Widespread use of poly(A) tail length control to accentuate expression of the yeast transcriptome

Abstract

Control of poly(A) tail length can affect translation and stability of eukaryotic mRNAs. Although well established for individual cases, it was not known to what extent this type of adjustable gene control is used to shape expression of eukaryotic transcriptomes. Here we report on microarray-based measurements of mRNA poly(A) tail lengths and association with the poly(A)-binding protein Pab1 in S. cerevisiae, revealing extensive correlation between tail length and other physical and functional mRNA characteristics. Gene ontology analyses and further directed experiments indicate coregulation of tail length on functionally and cytotopically related mRNAs to coordinate cell-cycle progression, ribosome biogenesis, and retrotransposon expression. We show that the 3'-untranslated region drives transcript-specific adenylation control and translational efficiency of multiple mRNAs. Our findings suggest a wide-spread interdependence between 3'-untranslated region-mediated poly(A) tail length control, Pab1 binding, and mRNA translation in budding yeast. They further provide a molecular explanation for deadenylase function in the cell cycle and suggest additional cellular processes that depend on control of mRNA polyadenylation.

FIGURE 1.

Fractionation of mRNA by poly(U)-Sepharose chromatography. (A) Total RNA from BY4741 cells was fractionated on poly(U) Sepharose by stepwise temperature increases (12°C–45°C, as indicated above the panels). Separately, all bound mRNA was eluted from poly(U) Sepharose in a single step (reference [45°C]). Aliquots of each mRNA fraction were end labeled with [³²P]-pCp, digested with RNases A and T1, and poly(A) fragments analyzed by denaturing PAGE. DNA markers are indicated on the left. (B) Schematic of the LM-PAT assay. (C) To test for a PCR bias toward shorter amplicons, total RNA was isolated from either wild-type cells (PAN2/CCR4) or a deadenylase mutant strain (Δpan2/ccr4-1). The cDNA synthesis step of the LM-PAT assay was carried out separately for each RNA sample. Both cDNA preparations were then mixed in proportions outlined above the panel, prior to PCR for MCD1 mRNA. (D) Temperature eluates were analyzed by LM-PAT with primers specific to the mRNAs indicated to the left of the panels. Parallel assays were done with RNA from wild-type (PAN2/CCR4) or mutant (Δpan2/ccr4-1) strains. PCR products were sized against a 100-bp DNA ladder to determine the range of poly(A) lengths for each mRNA (indicated on the right).

FIGURE 2.

Polyadenylation state array (PASTA) analysis. mRNA was fractionated as detailed in Figure 1A. (A) For high-resolution PASTA analyses, each temperature eluate (reverse-transcribed into Cy5-labeled cDNA) was compared against reference mRNA (ref. 45°C; Cy3-labeled cDNA). The lower panels show the mRNA elution profiles of candidate mRNA groups (SOM–short or SOM–long; averaged data from three biological repeat experiments). Profiles are colored by spot ratio for the 45°C elution. (B) For low-resolution PASTA analyses, pools of elution fractions were compared to each other by microarray (30°C, 35°C, and 45°C [Cy5-labeled cDNA] versus 12°C and 25°C [Cy3-labeled cDNA]). The graph displays the resulting data (averaged spot ratios from three biological repeats against mRNA frequency; high over low temperature pool). Outlined in purple are mRNA groupings chosen for further analysis (pool–short or pool–long). (C) The diagrams illustrate the selection of mRNAs on the basis of coherent tail length assignment in A and B. (D) Statistically significant relationships between PASTA assignment and characteristics of mRNAs in S. cerevisiae taken from Table 1 are visualized here in simplified form. Each parameter is represented as a black dot with connecting lines indicating relationships as follows: Solid lines represent correlations exposed by this study; dashed lines are correlations determined by previous studies (Graber et al. 2002; Wang et al. 2002; Arava et al. 2003; Hurowitz and Brown 2003; Beyer et al. 2004; MacKay et al. 2004); green lines represent positive correlations; red lines represent negative correlations; and black lines indicate no simple correlation. (Multiple connections indicate conflicting published information.)

FIGURE 3.

Confirmation of mRNA tail length distribution by LM-PAT assay. Shown are analyses of five PASTA–long candidates in A and five PASTA–short examples in B. For each case, assays were done in parallel with total RNA from wild-type, Δpan2/ccr4-1, Δpan2, and ccr4-1 strains (1, 3, 3, and 1 μg input, respectively). Range of poly(A)-tail lengths is shown to the right of the panels. The dashed box exemplifies quantification of PCR product size distribution by densitometry that was done for all lanes with wild-type RNA. A ratio of signal above to signal below the dashed line was calculated as an arbitrary measure of mRNA tail length distribution. This ratio was ≥1.5 for all examples in A and ≤0.5 for all cases in B.

FIGURE 4.

