The Saccharomyces cerevisiae acetyltransferase Gcn5 exerts antagonistic pleiotropic effects on chronological ageing

Gcn5 and Hda1 play opposing roles in starvation-induced HSP expression

To find the epigenetic factors that are involved in starvation-induced stress response, we screened all the viable HDAC and HAT mutants from the BY4741 deletion library, using the heat shock protein (HSP) reporters, pHSP26-HSP26-RFP/VFP (a member of small HSP) and pSSA3-RFP/VFP (a member of the HSP70 family) [21]. Among the HDAC mutants, only the hda1Δ, hda2Δ and hda3Δ deletants displayed enhanced expression of both reporters (Supplementary Figure 1A, 1B), indicating that the HDA1 HDAC complex negatively influences the starvation-induced HSP expression. The HDA1 complex consists of the catalytic Hda1 homodimer, which interacts with the Hda2-Hda3 heterodimer to form an active tetramer to deacetylate histones H2B, H3 and H4. Disruption of any of the three subunits has been shown to similarly abolish the catalytic activity of the HDA1 complex both in vitro and in vivo [23]. Hence, we only include the hda1Δ mutant for further investigation. Among the HAT mutants, removal of GCN5, encoding the catalytic subunit of the SAGA/ADA/SLIK complexes, led to significantly reduced levels of both reporters (Supplementary Figure 1C, 1D). Intending to ensure the same genetic background, the gcn5Δ and hda1Δ deletion mutants were regenerated from BY4742 cells. As shown in Figure 1A, 1B, starvation-induced HSP expression was dramatically reduced in the gcn5Δ mutants and significantly enhanced in the hda1Δ cells. Further removal of HDA1 has no impact on the levels of either reporter observed in the gcn5Δ mutants (Figure 1A, 1B), suggesting an epistatic relationship between them. Moreover, in comparison to pHSP26-VFP, starvation-induced pSSA3-VFP is expressed to a relatively higher level in WT or hda1∆ cells and is more severely reduced in the gcn5∆ or gcn5∆hda1∆ mutants (note the different scales in Y axis of Figure 1A, 1B), suggesting that the acetylation status has a more significant influence on the expression of SSA3 than that of HSP26. The gcn5Δ mutants were previously shown to have respiratory growth defects [24]. The moderate fermentative or more severe respiratory growth defects exhibited by the gcn5Δ cells were not rescued by HDA1 removal (Figure 1C). These data imply that the roles of the HDA1 complex in the regulation of starvation-induced gene expression are dependent on Gcn5.

Gcn5 functions in the HAT module of SAGA to promote HSP expression and respiratory growth

Gcn5 only acetylates nucleosomal histones efficiently when functioning within multi-subunit complexes. To find which complex (Supplementary Figure 2A) is responsible for the stress response and/or respiratory growth, we next screened all the viable mutants of the three complexes in the BY4742 library. Deletion of any of the HAT module subunits (GCN5, ADA2, NGG1, and SGF29) led to dramatic reduction of pSSA3-VFP (Figure 2A) and pHSP26-VFP (Figure 2B, except for ngg1Δ) and severe respiratory growth defects (Figure 2C), supporting that the integrity of the HAT module is essential to the activation of stress response and respiratory growth. However, removal of the ADA-specific subunits resulted in moderately reduced pSSA3-VFP expression only in ahc2Δ (Supplementary Figure 2B) but little impact on pHSP26-VFP expression (Supplementary Figure 2C) or respiratory growth (Supplementary Figure 2D). These data suggest that the ADA complex may play only a minor role in starvation-induced gene expression. In contrast, deletion of the SAGA-specific subunit SPT7 significantly reduced the expression levels of both reporters (Figure 2A, 2B) and respiratory growth on either glycerol or ethanol (Figure 2C), similar to those observed for gcn5Δ. Moreover, removal of SAGA/SLIK-specific SPT3 or the SAGA-specific SPT8 significantly enhanced the expression levels of the two reporters (Figure 2A, 2B), consistent with their roles in preventing SAGA from interaction with the transcriptional machinery [25, 26]. The SLIK complex contains a unique subunit Rtg2 [27] and a truncated version of the Spt7 modified by the proteinase Pep4 [28]. Removal of SLIK-specific RTG2 or PEP4 resulted in significant reduction of both reporters (Figure 2A, 2B). However, unlike gcn5Δ mutants, rtg2Δ and pep4Δ mutants displayed no observable respiratory growth defects (Figure 2C). Put together, these data suggest that Gcn5 is likely to function within the HAT module of the SAGA/SLIK complex to regulate the Msn2/4- and Gis1-mediated stress response. Meanwhile, Gcn5 may function in the HAT module of SAGA to promote respiratory growth.

