Balance between autophagy and cell death is maintained by Polycomb‐mediated regulation during stem cell differentiation

Autophagy is a conserved cytoprotective process, aberrations in which lead to numerous degenerative disorders. While the cytoplasmic components of autophagy have been extensively studied, the epigenetic regulation of autophagy genes, especially in stem cells, is less understood. Deciphering the epigenetic regulation of autophagy genes becomes increasingly relevant given the therapeutic benefits of small‐molecule epigenetic inhibitors in novel treatment modalities. We observe that, during retinoic acid‐mediated differentiation of mouse embryonic stem cells (mESCs), autophagy is induced, and identify the Polycomb group histone methyl transferase EZH2 as a regulator of this process. In mESCs, EZH2 represses several autophagy genes, including the autophagy regulator DNA damage‐regulated autophagy modulator protein 1 (Dram1). EZH2 facilitates the formation of a bivalent chromatin domain at the Dram1 promoter, allowing gene expression and autophagy induction during differentiation while retaining the repressive H3K27me3 mark. EZH2 inhibition leads to loss of the bivalent domain, with consequent ‘hyper‐expression’ of Dram1, accompanied by extensive cell death. This study shows that Polycomb group proteins help maintain a balance between autophagy and cell death during stem cell differentiation, in part, by regulating the expression of the Dram1 gene.


Introduction
Autophagy (macroautophagy) is an evolutionarily conserved catabolic process in which cytosolic components are sequestered by double-membraned vesicles called autophagosomes and transported to the lysosome for degradation [1,2]. The mechanism of autophagy includes steps, such as autophagy initiation, nucleation, autophagosome formation, fusion with the lysosome, and, ultimately, lysosomal degradation [1]. This involves the function of a distinct set of evolutionarily conserved genes called Autophagy-related genes (ATG) that form complexes with other proteins [3]. While being constitutively active at low levels, external stimuli, such as starvation, stress, etc., lead to autophagy induction. Autophagy modulation facilitates cellular homeostasis, metabolic turnover, and degradation of organelles and misfolded proteins [4]. Defects in autophagy are associated with numerous degenerative disorders, such as Alzheimer's Disease, Parkinson's Disease, and Amyotrophic Lateral Sclerosis, among others [5]. While autophagy was previously thought of as a process purely regulated by changes in cytosolic components, recent studies highlight the role of transcriptional regulation of autophagy genes in governing this pathway [6,7].
Epigenetic regulation, primarily mediated by modifications to chromatin and transcription factors, are major governing factors of gene transcription [8]. Autophagy genes are regulated at the chromatin level by epigenetic modifiers, such as BRD4 [9,10], G9A [11], CARM1 [12], and hMOF [13]. Additionally, transcription factors, such as TFEB, FOXO, FOXA, C/EBPb, ATF4, ZKSCAN3, E2F1, and others regulate the transcription of autophagy genes (reviewed in [14]). More recently, non-coding RNAs, such as miRNA, lncRNA, and circRNA, have also been implicated in autophagy gene regulation [15]. These studies point to the essential role of autophagy gene transcription in modulating the pathway. However, the focus of these reports is limited to unique regulatory mechanisms in isolated cell types, and a detailed study identifying regulators of autophagy genes is lacking. This lacuna is starker in the case of embryonic stem cells (ESCs), which have the unique property of selfrenewal and the ability to differentiate into three germ layers. Hence, these cells require an efficient homeostatic system and precisely regulated cellular and metabolic turnover pathways, such as autophagy [16]. While reports emphasize the importance of balanced autophagy levels in governing the ESC state [17][18][19], how these levels are maintained in the context of transcriptional regulation of autophagy genes in steadystate and cell state transition is less understood. This is especially important as autophagy has been implicated as an essential determinant for tissue differentiation and the self-renewal and differentiation potential of embryonic and adult stem cells [20,21]. It also has roles in vasculogenesis, angiogenesis, skeletal muscle maintenance, regeneration, and immune system homeostasis [22]. Furthermore, embryonic and adult development and cell differentiation are associated with cell death, with significant cross-talk between autophagy and cell death and the genes involved in both pathways [23]. In fact, autophagic cell death is crucial for normal development in some organisms [24].
