The Paf1 Complex Broadly Impacts the Transcriptome of Saccharomyces cerevisiae

Abstract

The Polymerase Associated Factor 1 complex (Paf1C) is a multifunctional regulator of eukaryotic gene expression important for the coordination of transcription with chromatin modification and post-transcriptional processes. In this study, we investigated the extent to which the functions of Paf1C combine to regulate the Saccharomyces cerevisiae transcriptome. While previous studies focused on the roles of Paf1C in controlling mRNA levels, here, we took advantage of a genetic background that enriches for unstable transcripts, and demonstrate that deletion of PAF1 affects all classes of Pol II transcripts including multiple classes of noncoding RNAs (ncRNAs). By conducting a de novo differential expression analysis independent of gene annotations, we found that Paf1 positively and negatively regulates antisense transcription at multiple loci. Comparisons with nascent transcript data revealed that many, but not all, changes in RNA levels detected by our analysis are due to changes in transcription instead of post-transcriptional events. To investigate the mechanisms by which Paf1 regulates protein-coding genes, we focused on genes involved in iron and phosphate homeostasis, which were differentially affected by PAF1 deletion. Our results indicate that Paf1 stimulates phosphate gene expression through a mechanism that is independent of any individual Paf1C-dependent histone modification. In contrast, the inhibition of iron gene expression by Paf1 correlates with a defect in H3 K36 trimethylation. Finally, we showed that one iron regulon gene, FET4, is coordinately controlled by Paf1 and transcription of upstream noncoding DNA. Together, these data identify roles for Paf1C in controlling both coding and noncoding regions of the yeast genome.

Keywords: Paf1 complex; RNA polymerase II; chromatin; histone modifications; noncoding RNA.

Figure 1

Deletion of PAF1 affects all Pol II transcript classes. (A–G) Volcano plots graphing statistical significance (y-axis) against expression change (x-axis) between paf1Δ trf4Δ and trf4Δ strains (KY2016 and KY2012, respectively) for the indicated Pol II transcript classes. In (G), snRNAs and snoRNAs are shown in red and black, respectively. Each point represents an individual transcript. Tiling array probe intensities were averaged over annotated regions using a custom Python script and an average log2 fold change and P value were calculated using the limma R package. The horizontal line indicates an FDR adjusted P value of 0.05 and the vertical lines indicate a 1.5-fold change in expression (log2 fold change of 0.58). Counts and percentages of differentially expressed transcripts shown here are listed in Table S3. (H) Heatmap of log2 fold change in expression between paf1Δ and WT strains (KY1702 and KY2276, respectively) for the 29 most affected snoRNA genes. The snoRNA gene bodies and regions 0–50, 50–100, and 100–150 bp downstream of their annotated 3′ ends are plotted and sorted by the 0–50 bp region.

Figure 2

Paf1 positively and negatively regulates antisense transcription. (A) Horizontally stacked bar graphs showing the percentage of each transcript class (listed in Table S2) found to overlap with a differentially expressed transcript identified in paf1∆ or paf1∆ trf4∆ strains by de novo analysis (counts and percentages listed in Table S4). (B) Vertically stacked bar graph plotting percentage of transcripts, identified in the de novo analysis, that overlap with mRNA coding regions on the sense or antisense strand. These data are also presented in Table S5. (C) Bar graph summarizing the overlap between differentially expressed antisense transcripts detected by the de novo analysis and previously annotated noncoding RNAs (see sums in Table S6 for counts).

Figure 3

Paf1 regulates many of its target loci at the transcriptional level. (A) Heatmaps plotting log2 fold-change in transcript levels detected by tiling array for paf1∆ vs. WT (KY1702 vs. KY2276) and paf1∆ trf4∆ vs. trf4∆ (KY2016 vs. KY2012) as well as a paf1∆ vs. WT NET-seq comparative analysis (Harlen and Churchman 2017). Previously annotated coding and noncoding transcripts were scaled so that each row in the heatmap represents a single transcript from transcription start site (TSS) to transcription end site (TES). (B) Pie charts showing the direction of change in NET-seq data (Harlen and Churchman 2017) for mRNAs that increased or decreased expression by at least 1.5-fold in the paf1∆ vs. WT comparison as measured by tiling array. Direction of change in NET-seq was determined by summing the reads in the first 500 bp of protein-coding genes in both WT and paf1∆ NET-seq datasets and calculating a fold-change (1.5-fold cutoff).

