Glucose starvation induces a switch in the histone acetylome for activation of gluconeogenic and fat metabolism genes

Abstract

Acetyl-CoA is a key intermediate situated at the intersection of many metabolic pathways. The reliance of histone acetylation on acetyl-CoA enables the coordination of gene expression with metabolic state. Abundant acetyl-CoA has been linked to the activation of genes involved in cell growth or tumorigenesis through histone acetylation. However, the role of histone acetylation in transcription under low levels of acetyl-CoA remains poorly understood. Here, we use a yeast starvation model to observe the dramatic alteration in the global occupancy of histone acetylation following carbon starvation; the location of histone acetylation marks shifts from growth-promoting genes to gluconeogenic and fat metabolism genes. This reallocation is mediated by both the histone deacetylase Rpd3p and the acetyltransferase Gcn5p, a component of the SAGA transcriptional coactivator. Our findings reveal an unexpected switch in the specificity of histone acetylation to promote pathways that generate acetyl-CoA for oxidation when acetyl-CoA is limiting.

Keywords: Gcn5p; Rpd3p; SAGA; acetyl-CoA; environmental stress response; fat metabolism; gluconeogenesis; glucose starvation; histone acetylation; transcription.

Figure 1.. Glucose starvation results in reduced intracellular acetyl-CoA, decreased amounts of bulk H3 acetylation, and redistribution of H3K9ac occupancy

(A) Experimental setup of glucose starvation model. Cells were grown to log phase in glucose (+D), before switching to medium without glucose (−D). After glucose starvation, glucose was replenished at a final concentration of 2% (−D→+D). (B) Intracellular acetyl-CoA levels were detected by LC-MS/MS. Two specific daughter fragments for acetyl-CoA (303, 159 Da) were quantified. The data are represented as mean ± SD (n = 3). (C) Immunoblots of bulk histone acetylation in response to glucose availability. After 60 min of glucose starvation, glucose was added back for another 5 or 30 min. (D) ChIP-PCR analysis of histone acetylation occupancy at TSS of RPL33B under three conditions: glucose-replete (+D), glucose starvation for 30 min (−D), and glucose replenished for 30 min (−D→+D). All signals were normalized to H3. The data are represented as mean ± SD (n = 3). (E) ChIP-seq data displaying genomic occupancy of H3K9ac ± 2 kb flanking TSS in two conditions: glucose-replete (+D) and glucose starvation for 30 min (−D). The conditions were the same for all sequencing experiments in this study. Genes shown in rows were sorted in descending order by signal intensity in each condition. The black lines denote a lack of reads that map to certain regions such as telomeres or the ends of chromosomes. (F) Differential analysis of ChIP-seq data showing genomic occupancy of H3K9ac from panel (E). The whole genome was classified based on the changes in H3K9ac signals upon glucose starvation (-D/+D): decreased (↓, colored in orange), increased (↑, colored in blue) or unchanged (-, colored in grey). Thresholds of 2-fold change (FC) and 0.05 false discovery rate (FDR) were considered significant. The numbers of genes in each category are denoted. In heatmaps, the subset with decreased H3K9ac was ranked by signal intensity from glucose-replete condition. The subsets with increased and unchanged H3K9ac were ranked by signal intensity from glucose starvation condition. Genes shown in each row are the same between two samples.

Figure 2.. H3K9ac and transcriptome changes are highly correlated under both glucose-replete and glucose starvation conditions

(A) Volcano plot of RNA-seq showing differentially expressed genes in WT cells upon glucose starvation (-D/+D). Thresholds of 2-FC and 0.05 FDR were considered significant. We referenced the clusters of genes reported to be associated with growth rate (Brauer et al., 2008). Genes that are positively correlated with growth rate are colored in orange; genes that are negatively correlated with growth rate are in blue. (B) Scatter plot depicting differential regulation upon glucose starvation from RNA-seq (y-axis) and H3K9ac ChIP-seq (x-axis) data. Orange dots denote genes repressed in both mRNA and H3K9ac; blue dots denote genes induced in both mRNA and H3K9ac. Black dots indicate significant genes that did not have positive correlation between changes in transcript levels and H3K9ac occupancy. (C) Venn diagram depicting the number of differentially regulated genes with overlap between RNA-seq and H3K9ac ChIP-seq. Arrows depict the transcripts that were independent of H3K9ac. The amount of genes exhibiting changes in H3K9ac is likely an overestimate due to the peak calling algorithm and short intergenic regions in the yeast genome, as reads overlapping multiple genes were not discarded and assigned to multiple genes (see STAR METHODS).