Determining whether changes in autophagy during stem cell differentiation correlate with changes in autophagy gene expression, and dissecting the mechanisms regulating these genes, would help open up potential avenues to identify druggable targets for autophagy modulation. This becomes pertinent given the background of misregulated autophagy genes implicated in degenerative disorders, tumorigenesis and cancer progression, genome instability, impaired immune signalling and pathogen degradation, among other disorders [25][26][27]. This study focuses on understanding the regulation of autophagy in mESCs in steady-state and during the early stages of differentiation. We observe induction of autophagy as mESCs differentiate, accompanied by the upregulation of a subset of autophagy genes. Using a small molecule epigenetic inhibitor screen, we identify chromatin modifiers such as BRD4, G9A, EED and EZH2 as repressors of autophagy in mESCs. We further determine a novel role for EZH2 in autophagy maintenance during mESC differentiation by regulating the expression of the autophagy gene, Dram1. Inhibition of EZH2 leads to the derepression of Dram1, which correlates with a shift from autophagy induction to caspase-independent cell death in differentiating mESCs. Thus, epigenetic regulation of Dram1 by EZH2 helps to maintain the autophagy-cell death balance during mESC differentiation. The data in this study point to the crucial role of chromatin-mediated regulation of autophagy genes in governing this pathway as cells differentiate and open new avenues for using epigenetic inhibitors as druggable solutions for autophagymediated disorders.

Autophagy is induced during RA-mediated mESC differentiation
To determine autophagy status in mESCs, we established a stable mESC line expressing the autophagosome membrane protein LC3, tagged with GFP and mCherry (henceforth called LC3-tagged-mESCs) [28]. Autophagosomes were identified by the presence of GFP as well as mCherry positive puncta, while autophagic flux was detected by mCherry only positive puncta, as GFP is denatured by the acidity of the lysosome when autophagosomes fuse with the lysosomes. Untreated mESCs showed fewer and diffused GFP and mCherry positive puncta, while induction of autophagy by the known autophagy activator rapamycin led to an increase in both GFP and mCherry positive puncta, indicating an increase in both autophagy initiation and flux. Treatment with the autophagosome-lysosome fusion inhibitor chloroquine, however, led to an increase in GFP positive puncta, consistent with the accumulation of autophagosomes, and block in autophagic flux (Fig. 1A,B). These cells, therefore, were reliable indicators of autophagy levels in mESCs at steady state as well as in response to autophagy modulators. We further treated these cells with retinoic acid (RA) to determine autophagy status during mESC differentiation along the neural lineage. The cells expressed neural markers such as Nestin and bIII tubulin and showed reduced expression of pluripotency markers such as Oct4 and Nanog after 3 days of RA treatment (Fig. 1C). Undifferentiated mESCs exhibited largely diffused GFP and mCherry, with few GFP or mCherry puncta visible (a higher number of mCherry puncta were observed compared to GFPpositive puncta), consistent with basal levels of autophagy in these cells. However, RA treatment resulted in increased formation of distinct GFP and mCherry positive puncta, indicating an increase in autophagy initiation and autophagic flux (Fig. 1D,E). We verified this increase in autophagy at the protein level by immunoblotting with an antibody against LC3. We observed an increase in LC3 II levels and LC3 II/LC3 I ratio upon RA treatment of mESCs, indicating induction of autophagy (Fig. 1F,G). We further treated mESCs, and RA-treated cells with the autophagic flux inhibitor chloroquine to understand the role of autophagy inhibition on mESC differentiation. We observed no significant changes in the expression of Oct4 or Nestin in chloroquine-treated cells (Fig. 1H), indicating that while induction of autophagy accompanies RAmediated differentiation, early differentiation is not affected by autophagy inhibition. To determine whether the induction of autophagy during mESC differentiation was accompanied by an increase in the expression of autophagy genes, we analysed published RNA-Seq data (GSE84401 [29]) to determine the expression levels of autophagy genes in mESCs, embryoid bodies (EBs), RA treated EBs and RA treated mESCs grown in N2B27 media. We found that out of 86 genes, 24 genes were significantly upregulated (> 2 fold) upon mESC differentiation ( Fig. 2A,B). Interestingly, the most highly up-regulated genes were regulators of autophagy rather than genes coding for constituents of the pathway itself. These include autophagy activators, such as Dram1, Ulk2, Gfi1b and Wipi1. Genes coding for antiapoptotic proteins such as BCL2 and GPSM1 were also upregulated in differentiated cells. We validated the upregulation of select genes in mESCs treated with RA by RT-qPCR. We saw a significant upregulation in the expression of genes, such as Lamp2, Gabarap, Ulk2, Dram1, and Wipi1, among others (Fig. 2C). Among the Atg genes, Atg4b, Atg4c, Atg7, and Atg10 were up-regulated in RA-treated mESCs. Our studies indicate that autophagy is induced in RA-mediated differentiation of mESCs, which is associated with an increased expression of a subset of autophagy genes.

