Poly(A)-tail profiling reveals an embryonic switch in translational control

Abstract

Poly(A) tails enhance the stability and translation of most eukaryotic messenger RNAs, but difficulties in globally measuring poly(A)-tail lengths have impeded greater understanding of poly(A)-tail function. Here we describe poly(A)-tail length profiling by sequencing (PAL-seq) and apply it to measure tail lengths of millions of individual RNAs isolated from yeasts, cell lines, Arabidopsis thaliana leaves, mouse liver, and zebrafish and frog embryos. Poly(A)-tail lengths were conserved between orthologous mRNAs, with mRNAs encoding ribosomal proteins and other 'housekeeping' proteins tending to have shorter tails. As expected, tail lengths were coupled to translational efficiencies in early zebrafish and frog embryos. However, this strong coupling diminished at gastrulation and was absent in non-embryonic samples, indicating a rapid developmental switch in the nature of translational control. This switch complements an earlier switch to zygotic transcriptional control and explains why the predominant effect of microRNA-mediated deadenylation concurrently shifts from translational repression to mRNA destabilization.

Figure 1. Global measurement of poly(A)-tail lengths

a, Outline of PAL-seq. For each cluster, the fluorescence intensity reflects the tail length of the cDNA that seeded the cluster. Although the probability of incorporating a biotin-conjugated dU opposite each tail nucleotide is uniform, stochastic incorporation results in a variable number of biotins for each molecule within a cluster. b, Median streptavidin fluorescence intensities for two sets of mRNA-like molecules with indicated poly(A)-tail lengths, which were added to 3T3 (circle), HEK293T (triangle), and HeLa (square) samples for tail-length calibration.

Figure 2. Poly(A)-tail lengths in yeast, plant, fly and vertebrate cells

a, Global tail-length distributions. For each sample, histograms tally tail-length measurements for all poly(A) tags mapping to annotated 3′ UTRs (bin size = 5 nt). Leftmost bin includes all measurements <0 nt. Median tail lengths are in parentheses. b, Intergenic tail-length distributions. For each sample, histograms tally average tail lengths for protein-coding genes with ≥50 tags (yeasts, zebrafish and Xenopus) or ≥100 tags (other samples). Median average tail lengths are in parentheses. c, Intragenic tail-length distributions for 10 genes sampling the spectrum of average tail lengths in 3T3 cells. d, Intragenic tail-length distributions. Heatmaps show the frequency distribution of tail lengths for each gene tallied in b. The color intensity indicates the fraction of the total for the gene. Genes are ordered by average tail length (dashed line). Results from the S. cerevisiae total-RNA sample are reported in this figure.

Figure 3. Transient coupling between poly(A)-tail length and TE

a, Relationship between mean tail length and TE for genes with ≥50 poly(A) tags from embryonic samples at the indicated developmental stages. For each stage, tail lengths and TEs were obtained from the same sample. MGC116473 and DDX24 fell outside the plot for X. laevis, stages 3–4, and LOC100049092 fell outside the plot for X. laevis, stages 12–12.5. b, Relationship between mean tail length and TE in the indicated cells, for genes with ≥50 (yeasts) or ≥100 (others) tags. With the exception of HeLa, tail lengths and TEs were from the same samples. Budding yeast YBR196C, YLR355C and YDL080C, fission yeast SPCC63.04.1, mouse-liver NM_007881 and NM_145470, HEK293T NM_001007026, NM_021058, and NM_003537 and HeLa NM_001007026 fell outside their respective plots.

Figure 4. No detectable intragenic coupling between poly(A)-tail length and TE

a, Global analysis of tail lengths across the polysome profile for 3T3 cells. UV absorbance indicates mean number of ribosomes bound per mRNA for each fraction from the sucrose gradient (top, fractions demarcated with vertical dashed lines). Boxplots show distributions of tail lengths in each fraction for all tags mapping to annotated 3′ UTRs (bottom). Boxplot percentiles are line, median; box, 25^th and 75^th percentiles; whiskers, 10^th and 90^th percentiles. The horizontal line indicates the overall median of the median tail lengths. b, Relationship between tail lengths and ribosomes bound per mRNA for mRNAs from the same gene. For each gene, the data from a were used to plot the mean tail length as a function of bound ribosomes. Log-log plots for 8 randomly selected genes with ≥50 poly(A) tags in ≥6 fractions are shown (left), with lines indicating linear least-squared fits to the data (adding a pseudocount of 0.5 ribosomes to the fraction with 0 ribosomes). The boxplot shows the distribution of slopes for all genes with ≥50 poly(A) tags in ≥4 fractions (right; n = 4,079; one-sided, one-sample Wilcoxon test; boxplot percentiles as in a).

Figure 5. The influence of miR-155 on ribosomes, mRNA and tails in the early zebrafish embryo

a, Relationship between changes in ribosome protected fragments (RPFs) and changes in mRNA levels after injecting miR-155. Changes observed between miRNA- and mock-injected embryos are plotted at the indicated stages for predicted miR-155 target genes (red, genes with ≥1 miR-155 site in their 3′ UTR) and control genes (gray, genes that have no miR-155 site yet resemble the predicted targets with respect to 3′ UTR length). To ensure that differences observed between 4 and 6 hpf were not the result of examining different genes, only site-containing genes and no-site control genes detected at both 4 and 6 hpf are shown for these stages. Lines indicate mean changes for the respective gene sets, with statistically significant differences between the sets indicated (*, P ≤0.05; **, P <10⁻⁴, one-tailed Kolmogorov–Smirnov test). Because injected miRNAs partially inhibited miR-430–mediated repression, genes with miR-430 sites were not considered. Data were normalized to the median changes observed for the controls. b, Relationship between RPF changes and mean tail-length changes after injecting miR-155. Tail-lengths were determined using PAL-seq, otherwise as in a. c, A developmental switch in the dominant mode of miRNA–mediated repression. The schematic (left) depicts the components of the bar graphs, showing how the RPF changes comprise both mRNA and TE changes. The compound bar graphs show the fraction of repression attributed to mRNA degradation (blue) and TE (green) for the indicated stage, depicting the overall impact of miR-155 (center; plotting results from a and b for genes with sites) and miR-132 (right, plotting results from Extended Data Fig. 8b for genes with sites). Slight, statistically insignificant, increases in mRNA for predicted targets resulted in blue bars extending above the axis. For samples from stages in which tail length and TE are coupled, a bracket adjacent to the compound bar indicates the fraction of repression attributable to shortened tails. Significant changes for each component are indicated with asterisks of the corresponding color (*, P ≤0.05; **, P <10⁻⁴, one-tailed Kolmogorov–Smirnov test).