Figure 4

Paf1 positively regulates many phosphate homeostasis genes. (A) Heatmap of expression differences observed in a paf1Δ strain (KY1702) relative to a WT strain (KY2276) at Pho4-responsive genes (Zhou and O’Shea 2011). (B–D) RT-qPCR analysis of phosphate gene expression in strains lacking (B) individual Paf1C subunits (KY1021, KY2271, KY2239, KY2243, KY2241, and KY2244), (C) histone modification enzymes (KY1683, KY2045, KY1952, KY938, KY914, KY934) or (D) H2B K123. In (D), RNA levels in the H2B K123R mutant (KY2167) were compared to the appropriate WT control strain (KY2027). Relative expression ratio is calculated using primer efficiency, normalization to the RNA polymerase III transcript SCR1 and a comparison to a WT strain (Pfaffl 2001). Error bars represent SEM and all statistically significant results are reported as asterisks (0.01 < P < 0.05 = *, 0.001 < P < 0.01 = **, 0 < P < 0.001 = ***). All P values were derived from a Student’s t-test between the mutant strain and WT. (E) Cumulative data from crosses between a paf1∆ strain and strains deleted for chromatin remodeling factors. Following tetrad analysis of the following crosses, growth defects of double mutants were determined: chd1∆ paf1∆ = KY583 × KY804; isw1∆ paf1∆ = KY3464 × KY901; isw2∆ paf1∆ = KY884 × KY804; snf2∆ paf1∆ = KY508 × KY804; arp8∆ paf1∆ = KY3460 × KY804; swr1∆ paf1∆ = KY3462 × KY972. (F) Northern blot analysis of SPL2 and VTC3 RNA levels. Strains used in this analysis were KY292, KY802, KY457, KY508, KY3465, KY3461, KY884, KY3463, KY972, and KY632. SCR1 serves as a loading control. (G) Genome browser view showing antisense transcription at the PHO84 locus. The browser view shows smoothed differential expression tracks (log2(paf1∆ trf4∆/trf4∆), 160 bp sliding window) with both SGD and de novo transcript annotations. Plus (+) and minus (−) symbols refer to DNA strand. The PHO84 gene is oriented right to left.

Figure 5

Paf1 represses iron homeostasis genes. (A) Heatmap of expression differences observed in a paf1Δ strain (KY1702) relative to a WT strain (KY2276) at Aft1 and Aft2 responsive genes involved in maintenance of iron homeostasis (Cyert and Philpott 2013). (B and C) RT-qPCR analysis of the indicated genes in strains lacking (B) individual Paf1C subunits (KY1021, KY2271, KY2239, KY2243, KY2241, and KY2244) or (C) genes in the Set2/Rpd3S pathway (KY307, KY914, KY1702, and KY1235). Calculation of the relative expression ratio and statistical testing were performed as in Figure 4. (D) Heatmaps of expression differences between mutant yeast strains and their respective WT strains in tiling array (this study) and NET-seq (Churchman and Weissman 2011; Harlen and Churchman 2017) datasets. (E) RT-qPCR results for iron homeostasis genes in strains lacking enzymes that catalyze Paf1C-associated histone modifications (KY1683, KY2045, KY1952, KY938, and KY934). (F) Genome browser view of the FIT3 locus showing H3 K36A and set2∆ RNA-seq data from Venkatesh et al. (2016) and H3 K36me3 occupancy data from Weiner et al. (2015).

Figure 6

The FET4 locus is regulated by Paf1 and transcription of ncDNA upstream of the coding region. (A) Diagram of the FET4 locus and the position of a transcription termination sequence (HIS3 TTS) inserted 400 bp upstream of the FET4 start codon to block CUT 794/793 transcription. (B) Northern analysis of FET4 mRNA, CUT 794/793 and SCR1 RNA (loading control) from WT, paf1Δ, trf4Δ, and paf1Δ trf4Δ strains without the inserted TTS (KY2276, KY1702, KY2012, KY2016) or with the TTS (KY3466, KY2846, KY2851, KY2845). (C and D) Quantification of northern blot results for FET4 and CUT 794/793 normalized to SCR1. Error bars represent SEM and all statistically significant results are reported as asterisks that represent P values from Students t-test as in Figure 4. (E) Diagram of the observed effects of PAF1 and CUT794/793 at the FET4 locus.