Autophagy genes are regulated by epigenetic pathways
As mESC differentiation led to an upregulation of autophagy genes, we sought to identify transcriptional regulators of these genes. Epigenetic regulatory mechanisms are crucial in determining transcriptional outcomes in response to stimuli. To identify the epigenetic regulators of autophagy, we used small molecule inhibitors (Table 1) targeting specific epigenetic pathways. We used a fluorescence-based autophagy detection kit (Cyto-ID, Enzo) to detect autophagy levels. We observed an increase in autophagy in mESCs treated with inhibitors such as GSK343 (EZH2), UNC1999 (EZH2), A-395 (EED), UNC0642 (G9a/GLP), A366 (G9a/GLP), and JQ1 (BET domain proteins; Fig. 3A); while treatment with other inhibitors did not show a significant change in autophagy (Fig. 3B). Treatment with chemically inactive analogues, such as UNC2400 (EZH2 Àve), A395N (EED Àve), ÀJQ1 (BET domain Àve), did not show significant changes in autophagy (Fig. 3A), indicating specific effects of certain epigenetic inhibitors on autophagy. We validated the FACS data by treating LC3-tagged-ES cells with GSK343, A366, and JQ1; and observed induction of autophagy as seen by an increase in GFP and mCherry positive puncta (Fig. 3C,D). Our screen identified the Polycomb group proteins EED and EZH2 as regulators of autophagy in mESCs. Inhibition of EZH2 using GSK343 and UNC1999 led to a significant induction of autophagy ( Fig. 3A,C,D). FACS analysis in LC3-tagged-ES cells revealed increased GFP and mCherry positive puncta upon treatment with GSK343. However, the increase of mCherry puncta was higher than that of the GFP puncta, indicating an effect on autophagic flux upon EZH2 inhibition (Fig. 3C). RT-qPCR analysis showed the misregulation of several autophagy genes upon treatment with small molecule epigenetic inhibitors (Fig. 3E). Genes, such as Becn1, Lc3a and Ulk2, were up-regulated upon treatment with all candidate inhibitors, while other genes responded in an inhibitor-dependent manner, indicating that different epigenetic molecules regulate the expression of distinct autophagy genes. Interestingly, genes, such as Lamp2, Dram1, and Ulk2, up-regulated after RA treatment ( Fig. 2), were also up-regulated upon inhibition of EZH2 as well as EED in mESCs (Fig. 3E), indicating a putative role for the Polycomb group proteins in the regulation of these genes during mESC differentiation. Taken together, these results indicate that in mESCs, autophagy is repressed by epigenetic modulators, such as G9a/GLP, EZH2, EED, and BET proteins, and inhibition of these pathways results in a concomitant upregulation of autophagy genes.