Poly(A) tail length positively correlates with Pab1 binding. (A) Western blot of cell extracts from the control (mock) or protein A-tagged Pab1 strain (tagged) probed with β-tubulin antibodies (Pab1-protA and tubulin bands indicated on the right). (B) Material eluted from IgG Sepharose by TEV protease cleavage was separated by 10% SDS-PAGE, followed by deep purple staining (positions of cleaved Pab1 and a copurifying protein [asterisk] are marked on the right). (C) RNA was isolated from Pab1-protA or mock purifications (Cy5-labeled cDNA) and compared with total RNA isolated from the corresponding yeast cultures (Cy3-labeled cDNA) by microarray analysis. mRNA enrichment, Pab1-protA over mock (derived from several biological repeats), is plotted against mRNA frequency. Dashed lines indicate positions of the 1000th and 5000th mRNA in enrichment ranking. (D) The graph displays the frequency of members of different PASTA groups occurring within the Pab1-protA enrichment ranking using a 200-gene sliding window.

FIGURE 5.

Gene ontology term enrichment within the PASTA lists. The PASTA—long and—short lists were uploaded into the GOstat tool, available at http://gostat.wehi.edu.au/ (Beissbarth and Speed 2004). GOstat was run with SGD as the gene ontology database, a requirement for a minimal GO path of three, and correction for multiple testing by false discovery rate (Benjamini). Over- or underrepresented GO terms are displayed only if the associated P-values were in the order of ≤10⁻⁴. Diagrams summarize association with terms from (A) molecular function, (B) cellular component, and (C) biological process ontologies. Box shading indicates approximate P-value (see legend in the center of figure).

FIGURE 6.

The regulation of poly(A) tail length within functional groups is linked to deadenylation kinetics. (A) The bar chart displays the distribution of PASTA assignments within indicated mRNA groups. (B, C) Total RNA was isolated from BY4147 cells in stationary phase and after replenishing with fresh media (time-points indicated above the panels). LM-PAT assays were done for six mRNAs as indicated within the panels. Controls with RNA from exponentially growing PAN2/CCR4 and Δpan2/ccr4-1 cells are also shown, and poly(A) tail lengths ranges are indicated to the right of the panels.

FIGURE 7.

poly(A) tail length control is important during the cell cycle. (A) PASTA lists were uploaded onto the FunSpec Web site (http://funspec.med.utoronto.ca/; default parameters; Robinson et al. 2002) and analyzed for enrichment of MIPS nonessential haploid deletion phenotype descriptors. The table lists all associations with a P-value of ∼ 10⁻³ or less; box shading indicates approximate P-value (see legend to the right of the table). No significant enrichment was found with the “unassigned” group. (B) DIC images (equal magnification) of the indicated yeast strains. Morphological changes are consistent with cell size-check point failure in ccr4-1 cells and a failure in the axial-to-isotropic growth switch in Δpan2 cells. Double mutants display a combination of these defects with increased cell size and body axis elongation. (C) LM-PAT assays were done on total RNA samples taken from wild-type cells before, and at indicated time-points after, release from mating factor synchronization. The HTA1, HHT1, ASH1, MYO1, and MFA1 mRNAs all fluctuate in transcript level during the time-course. LM-PAT data for RPL46(39) is shown in the bottom panel as a control with approximately constant mRNA level and adenylation state throughout the time-course. Although the MYO1 mRNA is clearly induced at 50 min after cell-cycle release, no long-tailed form is seen, possibly because deadenylation of MYO1 mRNA proceeds too quickly to be captured in our assays. Results with total RNA from unsynchronized wild-type, Δpan2/ccr4-1, Δpan2 and ccr4-1 strains (as in Fig. 3) and estimated poly(A) tail lengths are further shown to the right of the panels. The dashed boxes frame time-points where mRNA deadenylation following transcript induction is seen. Signals within the dashed boxes were quantified and partitioned into intensity derived from mRNA with an elongated or a minimal poly(A) tail (cut-off indicated by the dashed horizontal lines). This yielded the following long/short intensity ratios: (20, 30, 40 min) HTA1, 2.171, 0.901, 0.489; HHT1, 0.923, 0.312, 0.208; (60, 70, 80 min) ASH1, 1.413, 0.579, 0.573; MFA2, 2.287, 1.390, 0.991.

FIGURE 8.

The 3′-UTR drives adenylation state and protein expression level. GAL1-regulated GFP constructs harboring the indicated 3′-UTRs were maintained in BY4741 cells. Transcription of GFP was induced with galactose at time zero and repressed by addition of glucose at 10 min. (A) GFP-mRNA analyses at various time-points by LM-PAT assay (top) and semi-quantitative RT-PCR reaction using internal GFP primers (middle). The bottom panel shows input total RNA. (B) Western blot detection of GFP and GAPDH proteins by infrared fluorescent imaging. (C) Graph of corresponding GFP/GAPDH protein ratios over time (ratio for GFP-3′RPL10 at 40 min set to one).