Nutrient-driven histone acetylation underlies energy storage and mobilization

Abstract

In natural settings, energy storage and mobilization maintain a dynamic balance in response to recurrent overfeeding and fasting. Imbalanced energy storage and mobilization lead to a variety of metabolic dysfunctions. However, whether the metabolic status directly couples with epigenetic modifications and transcriptional outputs remains unclear. Here, we aimed to investigate the epigenetic mechanism underlying this adaptive balance and observed that, in an overfeeding state, increased glucose availability is associated with enhanced histone acetylation coinciding with acetyl-CoA production in an acyl-CoA short-chain synthetase 2 (ACSS2)-dependent manner, contributing to energy storage (e.g., lipogenesis); in contrast, in the fasting state, elevated d-β-hydroxybutyrate levels are associated with altered histone acetylation distribution and transcriptional programs, supporting a metabolic shift from anabolism to catabolism, such as fatty acid oxidation. In both overfeeding and fasting states, acetylated lysines in the histone require BRD4 to recognize and initiate transcriptional regulation. Inhibition of BRD4 leads to context-dependent phenotypic effects: it ameliorates non-alcoholic fatty liver disease (NAFLD) pathology induced by a high-fat diet, while it exacerbates hepatic steatosis in fasted mice or mice fed a ketogenic diet. Thus, these findings highlights that epigenetic regulation of energy storage and mobilization is closely linked to the availability of glucose, and ketone bodies. Moreover, our study revealed that modulation of ACSS2-associated pathway may represent a potential strategy for treatment of metabolic diseases, such as NAFLD.

Graphical abstract

Glucose-derived acetyl-CoA is associated with an “active” mode of histone acetylation, which coincides with the expression of genes involved in energy storage; in parallel, overfeeding is accompanied by chronic metaflammation. By contrast, BHB sustains a “passive” mode of histone acetylation and is associated with the expression of genes related to energy mobilization, while fasting is accompanied by a dampened acute inflammatory response. Both “active” and “passive” histone acetylation were recognized by BRD4 to execute transcription regulatory effects. BRD4 inhibition improved high-fat diet-induced nonalcoholic fatty liver diseases pathology but aggravated hepatic steatosis in fasted mice or mice fed with the ketogenic diet.

Highlights

• •d-β-hydroxybutyrate enhances histone “passive acetylation” to support energy mobilization.
• •Glucose enhances histone “active acetylation” to support energy storage.
• •ACSS2 is required for in vitro and in vivo lipogenesis.

1. Introduction

Feast or famine is a regular pattern for animals with respect to obtaining food in the natural environment. In a food-plentiful state, animals store energy, mainly in the form of fat, as much as they can, whereas in a food-scarce state, the stored energy is mobilized to sustain basic life activities for survival. Animals use glucose as regular fuel, but it is converted to ketone bodies (acetoacetate, d-β-hydroxybutyrate, and acetone) as an alternative fuel in carbon-scarce states [1,2]. This periodic shift in fuel metabolism between glucose and ketone bodies is vividly activated but disrupted in modern society due to the elimination of starvation [2].

Glucose and ketone bodies are analogous, as both are water-soluble small molecules, enabling them to pass through the blood–brain barrier [3]. Despite these similarities, glucose and ketone bodies differ significantly in their ability to produce nucleocytosolic acetyl-CoA, which is generated by ATP-citrate lyase (ACLY) or acyl-CoA synthase short-chain family member 2 (ACSS2), using glucose-derived citrate or acetate as the substrate [4,5]. This nucleocytosolic acetyl-CoA is normally used for lipogenesis and protein acetylation, such as histone acetylation. In contrast, ketone bodies cannot generate nucleocytosolic acetyl-CoA, as ketolytic enzymes, such as BDH1, are restricted to the mitochondrial compartment [1]. Traditionally, ketolysis-derived acetyl-Co-A has been considered as fuel of the tricarboxylic acid (TCA) cycle for mitochondrial energy production. Since ketone-derived carbon cannot be readily exported to nucleus due to the lack of citrate shuttle, it has been assumed that ketone bodies cannot support nucelocytosolic acetyl-CoA production and therefore, cannot sustain histone acetylation [[6], [7], [8], [9]]. Recent studies showed that acetylcarnitine can be shuttled between mitochondria and nucleocytosolic compartment and can serve as an alternative source of nucleocytosolic acetyl-CoA in cancer cell lines [10]. However, how animals maintain global histone acetylation and the fasting gene programs in response to starvation remains unclear. In 2013, Shimazu et al. reported that d-β-hydroxybutyrate (BHB) functions as an endogenous histone deacetylase (HDAC) inhibitor [11] to maintain histone acetylation in multiple organs, which might compensate for the reduced levels of nucleocytosolic acetyl-CoA available for protein acetylation. Thus, the acetylated modification catalyzed by acetyl-CoA and histone acetyltransferases (HAT) could be considered as “active acetylation”, whereas the acetylated modification enhanced by HDAC inhibition could be considered as “passive acetylation”. BHB protects against oxidative stress through HDAC inhibition–associated acetylation effect [11]; however, its broader evolutionary role as an HDAC inhibitor (HDACi) is still unknown.

Synchronizing gene expression with metabolic status is an optimal strategy for regulating energy storage and mobilization [9,12]. Thus, chromatin modification by metabolic intermediates, particularly acetyl-CoA, which sits at the intersection of many metabolic pathways, provides a useful paradigm for control of metabolic homeostasis [12]. In this study, we observed that active acetylation (glucose-mediated) and passive acetylation (BHB-mediated) are oppositely distributed at certain genomic loci, transactivating genes to support exogenous energy storage or endogenous energy mobilization. We defined the functions of these acetylation by blocking the recognition of different acetylation marks through inhibition of BRD4 (Bromodomain-containing protein 4), an acetylated histone “reader” protein. Blocking either active or passive acetylation results in opposing metabolic phenotypes under different energetic contexts. This study highlights the involvement of epigenetic modifications in linking environmental energy input with transcriptional output, suggesting that targeted manipulation of epigenetic modifications may offer novel opportunities for therapeutic development of in metabolic diseases.

2. Results

2.1. BHB-mediated HDAC inhibition is linked to reduced lipolysis in the white adipose tissue

First, the mice were subjected to fasting to induce body weight loss and ketogenesis. We observed that the body weight loss trajectory of the mice is not in a linear pattern in response to fasting. The body weight decreased by approximately 3.7% in the first 6 h, but the rate of decline gradually slowed (Figure 1A). This uneven reduction in body weight is not attributed to exercise-mediated energy expenditure, as fasting does not reduce and tends to increase locomotor activity [13]. Given that fasting causes a gradual decrease in blood glucose levels but increase blood ketone bodies, we performed a Pearson correlation analysis and demonstrated an inverse correlation between body weight loss velocity and blood BHB levels; the higher the blood ketone bodies level, the slower the body weight loss rate (Figure 1B).

Figure 1.

BHB-mediated HDAC inhibition suppresses lipolysis in the WAT. (A) The body weight loss and the blood BHB levels of the mice underwent starvation at the indicated time points (n = 5). (B) Pearson correlation analysis of the blood BHB levels and body weight loss rate. (C and E) Western blotting assay of H3K27ac in the white adipose tissue (WAT) of mice underwent fasting (C) or re-feeding (E) at the indicated time points (left); the quantification of the bands (right) (n = 5). (D and F) The blood glucose and BHB levels of mice underwent fasting (D) or re-feeding (F) at the indicated time points (n = 5). (G and H) The medium FFA levels of the adipocytes differentiated from 3T3-L1 (G) or C3H10T1/2 (H) with indicated treatment (n = 3–4). JQ35 (1 μM), BHB (20 mM), and Merck60 (1 μM) pretreated for 4 h, followed by isoproterenol (ISO, 10 μM) treatment for 3 h. (I) The medium FFA levels of WAT pretreated with BHB ± JQ35, followed by 2 h of ISO treatment. (n = 3). (J) qPCR assays of lipolytic genes (Hsl, Atgl) of the adipocytes differentiated from C3H10T1/2 with the indicated treatment (n = 3). (K and L) The body weight (K) and the blood BHB levels (L) of mice underwent fasting or fasting plus Merck60 (n = 5). (M and N) Pearson correlation analysis of the blood BHB levels with core body temperature (M) or heart rate (N). Data are shown as the mean ± SEM. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001 by unpaired two-tailed Student’s t-test (L), one-way (C, E, and G-J), or two-way ANOVA (D, F, and K) followed with Bonferroni’s multiple comparisons test.