EZH2 inhibition leads to caspase-independent, autophagy-associated cell death in differentiating mESCs Polycomb group proteins regulate numerous cellular pathways that play crucial roles in cellular homeostasis, development and differentiation, and the enzyme EZH2 is the primary determinant for establishing the repressive H3K27me3 modification of target sites. Given the specific roles of Polycomb proteins in stem cells and during neuronal differentiation [30][31][32][33], we focussed our further experiments on dissecting the role of the Polycomb enzyme EZH2 in autophagy regulation in mESCs and differentiation. Observation of cells under bright field showed that while mESCs treated with GSK343 did not exhibit significant changes in morphology, RAtreated cells exposed to GSK343 showed extensive cell death (Fig. 4A). Cell counting of the number of cells remaining adhered to the dish showed that compared to untreated mESCs, 49% of cells remained on day 3 of RA treatment, and this number dropped to 8% when cells were treated with RA_GSK343 (Fig. 4B). Treatment with GSK343 alone, however, did not significantly affect the survival of ES cells. While programmed cell death is a hallmark of neural differentiation in normal development, RA-mediated differentiation, and neuroblastomas [34][35][36][37], treatment with RA_GSK343 exacerbated this phenomenon, and we found no viable cells beyond day 4 of treatment. We limited our further treatment duration to 48 h as extensive cell death was seen beyond this point. To determine whether the cell death phenomenon involved apoptosis, we performed Annexin-PI staining followed by FACS analysis. We observed that consistent with previous reports [34], there was an increase in the percentage of Annexin-positive (Early apoptotic) and Annexin + PI-positive (Late apoptotic) cells upon RA treatment of mESCs. In RA_GSK343 treated cells, however, there was a significant increase in PI-positive/Annexin-negative as well as Annexin + PI-positive cells, while the percentage of only Annexin-positive cells remained essentially unchanged (Fig. 4C). The increase in PI-positive/Annexin-negative cells in RA_GSK343 treated cells indicates induction of non-apoptotic cell death. To determine whether the cell death was Caspase-dependent, we performed immunoblotting with a Caspase-3 antibody and saw no change in cleaved Caspase-3 levels ( Fig. 4D) in RA_GSK343 treated cells. We validated this by testing the effect of the pan-Caspase inhibitor, Z VAD-FMK, on RA_GSK343 treated cells, with etoposide-treated ES cells as a positive control for cell death. While the percentage of Annexin-positive and Annexin + PI-positive cells was reduced in Etoposide+Caspase inhibitortreated cells, we observed no significant changes in the presence of RA_GSK343 + Caspase inhibitor (Fig. 4E), further confirming that the cell death phenotype is Caspase independent. RT-qPCR using primers for proapoptotic genes also revealed no significant changes at the transcript level (Fig. 4F).
Autophagic cell death (Type II cell death) refers to non-apoptotic cell death accompanied by increased autophagy [38]. Because the causative relationship between autophagy and cell death is difficult to assess [39], the guidelines for the nomenclature of autophagic processes suggest that type II cell death can be classified as autophagy-mediated or autophagy-associated depending on whether an increase in autophagy accompanies the observed cell death and whether chemical or genetic inhibition of autophagy can rescue cell death [40,41]. To determine whether EZH2 inhibition in differentiating mESCs caused autophagic cell death, we assessed the levels of autophagy in RA_GSK343 treated cells. We observed that treatment with RA, GSK343 and RA_GSK343 led to robust induction of autophagy. However, RA_GSK343 treatment led to similar induction levels as individual treatments (Fig. 5A,B). This may indicate that the processes by which GSK343 and RA activate autophagy may overlap. The overlap between autophagy genes up-regulated in RA and GSK343 treated cells (Figs 2 and 3E) also points to a potential common autophagy activation mechanism in the two treatment conditions. To determine whether the cell death phenotype was autophagy-dependent, we treated the cells with the autophagy inhibitor Wortmannin. Wortmannin treatment led to a reduction in autophagy in RA_GSK343 treated cells (Fig. 5C). However, Annexin-PI staining showed a modest, statistically insignificant reduction in cell death upon Wortmannin treatment (Fig. 5D). To assess the effect of genetic inhibition of autophagy on cell death, we performed Annexin-PI staining in Beclin KO mESCs [42]. While the population had a higher percentage of Annexinpositive cells, cell death seen upon RA and RA_GSK343 treatment could not be rescued in Beclin KO cells (Fig. 5E). This indicates that while an increase in autophagy accompanies the cell death seen in RA_GSK343 treated cells, it may not wholly depend on canonical autophagy pathways. We further characterized the viable cells after Polycomb inhibition by performing RT-qPCR analysis using primers for pluripotency and differentiation genes. We found that while the expression of Oct4, Nanog, Nestin and b III tubulin was unchanged upon Polycomb inhibition in mESCs, Oct4 and Nanog expression was up-regulated in cells treated with RA_GSK343 (Fig. 5F). This is consistent with previous reports that suggest that Polycomb group proteins are dispensable for pluripotency maintenance but may have a role during differentiation [32,43,44].
