Histone acylation marks respond to metabolic perturbations and enable cellular adaptation

Abstract

Acetylation is the most studied histone acyl modification and has been recognized as a fundamental player in metabolic gene regulation, whereas other short-chain acyl modifications have only been recently identified, and little is known about their dynamics or molecular functions at the intersection of metabolism and epigenetic gene regulation. In this study, we aimed to understand the link between nonacetyl histone acyl modification, metabolic transcriptional regulation, and cellular adaptation. Using antibodies specific for butyrylated, propionylated, and crotonylated H3K23, we analyzed dynamic changes of H3K23 acylation upon various metabolic challenges. Here, we show that H3K23 modifications were highly responsive and reversibly regulated by nutrient availability. These modifications were commonly downregulated by the depletion of glucose and recovered based on glucose or fatty acid availability. Depletion of metabolic enzymes, namely, ATP citrate lyase, carnitine acetyltransferase, and acetyl-CoA synthetase, which are involved in Ac-CoA synthesis, resulted in global loss of H3K23 butyrylation, crotonylation, propionylation, and acetylation, with a profound impact on gene expression and cellular metabolic states. Our data indicate that Ac-CoA/CoA and central metabolic inputs are important for the maintenance of histone acylation. Additionally, genome-wide analysis revealed that acyl modifications are associated with gene activation. Our study shows that histone acylation acts as an immediate and reversible metabolic sensor enabling cellular adaptation to metabolic stress by reprogramming gene expression.

Fig. 1. Histone acylation is closely linked to metabolic conditions.

a Acetylation, propionylation, butyrylation, and crotonylation of lysine residues on histone tails. b Microarray data of glucose-deprived C2C12 myotube cells. Left; scatter plot with red dots representing genes with significantly altered expression (FC ≥ 2). Right panel: total heatmap (FC ≥ 2). C2C12 myotube cells were cultured in “Glc”, high-glucose DMEM for 24 h; “-Glc”, no glucose DMEM for 24 h; or “−/+Glc”, no glucose DMEM for 24 h followed by high-glucose DMEM for 12 h. c Top 5 DAVID annotated gene ontology groups for significantly altered genes under “-Glc” conditions relative to “Glc” conditions. d Heat map of the “muscle contraction” gene set from the microarray dataset. e WB with the indicated antibodies was used to identify histone modification.

Fig. 2. H3K23 acylation is reversibly modulated upon metabolic perturbation.

a WB of C2C12 myotubes grown under glucose-deprivation conditions. C2C12 myotubes were cultivated in glucose-deficient medium for the indicated period. Anti-H3K23 acylation-specific antibodies were used for protein IP, which were subsequently immunoblotted with anti-H3 antibody. b WB of H3K23 modifications using the indicated antibodies with cellular extracts obtained from C2C12 myotubes treated with/without 2DG (25 mM) for 24 h. c WB of H3K23 acylation in C2C12 myotubes grown in “Glc”, high glucose-DMEM for 24 h; “BSA”, no glucose DMEM for 24 h with fatty acid-free BSA; or “+OA”, no glucose DMEM for 24 h with oleic acid-conjugated BSA (200 μM oleic acid). d Relative quantitation of WB bands representative of the in vivo starvation experiment. “Control” (n = 4), fed ad libitum 24 h; “Starved” (n = 4), no food for 24 h; “Refed” (n = 5), starved for 24 h followed by refeeding for 6 h. Soleus muscles from each mouse were analyzed by WB using acylation-specific antibodies. H3K23 acylation and p-AMPK were normalized to their control counterparts, H3 and AMPK, respectively. The error bars indicate the SEM for each group.