Since white adipose tissue (WAT) is the major reservoir providing energy substrates for global organs through activating lipolysis during fasting, we examined whether increased ketone bodies suppress WAT lipolysis biology. Fasting-induced histone 3 lysine 27 site acetylation (H3K27ac), a mark associated with active gene promoters and enhancers increased at 6 h, and further elevated at 24 h (Figure 1C). The increased H3K27ac rapidly reduced upon re-feeding (Figure 1E). The change of histone acetylation in WAT closely correlates with blood BHB levels (Figure 1D,F). To examine the role of BHB-mediated HDACi on lipolysis, 3T3-L1 adipocytes were pretreated with BHB followed by isoproterenol (ISO) treatment. We observed that BHB significantly suppressed ISO-induced lipolysis. MERCK60, an HDAC1/2 inhibitor [14], mimics the suppressive action of BHB on lipolysis (Figure 1G). Bromodomain and extraterminal (BET) proteins function as key epigenetic “reader” to recognize acetylated histones [15]. We observed that JQ35, a novel BET proteins inhibitor developed in our lab, completely reversed BHB-suppressed lipolysis (Figure 1G). As expected, BHB suppressed lipolysis in both C3H10T1/2 adipocytes and ex vivo cultured white adipose tissues (Figure 1H,I). Moreover, BHB suppressed the expression of Hsl and Atgl, key enzymes catalyzing lipolysis, whereas JQ35 treatment recovered their expression (Figure 1J). In the in vivo study, MERCK60 treatment readily suppressed body weight loss without affecting ketone body levels (Figure 1K,L). In addition to the HDACi-mediated effect of BHB on lipolysis, we observed a significant negative correlation between blood BHB levels with core body temperature and heart rates (Figure 1M,N). As body temperature and heart rate are primarily controlled by sympathetic nervous system (SNS) activity, we consider that BHB-mediated energy conservation might be partly caused by the suppression of SNS activation, which is consistent with previous finding [16]. Taken together, these data revealed that BHB suppresses energy mobilization in response to carbon scarcity through its HDACi-dependent and -independent effects.

2.2. BHB-mediated HDAC inhibition accompanies a hepatic metabolic shift toward lipid oxidation

As JQ35 treatment reversed BHB-inhibited lipolysis in the in vitro studies, we performed an in vivo study to observe the effect of JQ35 treatment on fasting-induced lipolysis. The fasted mice treated with JQ35 had higher serum free fatty acids (FFA) levels than those treated with vehicle (Figure 2A,B). Increased lipolysis is expected to increase body weight loss; surprisingly, JQ35 treatment paradoxically retards it, suggesting complexity in the in vivo context (Figure 2C). Inhibition of fatty acid oxidation (FAO) might also elevate serum FFA levels; thus, we focused on the liver, the main organ responsible for FAO and ketogenesis in response to fasting [1]. Consistent with previous publications, we observed that bulk H3K27ac levels increased in the liver at 24 h and became more pronounced at 48 h of fasting [11] (Fig. S1A). Inhibition of histone acetylation recognition by JQ35 suppressed FAO-associated genes expression, suggesting that the fasting-induced hepatic gene program is bromodomain containing 4 (BRD4)- and probably BHB-mediated histone acetylation-dependent (Fig. S1B).

Figure 2.

BHB-mediated HDAC inhibition shifts hepatic anabolism toward lipid oxidation. (A) Experimental design. (B) Serum gross appearance and the FFA levels (n = 5). (C) Body weight loss of mice with the indicated treatment (n = 5). (D and E) ChIP-seq data displaying H3K27ac (D) or BRD4 (E) occupancy at two subsets of genes: H3K27ac and BRD4 enriched at the genomic loci of fasting-induced transcripts (energy mobilization associated genes, such as Cyp4a14), but dislodged from those loci of fasting-suppressed transcripts (energy storage-associated genes, such as Pcsk9). (ref. to Hsieh et al., 2022, Mol Cell for detail method of analysis). (F and G) GO enrichment analysis of the fasting-induced H3K27ac (F) and BRD4 (G) peak-associated genes. The reads number of the ChIP-seq data for H3K27ac (log2FC > 1 and P < 0.01) or BRD4 (log2FC > 1.2 and P < 0.01) were included as the fasting-induced peaks. (H and I) Gene tracks display that fasting induced the occupancy of H3K27ac and BRD4 at the genomic loci of energy mobilization-associated genes (Cyp4a14, Pck1) (H), but reduced at those loci of energy storage-associated genes (Pcsk9, Srebf1) (I). (J) The ratio of the liver to body weight (upper panel) and the liver triglycerides levels (lower panel) of mice with the indicated treatment (n = 5). (K-O) Body weight loss data (K), serum gross appearance and FFA levels (L), the blood glucose and BHB levels (M), the ratio of liver to body weight (N) and Liver triglycerides (O)of Brd4-flox or Brd4 hKO mice subjected to 24 h of fasting (n = 5). Data are shown as the mean ± SEM. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001 by unpaired two-tailed Student’s t-test (K-O) or one-way ANOVA followed with Bonferroni’s multiple comparisons test (B, C, and J).

To validate the hypothesis, we performed H3K27ac and BRD4 chromatin immunoprecipitation followed by sequencing (ChIP-seq) assays to determine whether fasting-induced histone acetylation links metabolic status to transcriptional outputs. Regardless of the treatment, H3K27ac and BRD4 signals were highly enriched at the promoter and intronic regions, consistent with their role as markers of active gene transcription [17,18] (Fig. S1C). Quantitative overlap analysis showed a strong colocalization BRD4 and H3K27ac in differentially regulated genomic loci (Fig. S1D). Correspondingly, by comparing the RNA-seq with H3K27ac and BRD4 ChIP-seq data, we observed a positive correlation between the transcriptome and H3K27ac/BRD4 signals, indicating that the changed occupancy of H3K27ac and BRD4 in response to fasting is functionally relevant to transcription (Figure 2D,E). Gene Ontology (GO) analysis of the differential peaks (log₂FC > 1, P < 0.01) associated genes revealed that fasting-induced H3K27ac and BRD4 signals were significantly enriched to lipid catabolic process (Figure 2F,G). To closely monitor the transcriptome-epigenome relevance, we analyzed the loci of these exemplary fasting-adaptive genes. The signals of H3K27ac and BRD4 markedly increased at the genomic loci of fasting-induced genes (Cyp4a14 and Pck1, relevant to fasting adaptation) but decreased at fasting-suppressed genes (Pcsk9 and Srebf1, relevant to feeding adaptation) (Figure 2H,I). These data support the role of epigenetic reprogramming in shaping transcriptional output.

Suppression of FAO might cause an accumulation of triglycerides (TG) in the hepatocytes. As expected, JQ35-treated mice developed livers with a fatty appearance and enhanced liver weight and triglyceride accumulation 24 h post-fasting (Figure 2J). Despite the metabolic disturbance in the liver, blood glucose and circulating ketone levels were not influenced by JQ35 treatment (Fig. S1E). We validated the findings in an alternative ketogenic model and observed that ketogenic diet (KD) feeding and JQ35 treatment replicated the phenotypes observed in the fasted mice, including increased serum FFA levels, liver weight, and hepatic TG accumulation (Fig. S1F–S1I). In contrast to the fasting model, the JQ35 treatment reduced food intake, leading to lower body weight and blood glucose levels in the KD model; notably, circulating ketone levels remained unaffected (Fig. S1J-S1L). The qPCR assay confirmed the downregulation of FAO-associated gene expression by JQ35 treatment (Fig. S1M). Canagliflozin has been shown to reduce hepatic steatosis through enhanced FAO and ketogenesis [19]. We observed that co-treatment with JQ35 completely reversed its anti-steatotic effects; it increased triglyceride accumulation in the liver (Fig. S1N and S1O). This finding raises concerns about the combination of SGLT2 and BET inhibitors, which are now developed as anticancer agents [20].

To validate the findings resulting from the global inhibition of BET bromodomain, we generated a hepatic-specific Brd4-knockout mouse (Brd4-hKO) model. A direct knockout of Brd4 in the liver causes embryonic lethality; however, tamoxifen-induced hepatic Brd4 knockout in adult mice is viable and appears normal (Fig. S1P). When subjected to fasting for 24 h, Brd4-hKO mice showed less body weight loss with enhanced serum FFA levels, suggesting a suppression of FAO in the liver (Figure 2K,L). Consistent with this assumption, the expression of FAO-associated genes was downregulated in the Brd4-hKO model compared to control mice (Fig. S1Q). Blood glucose levels tended to be lower in Brd4-hKO mice, while circulating BHB levels were comparable between genotypes (Figure 2M). Despite either Brd4 or Ppara knockout blocks FAO, Brd4-hKO did not affect circulating BHB levels, the phenotype was sharply different to that of Ppara knockout mice [21]. Similar to the phenotype observed in the fasting model, Brd4 knockout led to enhanced serum FFA levels and downregulation of expression of FAO-associated genes when the mice were fed with KD (Fig. S1R and S1T). In addition, Brd4-hKO mice exhibited increased body weight loss and reduced blood glucose levels due to less food consumption (Fig. S1S). Consistently, blood BHB levels were not influenced by Brd4 knockout (Fig. S1S). Although Brd4-hKO mice had larger livers in both fasting and KD-feeding models, hepatic TG levels were significantly lower than control mice, which is strikingly opposite to the increased TG accumulation in JQ35-treated mice (Figure 2N, O and Fig. S1U-S1V). In contrast, hepatic FFA levels in the KD-feeding model (Fig. S1W) were higher in Brd4-hKO than control mice, suggesting an impaired capability of fatty acids esterification. Indeed, we found that the expression of TG synthesis genes, such as Dgat2 and Gpat3 [22,23], was suppressed by Brd4 knockout but not by JQ35 treatment. Taken together, these data indicate that BHB-promoted histone acetylation couples with BRD4 to facilitate a shift in hepatic metabolism from anabolism toward lipid oxidation during fasting adaptation.