Within the HAT module of SAGA/SLIK complexes, Gcn5 acts as the catalytic subunit. Ada2 functions to promote the HAT activity of Gcn5, while Ngg1 plays a role in expanding the range of lysine residues that undergo acetylation to allow robust nucleosomal histone acetylation [29]. Deletion of NGG1 led to similarly reduced pSSA3-VFP as observed in the gcn5∆ mutants (Figure 2A) but had no impact on the expression of pHSP26-VFP (Figure 2B), suggesting that nucleosomal histone acetylation is necessary for the expression of SSA3 and may not be required for HSP26 induction. Furthermore, Ada2 is essential for the proper binding of Gcn5-containing complexes to histone H3 N-terminal tail [30]. The exact roles of Sgf29 remained unclear, although preliminary data indicated that Sgf29 can recognise and bind to H3K4me3, hence recruiting Gcn5-containing complexes to the N-terminus of histone H3 to stimulate histone acetylation [29]. Therefore, while deletion of GCN5 and NGG1 abolishes and significantly impairs the HAT activity respectively, loss of either ADA2 or SGF29 may completely prevent the binding of Gcn5-containing complexes to the N-terminus of histone H3. The more severely reduced expression of both reporters observed in ada2Δ and sgf29Δ (Figure 2A, 2B) cells may suggest that the SAGA/SLIK complex (even without the HAT activity) is necessary for optimal gene expression upon glucose starvation.

Gcn5 displays contradictory roles in the regulation of CLS

Next, we investigated the roles of Gcn5 in chronological lifespan extension by measuring the ability of stationary-phase cells to exit quiescence (colony-forming units, CFU). Simultaneously, cell viability was measured by nuclear DNA staining using Sytox Green [21] in cells that have lost their membrane integrity [31]. Comparing CFU with cell viability indicates the relative levels of viable cells capable of quiescence exit, or oppositely, the levels of cellular senescence. Compared to WT, the gcn5Δ mutants lost their colony-forming ability much faster in the early- to mid-stationary phase (≤24 days, Figure 3A). However, such CFU loss in gcn5Δ mutants was constant and more slowly than that seen in WT cells towards the late stationary phase (≥30days, Figure 3A) when the latter started to lose their CFU rapidly. In contrast, the gcn5Δ mutants had similar cell viability (>85%) as WT cells during the early- to mid-stationary phase (≤24 days, Figure 3B) and lost their viability faster than WT cells thereafter (≥30 days, Figure 3B). However, the absolute difference of cell viability between WT cells and the gcn5Δ mutants is within 10% until the end of the observational period (note that the lower limit on the Y axis is 60% in Figure 3B). Relative CFU to cell viability for the gcn5Δ mutants was reduced from 100% to ~70% in the first 24 days and at a similar rate from ~70% to ~50% between 24 and 48 days (Figure 3C). In contrast, relative CFU potential for WT cells was decreased from 100% to nearly 0 during the latter period (Figure 3C). These data indicate that Gcn5 plays contradictory roles in the regulation of CFU potential of the stationary phase cells. Removal of GCN5 results in early but slower rate of chronological ageing, thus a longer maximum lifespan than that of WT cells.

Compared to WT cells, the hda1Δ mutants did not display an improved ability to form colonies (Figure 3A), as opposed to that demonstrated by Yu et al. [32]. One possibility to account for the differences could be the dramatically different CFU of WT cells used in the two studies. Relative CFU of the WT cells in our study continued to increase until day 9 and did not decrease below 100% until day 21 (Figure 3A), whereas the relative CFU of WT cells in the study by Yu et al. did not increase during early stationary phase and started to decrease after day 10. Thus, the likely CFU improvement as a result of Hda1 removal [32] may not be revealed in our study. Nevertheless, Hda1 removal did lead to increased cell viability across the entire stationary phase tested (Figure 3B). Normalising CFU to cell viability indicated that Hda1 removal resulted in reduced ability of viable cells to exit quiescence towards the late-stationary phase (≥30 days, Figure 3C), indicating enhanced senescent cell population in the aged cultures. Further removal of HDA1 did not significantly influence the clonogenic survival (Figure 3A), cell viability (Figure 3B) or the CFU/viability ratio (Figure 3C) of the gcn5Δ mutants, further suggesting the epistatic relationship between GCN5 and HDA1 in CLS regulation.

