pubmed.ncbi.nlm.nih.gov

Metabolic labeling of RNA uncovers principles of RNA production and degradation dynamics in mammalian cells - PubMed

doi: 10.1038/nbt.1861. Epub 2011 Apr 24.

Affiliations

Metabolic labeling of RNA uncovers principles of RNA production and degradation dynamics in mammalian cells

Michal Rabani et al. Nat Biotechnol. 2011 May.

Abstract

Cellular RNA levels are determined by the interplay of RNA production, processing and degradation. However, because most studies of RNA regulation do not distinguish the separate contributions of these processes, little is known about how they are temporally integrated. Here we combine metabolic labeling of RNA at high temporal resolution with advanced RNA quantification and computational modeling to estimate RNA transcription and degradation rates during the response of mouse dendritic cells to lipopolysaccharide. We find that changes in transcription rates determine the majority of temporal changes in RNA levels, but that changes in degradation rates are important for shaping sharp 'peaked' responses. We used sequencing of the newly transcribed RNA population to estimate temporally constant RNA processing and degradation rates genome wide. Degradation rates vary significantly between genes and contribute to the observed differences in the dynamic response. Certain transcripts, including those encoding cytokines and transcription factors, mature faster. Our study provides a quantitative approach to study the integrative process of RNA regulation.

PubMed Disclaimer

Figures

Figure 1
Figure 1. Changes in transcription rates during the response of DCs to LPS

(a) Measuring transcription rates with short metabolic labeling. We used short metabolic labeling (10 min, red lines), and measured the expression of RNA-Total (blue) and RNA-4sU (red) for 254 ‘signature’ genes at 13 time points in 15 min intervals (rows) over the first 3 hours post-LPS stimulation. (b) Changes in RNA-4sU levels follow changes in pol-II binding and precede changes in total RNA levels. Shown are example time course profiles for selected genes for RNA-4sU expression (nCounter, red), RNA-Total expression (nCounter, blue) and pol-II binding at the promoter (ChIP, dashed gray). (c) Distinct temporal clusters of newly transcribed and total RNA. Shown are clusters of expression profiles (nCounter) for 254 ‘signature’ genes (rows) based on RNA-Total (left) and RNA-4sU (right) measurements across 13 time points (columns). Cluster I includes the control genes. Cluster numbers (I-VIII) are noted on right; representative member genes are noted on left. Purple: high relative expression; white: mean expression; pink: low relative expression. (d) Peak transcription precedes peak expression by 15-30 minutes. Shown are average profiles (Y axis) for RNA-4sU (red) and RNA-Total (blue) for each cluster at each time point (X axis), ordered by cluster numbers (cluster I topmost; cluster VIII bottommost). The size of each cluster is indicated in brackets. Pearson correlation coefficient (ρ) of the best time-lag correlation between transcription and expression is indicated on right, with the optimal time lag in square brackets.

Figure 2
Figure 2. Changes in transcription rate account for most expression changes; changes in degradation rate contribute to ‘peaked’ responses

(a) The ‘constant degradation’ and ‘varying degradation’ models. A first-degree dynamical model (formula, right) models the expression level of a gene (grey curve) as a function of transcription (black) and degradation (green) rates. Parameters include an ‘impulse’ model, for transcription (black curve), and either a constant function for degradation (‘constant degradation’ model, solid green line), or an ‘impulse’ model (‘varying degradation’ model, dashed green line). We fit them to our data (left, RNA-Total, blue, and RNA-4sU, red) by optimizing the likelihood function (separately per gene). We compare the model’s fit (black and grey curves) to the data (red and blue curves, respectively) and calculate the error. (b) The ‘constant degradation’ model fits the majority of genes well. Shown is the distribution of the log likelihood ratios between the ‘constant degradation’ and ‘varying degradation’ models. Dashed line indicates the threshold for rejecting constant degradation (p<0.01). (c) The percent of genes per cluster (numbered as in Fig. 1c) that reject the constant degradation model. (d) Varying degradation profiles estimated for the 44 genes that reject the ‘constant degradation’ model. Right: estimated degradation rates (relative rate; purple: high; pink: low) for the 44 genes (rows), clustered into 3 groups (A-C), across 12 time points (columns; excluding t=0 which is highly sensitive to noise due to low RNA levels). Asterisk: known regulators of RNA degradation (see Discussion). Left: mean degradation rate profile per cluster (bracket: number of genes in cluster). (e) Genes with peaked responses reject the ‘constant degradation’ model. Shown are two example genes (top: Cxcl1, bottom: Zfp36). For each, upper row: ‘constant degradation’ model fit (solid line) to the data (dashed line); lower row: ‘varying degradation’ model fit (solid line) to the data (dashed line). Left: expression level; middle: transcription rate; right: degradation rate (estimate only).