2.3. BHB-mediated HDAC inhibition is involved in BAT fasting biology

To maintain normal fasting adaptation, brown adipose tissue (BAT) shifts from using glucose to fatty acids for thermogenesis [24,25]. We first performed a western blot and confirmed that bulk H3K27ac is enhanced in the BAT after 24 h of fasting (Figure 3A). To determine H3K27ac distribution, we performed an H3K27ac CUT&Tag-seq assay. GO and KEGG analysis of the fasting-induced peak-associated genes showed the enrichment to lipid metabolic processes and autophagy (Figure 3B,C). We observed that H3K27ac enrichment significantly increased at key fatty acids catabolic gene loci, such as Cd36 (fatty acid uptake), Acsl1, Acot1 (fatty acid activation), Cpt1a, Cpt1b (fatty acid import to the mitochondrial matrix), and Acadl, Hadha (β-oxidation) (Figure 3D). The enrichment of H3K27ac at the loci of Ucp1, a gene for thermogenesis, is not significantly influenced by fasting, indicating the primary effects of BHB-mediated histone acetylation on metabolic, but not thermogenic, gene programs (Figure 3D).

Figure 3.

BHB-mediated HDAC inhibition sustains BAT fasting biology. (A) Western blotting assay of H3K27ac in the BAT of mice with or without 24 h of fasting (upper panel); the quantification of the bands (lower panel) (n = 4). (B and C) GO (B) and KEGG (C) enrichment analysis of fasting-induced peak-associated genes. The reads number of the CUT&Tag-seq data for H3K27ac (log2FC > 0.8 and P < 0.05) were included as the fasting-induced peaks. (D) Gene tracks display that fasting induced the occupancy of H3K27ac at the genomic loci of FAO associated genes, but not that of Ucp1. (E) Experimental design. (F-H) The core body temperature of the mice subjected to the indicated treatment (n = 4–5). (I) Glucose administration enhanced fasting-caused hypoglycemia (left) and rescued JQ35-suppressed thermogenesis (right) (n = 4–5). (J and K) The core body temperature, blood BHB and glucose levels of Brd4-flox or Brd4-hKO mice subjected to 24 h of fasting (J) (n = 5) or fed with ketogenic diet (K) (n = 4–5). Data are shown as the mean ± SEM. ∗P < 0.05, ∗∗∗∗P < 0.0001 by unpaired two-tailed Student’s t-test (A, I, J, and K) or two-way ANOVA followed with Bonferroni’s multiple comparisons test (F-I).

Next, we examined the influence of JQ35 treatment on the core body temperature of the mice subjected to fasting and cold exposure (Figure 3E) to indirectly evaluate the effect of fasting-induced histone acetylation on thermogenesis. A single dose of JQ35 was administered 1 h before cold exposure to avoid disturbing the fasting adaptive metabolism. We observed that the core body temperature of the fed mice was unaffected by JQ35 treatment (Fig. S2A), but it triggered a severe cold intolerance in fasted mice (Figure 3F). Likewise, JQ35 administration results in hypothermia in the KD-feeding and canagliflozin-treated mice (Figure 3G,H). Notably, glucose instead of intralipid or BHB restored JQ35-suppressed thermogenesis, indicating an inhibition of FAO rather than the thermogenic program per se. (Figure 3I and Fig. S2B-S2C). We observed that Brd4-hKO mice developed pathological hypothermia in response to fasting and KD-feeding at room temperature despite the comparable circulating BHB levels in both genotypes (Figure 3J,K), which indicates that apart from the effect by ketone bodies-mediated local histone acetylation, a liver-derived factor(s) could synergistically control thermogenesis. We observed that FGF21, a hepatokine induced by fasting that mediates lipolysis and thermogenesis [19,21,26], is downregulated in Brd4-hKO livers (Fig. S1Q and S1T). Administration of recombinant FGF21 rescued the loss of body temperature in fasted Brd4-hKO mice without increasing blood glucose or ketone levels (Fig. S2D). These data indicate that the liver might influence global metabolism through dual effects involving BHB-mediated HDAC inhibition and hepatokine output.

2.4. BHB-mediated HDAC inhibition promotes renal gluconeogenesis and anti-oxidation ability

Renal gluconeogenesis plays a key role in maintaining basic glycemia levels during prolonged fasting [27]. Whether BHB serves as a messenger through its HDACi effect to signal the kidney for compensatory gluconeogenesis is a fascinating question. Bulk H3K27ac levels increased markedly in the kidney 24 h after fasting, consistent with a prior report (Figure 4A) [11]. The ChIP-seq assay showed a significant induction of H3K27ac enrichment at gene loci of Pck1, but not G6pc, consistent with their transcriptional changes (Figure 4B,C). To examine the causal role of BHB-mediated histone acetylation on PCK1 expression, proximal tubule epithelial cells (HK2) were treated with BHB or MERCK60. In parallel, LY294002, a PI3K inhibitor known to relieve insulin-mediated suppression of gluconeogenesis, was included as a positive control for gluconeogenic gene activation [28]. Both treatments induced the expression of PCK1 but not G6PC, indicating that the HDAC inhibitory effect per se is sufficient to upregulate PCK1 expression (Figure 4D,E). BHB and MERCK60 also induced the expression of PCK1 in Caco-2, a human colon carcinoma cell line, suggesting a common regulatory effect on extrahepatic gluconeogenesis (Figure 4D,E).

Figure 4.

BHB-mediated HDAC inhibition promotes renal gluconeogenesis and anti-oxidation ability. (A) Western blotting assay of H3K27ac in the kidney of mice with or without 24 h of fasting (upper panel); the quantification of the bands (lower panel) (n = 5–7). (B and C) Gene tracks displayed the differential occupancy of H3K27ac at the genomic loci of Pck1 (massively increase) and G6pc (slightly increase) (B), which matches with the differential transcriptional output of the genes (C) (n = 5). (D and E) Western blotting (D) and qPCR (E) assay of HK2 and Caco-2 cells treated with LY294002 (1 μM), BHB (20 mM), or Merck60 (1 μM) for 24 h (n = 4). (F) Transcription factor motif enrichment analysis of the kidney ChIP-seq datasets. (G and H) Gene tracks displayed the occupancy of H3K27ac at the genomic loci of anti-oxidative genes (Cat, Gstt2, Gpx3, Hnf4a) (G), and JQ35 treatment suppressed the expression of these genes (H) (n = 5). Data are shown as the mean ± SEM. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001 by unpaired two-tailed Student’s t-test (A, C, and H) or one-way ANOVA followed with Bonferroni’s multiple comparisons test (E).

BHB exhibits an antioxidation effect in the kidney [11]. Transcription factor motif enrichment analysis of the kidney ChIP-seq datasets revealed high enrichment in FOXO3 and HNF4A binding elements of these fasting-induced peaks, suggesting an activated antioxidative transcriptional program (Figure 4F). Visualization analysis showed that fasting significantly increased the H3K27ac signal at classical antioxidation-related gene loci (Cat, Gpx3, Gstt2, Hnf4a) in the kidney (Figure 4G), and the expression of these genes was suppressed by JQ35 treatment (Figure 4H). Since hypoglycemia aggravates oxidative stress by reducing pentose phosphate pathway flow (which generates NADPH) and increases reactive oxygen species (ROS) generation via enhanced FAO and mitochondrial metabolism [29], the anti-oxidation effect of BHB might be an evolutionary consequence to counterbalance the promoted oxidative stress in response to suppressed glycometabolism.

In summary, above findings reveal that BHB regulates fasting-adaptive metabolism through its HDACi effect, which is characterized by the limitation of lipolysis and a shift from an anabolic to a more catabolic state (FAO). The evolutionary purpose of BHB-mediated HDAC inhibition could be aimed at reprogramming the transcription to match a metabolic status that favors animals to overcome famine.

2.5. Glucose-mediated histone acetylation promotes lipogenesis

Fasting follows by re-feeding, which causes a rapid increase in blood glucose levels and a decrease in ketone bodies (Figure 1F). Altered metabolic status is expected to shift away from energy mobilization (FAO) to storage (lipogenesis), which could be accompanied by global epigenetic reprogramming. To test this, we employed in vitro models to examine whether adipogenic induction is accompanied with and controlled by histone acetylation. When 3T3-L1 pre-adipocytes were subjected to a classical adipogenic induction (dexamethasone, isobutylmethylxanthine, and insulin, hereafter referred to as MDI), histone acetylation at multiple sites on H3 increased as early as 3 h and remained elevated until the end of the experiment (day 6) (Fig. S3A). By re-analyzing the publicly available ChIP-seq dataset (GSE20752), we observed that H3K27ac signals were significantly enriched at the genetic loci of proadipogenic transcription factors (TFs) (Cebpa, Pparg) and mature adipocyte markers (Fabp4) in both 3T3-L1 and human adipose stromal cells (hASC) (Fig. S3B and S3C). We observed that glucose, but not other energy sources, such as glutamate, fatty acids, or BHB, supported MDI-induced H3K27ac and adipogenesis (Fig. S3D–S3F). To unveil a causal role of histone acetylation in adipogenesis, 3T3-L1 pre-adipocytes were treated with A485, a selective and potent inhibitor of the histone acetyltransferases CBP and p300 [30]. MDI-induced H3K27ac, lipogenic transcriptional program, and adipogenesis were suppressed by A485 treatment (Fig. S3G–S3I). These effects were readily reversed by MERCK60, which recovered histone acetylation (Fig. S3G–S3I). While BHB cannot substitute for glucose to sustain adipogenesis, it reversed A485-inhibited adipogenesis and H3K27ac in a high-glucose medium, reinforcing its role as an HDACi that regulates energy homeostasis (Fig. S3J and S3K). These findings revealed that glucose contributes to energy storage in a histone acetylation-dependent manner in the in vitro adipogenic models.