Gcn5 transcriptionally promotes metabolic reprogramming and inhibits ribosome biogenesis upon glucose starvation

Intending to find how Gcn5 may regulate CLS extension contradictorily, we first conducted RNA-seq analyses of the transcriptome isolated from exponentially-growing (EXP, glucose-replete) and early post-diauxic shift (PDS, glucose-depleted) cells. Differential expression (DE) analysis identified 594 genes whose transcription is significantly regulated (>2-fold) in glucose-depleted gcn5Δ cells. Surprisingly, ~42% of these DE genes were upregulated and the rest downregulated in PDS gcn5Δ cells (Figure 4A), suggesting that Gcn5 is implicated in transcription activation (clusters 1, 2, 3, 4 and 5, Figure 4A and Supplementary Table 1) as well as transcriptional repression (clusters 6 and 7, Figure 4A and Supplementary Table 1). Apart from cluster 1, the hda1Δ mutants displayed similar gene expression patterns as WT cells in all other clusters and the expression of these DE genes were similarly regulated in the gcn5Δ and gcn5Δhda1Δ mutants (Figure 4A), supporting that the roles of HDA1 in glucose starvation-induced gene expression are largely dependent on GCN5.

Enriched GO terms, motifs in the promoter regions [33] and potential transcription factors (TFs) regulating their expression [34] were identified for each gene cluster (Figure 4A). GO analysis indicated that ~50% and ~40% of the 201 genes in cluster 6 were involved in cytoplasmic translation and ribosome biogenesis (RiBi) respectively (Figure 4A), indicating that GCN5 is necessary for global repression of translation in response to glucose depletion. Promoter analysis revealed that the most significantly overrepresented motif is the consensus sequence targeted by Rap1, the transcription factor required for expression of ribosomal protein genes (RPGs) and RiBi genes [35]. Maximal expression of RPGs also requires RPG-specific transcription activators Fhl1 and Ifh1, the latter of which has been shown to be rapidly released from target promoters under the stress conditions or TORC1 inhibition to downregulate RPG expression [35]. In this regard, Gcn5-mediated acetylation of Ifh1 negatively modulates the transcriptional activity of Ifh1 [36, 37]. Thus, removal of Gcn5 would lead to insufficient inhibition of Ifh1 and hence moderate expression of RPG and RiBi genes in glucose-depleted cells (cluster 6). Furthermore, a number of genes implicated in the glycolysis pathway, including HXT2, HXT3, HXK2, PFK27, TDH2, GPM1, ENO2 and CDC19, were also found in cluster 6. Intriguingly, the transcription of these glycolytic genes also requires Rap1 [38, 39]. These data suggest that Gcn5 may target transcription factors directly to repress Rap1-mediated gene expression in response to glucose depletion.

Similar bioinformatic analyses indicated that Gcn5 promotes starvation-induced stress response and carboxylic acid metabolic process (clusters 3 and 4 in Figure 4A). Among the top 20 genes of cluster 3 (Table 1), many encode chaperones or hydrophilins (GRE1, HSP32, HSP33, SIP18, SNO4 and SPG1) that are necessary for survival under stress conditions and a few are implicated in peroxisomal and mitochondrial metabolisms (CYB2, POX1, FOX2 and SHH4). We and others have previously demonstrated that starvation- or rapamycin treatment-induced expression of heat shock proteins and hydrophilins, such as GRE1 and those reporter genes (SSA3 and HSP26) used to identify GCN5 and HDA1 in Figure 1 were mediated by the stress response transcription factors Msn2/4 and the post-diauxic shift factor Gis1 [40–45]. Indeed, consensus sequences targeted by Msn2/4 and Gis1 were significantly enriched in the promoter regions of cluster 3 genes (Figure 4A), supporting that Gcn5 is necessary to activate starvation-induced stress response mediated by both Msn2/4 and Gis1. Similarly, motifs targeted by the carbon source-responsive (CSRE) transcription factors Adr1 and Cat8 were overrepresented in the promoter sequences of cluster 4 genes (Figure 4A). Among the group of 53 genes that were activated by Cat8 and/or Adr1 [46], 16 were found in cluster 4 (Figure 4B), including those involved in gluconeogenesis (FBP1 and PCK1) and acetyl-CoA synthesis (ACS1). Cluster 4 genes also include those implicated in the TCA cycle (ACO1, IDP1, KGD1, LSC2, FUM1), the PDH bypass (ADH2, ALD6 and ACS1) (Supplementary Figure 3) and in oxidative phosphorylation (Supplementary Figure 4).