Figure 3
Figure 3. Genome-wide analysis of RNA transcription and degradation rates using RNA- and 4sU-Seq

(a) Experiment overview. We isolated RNA-4sU (after 45 min of 4sU labeling, red) and polyA+ RNA-Total (blue) at 1h intervals (rows) over the first 6 hours of the response of DCs to LPS stimulation, and used massively parallel sequencing to measure RNA levels. (b) 4sU-Seq captures a broader representation of transcripts compared to polyA+ RNA-Seq. Shown is the fraction of reads in RNA-4sU-Seq libraries (left) and polyA+ RNA-Seq libraries (right), across several annotation categories. Only reads that are mapped to a unique location in the genome or to rRNA are considered. (c) Distribution of predicted constant mRNA half-lives for the 9,448 genes expressed during the first 6 hours of the response to LPS stimulation that do not reject the ‘constant degradation’ model. Dashed lines distinguish 10 deciles (A-J, 10% increments, 35 transcripts with >200min half-life are included in the last decile).

Figure 4
Figure 4. Genome-wide analysis of RNA processing rates

(a) Using 4sU-Seq data to study RNA processing. Sequencing reads in the 4sU-Seq libraries originate from either pre-mRNA (U; purple) or mature mRNA (M; light blue). While mRNA reads map only to exons, the pre-mRNA reads map to both exons and introns. We estimate newly transcribed pre-mRNA expression by the RPKM of a gene’s introns alone, and overall newly transcribed RNA expression (pre-mRNA + mRNA) by the RPKM of a gene’s exons. (b) An overview of the ‘constant degradation and processing’ model. The model expands on our ‘constant degradation’ model (Fig. 2a) by adding a constant processing rate (right; orange). We fit the model parameters to our data (left; mRNA-4sU, dashed red, and pre-mRNA-4sU, dashed purple) by optimizing the likelihood function (separately per gene) and using the degradation rates predicted by the ‘constant degradation’ model. (c) Distribution of predicted constant processing rates for 3,011 genes with exonic and intronic expression during the first 6 hours of the response to LPS stimulation. Dashed lines distinguish 5 quintiles (a-e, 20% increments), and transcripts with >30min half-life are added to the last bin. Pre-mRNA half-lives for illustrative genes are denoted in each bin. (d) Transcripts with low or high pre-mRNA half-lives are enriched in functional categories, clusters, exon structures or transcript lengths. Shown are the enrichments (P-value, hypergeometric test, grey), of the overlap between the genes in each of the half-life bins in (c) (A-E, columns) and each tested category (rows). Only categories with at least one significant enrichment are shown.

Comment in

  • Transcription: getting close to the action.

    Muers M. Muers M. Nat Rev Genet. 2011 Jun;12(6):382. doi: 10.1038/nrg3008. Epub 2011 May 17. Nat Rev Genet. 2011. PMID: 21577224 No abstract available.

  • The dynamic RNA world.

    de Souza N. de Souza N. Nat Methods. 2011 Jul;8(7):536. doi: 10.1038/nmeth0711-536. Nat Methods. 2011. PMID: 21850735 No abstract available.

Similar articles

Cited by

References

    1. Kim HD, Shay T, O’Shea EK, Regev A. Transcriptional regulatory circuits: predicting numbers from alphabets. Science. 2009;325:429–432. - PMC - PubMed
    1. Wang RS, Zhang XS, Chen L. Inferring transcriptional interactions and regulator activities from experimental data. Mol Cells. 2007;24:307–315. - PubMed
    1. Raghavan A, et al. Genome-wide analysis of mRNA decay in resting and activated primary human T lymphocytes. Nucleic Acids Res. 2002;30:5529–5538. - PMC - PubMed
    1. Shalem O, et al. Transient transcriptional responses to stress are generated by opposing effects of mRNA production and degradation. Mol Syst Biol. 2008;4:223. - PMC - PubMed
    1. Amit I, et al. A module of negative feedback regulators defines growth factor signaling. Nat Genet. 2007;39:503–512. - PubMed

Publication types

MeSH terms

Substances