The Enrichment of miRNA-Targeted mRNAs in Translationally Less Active over More Active Polysomes

Abstract

miRNAs moderately inhibit the translation and enhance the degradation of their target mRNAs via cognate binding sites located predominantly in the 3'-untranslated regions (UTR). Paradoxically, miRNA targets are also polysome-associated. We studied the polysome association by the comparative translationally less-active light- and more-active heavy-polysome profiling of a wild type (WT) human cell line and its isogenic mutant (MT) with a disrupted DICER1 gene and, thus, mature miRNA production. As expected, the open reading frame (ORF) length is a major determinant of light- to heavy-polysome mRNA abundance ratios, but is rendered less powerful in WT than in MT cells by miRNA-regulatory activities. We also observed that miRNAs tend to target mRNAs with longer ORFs, and that adjusting the mRNA abundance ratio with the ORF length improves its correlation with the 3'-UTR miRNA-binding-site count. In WT cells, miRNA-targeted mRNAs exhibit higher abundance in light relative to heavy polysomes, i.e., light-polysome enrichment. In MT cells, the DICER1 disruption not only significantly abrogated the light-polysome enrichment, but also narrowed the mRNA abundance ratio value range. Additionally, the abrogation of the enrichment due to the DICER1 gene disruption, i.e., the decreases of the ORF-length-adjusted mRNA abundance ratio from WT to MT cells, exhibits a nearly perfect linear correlation with the 3'-UTR binding-site count. Transcription factors and protein kinases are the top two most enriched mRNA groups. Taken together, the results provide evidence for the light-polysome enrichment of miRNA-targeted mRNAs to reconcile polysome association and moderate translation inhibition, and that ORF length is an important, though currently under-appreciated, transcriptome regulation parameter.

Keywords: 3′-untranslated regions (UTR); DICER1; microRNA (miRNA); more-active heavy polysome; polysome; polysome profiling; translationally less-active light polysome.

Figure 1

MiRNA-targeted mRNAs’ shift toward translationally more-active heavy polysomes in isogenic DICER1 mutant cells, i.e., suggesting enrichment in less-active light polysomes in WT cells, and its inconsistence with the raw light to heavy polysome mRNA abundance log-ratio in WT cells. (A): Scatter plots of mRNA abundance log-ratios versus mRNA 3′-UTR miRNA binding site counts (WT: wildtype cells; MT: mutant cells; H: heavy polysome; L: light polysome). Mean log-ratio of genes with corresponding binding site counts were plotted. Binding site count is in log scale, with one added to each count to avoid the log₂(0) error. LOESS regression trend curves are also shown. (B): Comparative boxplots of light to heavy polysome abundance log-ratios of mRNAs with more than 50 binding sites and those with no identified site in WT cells (“WT > 50 Sites” and “WT No Site”) as well as MT cells (“MT > 50 Sites” and “MT No Site”). The t-test and F-test p-values for WT cells are shown.

Figure 2

Open reading frame (ORF) length as a determinant of light to heavy polysome mRNA abundance log-ratio. A scatter plot of the log-ratio versus ORF length is shown for WT (A) and mutant cells (B). LOESS regressions (log-ratio versus ORF length) are used to illustrate the relationship, with predicted values shown as blue datapoints. The R² value of the regression is also shown. In (A), the genes with more than 50 miRNA binding sites in their mRNA 3′-UTR are highlighted as red data points.

Figure 3

Comparative boxplots of light to heavy polysome mRNA abundance log-ratio. (A): A boxplot to compare raw log-ratio (WT and MT) and ORF-length-adjusted log-ratio (Adj.-WT and Adj.-MT), i.e., the residuals of LOESS regressions in Figure 2, both illustrating reduced dispersion in mutant cells. That is, DICER1 disruption reduces raw log-ratio’s value range and makes ORF length a more powerful determinator of the log-ratio in mutant cells. (B): Boxplots of the adjusted log-ratio to compare mRNAs with more than 50 3′-UTR miRNA binding sites, mRNAs with no sites and the means of random mRNA samples in WT cells. A more significant difference between miRNA-targeted and untargeted mRNAs is observed than in Figure 1B; the t- and F-test p-values are shown.

Figure 4

Relationship between light to heavy polysome mRNA abundance log-ratio and miRNA binding site count viewed in the context ORF length. A LOESS regression (log-ratio versus ORF length plus log₂(binding site count)) was performed for WT cells. A scatter plot of the predicted log-ratio versus ORF length is shown, with the data points color-coded by the log₂(binding site count) values. The vertical bar denotes mean ORF length.

Figure 5

MiRNA-targeted mRNAs tend to have longer ORFs. The scatter plots display mean ORF length versus corresponding 3′-UTR miRNA binding site counts (A) or the site counts normalized by ORF length in KB (B), with both axes in log scale. The smooth curve denotes the trend computed by a LOESS regression. The horizontal bar denotes the mean log₂(ORF length).

Figure 6

ORF length adjustment improves correlation between light to heavy polysome mRNA abundance log-ratio and miRNA binding site counts. For x-axis, all site counts were increased by one to avoid the log₂(0) error. Panel (A) compares raw log-ratio and ORF-length-adjusted log-ratio in WT cells. Panel (B) compares the ORF-length-adjusted log-ratio and the adjusted log-ratio subtracted by the adjusted log-ratio of mutant cells. Upon the subtraction, a nearly perfect linear correlation is observed.