To further confirm that Gcn5 is involved in Cat8/Adr1-dependent gene expression, Fbp1-GFP fusion protein under the control of the endogenous FBP1 promoter was expressed in WT and mutant cells grown to glucose starvation. As demonstrated previously [46], derepression of Fbp1-GFP in response to glucose depletion was dependent on CAT8 (comparing 24h/48h to 6h, Figure 4C). Compared to WT cells, the induced levels of Fbp1-GFP were slightly higher in the hda1Δ mutants but nearly abolished in the gcn5Δ single or gcn5Δhda1Δ double mutants (Figure 4D), confirming that Gcn5 is essential to the activation of Cat8/Adr1-dependent gene expression in response to glucose starvation, whereas Hda1 has only a marginal impact. Transcription activation of Cat8-dependent genes is mediated by the energy-sensing Snf1 complex in glucose-depleted cells [47, 48]. Furthermore, H3 phosphorylation by Snf1 has been shown to precede SAGA recruitment and chromatin remodelling to activate Cat8- and Adr1-dependent transcription [49–51]. Put together, these and our data suggest that in response to glucose depletion, Gcn5 may function to repress the expression of the translational machinery and as the catalytic subunit of the SAGA/SLIK complexes, to remodel chromatin to activate a number of regulons: the stress response program mediated by Msn2/4 and Gis1, and the respiratory metabolisms promoted by the Snf1 complex and its downstream activators Cat8 and Adr1.

Gcn5 enables redox homeostasis and stress resistance to promote CLS extension

Next, we assayed the phenotypes of the gcn5Δ mutants which may be associated with CLS extension. Early-stationary phase gcn5Δ cells displayed severe defects in resisting exogenous oxidative stress (Figure 5A), corresponding to their significantly increased intracellular ROS levels (Figure 5B). In comparison, ROS accumulation was insignificantly reduced in the hda1Δ mutant (Figure 5B). Further removal of HDA1 did not influence the levels of intracellular ROS (Figure 5B) or the defects to resist oxidative stress in gcn5Δ cells (Figure 5A). In agreement with enhanced intracellular ROS accumulation, the transcript levels of mitochondrial superoxide dismutase SOD2 and catalase CTA1 were significantly impaired in starved cells bearing gcn5Δ (Supplementary Figure 5), suggesting that the anti-oxidant defence system is impaired in the gcn5Δ mutants. Similarly, early-stationary phase gcn5Δ and gcn5Δhda1Δ mutants display minor defects in heat resistance (Figure 5A) and marginally reduced trehalose and glycogen accumulation (Figure 5C). These data suggest that Gcn5 is necessary to promote CLS extension through its roles in the expression of molecular chaperones (Figure 1 and Table 1), the anti-oxidant defence system and optimal accumulation of trehalose to maintain redox homeostasis and resistance to stress conditions.