EZH2 targets Dram1 to regulate autophagy and cell death As treatment with RA and GSK343 led to an increase in autophagy and RA_GSK343 treatment led to an increase in autophagy along with cell death, we sought to dissect the possible mechanism that regulated this phenotype. GSK343 is a selective inhibitor of the Polycomb group enzyme EZH2 that establishes the H3K27me3 mark to repress target genes [45,46]. To determine whether the induction of autophagy and cell death in RA_GSK343 treated cells was associated with changes in autophagy gene expression, we compared the genes up-regulated in RA and GSK343 treatment (Figs 2 and 3E). We identified genes such as Lamp2, Gabarap and Dram1, up-regulated in RA and GSK343 treated cells. The gene coding for the DNA damageregulated autophagy modulator 1 (DRAM1) was of particular interest, as this protein is shown to activate autophagy and cell death [47][48][49] and could potentially serve as a regulator of both phenotypes observed in our experiments. Dram1 was up-regulated in mESCs treated with RA and GSK343; however, RA_GSK343 treatment led to a further compounded upregulation of Dram1, many times higher than individual  treatments (Fig. 6A). We used a siRNA-mediated Dram1 knockdown approach to determine the role of Dram1 in the induction of autophagy and cell death.
We saw a significant decrease in Dram1 expression in siRNA-treated cells (Dram1 KD; Fig. 6B). We then characterized the effect of Dram1 knockdown on autophagy and cell death levels in mESCs and differentiating cells. The cytoID assay revealed that autophagy levels remained unchanged in control, RA and RA_343 treated Dram1 KD cells. However, a significant reduction in autophagy was seen in GSK343 treated mESCs (Fig. 6C). This indicates that the autophagy induction seen in RA-treated cells may not depend on Dram1. This data also suggest that while autophagy induction is seen in RA_GSK343 treated cells, knockdown of Dram1 is insufficient to reverse this induction and may point to other regulatory modules. Other upstream autophagy regulators, such as Gabarap and Lamp2, which were also up-regulated in RA-treated cells, may contribute to the induction of autophagy.
To determine the effect of Dram1 depletion on cell death, we performed Annexin-PI staining. Interestingly, we observed that Dram1 KD could partially rescue cell death only in RA_GSK343 treated cells, where we detected a significant reduction of Annexinnegative/PI-positive and Annexin + PI-positive cells (Fig. 6D). This effect was not seen in other treatment conditions. Our results point to distinct effects of Dram1 KD on autophagy and cell death in mESCs and differentiating cells which may indicate that Dram1 plays a cell state-specific role in regulating autophagy vs. cell death. This regulation may be mediated by the Polycomb complex, as Dram1 upregulation by the EZH2 inhibitor primarily affected autophagy in mESCs, and cell death in differentiating cells.
Dram1 has been identified as a p53-dependent regulator of autophagy and apoptosis [48]. p53 activation is mediated by numerous modes, such as posttranslation modifications, e.g. phosphorylation and acetylation, DNA binding, and transcriptional activation [50]. As polycomb group proteins regulate gene transcription, p53 could be a potential target for GSK343-mediated gene activation. To determine whether GSK343 activated Dram1 in a p53-dependent manner, we measured the levels of p53 transcript in our cells. We observed no significant changes in p53 in any of our treatments (Fig. 7A). This indicates that EZH2 may directly regulate Dram1 transcription. To determine whether EZH2 is enriched at the Dram1 promoter, we analysed published ChIP-Seq data (GSE89929) [51] and observed enrichment of EZH2 and the H3K27me3 mark at the Dram1 promoter in mESCs. In contrast, Lamp2, which was also up-regulated in RA-treated cells, remained devoid of both EZH2 and H3K27me3 enrichment (Fig. 7B). This indicates that EZH2 selectively targets the Dram1 promoter. The upregulation of Lamp2 upon GSK343 treatment may be attributed to an indirect mechanism involving upstream regulators of the gene. As our studies indicated that Dram1 upregulation leads to distinct effects on autophagy and cell death in mESCs vs. differentiating cells, we analysed the enrichment of activating and repressive chromatin modifications on the Dram1 promoter in these cells. ChIP-Seq data (GSE135318) [52] revealed that the Dram1 promoter is enriched with primarily the H3K27me3 mark in mESCs. In neural progenitor cells (NPCs), however, we observed the enrichment of H3K27me3 and H3K4me3 at the Dram1 promoter (Fig. 7C). The increase in the active mark H3K4me3 in NPCs is consistent with the increased expression of Dram1. This pattern of chromatin modifications, called the bivalent domain, is commonly present on promoters of repressed genes that are poised for activation in stem cells [53]. We validated this in mESCs, and RA-treated cells by ChIP-qPCR and