The liver is the major site for de novo lipogenesis during energy-excess conditions. To examine epigenetic changes in response to the transition from fasting to re-feeding, we performed H3K27ac and BRD4 ChIP-seq assays, which revealed a remarkable genomic redistribution of these marks upon re-feeding. Although re-feeding causes a sharply decrease in bulk H3K27ac amounts (Fig. S3L), H3K27ac occupancy at some genomic loci markedly increased. A total of 7014 differential H3K27ac peaks (log₂FC > 1, P < 0.01) have been identified, with 2517 peaks up and 4497 down (Figure 5A). Notably, genome-wide BRD4 enrichment exhibited a pattern highly similar to that of H3K27ac, consistent with the fasting model (Figure 5B). Functional enrichment analysis revealed that genomic loci showing reduced H3K27ac and BRD4 signals upon re-feeding were significantly associated with lipid catabolic processes (Fig. S3M and S3N). In contrast, genes linked to increased H3K27ac and BRD4 signals were predominantly enriched in lipid and steroid anabolic processes, such as Srebf1, Pcsk9, Fasn, Hmgcr, Acaca, and Gck (Figure 5C and Fig. S3O). Meanwhile, genes associated with decreased H3K27ac and BRD4 signals were mainly involved in fatty acid catabolic and gluconeogenic processes, such as Pck1 and Cyp4a14 (Figure 5D). Consistently, qPCR analysis revealed that re-feeding upregulated the expression of lipogenic genes (Figure 5E). Thus, these data suggest that re-feeding-induced redistribution of genomic H3K27ac and BRD4 might support an energy-storage gene program, which is exactly contrary to fasting-induced epigenetic changes. Supported by this notion, JQ35 treatment suppressed the expression of energy-storage-associated genes, such as Fasn, Hmgcr, Pcsk9, Gck and Dgat1, but not catabolism-associated genes, such as Cpt1a and Cyp4a14 (Figure 5F).

Figure 5.

Glucose-mediated histone acetylation promotes lipogenesis. (A-D) ChIP-seq data showed that re-feeding induced a significant redistribution of H3K27ac and BRD4 in the whole genome (A and B) with an increased occupancy at the genomic loci of lipids synthesis genes (Srebf1, Pcsk9) (C) but a decreased occupancy at those of fatty acids catabolic (Cyp4a14) and gluconeogenic genes (Pck1) (D). (E and F) qPCR assay showed that re-feeding induced the expression of glucose utilization (Gck) and lipid synthesis genes (Pcsk9, Hmgcr, Dgat1, Srebf1) in the liver (E), while the expression was suppressed by JQ35 treatment (F) (n = 4). (G and H) Brd4-flox or Brd4-hKO mice housed in a thermoneutral environment (30°C) feeding with high-fat diet. Hepatic Brd4 knockout suppressed body weight gain (G) without influence food intake (H) (n = 5–6). (I-L) ITT assays (I) , liver weight (J), the liver gross appearance, HE and Oil Red O staining (K), and TG levels (L) of Brd4-flox or Brd4-hKO mice that were subjected to HFD feeding with housing at 30°C (n = 5–6). (M) qPCR assay of lipids anabolism- (Cd36, Pparg, Dgat1) and VLDLs secretion/metabolism- (Mttp, Apoc3) associated genes in the liver of Brd4-flox or Brd4-hKO mice (n = 5–6). (N) The serum ALT levels were lower in Brd4-hKO than that of control mice (n = 5–6). (O-Q) WAT weight (O) and serum TG (P) and FFA (Q) levels of Brd4-flox or Brd4-hKO mice under HFD feeding (n = 5–6). Data are shown as the mean ± SEM. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001 by unpaired two-tailed Student’s t-test (E, F, H, J, and L-Q) or two-way ANOVA followed with Bonferroni’s multiple comparisons test (G and I).

Non-alcoholic fatty liver disease (NAFLD) reflects an imbalance in energy metabolism, including an increase in lipogenesis, a decrease in FAO, or a reduction in triglyceride export as very low-density lipoproteins (VLDLs) [31]. Housing mice in a thermoneutral environment (30 °C) with a high-fat diet leads to an accelerated NAFLD phenotype [32]. Under this condition, Brd4-hKO mice have decreased body weight gain (Figure 5G). Despite this difference, they appeared healthy, with normal behavior and a comparable amount of food intake to that of control mice (Figure 5H). Since Brd4-hKO mice were intolerant to prolonged fasting, we performed an insulin tolerance test to assess their metabolic health. Brd4 knockout slightly promoted insulin sensitivity, suggesting an improved metabolic consequence (Figure 5I). Control mice displayed a fatty liver phenotype at gross and histological levels, but it was significantly attenuated in Brd4-hKO mice, as quantified by liver weight, HE and Oil Red O staining, and TG levels (Figure 5J-L). The expression of lipogenic gene was lower in Brd4-hKO mice (Figure 5M). Surprisingly, Brd4-hKO mice had lower serum alanine aminotransferase (ALT) levels than control mice, suggesting a mitigated, rather than deteriorated, liver injury (Figure 5N). Thus, these data suggest that hepatic Brd4 knockout protects against NAFLD pathogenesis. The liver synthesizes triglycerides in response to overfeeding and transports them as VLDLs to adipose tissue for storage [31]. We observed that hepatic Brd4 knockout caused a decrease in adiposity, as evidenced by the lower body and WAT weight ((Figure 5G,O). In addition, Brd4-hKO mice had lower serum triglyceride levels, leaving serum FFA levels unaltered (Figure 5P,Q). Correspondingly, the mRNA levels of Mttp and Apoc3 genes involved in VLDL assembly, secretion, and clearance [31] were significantly downregulated by Brd4 knockout (Figure 5M). Taken together, these results suggest that BRD4 is involved in both BHB-mediated energy catabolism during fasting and glucose-driven anabolic programs during overfeeding.

2.6. ACSS2 is required for in vitro and in vivo lipogenesis

Nucleocytosolic acetyl-CoA, a substrate for histone lysine modification, is primarily produced by ACLY from citrate and ACSS2 from acetate [4,5,33]. To elucidate which enzyme is responsible for glucose-mediated lipogenesis and histone acetylation, we first examined the inhibition of ACLY and ACSS2 in an in vitro adipogenic model. Chemical inhibition of both ACSS2 and ACLY reduced H3K27ac, but the effect was more pronounced with ACSS2 inhibition, and only ACSS2 inhibition suppressed MDI-induced adipogenesis (Figure 6A,B). Similarly, RNA interference of Acss2, but not Acly, strongly suppressed H3K27ac and adipogenesis (Figure 6C,D). To explore the relevance of these findings in primary cells, stromal vascular fractions (SVF) were isolated from Acss2^fl/fl subcutaneous adipose depots, and the cells were infected with Ad-GFP or Ad-Cre, followed by adipogenic induction. Consistent with our observations in 3T3-L1, Acss2 knockdown reduced TG accumulation and H3K27ac levels (Figure 6E,F). To examine the effects of ACSS2 versus ACLY on in vivo lipogenic gene expression, mice were fed with fructose:glucose water [34], which induced a rapid increase in hepatic glycolytic and lipogenic gene expression. Unexpectedly, these gene’s expression was suppressed by both ACSS2 and ACLY inhibition (Figure 6G,H).

Figure 6.