Gcn5 is required for acetyl-CoA synthesis and histone acetylation on target gene promoters

To confirm that Gcn5/SAGA may be recruited to regulate respiratory metabolisms, we next interrogated the levels of histone acetylation at the target gene promoters through chromatin immunoprecipitation (ChIP) analysis. Histone H3 lysine 9 (H3K9) is a well-known target of Gcn5 within the SAGA complex [52, 53]. Overall H3K9ac levels were decreased by two-fold in WT cells transitioned from glucose-replete (EXP) to glucose-depleted conditions (PDS, Figure 6A). In comparison, H3K9ac levels between the two conditions remained similar at the promoter regions of FBP1 (Figure 6B) and ACS1 (Figure 6C), or significantly increased at the PCK1 promoter (Figure 6D) in PDS cells. Strikingly, H3K9ac levels at any of the promoter regions were significantly reduced in the gcn5Δ deletants at the glucose-replete condition and more severely decreased upon glucose depletion (Figure 6B–6D). Recently, Gcn5/SAGA was shown to interact with Cat8 and Adr1 transcription factors to activate gluconeogenic and fat metabolism genes in cells subject to acute glucose starvation [54]. Our data suggest that Gcn5/SAGA is not only necessary for the basal levels of genome-wide H3K9 acetylation but also recruited to maintain or to enhance H3K9ac levels at the target genes in response to glucose depletion and when cells were grown on non-fermentable carbon sources. Furthermore, in contrast to fat metabolisms activated under acute glucose withdrawal [54], the PDH bypass pathway may be important for the provision of Acetyl-CoA in cells transitioned from fermentative growth to stationary phase (see next paragraph).

We next assessed the metabolome extracted from glucose-depleted cells using LC-MS/MS. In comparison to WT, the levels of the TCA cycle intermediates apart from aconitate and oxaloacetic acid (OAA) were not significantly changed (Figure 6E), suggesting that the flux through the TCA cycle was largely maintained despite transcription downregulation of most of the TCA cycle genes (Supplementary Figure 3). The significant increase of OAA (Figure 6E) and the corresponding decrease of phosphoenolpyruvate (PEP), glucose-6P and glucose-1P together, and UDPG (Figure 6F) support that Gcn5 is necessary for the activation of gluconeogenesis (Figure 4 and Supplementary Figure 3). Furthermore, the levels of acetyl-CoA were severely reduced in the gcn5Δ mutants depleted for glucose (Figure 6F), in agreement with transcription downregulation of the cytosolic PDH bypass leading to acetyl-CoA synthesis (Supplementary Figure 3). These findings lend further support to the conclusion from the transcriptome analysis that Gcn5 is required for the Cat8-/Adr1-dependnet gluconeogenesis and PDH bypass pathways to enable metabolic reprogramming in response to glucose starvation. ACS1 encodes the aerobic isoform of acetyl-CoA synthase necessary to maintain nucleo-cytosolic acetyl-CoA levels and respiratory growth [55, 56]. These and our data also imply that Gcn5 may lie at the centre of a feed-forward loop to promote acetyl-CoA synthesis, histone acetylation, the stress response program and respiratory metabolisms.

Global H3K9 acetylation mediated by Gcn5 and Hda1 is correlated with chronological senescence

To determine the connection between Gcn5-mediated histone acetylation and its negative roles in CLS regulation (Figure 3A), we next assessed the global H3K9ac and H3K14ac levels in cells transitioned into the stationary phase. The levels of H3K9ac, which were previously shown to decrease by ~50% in WT cells in response to glucose depletion (Figure 5A), were further but more modestly reduced during the transition into the early-stationary phase (Figure 7A, 7B). Compared to WT cells, GCN5 deletion led to ~2-fold decrease of H3K9ac in glucose-replete (EXP) condition, ~3.5-fold reduction upon glucose depletion (PDS) and remarkably, ~14-fold decline during the transition into stationary phase (Figure 7A, 7B). These data further support that Gcn5 plays a feed-forward role in promoting acetyl-CoA synthesis and hence H3K9 acetylation. Previously, the HDA1 complex was reported to deacetylate K9, K14, K18, K23, and K27 on histone H3 that are acetylated by Gcn5 [57]. Indeed, global H3K9ac levels were increased in the hda1Δ mutant cells by ~40% in the glucose-replete (EXP) conditions and more drastically by ~100% upon glucose depletion (PDS) and during the transition into stationary phase (Figure 7A, 7B). HDA1 removal from the gcn5Δ mutants, however, only marginally enhanced the levels of H3K9ac at the glucose-replete conditions and had no impact on H3K9ac levels in glucose-depleted gcn5Δ mutants (Figure 7A, 7B). In contrast to H3K9ac, H3K14ac levels were modestly reduced in the gcn5Δ or gcn5Δhda1Δ mutants, and not significantly enhanced in the hda1Δ cells across the three growth phases (Figure 7C). These data support that the HDA1 complex regulates H3K9 acetylation in a manner dependent on Gcn5.

