Dynamic transcriptome analysis measures rates of mRNA synthesis and decay in yeast

Abstract

To obtain rates of mRNA synthesis and decay in yeast, we established dynamic transcriptome analysis (DTA). DTA combines non-perturbing metabolic RNA labeling with dynamic kinetic modeling. DTA reveals that most mRNA synthesis rates are around several transcripts per cell and cell cycle, and most mRNA half-lives range around a median of 11 min. DTA can monitor the cellular response to osmotic stress with higher sensitivity and temporal resolution than standard transcriptomics. In contrast to monotonically increasing total mRNA levels, DTA reveals three phases of the stress response. During the initial shock phase, mRNA synthesis and decay rates decrease globally, resulting in mRNA storage. During the subsequent induction phase, both rates increase for a subset of genes, resulting in production and rapid removal of stress-responsive mRNAs. During the recovery phase, decay rates are largely restored, whereas synthesis rates remain altered, apparently enabling growth at high salt concentration. Stress-induced changes in mRNA synthesis rates are predicted from gene occupancy with RNA polymerase II. DTA-derived mRNA synthesis rates identified 16 stress-specific pairs/triples of cooperative transcription factors, of which seven were known. Thus, DTA realistically monitors the dynamics in mRNA metabolism that underlie gene regulatory systems.

Figure 1

Dynamic transcriptome analysis in yeast. (A) Scheme of metabolic mRNA labeling in yeast. Nascent mRNA is labeled with 4-thiouridine (4sU) and thiol-specifically biotinylated following cell lysis and preparation of total cellular RNA. Fractionation into pre-existing and nascent mRNA is achieved with streptavidin beads. (B) The yeast transcriptome is undisturbed by expression of the human nucleoside transporter hENT1. The volcano plot compares mRNA levels after 6 min labeling versus wild-type cells without labeling. Each dot corresponds to one gene, the x axis displays the log₂(fold) of that gene, the y axis represents the multiple testing adjusted P-value (see Supplementary Section 12.5). In all, 17 genes showed a significant change in mRNA levels (adjusted P-value <5%), only three of which were at least twofold. The results for the other tested labeling periods are given in the Supplementary Section 12.6. (C) 4sUTP is modeled into the crystal structure of a Pol II transcribing complex (PDB code: 116H). The thiol group at position 4 can form a hydrogen bond with the DNA template strand (blue). Nascent RNA is in red.

Figure 2

Determination of mRNA synthesis and decay rates. (A) Labeling bias. A plot of the number of uridines (x axis) versus the log ratio of labeled and total mRNA levels revealed that the labeling efficiency varies with the uridine content of the gene. As a consequence, synthesis rates for short transcripts are systematically underestimated. Modeling of the bias function (Material and methods, determination of mRNA synthesis and decay rates) showed that ∼1 out of 200 uridines is replaced by 4sU. (B) Center, scatter plot of the mRNA half-lives and synthesis rates for exponentially growing yeast cells. Colored points belong to the indicated gene sets (green, ribosomal biogenesis genes; violet, ribosomal protein genes; red, stress genes; dark gray, transcription factors (TFs). Assuming Gaussian distributions, ellipses show the 75% regions of highest density for the respective sets. Histograms along the x axis resp. y axis show the global half-life resp. synthesis rate distribution (light gray) and the half-life resp. synthesis rate distribution of the TFs (dark gray). Overall half-lives and synthesis rates are uncorrelated (Spearman correlation 0.06), however some gene groups behave differently (correlations: Ribosomal protein genes (Rp) −0.79, Ribosomal biogenesis genes (RiBi) 0.35, ISA stress module genes 0, TFs 0.07). (C) Gene Ontology (GO) analysis of the short-lived mRNAs (lower 25% of the half-life distribution). The 10 most significant categories are displayed, sorted from bottom (most significant) to top (Materials and methods). Red line, proportion of short-lived transcripts in the whole population (25% by construction). The number of short-lived transcripts in the resp. GO category is given relative to the GO category size (green bar) and relative to the number of short-lived transcripts (black line). Dashed line, relative size of the GO set in the whole population. (D) GO analysis of the long-lived mRNAs (upper 25% of the half-life distribution), analogous to (C).