ACSS2 is required for in vitro and in vivo lipogenesis. (A and B) 3T3-L1 preadipocytes were pretreated with ACLY (5 μM) or ACSS2 (5 μM) inhibitors, followed by adipogenic induction. Oil Red O staining (day 8 after adipogenic induction) (A, n = 3) and western blotting assay (day 1 after adipogenic induction) (B, n = 3) were performed at indicated times. (C and D) 3T3-L1 preadipocytes (at the cell confluence of 50–60%) were transfected with Acly or Acss2 siRNA. Two days after the cells reaching at 100% confluence, they were subjected to adipogenic induction. Oil Red O staining (day 8 after adipogenic induction) (C, n = 3) and western blotting assays (day 1 after adipogenic induction) (D, n = 3) were performed at indicated times. (E and F) Stromal vascular fractions (SVFs) were isolated from Acss2^fl/fl subcutaneous adipose depots and the cells were infected with Ad-GFP or Ad-Cre, followed by adipogenic induction. Oil Red O staining (day 8 after adipogenic induction) (E, n = 3) and western blotting assays (day 2 after adipogenic induction) (F, n = 3) were performed at indicated times. (G and H) qPCR assay of glucose metabolism (G6pd, Me1, Pklr) (G) and fatty acid synthesis genes (Fasn, Chrebpb, Acaca) (H) in the liver of mice fed with fructose and glucose water and treated with ACLYi (50 mg/kg, i.p., bid), ACSS2i (40 mg/kg, i.p., bid), or both inhibitors (n = 4–5). (I) Western blotting assay of ACSS2 in the liver of Acss2-flox or Acss2-hKO mice (upper panel); the quantification of the bands (lower panel) (n = 4). (J) GO enrichment analysis of down-regulated genes in the liver of Acss2-hKO versus Acss2-flox mice subjected to 24 h fructose and glucose water exposure. (K) qPCR assay of Acss2 and fatty acid synthesis genes (Acaca, Chrebpb, Fasn) in the liver of Acss2 flox and Acss2-hKO mice subjected to 24 h of fructose and glucose water exposure (n = 4). (L) ChIP was performed using anti-H3K27ac antibody or control IgG, and DNA was amplified with specific primers to detect the target genes (Pklr and Mlxipl) (n = 4). (M) The ratio of liver to body weight (left) and the liver TG levels (right) of Acss2-flox or Acss2-hKO mice subjected to 24 h of fructose and glucose water exposure (n = 4). Data are shown as the mean ± SEM. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001 by unpaired two-tailed Student’s t-test (I, K, and M) or one-way ANOVA followed with Bonferroni’s multiple comparisons test (G, H, and L).

To specifically delete Acss2 in the liver, Acss2^fl/fl mice were crossed with Alb-CreERT2 mice. Acss2 expression was significantly abolished at mRNA and protein levels (Figure 6I,K). Bulk RNA sequencing data showed that Acss2 knockout-suppressed genes were significantly enriched to lipid biosynthetic, steroid metabolic, and fatty acid metabolic processes when the mice were subjected to fructose:glucose consumption (Figure 6J). The key lipogenic genes, such as Acaca, Chrebpb, and Fasn, decreased in Acss2-hKO mice (Figure 6J,K). As expected, H3K27ac enrichment at the genomic loci of Pklr and Mlxipl was suppressed by Acss2 knockout (Figure 6L). Consistent with the transcriptional suppression of lipogenic genes, Acss2-hKO mice fed with fructose:glucose water displayed lower liver weight and milder fatty liver appearance with reduced hepatic TG content than the control mice (Figure 6M). Thus, in vitro and in vivo data indicate that ACSS2-mediated histone acetylation plays a key role in regulating energy storage.

2.7. Hepatic Acss2 knockout attenuates NAFLD pathogenesis

Given the crucial role of ACSS2 in hepatic TG synthesis, we sought to examine whether Acss2 knockout alleviates NAFLD pathology and improves metabolic health. Although Acss2 deficiency displayed no distinct phenotype compared to WT mice when subjected to a Western diet (high fat [42% kcal], high sucrose [30% kcal] diet with 0.2% cholesterol) [35]. The Acss2-hKO mice gained significantly less body weight than control mice (Figure 7A) in the aforementioned NAFLD model. Both genotypes appeared healthy and had comparable food consumption in this condition (Figure 7B). To assess metabolic health, we performed insulin and glucose tolerance tests and found that Acss2 knockout increased insulin sensitivity without improving glucose tolerance (Figure 7C–F). Blood glucose levels at 60 min were higher than those at 30 min in Acss2-hKO mice, suggesting a potential impairment in hepatic glucose metabolism. This phenotype aligned with the reduced hepatic lipogenesis associated with glucose (Figure 6).

Figure 7.

Hepatic Acss2 knockout attenuates NAFLD pathogenesis. (A and B) Acss2-flox or Acss2-hKO mice were subjected to HFD feeding with housing at thermoneutral environment (30°C) at indicated times. Hepatic Acss2 knockout reduced body weight gain (A) without influence food intake (B) (n = 5–6). (C-F) ITT (C and D) and GTT (E and F) assays of Acss2-flox and Acss2-hKO mice (n = 5–6). (G-I) The ratio of liver to body weight (G), the liver gross appearance, HE and Oil Red O staining (H), and TG levels (I) of Acss2-flox and Acss2-hKO mice that were subjected to HFD feeding with housing at 30°C (n = 5–6). (J) The serum ALT levels of Acss2-flox and Acss2-hKO mice (n = 5–6). (K) qPCR assay of lipids anabolism- (Cd36, Pparg, Dgat1) and VLDLs secretion/metabolism- (Mttp, Apoc3) associated genes in the liver of Acss2-flox and Acss2-hKO mice subjected to HFD feeding with housing at 30°C (n = 5–6). (L-N) Serum TG (L) and FFA (M) levels and the ratio of WAT to body weight (N) of Acss2-flox and Acss2-hKO mice (n = 5–6). Data are shown as the mean ± SEM. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001 by unpaired two-tailed Student’s t-test (B, D, F, G, and I-N) or two-way ANOVA followed with Bonferroni’s multiple comparisons test (A, C, and E).

Acss2-hKO mice had lower liver weight and noticeably mitigated fatty liver appearance at both gross and histological levels compared to control mice (Figure 7G,H). Biochemical analysis revealed a significant decrease in hepatic TG content in the Acss2-hKO mice (Figure 7I). Acss2 knockout reduced serum ALT levels, suggesting mitigation of liver damage (Figure 7J). qPCR assay revealed that Acss2-hKO mice have lower expression of lipogenic genes in the liver (Figure 7K). Consistent with the changed transcriptional program, we observed that serum TG, FFA levels, and the weight of WAT were reduced in Acss2-hKO mice (Figure 7L-N), which might have resulted from decreased hepatic TG synthesis and export, similar to what was observed in Brd4-hKO mice. Although Brd4 knockout improves NAFLD pathogenesis, mice deficient in Brd4 are fasting-intolerant, hampering it as an optimal therapeutic target in the liver. However, Acss2-hKO mice are well-tolerant to fasting and re-feeding (Fig. S4A–S4D). These findings suggest that ACSS2 may represent a promising target for treating liver diseases, like NAFLD.

3. Discussion

This study demonstrated that histone acetylation functions as a metabolic tag to regulate energy storage and mobilization in alignment with recurrent overfeeding and fasting in natural settings. In the presence of excess dietary carbohydrates, glucose-derived acetyl-CoA fuels “active” histone acetylation that facilitates energy storage-associated gene expression. On the contrary, in a state of carbohydrate deficiency, the liver generates ketone bodies that function as endogenous HDAC inhibitors to sustain “passive” histone acetylation, contributing to energy mobilization-associated gene expression. Both “active” and “passive” histone acetylation were recognized by BRD4 to regulate transcription.

3.1. BHB-mediated histone acetylation epigenetically regulates systemic fasting adaptation

Ketone bodies function as a surrogate for glucose in the carbohydrate-deficient state, as they are catabolized in the mitochondria of extrahepatic tissues, like the brain, to acetyl-CoA that fuels the TCA cycle for ATP generation [1]. The classic idea considered that glucose, but not BHB, supports nucleocytosolic acetyl-CoA synthesis for histone acetylation, although the viewpoint was challenged by some new findings [10]. Recently, an elegant study performed in the yeast model reported that histone acetylation marks redistribute from growth-promoting genes to gluconeogenic and fat metabolism genes in response to glucose deprivation, which is essential for yeast survival [9]. Here, we reported a similar regulatory mechanism applicable to mammals that animals have evolved to sense BHB as an HDAC inhibitor to promote histone acetylation and regulate systemic metabolism.

BHB-mediated histone acetylation was enriched at specific genomic loci that sustained FAO- and gluconeogenesis-associated gene expression. This epigenetic reprogramming is of physiological importance, as inhibition of BRD4, a “reader” of the histone acetylation code [36], causes metabolic disturbances, such as hepatic steatosis and cold intolerance. Additionally, it may have pathological relevance, as metabolic switching from FAO to glycolysis has been widely implicated in various pathological processes, such as inflammation and cancer biology. BHB-promoted FAO, through intermittent fasting, has been shown to antagonize metabolic inflammation and cancer in animal and human studies [37,38]. Beyond its role as an endogenous HDAC inhibitor promoting histone acetylation, BHB can also serve as a metabolic substrate for histone lysine β-hydroxybutyrylation (Kbhb), representing an alternative epigenetic regulation during fasting [39]. Therefore, BHB might regulate chromatin through two mechanistically distinct, but potentially coexisting manners to support fasting adaptation (Fig. S4E). In addition, while BHB-promoted histone acetylation was validated in previous studies [11] and in our model, opposing findings have also been reported [39,40], highlighting that its functions are potentially context- and cell type-dependent. Moreover, FAO- and ketolysis-generated acetyl-CoA might also support the nucleocytosolic acetyl-CoA pool through an acetyl-carnitine shuttle [10] in the fasting model, which is outside of the scope of our current study, and will be further explored.