Global H3K9ac levels in transition-phase cells were found to inversely correlate with the percentage of viable cells capable of quiescence exit at the late-stationary phase (Figure 7D) when the majority of the ageing cells remained viable (Figure 3B), suggesting that decreased H3K9 acetylation in gcn5Δ may prevent chronologically ageing cells from entering into senescence, thus retaining their CFU ability (Figure 3C). Conversely, Hda1 removal led to enhanced acetylation, accompanied by a significant decrease of clonogenic survival among the viable populations towards the late-stationary phase (Figures 3C, 7D). Put together, these data suggest that the readouts of Gcn5-mediated histone acetylation (stress resistance and redox homeostasis) and histone acetylation status itself have opposing roles in CLS regulation. These two characteristics together ultimately determine the dynamics of lifespan in non-dividing cells of yeast (Figure 3A).

Strains and plasmids

Strains carrying single-gene deletions were obtained directly from the BY4741/BY4742 mutant libraries (Open Biosystems). Strains carrying deletions of GCN5 and/or HDA1 were regenerated by PCR-mediated gene replacement using drug resistant markers. Fluorescent reporter plasmids, pHSP26-HSP26-RFP/VFP and pSSA3-RFP/VFP, were constructed previously [21]. The FBP1-GFP expression plasmid was constructed by cloning FBP1-GFP coding sequences between the FBP1 promoter and the ADH1 terminator in pRS425.

Fluorescent reporter assays

Cells bearing reporter constructs were grown in appropriate synthetic dropout medium overnight, followed by inoculation into 5ml fresh dropout medium (0.6% glucose) in 50ml falcon tubes to achieve a starting OD_600nm of 0.2. At 6-, 12-, 24-, and 48-hour post-inoculation, around 1x10^7 cells were harvested by centrifugation and resuspended in 1ml of 1x PBS (pH7.4) for fluorescence quantification using BD LSRFortessaTM (BD Biosciences, USA) or Attune Nxt (Thermo Fisher Scientific, USA) instruments. RFP was measured using a 561nm excitation laser and a 610/20nm (BD LSRFortessaTM) or 620/15nm (Attune Nxt) emission filter. VFP was measured using a 488nm excitation laser and a 530/30nm emission filter. The mean fluorescence level was calculated from readings of at least 10,000 cells using the FlowJo software (version 10.2).

Determination of storage carbohydrates, ROS and stress resistance

Intracellular trehalose and glycogen, ROS levels and spotting assays to measure stress resistance displayed by the early-stationary phase cells (72 hours incubation in YPD medium) were determined as described previously [21]. Cell growth was measured either by the above described spotting assay on YPD (2% glucose) or YPGE (3% glycerol and 1% ethanol) or using a plate reader (BMG Labtech, USA).

Chronological lifespan assays

CLS was determined by clonogenicity to access the ability of stationary phase cells to exit from quiescence to form colonies (colony forming units, CFU), and by cell viability to measure the population of cells remaining alive in the stationary phase cultures. Yeast cells were inoculated in YPD medium and incubated for 3 days to early-stationary phase (CLS day 0). Clonogenicity was measured at day 0 and thereafter every 3 or 6 days. The number of CFU was scored by spreading cells onto YPD plates in triplicates after serial dilutions (1:20, 1:50, and 1:100). Relative CFU was calculated by normalising the mean CFU to that at day 0. Cell viability was determined by flow cytometry analysis of SYTOX Green-stained cells, as described previously [21].