Fig. 3. Impact of direct perturbation of Ac-CoA pools on histone acylation.

a Schematic presentation depicting the flux of metabolites mediated by ACLY, CRAT, and ACSS2, which contributes to Ac-CoA/SCA-CoA pools that support nuclear histone modification. b Relative cell count (%) of the indicated siRNA-transfected C2C12 myoblasts compared to the percentage of siCtl. The cells were analyzed 3 days after siRNA treatment. The error bars indicate the SEM (n = 5). Statistical significance was analyzed based on comparison to siCtl, *p < 0.05; **p < 0.01. c WB of total lysates obtained from C2C12 myoblasts transfected with the indicated siRNAs. d–e Heat maps of metabolite fold changes. C2C12 myoblasts were transfected with the indicated siRNAs for 72 h, and metabolite abundance was measured using GC–MS; each column indicates an individual sample (n = 3). Glycolysis, PPP, and TCA cycle intermediates (d) and fatty acids (e) are shown.

Fig. 4. Cellular levels of Ac-CoA and the Ac-CoA/CoA ratio modulate histone acylation.

a Bar plot of intracellular concentrations of CoA, Ac-CoA, Pr-CoA, and Bu-CoA in C2C12 myoblasts. The relative ratios of Ac-CoA/CoA (b), Pro-CoA/CoA (c), and Bu-CoA/CoA (d) were measured using LC–MS. Metabolic enzymes in C2C12 myoblasts was knocked down with specific siRNAs. Metabolome analysis is indicated in two independent experiments (1^st, 2^nd); the error bars indicate the mean SEM (n = 3 per group); *p < 0.05; **p < 0.01 compared to siCtl. e WB of siRNA-treated C2C12 myoblasts. Each supplement (Ac: acetate, Cr: crotonate, Bu: butyrate, and Pr: propionate) was added to the culture medium (10 mM for 24 h). f WB of the siRNA-treated C2C12 myoblasts. Butyrate (100 μM) was added to the medium for the indicated times; *p < 0.05, **p < 0.01.

Fig. 5. Transcriptome analysis reveals a cohort of genes sensitive to metabolic perturbation.

a Venn diagram showing differentially expressed genes upon Acly-, Acss2-, or Crat-knockdown in C2C12 myoblasts (relative to the level of siCtl, |FC | ≥ 1.5). Numbers represent the number of commonly upregulated or downregulated (underlined) genes following siRNA treatment. The values in brackets represent gene numbers differently regulated within paired groups. b Venn diagram showing the number of downregulated genes from Acly-, Acss2-, and Crat-knockdown C2C12 myoblasts (relative to the level of the siCtl, |FC | ≥ 1.5). c GO analysis of 204 commonly downregulated genes (|FC | ≥ 1.5). The top 5 biological functions are shown in order of ascending p-values. d Heat map of differentially expressed genes associated with DNA transcription (categorized into initiation, elongation, and termination). e GO analysis of significantly upregulated genes (|FC | ≥ 1.5). The top five biological functions are shown in order of ascending p-value. Each siRNA-mediated knockdown sample was compared with the siCtl.

Fig. 6. H3K23Ac marks are associated with active transcription and undergo remodeling following Acly knockdown.

a siCtl-transfected C2C12 myoblast cells were analyzed by ChIP-seq with the indicated antibodies. The average ChIP signals of the indicated histone modification were aligned to the TSSs of expressed genes and grouped into four quartiles: Q4, 75–100% (high expression); Q3, 50–75%; Q2, 25–50%; Q1, 0–25% (low expression). The list of differentially expressed genes is from the RNA-seq data. b ChIP-seq signals of the indicated H3K23 acyl marks ±1000 bp from the center of the H3K27Ac peaks. The H3K27Ac data was obtained from the GSE37525 dataset. H3K27Ac peaks found within −1000 bp ~ +500 bp of a TSS were classified as promoter peaks (Prom, brown). All the other peaks were classified as putative enhancers (Enh, orange). c Metagene profiles of the indicated histone modifications of control or Acly knockdown samples were aligned to the TSS of expressed genes. d Genome browser view of the H3K23Ac coverage of the Ccp110 gene locus. e Acyl modification of the Ccp110 promoter region was analyzed by ChIP-qPCR with anti-H3K23 acylation-specific antibodies. The error bars indicate SEM between duplicate experiments; *p < 0.05, **p < 0.01.