Infrequently transcribed long genes depend on the Set2/Rpd3S pathway for accurate transcription

Abstract

The presence of Set2-mediated methylation of H3K36 (K36me) correlates with transcription frequency throughout the yeast genome. K36me targets the Rpd3S complex to deacetylate transcribed regions and suppress cryptic transcription initiation at certain genes. Here, using a genome-wide approach, we report that the Set2-Rpd3S pathway is generally required for controlling acetylation at coding regions. When using acetylation as a functional readout for this pathway, we discovered that longer genes and, surprisingly, genes transcribed at lower frequency exhibit a stronger dependency. Moreover, a systematic screen using high-resolution tiling microarrays allowed us to identify a group of genes that rely on Set2-Rpd3S to suppress spurious transcripts. Interestingly, most of these genes are within the group that depend on the same pathway to maintain a hypoacetylated state at coding regions. These data highlight the importance of using the functional readout of histone codes to define the roles of specific pathways.

Figure 1.

The Set2–Rpd3 pathway controls global histone acetylation at ORFs. (A) Longer genes depend on the Set2–Rpd3S pathway to maintain the acetylation status at ORFs. ChIP–chip was performed using high-resolution tiling arrays manufactured by Agilent Technologies. The log2 ratio of acetylation of H4 (AcH4) in Δset2 over AcH4 in wild type were subjected to a modified average gene analysis originated by the Young laboratory (see Materials and Methods for details). All genes were divided into eight subclasses based on the length of their coding regions. The averages of each subclass were plotted. The number of genes in each class is indicated within the parentheses behind the gene categories in the white box. (B) K36 trimethylation is preferentially enriched at longer ORFs. The similar class-average analysis was used to show the distribution pattern of K36me3 on average genes based on Pokholok et al. (2005). (C) Identification of genes that require Set2 to suppress hyperacetylation at ORFs. The matrix generated from the average gene analysis that covers the ORF region was subjected to the K-mean cluster analysis using TM4 microarray suite software (Saeed et al. 2006). The bottom two clusters (∼25% of genome) displayed significant AcH4 increase upon deletion of SET2. (D) The ORF length of the genes that are within the bottom two clusters from the above analysis (in C) is directly compared with that of the entire genome in a histogram analysis (Microsoft Excel).

Figure 2.

Infrequently transcribed genes are sensitive to disruption of the Set2–Rpd3S pathway. Average gene analysis was performed essentially as described in Figure 1, except that genes were divided into five subgroups based on their transcription rates (Holstege et al. 1998). The averages of each subclass are plotted.

Figure 3.

Patterns of global acetylation increase at coding regions caused by specific mutations in the Rpd3S complex are similar to those seen in SET2 deletion mutants. ChIP–chip experiments were carried out using yeast strains bearing either eaf3Δchd (YBL619) or rco1Δphd (YBL632) mutations. Microarray analyses were performed exactly as described in Figures 1 and 2.

Figure 4.

Genome-wide identification of genes that rely on the Set2–Rpd3S pathway to suppress cryptic internal initiation. Poly(A)⁺ mRNA prepared from wild type and Δset2 was chemically labeled with fluorescent dyes and then subjected to competitive hybridization on tiling microarrays containing oligos representing either the antisense strand or the sense strand of coding regions. (A) Validation of using high-density tiling arrays to identify the cryptic transcripts caused by deletion of SET2. The STE11 locus, which showed the intragenic phenotype following deletion of SET2 in a previous study (Carrozza et al. 2005), displayed the expected changes. The level of mRNA transcript was unchanged at the 5′ end of the ORF, but drastically increased at the 3′ ORF. (B) An example of a candidate gene from the microarray analysis. The log2 ratios of relative mRNA levels in Δset2 and wild-type strains are displayed in the genome browser format. The distribution pattern of H3K36 trimethylation based on Pokholok et al. (2005) is shown in the bottom panel. (C) The novel candidates are confirmed through Northern blot with the probes corresponding to different regions of the given ORF. (D) Identification of novel genes that rely on the Set2–Rpd3S pathway to suppress the internal transcripts initiated only from the antisense strand. (E) An example of a candidate gene from the microarray analysis that contains cryptic transcripts from both sense and antisense strands. (F) Northern blot confirmation.

Figure 5.

Functional readout of the histone K36 methylation code. (A) Acetylation changes were used as the readout for K36 methylation. The K-mean cluster analysis revealed that in the Δset2 mutant, 621 genes possess internally initiated mRNA transcripts that are transcribed in the same direction as full-length transcripts, and 494 genes have cryptic transcripts that use the antisense strand as a template. The Venn diagram analysis (http://www.pangloss.com/seidel/Protocols/venn.cgi) of these genes and genes that are dependent on Set2–Rpd3S to maintain acetylation status (1685 genes) demonstrates an obvious overlap. (B) The enrichment level of K36 methylation was used as the readout for K36 methylation. The data set of the relative enrichment of K36me3 at coding regions (Pokholok et al. 2005) was subjected to K-mean Cluster analysis as described in Figure 1C. The two clusters that displayed the highest K36 trimethylation levels were chosen for subsequent Venn diagram analysis. The level of K36me3 at ORFs does not reflect well the functional dependence of genes on the Set2–Rpd3S pathway.