Poised RNA polymerase II changes over developmental time and prepares genes for future expression

Abstract

Poised RNA polymerase II (Pol II) is predominantly found at developmental control genes and is thought to allow their rapid and synchronous induction in response to extracellular signals. How the recruitment of poised RNA Pol II is regulated during development is not known. By isolating muscle tissue from Drosophila embryos at five stages of differentiation, we show that the recruitment of poised Pol II occurs at many genes de novo and this makes them permissive for future gene expression. A comparison with other tissues shows that these changes are stage specific and not tissue specific. In contrast, Polycomb group repression is tissue specific, and in combination with Pol II (the balanced state) marks genes with highly dynamic expression. This suggests that poised Pol II is temporally regulated and is held in check in a tissue-specific fashion. We compare our data with findings in mammalian embryonic stem cells and discuss a framework for predicting developmental programs on the basis of the chromatin state.

Figure 1. Experimental strategy for tissue-specific time course analysis of chromatin and transcription

(A) Overview of the experimental procedure. Embryos expressing GFP in a tissue of interest are dissociated into single cells. GFP-positive cells are isolated by FACS and analyzed by ChIP-seq and mRNA-seq. Since mef2-Gal4 driven GFP expression in the embryonic musculature is only apparent from 6 h AEL on, earlier events were studied using Toll10b mutant embryos, which consist of mesodermal precursor cells (visualized by twist in situ hybridization). See also Figure S1. (B) Genome browser snapshot of the dynamic changes in Pol II occupancy around the twi and Act57B genes during muscle development. Note that the changes correlate with changes in gene expression observed by in situ hybridization and mRNA-seq (C). (C) Heat map of the mRNA-seq data (left). The time course data from replicate experiments for 12,786 individual genes were clustered by Euclidian distance. The color scale reflects their expression levels shown in RPKM. Based on spike-in mRNA we estimate an RPKM value of 1 to correspond to 0.5 – 1 transcript/cell (see Extended Experimental Procedures). The timing of expression of well-characterized muscle genes (right) is consistent with the function of these genes. See also Figure S1.

Figure 2. Recruitment of poised Pol II is dynamically regulated during development

(A) Definitions of Pol II states. Stalled Pol II is defined by a high ratio of Pol II occupancy at the transcription start site (TSS) over the Pol II occupancy in the transcription unit (TU) of a gene. Poised Pol II is defined by high Pol II_TSS enrichment (top 20^th percentile) and low expression as measured by mRNA-seq (RPKM < 10). (B) Recruitment of poised Pol II is dynamically regulated during the time course. The Pol II occupancy at the TSS is shown across time for all genes (n= 1434) that have poised Pol II in at least one of the time points of the time course. 60 % of these genes (n=867) remain occupied by Pol II at all times, while 40 % switch between being Pol II occupied or not. Most of the genes that switch states lack Pol II occupancy at the first time point and gradually gain Pol II occupancy during the time course (opening set, n=502), while only few genes (closing set, n=65) are initially occupied by Pol II and subsequently lose it. (C) DNase hypersensitivity (DHS) data from whole embryos at three time windows (data from (Thomas et al., 2011) show increased DNase accessibility over time, consistent with promoter opening and Pol II recruitment. (D) Fraction of poised genes induced at another time point (precision rate) for each reference time point (gray box; n is the total number of poised genes). Note that poised genes from the constant set are expressed both in the past and future, while those from the opening set tend to be induced in the future only. (E) The same calculation as in (D) was done for genes lacking Pol II at the reference time point. (F) The ratio between the two percentages in D and E is the relative predictive value, which indicates how much more likely poised genes are activated than control genes without Pol II. Asterisks indicate p < 0.05. See also Figure S2.