observed that while the Dram1 promoter was marked with H3K27me3 in mESCs, a bivalent domain (H3K27me3 and H3K4me3) is seen in RA treated cells. Upon GSK343 treatment, H3K27me3 is reduced, leaving high enrichment of H3K4me3 in RA_GSK343 cells (Fig. 7D,E),  which correlates with the compounded increase in Dram1 expression. Our results indicate that in mESCs, Polycomb group proteins repress Dram1 by establishing H3K27me3 at the promoter. As cells differentiate, activating marks such as H3K4me3 are deposited, contributing to the upregulation of Dram1 and induction of autophagy. However, retention of the H3K27me3 mark on the promoter establishes a bivalent domain, possibly limiting the expression of Dram1. Treatment of differentiating cells with the EZH2 inhibitor GSK343 abolishes the H3K27me3 mark, and this coupled with the activating H3K4me3 modification, results in a 'hyper-expression' of Dram1 and extensive cell death (Fig. 7F). Thus, Polycomb-mediated regulation of the Dram1 gene helps to regulate autophagy in mESCs and cell death during early RA-mediated differentiation.

Discussion
This study systematically analyses autophagy regulation in mESCs and early neural differentiation. The increase in autophagy seen during RA-mediated differentiation of mESCs is consistent with previous studies showing autophagy induction during differentiation in the mouse olfactory bulb and RA-treated mouse neuroblastoma cells [54,55]. Several autophagy genes were up-regulated in differentiating cells. Interestingly, the gene coding for the lysosomal protein LAMP2 was one of the most up-regulated autophagy genes upon mESC differentiation, consistent with previous reports [19]. GABARAP has been shown to induce autophagy via ULK1 activation [56]. DRAM1 regulates autophagic flux by enhancing lysosomal acidification in an mTOR-dependent pathway [47,57]. Our results align with previous reports, which show that autophagy is induced during the differentiation of human ESCs, and neural and cardiac stem cells [54,58,59]. However, autophagy is reported to be reduced during the differentiation of haematopoietic, dermal and epidermal stem cells [60]. The difference in trends between different stem cell types points to a context-specific role of autophagy in self-renewal and differentiation [60,61]. It would be interesting to determine autophagy levels in mESCs subjected to directed differentiation along different lineages. Blocking autophagic flux using chloroquine did not alter mESC differentiation. However, it should be noted that chloroquine treatment beyond 8 h resulted in cell death and precluded long-term determination of differentiation. Targeted knock-out and overexpression experiments using the autophagy genes upregulated in RA-treated cells would help elucidate the causative effect of autophagy induction in mESC differentiation. The small molecule epigenetic inhibitor screen identified G9A/GLP, EZH2 and BET domain proteins as autophagy repressors in mESCs. G9A/GLP has been reported as a repressor of autophagy in HeLa cells, MEFs, and various cancer cell lines [11,62,63]. In contrast, Drosophila G9A activates starvation-induced autophagy by activating the autophagy gene Atg8a [64], indicating a context-specific role for the enzyme. It would be interesting to compare targets of G9A/ GLP in different cell/model systems to determine the downstream effectors of autophagy regulation. The role of JQ1 as an activator of autophagy has been reported in studies on cancer cell lines in viral infections and recovery after spinal cord injuries [65][66][67][68]. Our data is consistent with the role of JQ1 as an activator of autophagy in mESCs. Our study presents the Polycomb group proteins EZH2 and EED as novel regulators of autophagy in mESCs and indicates that EZH2 targets Dram1 to regulate its expression. Our data also suggest that RA-mediated differentiation of ES cells leads to an induction of autophagy and apoptosis. However, treatment with an EZH2 inhibitor in differentiating cells leads to severe cell death, independent of the caspase pathway. While associated with autophagy, this function may not be dependent on autophagy induction, as inhibition of autophagy was unable to rescue the severe cell death phenotype. It is tempting to hypothesize that the balance between autophagy and cell death is maintained in stem cells and early differentiation by regulating Dram1 expression levels, which are governed by epigenetic pathways. The mechanism causing the shift between autophagy and cell death remains to be understood. DRAM1 is a lysosomal protein that regulates cell death by increasing the levels and lysosomal localization of the pro-apoptotic protein BAX in a transcription-independent manner [49]. While our results did not show a change in the transcript levels of pro-apoptotic genes in RA_GSK343 treated cells (Fig. 4F), Dram1 upregulation may increase BAX localization to lysosomes which inhibits its degradation and enhances cell death.