Protein isolation and Western blotting

Cells grown to exponential (6h), early post-diauxic shift (14h) and during the transition to stationary phase (24h) in YPD medium were harvested by centrifugation, washed once with ice-cold lysis buffer (20mM HEPES, pH 7.5, 300mM NaCl, 10% (v/v) glycerol), snap-frozen with liquid nitrogen and stored at -80° C until cell lysis. Cells were lysed in lysis buffer containing 1x cOmplete™ proteinase inhibitor cocktail (Roche, Switzerland) by beads beating with acid-washed glass beads (425~600μm, Sigma-Aldrich, USA) using a FastPrep-24™ 5G instrument (MB Biochemical). Following 10 minutes of centrifugation at 4° C to remove pellet, total protein concentration was measured with Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific) and diluted to a concentration of 2mg/ml with lysis buffer. Proteins were denatured by incubating at 95° C for 10 minutes with 2x Laemmli buffer, and 20μg was applied to a 15% Tris-glycine gel for separation. The transfer process was performed using the Trans-Blot Turbo transfer system (Bio-Rad Laboratories, USA) with a 0.2μm PVDF transfer pack (Bio-Rad Laboratories). Membranes were incubated with 6% (w/v) BSA in TBST solution (0.1% Tween 20) for 2 hours at room temperature, followed by primary antibody incubation overnight at 4° C and subsequent secondary antibody incubation for 1 hour at room temperature. The membrane was scanned using an Odyssey® CLx imaging system (Li-COR, USA) and densitometry analysis of protein bands was performed using the Image Studio software (Version 5.2.5, Li-COR). Primary antibodies used were Anti-H3K9ac (1:5,000, 07-352, Upstate®), Anti-H3K14ac (1:5,000, 07-353, Upstate®) and Anti-H3 (1:5,000, ab1791, Abcam). Secondary antibodies used were IRDye® 680RD Goat anti-Rabbit IgG Secondary Antibody (1:20,000, 925-68071, Li-COR).

Chromatin immunoprecipitation (ChIP)

Cells grown to exponential and early post-diauxic shift phases in YPD medium were crosslinked with 1% formaldehyde at 30° C for 20 minutes. Crosslinking reaction was stopped by adding 2.5M glycine to a final concentration of 125mM and incubated for 5 minutes. Cells were harvested by centrifugation and washed once with FA lysis buffer (50mM HEPES-KOH, pH 7.5, 150Mm NaCl, 1mM EDTA, pH 7.6, 1% (v/v) Triton X-100, 0.1% (w/v) sodium deoxycholate and 0.1% (w/v) SDS), snap-frozen and stored at -80° C. Cell lysis was performed in cold FA lysis buffer containing cOmplete™ proteinase inhibitor cocktail (Roche) by beads beating with acid-washed glass beads (425~600μm, Sigma-Aldrich) using a FastPrep-24™ 5G instrument (MB Biochemical). Chromatin was sheared by sonication on a Bioruptor Plus (Diagnode) at the high power setting for 24 cycles (30sec ON/OFF). After centrifugation for 10 minutes at 10,000x g, 4° C, supernatants were transferred to fresh tubes. A similar amount of chromatin equivalent to 150μg of total protein was used for immunoprecipitation (IP) with 3μg of anti-H3K9ac (07-352, Upstate®) or anti-H3 (ab1791, Abcam) overnight at 4° C on a rotation wheel. 40ul of pre-equilibrated Protein A/G Sepharose® (ab193262, Abcam) beads were added to each IP sample and incubated for 5 hours at 4° C. Subsequently, beads were washed thrice with FA lysis buffer, twice with wash buffer 1 (FA lysis buffer containing 0.5M NaCl), and twice with wash buffer 2 (10mM Tris-HCl, pH 8.0, 0.25M LiCl, 1mM EDTA, 0.5% (v/v) NP-40, 0.5% (w/v) sodium deoxycholate), then eluted overnight at 65° C with 130μl of 1% SDS-TE buffer (100mM Tris-HCl, pH 8.0, 10mM EDTA, 1% (v/v) SDS) with 5μg/ml RNase A (Roche). Simultaneously, 30μl of 1% SDS-TE buffer with 5μg/ml RNase A (Roche) was added to the input sample and incubated overnight at 65° C. IP and input samples were treated with 100μg of proteinase K (Roche) for 1 hour at 55° C, then purified with the QIAquick PCR purification kit (Qiagen). Purified DNAs were used as template to quantify the targets bound by anti-H3K9ac or anti-H3 using the QuantiFast SYBR® Green PCR Kit (Qiagen) on Rotor-Gene 6000 (Corbett Research, UK). ChIP signals were normalised to input to generate percent input for subsequent analysis.