Figure 3

Global perturbation of mRNA synthesis and decay upon osmotic stress. (A) Design of the time series experiment. Each time window () corresponds to one sample, left end marks the start of the 4sU labeling, right end marks the time of mRNA extraction. Upper panel shows the drop in labeling efficiency from roughly one 4sU in 200 uridines to one 4sU in 400 uridines during the osmotic stress response. (B) Increased sensitivity and temporal resolution of DTA compared with standard transcriptomics. Gray, time course of the total mRNA fraction of the Hog1-induced genes (Capaldi et al, 2008). Red, time course of the synthesis rates of the same gene set. The solid lines represent the time course of the median, the shaded bands are the central 95% regions. In contrast to the monotonically increasing total mRNA time course, the synthesis rates clearly show three response phases. (C) Expression changes of the five clusters (up, up-even, even, down-even, down—see Materials and methods) that were defined in a normalization-independent manner. The box plot shows synthesis rate folds (30 versus 0 min). (D) Dynamics of synthesis and decay rates in the osmotic stress time series. Each diagram corresponds to one time point. Each point corresponds to one gene, which is colored according to its affiliation with one of the clusters in (C). Ellipses show the 75% regions of highest density within each cluster, assuming Gaussian distributions. The shape of the ellipses indicates the correlation structure within a cluster.

Figure 4

Analysis of transcription factor (TF) activity. (A) For each TF, the proportion of induced/repressed genes in the set of the TF targets is compared with that in the TF target complement set. TFs with less than 20 targets were filtered out beforehand. Log odds ratios for the induced (up cluster, red) and repressed (down cluster, blue) genes are plotted on the left hand side. The right side shows the negative decadic log P-value of the corresponding Fisher tests (red for the induced genes, blue for the repressed genes). A P-value of 10⁻⁵ was considered significant, and only these findings are shown. (B) Time series of TF activity. Each line represents the time course of the 90% quantiles (solid lines) resp. 10% quantiles (dotted lines) of the synthesis rates of the target genes of a TF. The 90% quantiles of the activatory TFs from (A) are shown in red, the 10% quantiles of the repressive TF is shown in blue.

Figure 5

Transcription factor cooperativity. Excerpt from the synthetic interaction matrix (SI). Colors from blue (negative) to yellow (positive) map the strength of an interaction, as quantified by our interaction score. Each row shows all interaction scores of one TF with all other TFs. Pairs of triples of TFs having highly correlated interaction profiles are grouped together. Groups that could be validated in the literature are shaded in green.

Figure 6

Pol II gene occupancy predicts mRNA synthesis. The vectors of mean Pol II occupancies on transcribed regions were calculated from ChIP-chip data at 0, 12, and 24 min after salt stress and compared with the vectors of total mRNA levels (left), labeled mRNA (middle), and synthesis rates (right) at each time point of the osmotic stress time course experiment. The pair-wise Spearman correlation values are represented by color-coded squares.

Figure 7

Global Pol II redistribution upon osmotic stress. Dynamics of the mean Pol II occupancy profiles of the up-, even- and down clusters from Figure 3C. Only genes of at least 1000 base pairs (bps) are considered. Profiles are taken after 0, 12, and 24 min of osmotic stress (light blue, blue, and dark blue lines). Vertical dotted lines are drawn at the transcription start site (TSS) and the polyadenylation site (pA) site. The plot shows the ±500 bp regions around the TSS and the pA sites, together with the intermediate part, which has been rescaled to a length of 400 (Supplementary section 16). The Pol II profiles of all genes and all clusters are displayed in (Supplementary Figure S26).