BHB has been shown to protect the kidney and brain against oxidative damage [11]. In the present study, BHB-mediated histone acetylation was enriched at the genomic loci of some classical antioxidative genes, such as Cat and Gpx3. Glucose is a natural antioxidant that functions through multiple mechanisms, such as NADPH production via the pentose phosphate pathway. In physiological conditions, increased blood ketone body levels usually imply a low glucose level. Thus, BHB-mediated histone acetylation promotes antioxidative gene expression, as it counteracts oxidative stress caused by the absence of glucose. These findings suggest that glucose and BHB function as “antagonists” in their roles to regulate energy flow (energy storage versus mobilization), but as “analogs” in their biological effects such as ATP generation, histone acetylation, and antioxidation (Fig. S4E).

3.2. Glucose-mediated histone acetylation epigenetically regulates systemic overfeeding adaptation

In an energy-sufficient state, glucose-derived acetyl-CoA fuels ATP production in the mitochondria and supports histone acetylation and fatty acid synthesis in the nucleocytosolic chambers. We have previously shown that BRD4 cooperates with master TF networks to regulate adipogenesis [41]. In this study, we observed that histone acetylation plays a causal role in contributing to DNL in the in vitro and in vivo models. Fasting, followed by refeeding, causes a rapid transition from FAO to lipogenesis, which is accompanied by genomic redistribution of H3K27ac and BRD4 in the liver chromatin. In sharp contrast to the phenotype observed in fasted Brd4-hKO mice, hepatic Brd4 knockout ameliorated NAFLD pathogenesis.

Nucleocytosolic acetyl-CoA is primarily produced by ACLY from citrate and ACSS2 from acetate [4,5,33]. We observed that ACSS2, but not ACLY, regulated adipogenesis-induced histone acetylation and lipid accumulation in 3T3-L1 cells, which differs from a previous report [42]. In the fructose:glucose consumption model, genetic deletion of hepatic Acss2 resulted in downregulated lipogenic gene expression and reduced TG accumulation, suggesting a crucial role of ACSS2 in modulating both in vitro and in vivo lipogenesis. We observed that Acss2 knockout attenuates NAFLD pathogenesis that mimics the phenotype observed in Brd4-hKO mice. However, unlike Brd4 knockout mice, deletion of Acss2 in the liver did not impair fasting adaptation when the mice were subjected to repeated fasting/re-feeding cycles, suggesting that ACSS2 may be a relatively safe target for the treatment of liver diseases.

In summary, this study highlights histone acetylation as a central hub linking nutrient availability to transcriptional output. These findings not only advance our understanding of feast-famine biology but also suggest potential strategies for targeting the metabolic–epigenetic axis in obesity, NAFLD, and cancer.

4. Materials and methods

4.1. Cell lines and reagents

All cell lines were grown at 37 °C with 5% CO₂. 3T3-L1 were cultured in DMEM (ATCC 30–2002) with 10% fetal bovine serum (FBS, Gibco 10099141) and 100 U/mL penicillin, 100 μg/mL streptomycin. C3H10T1/2 were cultured in BME (Gibco 21010046) with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin. HK2 cells were cultured in DMEM (BI 010521ACS) with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin. Caco-2 cells were cultured in MEM (Procell, PM15041) with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin. Key reagents and chemicals used in this study are listed in Supplementary Table S1.

4.2. Animals

All animal protocols were conducted according to the National Institutes of Health Guide for the Care and Use of Laboratory Animals, and experimental protocols were approved by the animal care and use committee of Nanchang University (Approval number: CDYFY-IACUC-202208QR010). Eight-week-old C57BL/6J wild-type male mice were purchased from Gempharmatech Co, Ltd (Nanjing, Jiangsu, China). Before starting the experiment, all the mice were acclimatized for a week. All experimental mice were housed under controlled conditions with temperature of 22–24 °C and humidity around 50% with a 12:12 light dark cycle. Food and water were available ad libitum for all animals. To establish a NAFLD model, twelve-week-old male mice were housed at thermoneutral environment (30 °C) with 60% kcal high-fat diet (D12492, Research Diets) feeding, which leads to an accelerated NAFLD phenotype [32]. To test the acute hepatic lipogenesis, mice were provided with 15% (w/v) fructose:15% (w/v) glucose drinking water for 24 h, and orally gavaged with 2.0 g kg⁻¹ fructose plus 2.0 g kg⁻¹ glucose 1 h before being euthanized, as described in previous study [34]. To keep the sugar-containing water fresh enough, the drinking water was replaced every day.

4.3. Generation of Brd4 and Acss2 hepatic KO mice

Brd4^f/f mice were established as we previously reported [43]. Acss2^f/f mice were engineered with a CRISPR/Cas9 system in C57BL/6J genetic background. A donor vector containing flox sites flanking exon 5–7 was generated and introduced into recipient mice. To generate hepatic specifically knockout mice, Brd4^f/f and Acss2^f/f mice were crossed with Alb-CreERT2 (+) mice (Cyagen, I001003), respectively, to generate Brd4 or Acss2 floxed, Alb-CreERT2 mice. Tamoxifen administration followed by the protocol provided by Jackson Lab. Briefly, tamoxifen (75 mg/kg body weight, dissolved in corn oil) was injected intraperitoneally once every 24 h for a total of 5 consecutive days. After the final injection, the mice were observed for another 7 days before the further treatment.

4.4. Dietary intervention

To test the effects of fasting biology, eight-week-old male mice were deprived of food for the indicated times with ad libitum access to water. Canagliflozin has been shown to induce a fasting-like response and ameliorate hepatic steatosis [19]. To test the role of JQ35 on canagliflozin-mediated effects, 180 mg of canagliflozin (MERCK) was mixed in 1 kg of the high-fat diet, which translated to approximately 30 mg/kg/day of the drug. The mice were fed with the drug-containing diet for 3 days, followed by a single dose of JQ35 treatment. The ketogenic and control diets were purchased from Jiangsu Xietong Pharmaceutical Bio-engineering Co., Ltd. The mice were fed with the ketogenic and control diets, as indicated in the text.

4.5. In vivo suppression of BRD4 by JQ35

JQ35 is a newly developed BRD4 inhibitor in our lab. JQ35 is a water-soluble and blood–brain barrier-impermeable chemical molecule. Eight-week-old male mice were intraperitoneally injected with JQ35. The concentration and frequency of injection were determined according to different experimental purposes, as indicated in the text.

4.6. Cold exposure

Cold exposure assay was performed as reported previously [44]. Briefly, eight-week-old mice were housed individually in cages at 4 °C in the presence or absence of different foods. Core body temperature was monitored hourly. To avoid the impact on fasting adaptive metabolism, a single dose of JQ35 (50 mg/kg) was administered 1 h before cold exposure. At the end of the experiment, fat tissues were collected for further analysis. To test the role of different nutrients on JQ35-suppressed thermogenesis during the fasting conditions, isocaloric and isovolumetric amounts of glucose, BHB, or intralipid (8 kcal/kg of the nutrients in 125 μL volume) were administered intraperitoneally immediately before cold exposure.

4.7. FGF21 administration

A recombinant FGF21 rescue experiment was performed as described previously [26]. Briefly, eight-week-old Brd4-hKO male mice were fasted at room temperature, which causes hypothermia, and FGF21 was injected twice a day (10 ng/mouse, retro-orbital injection). Blood glucose level and rectal temperature were monitored, and the mice were sacrificed at 52 h after starvation.

4.8. Glucose tolerance test (GTT) and insulin tolerance test (ITT)

Mice were fasted overnight (GTT) or 4 h (ITT) followed by intraperitoneally injected with glucose (1 g kg⁻¹, GTT) or insulin (0.75 U/kg, ITT), and the levels of tail vein blood glucose were monitored at indicated time points.

4.9. Serum parameter measurement

Biochemical parameters, including serum and hepatic triglycerides (AKFA003M, BOXBIO), FFA (E-BC-K792-M, Elabscience), and ALT (E-BC-K235-M96T, Elabscience) were measured using commercial kits according to the manufacturer’s instructions.

4.10. Lipolysis analysis of murine adipose tissue and adipocytes

Ex vivo lipolysis of murine adipose tissue was performed as described previously [45]. Concisely, WAT was dissected and cut into small pieces, placed in a 24-well plate, and pretreated with Krebs–Ringer Bicarbonate buffer (KRBH) FFA-free BSA buffer containing JQ35 (1 μM), BHB (20 mM), or Merck60 (1 μM) for 4 h, followed by ISO (10 μM) treatment for 2 h at 37 °C incubator. Next, the plate was processed at 65 °C for 10 min to terminate the reaction, and the levels of FFA in the medium were measured by the commercial kit (E-BC-K792-M, Elabscience). To determine lipolysis in 3T3-L1 and C3H10T1/2 adipocytes, the cells were pretreated with JQ35 (1 μM), BHB (20 mM), and MERCK60 (1 μM) for 4 h, followed by treatment with ISO (10 μM) for 3 h. Subsequently, the levels of FFA in the medium were measured using the commercial kit (E-BC-K792-M, Elabscience).