RNA-seq analysis

Total RNA was isolated from exponentially growing and early post-diauxic phase cells using a RNeasy Mini kit (Qiagen). PolyA-enriched RNA samples were sequenced on Illumina Hiseq platform by Novogen UK limited. The raw transcriptome data are available at Biostudies (https://www.ebi.ac.uk/biostudies/arrayexpress) under the accession number: E-MTAB-11046. After quality check with FastQC (version 0.11.5), RNAseq data were mapped to the S. cerevisiae reference genome R64-1-1 using HISAT2 (version 2.1.0) [86]. Quality of reads alignment was further checked with Picard Tools (version 1.96) and MultiQC (version 1.8), followed by quantification of reads overlapping coding genes by the featureCounts programme [87]. DESeq2 (version 1.24.0) analyses [88] were performed within R (version 3.6.1), comparing the changes in gene expression. Hierarchical clustering was performed using the ComplexHeatmap package (version 2.4.2) [89]. Visualisation of expression of genes within a KEGG pathway was generated by the Pathview package (version 1.28.0) [90].

Metabolites extraction and targeted metabolite analysis (LC-MS/MS)

Intracellular metabolites were extracted following a protocol adapted from previous publications [91, 92]. Cells were grown in YPD medium (2% glucose) until early-post diauxic shift phase. A culture containing cells equivalent to 3 units of OD_600nm was aliquoted and injected into a 15ml tube containing 5 volumes of pure methanol precooled to -40° C (final methanol concentration of 83%) for quenching, followed by centrifugation to pellet cells. At the same time, a replicate cell sample was harvested for protein quantification. A 60 μL of an internal standard mix at 10 μM (AMP ₁₃C¹⁰, ₁₅N⁵; ATP ₁₃C¹⁰, ₁₅N⁵; Glutamate U¹³C, U¹⁵N; Leucine-d₁₀, Phenylalanine-d₅, Proline U¹³C, U¹⁵N; and Valine-d₈), was added to cell pellets, followed by resuspension in 2ml of -20° C extraction buffer (methanol:acetonitrile:water, 40:40:20). Metabolites were extracted from cells via sonication in a 4° C water bath for 30 minutes and four freeze-thaw cycles between -20° C and -80° C. The mixture was transferred into a new 2ml screw-cap tube for centrifugation at 16,100x g for 15 minutes at 4° C to pellet the cells. The supernatant was aliquoted into a fresh 2ml screw-cap tube to dry under a flow of nitrogen and stored at -80° C until analysis.

The dried metabolites were reconstituted in 100μL of 10 mM ammonium carbonate in acetonitrile:water mix (7:3). Samples were injected onto a UHPLC⁺ series coupled to a TSQ Quantiva mass spectrometer (Thermo Fisher Scientific, USA) using a heated ESI source. The electrospray voltage was set to 3.5 kV and 2.5 kV for positive and negative ionisation modes, respectively. Sheath gas was set at 52 (arbitrary units), auxiliary gas at 16, sweep gas at 2, and ion transfer tube at 356° C. Samples were analysed using two different chromatographic methods, normal phase, and reverse-phase analysis. For normal phase analysis, the samples were analysed with a BEH-amide HILIC (150 x 2.1 mm 1.7 μm) column at 30° C. The mobile phase consisted of: (A) 0.1% of ammonium carbonate and (B) acetonitrile and was pumped at a flow rate of 0.6 mL/min. The gradient was programmed as follows: 80 % of B for 1.50 min followed by a linear decrease from 80 % to 40 % of B for 3.5 min and finally returned to initial conditions. For reverse phase analysis: Samples were dried and reconstituted in 10 mM ammonium acetate solution and analysed with an ACE C18 PFP (150 x 2.1 mm 5 μm) column at 30° C. The mobile phase consisted of (A) 0.1 % formic acid in water and (B) 0.1 % formic acid in acetonitrile, pumped at 0.5 mL/min. The gradient was programmed as follows: 0 % of B for 1.60 min followed by a linear increase from 0 % to 30 % of B for 4 min and 90 % by 4.5 min, held for 1 min and then returned to initial conditions.

Calculated masses and mass fragments of the calculated compounds are reported in Supplementary Table 2. Data were processed using Xcalibur software (Thermo Fisher Scientific), and peak intensities of target metabolites were normalised to an appropriate internal standard and then to protein content, which was quantified following the same protocol as described for western blotting.