Figure 3. Poised Pol II predicts stage-specific but not tissue-specific gene expression

(A) Comparison of 10–12 h Pol II ChIP-seq and mRNA-seq between muscle and neurons for genes poised in 10–12h muscle. All genes poised in the 10–12 h muscle sample also have Pol II bound in 10–12 h neurons. Many (49%) of the genes that are poised in 10–12 h muscle are not only bound by Pol II in 10–12 h neurons, but are also expressed. (B) Analysis of in situ expression of the genes of both the constant set and the opening set. As expected, genes in the constant set are enriched for all developmental stages, whereas the opening set genes are expressed late (stages 13–16) during development. Neither gene set is muscle-specific but instead also enriched for expression in the central nervous system, as well as epithelial tissues. (C) Relative predictive values for poised Pol II in the constant set and opening set for tissue-specific samples and whole embryos. Genes with poised Pol II in muscle reference time points are expressed not only in muscle but also in neurons or whole embryo, arguing that the recruitment of Pol II is not tissue-specific. Note that the values for the constant set are strongly stage-specific during the whole embryo time course and that the opening set is not expressed in entire early embryo. For the calculation, see Figure 2. Asterisks indicate significance (p < 0.05) and the dashed box emphasizes the reference time point. See also Figure S3.

Figure 4. Different core promoters associated with distinct Pol II occupancy behavior

(A) Enrichment of core promoter elements in different gene groups. The asterisk indicates that the enrichment (yellow) or depletion (black) is significant (p < 0.05). First, housekeeping genes, as defined by broad expression throughout the embryo based on in situ hybridizations (Tomancak et al., 2007), are enriched for Ohler1, Ohler6, Ohler7 and DRE, which are found at dispersed promoters. Note that maternally expressed genes count here as ‘housekeeping genes’ although they may not be expressed in the embryo. Second, genes that have the disposition for poised Pol II (opening set, constant set) are enriched for GAGA, Inr, DPE, PB and MTE, which are found at focused promoters. The constant set is also enriched for some dispersed promoter elements, perhaps because of its higher average expression (not shown). The poised regulated and non-poised regulated sets are genes induced at the last time point with or without prior poised Pol II. Third, the non-poised regulated genes are enriched for Inr and TATA, thus have a different core promoter configuration that supports focused transcription. The TATA-enriched genes are depleted for housekeeping and developmental functions (not shown). (B) The average nucleosome profile (top) and predicted nucleosome occupancy (middle) are different between the three promoter classes. For each class, the nucleosome profile as measured by MNase-seq was analyzed in the presence and absence of Pol II (shown at bottom). Only genes with poised Pol II tend to have a prominent promoter nucleosome in the absence of Pol II but not when Pol II is present (asterisk, p-value < 10⁻⁴⁸ with Wilcoxon rank sum test at +16 bp). Furthermore, housekeeping genes tend to have higher occupancy at the first nucleosome (asterisk, Wilcoxon rank sum test at +151 bp) than poised genes (p-value < 10⁻⁴⁰) or TATA-enriched genes (p-value < 10⁻¹⁵). The average predicted nucleosome occupancy for each gene group was calculated based on Kaplan et al. (2009). See also Table S2.

Figure 5. PcG repression predicts tissue-specific repression

(A) Genes with differential H3K27me3 levels between muscle and neurons are differentially expressed between these tissues. mRNA-seq data from muscle cells or neuronal cells at 14–17h are shown as box plots (log2 RPKM) with whiskers as interquartile ranges. Genes that either have higher H3K27me3 in muscle (dark blue) or neurons (light blue) are differentially expressed (Scheirer-Ray-Hare test, p-value < 0.018). (B) Based on in situ hybridizations, the differential H3K27me3 sets significantly overlap (y axis shows percent overlap) with genes that are expressed in either only muscle or only neurons at any time point during embryogenesis (Fisher exact test p < 0.02). (C) Example of genes with differential H3K27me3 levels between muscle and neurons: twist (twi) has higher H3K27me3 enrichment in neurons (left) and shaven (sv) has higher H3K27me3 enrichment in muscle (right). Enrichment is shown as H3K27me3 reads over input reads, smoothened over 100 bp windows. (D) In muscle cells, genes with high levels of H3K27me3 (top 2.5% of all genes) over the transcription unit are less likely to be induced in the future as compared to genes without H3K27me3 (left). Such negative predictive values (blue) are not found when we predict gene expression in neuronal tissue (middle) or the whole embryo (right), indicating that the repression is tissue-specific. Asterisks indicate p < 0.05; dashed outlined boxes highlight the reference time points.