4.11. Histology

For hematoxylin and eosin staining (H&E staining), the livers were fixed in a 4% paraformaldehyde solution and embedded in paraffin. The tissues were cut into 5 μm-thick slices and stained with eosin for 3 min and hematoxylin for 25 s. Afterward, they were washed three times with phosphate-buffered saline (PBS) for 5 min each and photographed under an optical microscope. For Oil Red O staining, tissues were fixed in a 4% paraformaldehyde solution overnight, dehydrated by titrating in sucrose (10%, 20%, and 30%), and embedded in OCT by freezing on dry ice. The tissues were cut into 5 μm-thick slices using a freezing microtome. Slices were rewarmed at room temperature for 10 min and stained with Oil Red O for 20 min. After washing with PBS, the sections were stained with hematoxylin for 20 s and observed under a light microscope.

4.12. Isolation of stromal vascular fractions (SVFs) from subcutaneous adipose tissue

Stromal vascular fractions of adipose tissue were isolated according to the published protocol [46] with slight adjustments. Briefly, type II collagenase (Sigma, C2-BIOC) was dissolved in DMEM at 2 mg/mL. Later, the collagenase-containing fluid was filtered using a 0.2 μm sterile filter and pre-warmed at 37 °C. The freshly isolated subcutaneous adipose tissue was washed in sterile PBS to remove blood, and the tissues were cut into 0.5–1 mm pieces in a petri dish, suspended in collagenase-containing fluid, and shaken quickly in a 37 °C water bath for approximately 10 min until no obvious tissue particles were observed. Thereafter, the same volume of PBS containing 2% FBS was added to terminate the digestion. Undigested tissues were filtered using a 100 μm cell filter, followed by centrifugation at 800 g for 5 min, and the supernatant was discarded. Finally, 5 mL PBS was added to wash the cells, then centrifuged at 800 g for 5 min. The supernatant was discarded, and the cell pellet was resuspended for the following adipogenic induction.

4.13. Adipogenic differentiation assays

3T3-L1 pre-adipocytes were cultured in a DMEM medium containing 10% FBS. Two days after confluence, cells were induced to differentiate into adipocytes with adipogenic induction medium (M2) containing 10% FBS, 3-isobutyl-1-methylxanthine (IBMX, 500 μM), dexamethasone (1 μM), and insulin (0.5 μg/mL) for 48 h. Next, M2 was replaced by a maintenance medium (M3) containing 10% FBS and insulin (0.5 μg/mL). The M3 medium was changed every 3 days for 8–10 days until the pre-adipocytes differentiated into mature adipocytes. C3H10T1/2 was cultured in BME (Gibco 21010046) with 10% FBS. Two days after confluence, cells were induced to differentiate into adipocytes with induction medium (M2) containing 10% FBS, IBMX (500 μM), dexamethasone (1 μM), indomethacin (0.125 nM), and insulin (20 nM) for 48 h. Then, the differentiation induction medium was replaced with a maintenance medium (M3) containing 10% FBS, and insulin (20 nM). Following this, M3 was changed every 3 days for 8–10 days. SVFs were cultured in a DMEM medium containing 10% FBS. Two days after confluence, cells were induced to differentiate with adipogenic induction medium (M2) containing 10% FBS, insulin (5 μg/mL), indomethacin (125 μM), dexamethasone (2 μg/mL), IBMX (0.5 mM), and rosiglitazone (0.5 μM) for 48 h. Then, M2 was replaced by a maintenance medium (M3) containing 10% FBS, insulin (5 μg/mL). Subsequently, M3 was changed every 3 days for 8–10 days until mature adipocytes were induced.

4.14. Oil Red O staining of the cultured adipoctyes

After the mature adipocytes were successfully induced, the cell culture medium was removed, washed two times with PBS, and fixed with paraformaldehyde (4%) for 20 min. The fixative was removed, and the cells were washed two times with distilled water. Thereafter, the cells were washed with 60% isopropanol for 5 min, and the Oil Red O working solution was added to cover the cells for 15 min. The staining solution was discarded, and the cells were rinsed with distilled water 2–5 times. Then, it was stained with hematoxylin for 30 s. Finally, hematoxylin was discarded, and cells were rinsed again with distilled water 2–5 times. PBS was added to cover the cells, and they were observed under an optical microscope.

4.15. siRNA transfection

All siRNA transfections were performed using Lipofectamine® RNAiMAX (Invitrogen) according to the manufacturer’s manuals. siRNAs targeting Acss2 (GGATTGATGACATGCTCAA) or Acly (GGATGACATTTCCTATGTT) were purchased from Ribobio (Guangzhou, China).

4.16. Adenovirus infection

When the cultured SVF reached 50–60% confluence, the original medium of the cells was removed and replaced with half the volume of fresh medium, and then adenovirus was added and gently mixed. Four hours after infection, the normal culture volume of the medium was supplemented. At 10–16 h post-infection, the culture medium containing the virus was removed, the fresh complete culture medium was replaced. The infection efficiency was assessed by the expression of GFP, which was observed using a fluorescence microscope 48 h after infection.

4.17. Chromatin immunoprecipitation (ChIP) assay and sequencing

The mice were fasted or refed at the indicated time points. For ChIP-seq, tissues from 5 mice per experimental group were pooled to generate a SINGLE composite sample to minimize individual biological variability [47]. ChIP was performed using a SimpleChIP® Plus Enzymatic Chromatin IP Kit (9005; Cell Signaling Technology) with antibodies against H3K27ac (5 μg/ChIP, activemotif, 91193), BRD4 (5 μg per ChIP, Cell Signaling Technology, 83375S), or normal rabbit IgG as control according to the manufacturer’s procedures. DNA quality was assessed by an Agilent 2100 Bioanalyzer and Qubit. Only samples passing the commercial technical quality control standards were subjected to library construction and sequencing. Before sequencing, ChIP-qPCR were performed on representative gene loci (e.g., Pck1, Cyp4a14) to verify that the immunoprecipitation (IP) enrichment was successful (enrichment > IgG) and that fasting or refeeding effectively changed H3K27ac or BRD4 enrichment. Immunoprecipitated DNA was processed for sequencing by Novogene (Beijing, China). Briefly, the ChIP-Seq libraries were prepared by end-repairing and A-tailing DNA fragments, followed by adapter ligation, size selection, and PCR amplification. Library quality was assessed using the Agilent 5400 system and qPCR before pooling and sequencing on Illumina platforms (PE150). For bioinformatics analysis, raw sequencing data in FASTQ format underwent quality control using fastp, where reads with adapter contamination, high N-content (>10%), or excessive low-quality bases (Phred score <5 in >50% of bases) were removed. Clean reads were then mapped to the mouse reference genome (GRCm38/mm10) using BWA, with RefSeq gene annotations guiding downstream analyses.

4.18. Quantitative analysis of peak and gene overlap

The overlap between the annotated gene lists for differential BRD4 and H3K27ac peaks was quantified using custom R scripts. Gene lists were first filtered for uniqueness, and the number of common genes was determined using the base R intersect function. The statistical significance of the overlap was assessed using Fisher’s exact test. A 2x2 contingency table was constructed for each comparison, with the total number of protein-coding genes in the mouse genome (approximately 22,000) used as the background. A P value < 0.05 was considered statistically significant. Venn diagrams were generated using the VennDiagram package in R.

4.19. ChIP-qPCR assay

Liver samples from Acss2-flox and Acss2-hKO mice subjected to 24 h of fructose and glucose water exposure (n = 4) were harvested for ChIP assay using a SimpleChIP® Plus Enzymatic Chromatin IP Kit (9005; Cell Signaling Technology) with antibodies against H3K27ac (5 μg/ChIP, activemotif, 91193), or the control normal rabbit IgG according to the manufacturer’s procedures. Following immunoprecipitation, both the pulled-down DNA and input DNA were analyzed by qPCR using primers (Table S2) targeting the specified genomic sequences.

4.20. CUT&Tag-seq assay & data analysis

The mice were subjected to normal feeding or 24-h fasting, and the BAT samples from five mice in each group were pooled to generate one composite sample per condition. The CUT&Tag assay was performed following the protocol provided by the manufacturer (TD903-01; Vazyme). A total of 200,000 nuclei isolated from BAT (n = 5 per group) were pooled for the assay. For chromatin profiling, 2 μg of H3K27ac antibody (91193, ActiveMotif) was used. Libraries were amplified (16 cycles), validated, and sequenced (Illumina HiSeq/Novaseq, 2 × 150 PE). Raw reads were quality-filtered (Q20/Q30/GC content) and aligned to the mouse reference genome (GRCm38/mm10) using BWA. Peaks were called (MACS2, q < 0.05), normalized (CPM), and visualized (IGV). Differential H3K27ac signals at FAO-related genes (e.g., Cd36, Acot1, Acadl) were analyzed using the R packages ChIPseeker (version 1.34.1) and GenomicFeatures (version 1.50.4).

4.21. RNA isolation, qRT-PCR and RNA-seq assays

Total RNA was isolated using TRIzol (Thermo). Then, the RNA was reverse transcribed into cDNA using the RevertAid Master Mix(M1632), fluorescence quantitative PCR was performed using SYBR ® Green Pro Taq HS premix kit II (containing ROX). The primers used for qPCR were listed in Table S3. RNA-seq assay was performed as described in our previous publication [48].

