mRNA decay. To ask if this phenomenon also occurs in humans, we isolated RNA stability effects due to coding regions using the human ORFeome collection. We find that many open reading frame (ORF) characteristics, such as length and secondary structure, fail to provide explanations for how coding regions alter mRNA stability, and, instead, that the ORF relies on translation to impact mRNA stability. Consistent with what has been seen in other organisms, codon use is related to the effects of ORFs on transcript stability. Importantly, we found instability-associated codons have longer A-site dwell times, suggesting for the first time in humans a connection between elongation speed and mRNA decay. Thus, we propose that codon usage alters decoding speeds and so affects human mRNA stability.

Keywords: codons; mRNA stability; post-transcriptional regulation; translation elongation.

FIGURE 1.

Coding sequences regulate mRNA stability in human cells. (A) Schematic of ORFeome pool creation. The ORFeome collection contains ∼15,000 full-length coding regions in a lentiviral plasmid background, where each ORF is flanked by invariant UTRs and also contains a carboxy-terminal V5 tag. Pools of ∼3000 ORFeome clones from the collection were used to make lentivirus, which was then used to infect HEK293T cells and generate pools of stable cell lines. After selection, mRNA half-lives were measured through an approach-to-equilibrium 4SU-labeling experiment, giving stabilities for endogenous and ORFeome-derived transcripts. (B) Table summarizing the number of endogenous and ORFeome mRNAs passing each step in the processing pipeline. Endogenous and ORFeome transcripts were first examined in the steady state libraries. ORFeome transcripts were required to be expressed >3 times as much in the appropriate cell line than in the other. Once classified as “endogenous” or “ORFeome,” mRNA half-lives were calculated from the metabolic labeling experiment. Total endogenous mRNA half-lives correspond to the number of mRNAs with at least one measured half-life; if more than one half-life was measured, the arithmetic mean was used. (C) Coding regions change mRNA stability. Box-and-whisker plots of the stabilities of endogenous (End., in gray) and ORFeome mRNAs (in blue). Line represents median, box demarcates second and third quartiles, points are outliers. (D) ORFeome mRNAs show as much variability in stability as endogenous mRNAs. Plotted are the density distributions of median-centered stabilities of endogenous and ORFeome mRNAs (in gray and blue, respectively).

FIGURE 2.

Inhibiting ribosome loading reduces the effects of the coding region on mRNA stability. (A) 4EGI-1 treatment inhibits translation. Shown are A254 traces from sucrose density gradients of lysates from HEK293T cells treated with DMSO (gray) or 4EGI-1 (orange). (B) 4EGI-1 treatment reduces translation, as measured by puromycin incorporation. Cells were treated with DMSO, 4EGI-1, or cycloheximide (CHX) for the indicated times. To measure ongoing translation, cells were then pulsed with puromycin and harvested. Cell lysates were separated by SDS-PAGE electrophoresis, and western blotting was performed, probing against puromycin and α-tubulin (as a loading control). (C) Translation inhibition destabilizes endogenous mRNAs. Plotted are boxplots of half-lives for endogenous HEK293T mRNAs with DMSO or 4EGI-1 treatment (in gray and orange, respectively). The line represents median half-life, the box demarcates second and third quartiles, and points are outliers. (D) Translation inhibition has a minor effect on the variation in stability for endogenous mRNAs. Plotted are density distributions of median-centered half-lives for endogenous HEK293T in cells treated with DMSO or 4EGI-1 (in gray and orange, respectively). (E) Translation inhibition destabilizes ORFeome-derived mRNAs. As in C, except for ORFeome mRNAs. DMSO, in blue; 4EGI-1, in orange. (F) Translation inhibition reduces the variation in stability for ORFeome-derived mRNAs. As in D, except for ORFeome mRNAs. DMSO, in blue; 4EGI-1, in orange. See also Supplemental Figure S2 and Supplemental Tables S1, S2.

FIGURE 3.

Codon use corresponds to mRNA stability. (A) Codons are differentially associated with stability. Shown are average Spearman correlations, for each nonstop codon, of their frequency with mRNA stability (codon stability coefficient; CSC) for endogenous HEK293T mRNAs. The line corresponds to a standard deviation as calculated from the four measurements. (B) As in A, except for ORFeome mRNAs. (C) Endogenous and ORFeome mRNAs have similar CSCs. Plotted are the average CSC values for endogenous mRNAs compared to ORFeome mRNAs.

FIGURE 4.

Codons associated with instability are translated more slowly. (A) Endogenous HEK293T CSCs weakly correspond with pause scores. Using HeLa ribosome profiling (Arango et al. 2018), elongation speeds were calculated for each codon in the A site, and then codons were divided into three groups (slow in orange; neutral in green; fast in blue). Shown are boxplots and points for the corresponding CSC values as determined by endogenous HEK293T mRNAs. (B) As in A, except for ORFeome-derived CSCs. (C) Comparison of ORFeome CSCs and elongation speeds. Four independent sets of ORFeome CSC values were compared with elongation speeds derived from five different ribosome profiling experiments (Eichhorn et al. 2014; Arango et al. 2018). Plotted are the resulting Spearman correlations for each of the 20 comparisons with those including NIH3T3 elongation speeds shown in gray. Significance was determined by the Wilcoxon test. (D) ORFeome stability scores better correlate with elongation speeds than endogenous scores. Similar to C, four independent sets of CSC values derived from matched endogenous and ORFeome mRNAs were compared with the four values of human elongation speeds (from HeLa rep 1, HeLa rep 2, HEK293T, and U2OS ribosome profiling experiments). Shown are boxplots and points for the Spearman correlations from the 16 resulting comparisons for endogenous (gray) and ORFeome (red) CSC values. Significance was determined by the Wilcoxon test. (E) Published ORFeome CSC values derived in HEK293T and K562 cells (Wu et al. 2019b) were compared with elongation speeds determined from ribosome profiling in HeLa, HEK293T, U2OS, and NIH3T3 cells. Plotted are the corresponding correlations for each of the 10 comparisons; comparisons with NIH3T3 scores are shown in gray. Significance was determined by the Wilcoxon test. (F) Published ORFeome CSCs correspond better to elongation speeds than endogenous CSCs. Spearman correlations between each of the human-line elongation speeds and the published endogenous and ORFeome CSC values (Wu et al. 2019b) were calculated. Correlations with endogenous CSCs are shown in gray; with ORFeome CSCS, in red. Significance was determined by the Wilcoxon test. (G) A-site speeds correspond better to stability scores than P- and E-site scores. P- and E-site scores were calculated for the four ribosome profiling data sets from human cells (HeLa rep 1, HeLa rep 2, HEK293T, and U2OS), and compared with the ORFeome CSC values. Plotted are the Spearman correlations for the 16 comparisons for A-site scores (in red), P- and E- site scores (in gray). Significance was determined by the Wilcoxon test.