Figure 3.. H3K9ac is enriched at the TSSs of genes required for gluconeogenesis and fat metabolism upon glucose starvation

(A) ChIP-seq data displaying H3K9ac occupancy at two subsets of genes: growth-promoting genes (transcripts down-regulated upon starvation) and starvation-induced genes (transcripts up-regulated upon starvation). Metagene profile shows the average H3K9ac signal at either subset: growth-promoting genes in orange; starvation-induced genes in blue. Growth-promoting genes in heatmaps were ranked by signal intensity from glucose-replete condition (+D), while starvation-induced genes were ranked by signal intensity from glucose starvation (−D). Genes shown in rows are the same between samples. (B) Gene ontology (GO) analysis of enriched biological processes for starvation-induced genes with most increased H3K9ac occupancy upon glucose starvation. See also Table S1. (C) Consolidated enriched biological processes derived from GO analysis presenting major routes of carbon metabolic pathways in yeast under glucose starvation. Genes highlighted in blue were highly induced both in mRNA levels and H3K9ac signals upon glucose starvation. (D) Genome browser view showing H3K9ac occupancy (in black) and mRNA levels (in pink) at starvation-induced genes under glucose-replete (+D) or glucose starvation (−D) conditions. The divergence of two subsets were differentiated based on the dependence on H3K9ac. H3K9ac-dependent genes involved in gluconeogenic and fat metabolism are shown in blue; H3K9ac-independent genes in black. Arrowheads indicate peaks of H3K9ac that reside at the TSS of genes of interest. (E) ChIP-PCR analysis of histone acetylation occupancy at TSS of CAT2 under three conditions: glucose-replete (+D), glucose starvation for 30 min (−D), and glucose replenished for 30 min (−D→+D). All signals were normalized to H3. The data are represented as mean ± SD (n = 3).

Figure 4.. gcn5Δ cells lose H3 acetylation and exhibit impaired induction of gluconeogenic and fat metabolism genes upon glucose starvation

(A) Immunoblots of bulk histone acetylation in WT and gcn5Δ cells following glucose starvation. Note that gcn5Δ cells exhibited a delayed loss in H4 acetylation, which could be due to cooperative interactions with other H4 HATs (Li and Shogren-Knaak, 2009). (B) ChIP-seq data displaying genomic occupancy of H3K9ac in WT and gcn5Δ cells in glucose-replete (+D) or glucose starvation (−D) conditions. For each condition, genes were ranked in the order of decreasing signals in WT cells. Genes shown in rows are the same between WT and gcn5Δ cells. (C) Volcano plot of ChIP-seq data showing the differential occupancy of H3K9ac in gcn5Δ cells compared to WT cells. Thresholds of 2-FC and 0.05 FDR were considered significant. Gluconeogenic and fat metabolism genes are colored blue. Ribosomal protein genes are colored orange. (D) Volcano plot of RNA-seq data showing differential gene expression in gcn5Δ cells compared to WT cells in glucose-replete (+D) or glucose starvation (−D) conditions. Thresholds of 2-FC and 0.05 FDR were considered significant. Gluconeogenic and fat metabolism genes are colored in blue. (E) Venn diagrams showing the numbers of genes induced by glucose starvation in WT cells (starvation-induced transcripts) or repressed in gcn5Δ cells under glucose starvation.