Figure 6. Behavior of the balanced state during Drosophila embryogenesis

(A) Sequential ChIPs against H3K27me3 and then Pol II confirm the co-occurrence of Pol II and H3K27me3 at genes (top panel). Co-occurrence results in higher reChIP enrichment compared to the first ChIP (measured over input and normalized to an intergenic control region). Control regions enriched for Pol II but without H3K27me3 (Act5C-TSS, RpL19-TSS) show no significant enrichment in either ChIP or reChIP. Regions with H3K27me3 enrichment that lack Pol II (hbn-up, gcm2-up) show a decrease after the Pol II reChIP. At balanced genes (opa-TSS, ind-TSS), the reChIP enrichment is increased relative to the K27me3 ChIP enrichment. In contrast, an increase is not observed using either H3K4me3 (middle panel) or FLAG antibody (bottom panel) for the reChIP. Means for two to seven independent biological replicates are shown, error bars refer to the SEM. Sequential ChIPs were performed in wild type embryos at 2–4 h AEL since the assay requires large amounts of cells but not tissue homogeneity. Asterisks indicate p value < 0.05, a triple asterisk indicates p value < 0.001 (t-test). (B) To analyze the global co-occupancy of Pol II and H3K27me3, we sequenced the single H3K27me3 ChIP and the H3K27me3-Pol II reChIP. The results confirm at a genome-wide level that the additional enrichment at the TSS of genes after the second ChIP (y axis = log₂ Pol II ChIP − log₂ H3K27me3 ChIP) strongly depends on whether Pol II is present at the gene (classified as no Pol II, low Pol II and high Pol II based on an individual Pol II ChIP). Shown is a box plot with the whiskers representing the interquartile ranges and the circles showing outliers. The vast majority of genes with high Pol II show enrichment after the reChIP (p < 10⁻³² in t-test). (C) Genome browser snapshot of a ~ 320 kb genomic region encompassing the balanced genes inv and en (gray box). (D) Timecourse analysis of genes that have both Pol II and H3K27me3 enrichment at the first time point. The heatmap (left) represents the relative enrichments of Pol II and K27me3 as well as their respective relative expression levels (0 = no enrichment/expression, 1 = highest enrichment/expression). The line graphs (right) show the median levels for each of the four clusters in the heatmap. Note that expression levels decrease over time and correlate positively with Pol II occupancy and negatively with H3K27me3 enrichment. See also Figure S4 and Table S3.

Figure 7. The balanced state is similar to the bivalent domain in mammals

(A-D) The lack of bivalent domains in Drosophila can be explained by the absence of detectable levels of H3K4me3 at poised genes. (A) Significant H3K4me3 levels are detected at paused and expressed genes (twi example, top) but not poised genes (Skl example, bottom). (B) Average gene analysis of Pol II (blue) and H3K4me3 (green) at expressed (solid lines) or poised (dashed lines) genes in 10–12 h Drosophila muscle cells confirms that poised genes in Drosophila lack significant levels of H3K4me3 (p < 10⁻¹⁰⁴ for H3K4me3 enrichment between expressed and poised genes at base position +163; Wilcoxon rank sum test). (C). In mouse embryonic stem cells, average gene analysis of Pol II (blue) and H3K4me3 (green) at expressed (solid lines) or poised (dashed lines) genes shows that poised mouse genes have high levels of H3K4me3 (p < 10⁻⁶ for H3K4me3 enrichment between expressed and poised genes at base position +130; Wilcoxon rank sum test). (D) The presence of CpG islands in mouse promoters does not explain the high levels of H3K4me3 at poised genes as shown by the average gene analysis of Pol II (blue) and H3K4me3 (green) at poised mouse genes with (solid lines) or without (dashed lines) CpG island promoters. (E) The balanced state behaves similarly to the bivalent state during mouse ES cell differentiation induced by retinoic acid. The relative predictive values for future gene expression were calculated as described in Figure 2F. Both poised Pol II and H3K4me3 are positive predictors for future gene expression but there are more genes with H3K4me3 and the predictions for poised Pol II are more stage-specific. H3K27me3 is a negative predictor until ~8 h and a positive predictor starting ~24 h. Balanced genes (Pol II above background and H3K27me3, n = 445) and bivalent genes (H3K4me3 and H3K27me3, n = 483) highly overlap (n = 340). See also Figure S5.