4.22. Western blot assay

Western blotting assay was performed according to the standard protocol. Primary antibodies were used as listed in Table S4. H3K27ac was assessed as a representative marker of active chromatin. Anti-mouse (Abcam, ab6708, 1:5000) and anti-rabbit (Abcam, ab6702, 1:5000) were used as secondary antibodies. Blots were visualized by enhanced chemiluminescence (Thermo Fisher, Invitrogen iBright FL1500, A44241CFR). Band intensities were quantified using ImageJ software and normalized to the corresponding loading control to correct for variations in protein loading. The resulting ratios were expressed relative to the mean of the control group and are presented in arbitrary units.

4.23. Statistical analysis

Data are expressed as mean ± SEM, as indicated. A two-tailed Student’s t-test was used to compare the mean between the two groups, and one-way or two-way ANOVA was used to compare the mean between multiple groups. Pearson correlation analysis was used to quantify the linear relationship between blood BHB levels and body weight, core body temperature, and heart rate. P < 0.05 was considered statistically significant and marked with an asterisk: ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001.

CRediT authorship contribution statement

Linyun Chen: Writing – original draft, Visualization, Data curation. Lingyan Zhu: Supervision, Formal analysis. Huabing Xiao: Data curation. Xiaotao Wang: Investigation. Fan Xia: Validation. Zhichao Wang: Validation. Long Wu: Visualization. Dayu Wu: Validation. Qi Liu: Visualization. Junyun Cheng: Visualization. Jun Qi: Resources, Formal analysis. Qiong Duan: Writing – review & editing, Writing – original draft, Funding acquisition.

Funding

This work was supported by the National Natural Science Foundation of China (Grant No: 81960168, 82460077, 82160155, 81860153) and the Natural Science Foundation of Jiangxi Province (Grant No: 20212ACB206012).

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Dr. Qi is a scientific co-founder of Epiphanes, and consultant for Talus, Inc. All other authors declare no competing interests.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Multimedia component 2

Figure S1. BHB-mediated HDAC inhibition shifts hepatic anabolism toward lipid oxidation (A) Western blotting assay of H3K27ac in the liver of mice with or without 24 h of fasting (upper panel); the quantification of the bands (lower panel) (n = 3). (B) qPCR assay of the genes associated with FAO and ketogenesis in the liver of mice with the indicated treatment (n = 5). (C) Annotation of the genomic distribution of H3K27ac and BRD4 peaks. (D) Venn diagrams illustrating the overlap of target genes associated with differential BRD4 and H3K27ac peaks in the Fasting vs Feeding. (E) The blood glucose and BHB levels of mice with the indicated treatment (n = 5). (F) Experimental design. (G-J) The serum FFA levels (G), ratio of liver to body weight (H), liver triglycerides levels (I), and the body weight loss (J) of mice with the indicated treatment (n = 4). (K and L) The blood glucose (K) and BHB levels (L) of the mice underwent indicated treatment (n = 4). (M) qPCR assay of the genes associated with FAO and ketogenesis in the liver of mice with the indicated treatment (n = 4). (N) Experimental design. (O) The liver triglycerides levels of mice with the indicated treatment (n = 5). (P) Western blotting assay of BRD4 in the liver of Brd4-flox and Brd4-hKO mice (upper panel); the quantification of the bands (lower panel) (n = 4–6). (Q) qPCR assay of the genes associated with FAO and ketogenesis in the liver of Brd4-flox and Brd4 hKO mice with 24 h of fasting (n = 5). (R and S) The serum FFA levels (R) and body weight loss, blood glucose and BHB levels (S) of Brd4 flox or Brd4-hKO mice fed with ketogenic diet (n = 4–5). (T) qPCR assay of the genes associated with FAO and ketogenesis in the liver of Brd4-flox or Brd4-hKO mice fed with ketogenic diet (n = 4–5). (U-W) The ratio of liver to body weight (U), liver triglycerides (V) and FFA (W) levels of Brd4-flox or Brd4-hKO mice fed with ketogenic diet (n = 4-5). Data are shown as the mean ± SEM. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001 by unpaired two-tailed Student’s t-test (P-W) or one-way ANOVA followed with Bonferroni’s multiple comparisons test (A-B, E,G-M, and O).

Multimedia component 3

Figure S2. BHB-mediated HDAC inhibition sustains BAT fasting biology (A-C) The core body temperature of mice with the indicated treatment (n = 5–6). (D) Hepatic Brd4 knockout compromised thermogenesis for the mice subjected to fasting at room temperature, and that was rescued by FGF21 administration; and the blood glucose and BHB levels of mice underwent the indicated treatment (10 ng/mouse, n = 6). Data are shown as the mean ± SEM. ∗P < 0.05, ∗∗P < 0.01 by one-way (D) or two-way ANOVA (A-C) followed with Bonferroni’s multiple comparisons test.

Multimedia component 4

Figure S3. Glucose-mediated histone acetylation promotes lipogenesis (A) Western blotting assay of the acetylation of histone 3 at multiple lysine residues in the 3T3-L1 preadipocytes subjected to adipogenic differentiation at indicated times (n = 3). (B and C) Gene tracks displayed the occupancy of H3K27ac at the genomic loci of classical adipogenic genes (Cebpa, Pparg, and Fabp4) in 3T3-L1 (2 days after adipogenic induction) (B) or hASC (3 days after adipogenic induction) (C). Visulized using publicly available ChIP-seq datasets (GSE20752), and the analyses were performed on the mm9 (3T3-L1) and hg18 (hACS) reference genome sequences. (D and E) 3T3-L1 preadipocytes were subjected to adipogenic differentiation in the medium supplemented with indicated concentrations of glucose. Oil Red O staining (day 8 after adipogenic induction) (D) and western blotting assays (24 h after adipogenic induction) (E) (n = 3). (F) Oil Red O staining was performed for 3T3-L1 cells subjected to 8 days of adipogenic differentiation in medium containing high glucose or different concentrations of glutamine (Gln), FFA and BHB. (G-I) Oil Red O staining (day 8 after adipogenic induction) (G, n = 3), western blotting assays (day 1 or day 8 after adipogenic induction) (H, n = 3) and qPCR (day 4 after adipogenic induction) (I, n = 6), and of 3T3-L1 cells treated with A485 (1 μM) ± Merck60 (0.5 μM), followed by adipogenic induction. (J and K) Oil Red O staining (day 8 after adipogenic induction) (J, n = 3) and western blotting assay (day 1 after adipogenic induction) (K, n = 3) of 3T3-L1 cells treated with A485 ± BHB, followed by adipogenic induction. (L) Western blotting assay of H3K27ac in the Liver of mice with or without 24 h of fasting (upper panel); the quantification of the bands (lower panel) (n = 3–4). (M and N) GO enrichment analysis of the re-feeding-suppressed peak-associated genes. The reads number of the ChIP-seq data for H3K27ac (M) (log2FC > 1 and P < 0.01) or BRD4 (N) (log2FC > 1 and P < 0.01) were included as the re-feeding-suppressed peak. (O) Gene tracks displayed the occupancy of H3K27ac and BRD4 at the genomic loci of lipids synthesis genes (Fasn, Hmgcr, Acaca) and glucose utilization (Gck). Data are shown as the mean ± SEM. ∗P < 0.05, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001 by unpaired two-tailed Student’s t-test (L) or one-way ANOVA followed with Bonferroni’s multiple comparisons test (I).

Multimedia component 5

Figure S4. Chemical or genetic inhibition of ACSS2 did not impair mouse adaptation to fasting and re-feeding. (A and B) Blood glucose and BHB levels, body weight loss, and core body temperature of the mice treated with PBS or ACSS2 inhibitor. For fasting experiment, mice were treated with PBS or ACSS2 inhibitor every 8 h (40 mg/kg, i.p.), and the blood glucose and BHB levels, body weight loss, and core body temperature were measured after 24 h of fasting (A); for re-feeding experiments, mice were fasted for 23 h and followed by ACSS2 inhibitor administration 1 h before re-feeding. Blood glucose levels, body weight gain, and core body temperature were measured 3 h after re-feeding (B) (n = 4). (C) Blood glucose and BHB levels, body weight loss, and core body temperature of Acss2-flox and Acss2-hKO mice with 24 h of fasting (n = 4). (D) Blood glucose levels, body weight gain, and core body temperature of Acss2-flox and Acss2-hKO mice were measured after 3 h of re-feeding (n = 4). (E) Schematic depicts that glucose and BHB display equivalent or opposite effects. Both glucose and BHB display the similar biological effects as regarding to ATP generation, histone acetylation, and anti oxidation, but opposite effects to fatty acids synthesis versus oxidation. In addition, BHB can serve directly as a substrate to support histone Kbhb. Ac-CoA indicates acetyl-CoA; Kbhb, β hydroxybutyrylation; FAS, fatty acid synthesis; FAO, fatty acid oxidation. Data are shown as the mean ± SEM. ∗P < 0.05 by unpaired two-tailed Student’s t-test (A-D).

Data availability

Data will be made available on request.