Figure 5.. SAGA binds to different groups of genes and transcription factors depending on glucose availability

(A) ChIP-seq data displaying SAGA occupancy (assessed by Gcn5p-HA) at two subsets of genes upon glucose starvation. Metagene profile shows the average signal of SAGA binding at either subset: growth-promoting genes in orange; starvation-induced genes in blue. Growth-promoting genes in heatmaps were ranked by signal intensity from glucose-replete condition (+D), while starvation-induced genes were ranked by signal intensity from glucose starvation (−D).Genes shown in rows are the same between two conditions. (B) Volcano plot of SAGA ChIP-seq depicting the differential binding of SAGA in cells upon glucose starvation. Thresholds of 2-FC and 0.05 FDR were considered significant. Ribosomal protein genes are colored in orange. Gluconeogenic and fat metabolism genes are colored in blue. (C) Genome browser view showing co-occupancy of SAGA and H3K9ac at growth-promoting genes (orange) or gluconeogenic and fat metabolism genes (blue). Orange and blue arrowheads indicate the peaks of H3K9ac that reside at the TSS of the genes of interest. Black arrowhead indicates an H3K9ac peak independent of Gcn5p/SAGA. (D) Interaction of SAGA with different transcription factors (TFs) in glucose-replete (+D) or glucose-starvation conditions (−D). Endogenously Flag-tagged Spt7p was used to immunoprecipitate the SAGA complex. Co-immunoprecipitation of HA-tagged TFs were detected by immunoblotting. Empty arrowheads indicate the interaction between SAGA and growth-specific TFs (orange). Solid arrowheads indicate the interaction between SAGA and starvation-specific TFs (blue).

Figure 6.. rpd3Δ cells exhibit a retention of H3K9ac at growth-promoting genes and a reduction of H3K9ac at gluconeogenic and fat metabolism genes during glucose starvation

(A) Immunoblots of bulk histone acetylation in WT and rpd3Δ cells upon glucose starvation. (B) ChIP-seq data displaying genomic occupancy of H3K9ac in WT and rpd3Δ cells under glucose-replete (+D) or glucose starvation (−D) conditions. For each condition, genes were ranked in the order of decreasing signals in WT cells. Genes shown in rows are the same between WT and rpd3Δ cells. (C) ChIP-seq data displaying H3K9ac occupancy at two subsets of genes in WT and rpd3Δ cells under glucose starvation condition. Metagene profile shows the average H3K9ac signal from each subset of genes: orange line denotes growth-promoting genes; blue line denotes starvation-induced genes. Growth-promoting genes in heatmaps were ranked by signal intensity from glucose-replete condition (+D), while starvation-induced genes were ranked by signal intensity from glucose starvation (−D). Genes shown in rows are the same among all samples. (D) Volcano plot of ChIP-seq data showing differential occupancy of H3K9ac in rpd3Δ cells compared with WT cells. Thresholds of 2-FC and 0.05 FDR were considered significant. Ribosomal protein genes are colored orange. Gluconeogenic and fat metabolism genes are colored blue.

Figure 7.. Rpd3p and SAGA are required for proper H3K9ac refocusing onto gluconeogenic and fat metabolism genes following glucose starvation

(A) Cells of the indicated genotypes were grown in either glucose-replete (+D) or glucose starvation conditions for 2 hr (−D) prior to spotting on plates containing the indicated carbon sources: glucose (SD), ethanol (SE), glycerol (SG), and oleate (STYO). Note that gcn5Δ cells exhibited a severe growth defect on all carbon sources, consistent with its role in activating expression of either growth-promoting genes (glucose) or gluconeogenic/fat metabolism genes (-glucose), while rpd3Δ cells exhibited normal growth on glucose, but slower growth on both ethanol and oleate, consistent with reduced expression of gluconeogenesis and fat metabolism genes in this mutant. (B) Graphic model depicting the reallocation of histone acetylation marks following glucose starvation. Acetyl groups (red circles) can conjugate with CoA as acetyl-CoA (Ac-CoA), decorate histone tails as histone acetylation, or be released from histone tails as free acetate. In glucose-replete conditions (D), ample amounts of acetyl-CoA enable the Gcn5p-containing SAGA complex to target growth-promoting genes, acetylate histones on H3, and activate gene expression that promotes cell growth. Upon glucose starvation (−D), Rpd3p deacetylates growth-promoting genes, releasing acetate. The free acetate may be recycled in the form of acetyl-CoA for subsequent histone acetylation. Instead of growth-promoting genes, SAGA now switches its targets to genes required for gluconeogenesis and fat metabolism, in coordination with starvation-specific TFs. The expression of these genes promotes peroxisomal activity, fat metabolism and the subsequent production of acetyl-CoA.