Additionally, while the focus of our study has been on the regulation of Dram1 by EZH2, one cannot discount the possibility that GSK343 treatment would cause global transcriptional changes, indirectly affecting our phenotype. Polycomb group proteins perform regulatory functions in myriad cellular pathways [46], and the cell death phenotype may result from alternative regulatory pathways. ChIP-Seq and RNA-Seq experiments using RA_GSK343 treated cells would

Target
Forward primer Reverse primer Experiment TTCCTAGTTGCGCCGATTGT AAACAGTCTTGGGGCAGCAT ChIP-QPCR Dram1 esiRNA GGGCGGGTTGATTGCCTGTGCTTCACTC GGGCGGGTATGCCTCGTCTTCTTTGCAT Dram1KD    shed light on the genome-wide effects of the inhibitor treatment on EZH2 occupancy and gene expression and enable an accurate dissection of the balance between autophagy and cell death. Our study identifies small molecule epigenetic drugs that activate autophagy in mESCs. It should be noted that epigenetic pathways have many roles in numerous cellular processes [69], and using epigenetic inhibitors as specific regulators of autophagy may have off-target effects. Epigenetic inhibitors have emerged as promising therapeutic alternatives to cytotoxic drug regimens [70]. Thus, the identification of novel regulatory mechanisms for cytoprotective pathways, such as autophagy, not only provides an enhanced understanding of the role of autophagy in stemness and differentiation but also facilitates new avenues for developing druggable treatment paradigms to combat autophagy-mediated disorders.

Materials and methods
Cell culture and differentiation

Epigenetic inhibitor treatment
Small molecule inhibitors for epigenetic pathways (Table 1) were a kind gift from the Structural Genomics Consortium. The drugs and their inactive analogues (negative controls) were dissolved in DMSO, and working concentrations and treatment durations were optimized using standard protocols. The final treatment conditions are described in Table 1.

Confocal microscopy
The mCherry-GFP-LC3 mESCs were used to detect autophagy using confocal microscopy with a Nikon Eclipse Ti2 microscope using the 649 objective. The GFP and mCherry puncta were quantified using FIJI (IMAGEJ). After removal of background and setting appropriate intensity thresholds, the number of GFP and mCherry puncta were counted, and puncta per cell were calculated using DAPI as a nuclear marker.

Immunoblotting
Total proteins were extracted from cells using RIPA buffer containing proteinase inhibitors on ice, followed by centrifugation at 10 000 g for 20 min at 4°C. Protein concentration was measured using Bradford's reagent. Total protein was subjected to SDS/PAGE under reducing conditions, followed by transfer onto a PVDF membrane. The membrane was blocked using 5% BSA in Tris-buffered saline (TBS). Post blocking, the membrane was incubated at 4°C overnight with the appropriate primary antibody. After 3 9 10 min wash in 19 TBS containing 0.1% Tween-20 (TBS-T), the membranes were incubated with an HRP-conjugated secondary antibody for 1 h at room temperature. The membranes were developed using a Femto substrate development kit (34094; Invitrogen, Waltham, MA, USA), and images were captured using a chemi-doc system (AI600; GE Healthcare, Chicago, IL, USA). The following primary and secondary antibodies were used in the study: aTubulin (T5168

qRT PCR
Trizol (15596018; Life Technologies, Calrsbad, CA, USA) was used to isolate total RNA from cells. After DNAse treatment, RNA was quantified using a NanoDrop spectrophotometer. Verso cDNA synthesis kit (AB-1453; Thermo Scientific) was used to generate cDNA. Diluted cDNA was used for qRT-PCR using gene-specific primers ( Table 2) and Power SYBR Green PCR Master Mix (A25742; Applied Biosystems, Waltham, MA, USA). Gapdh was used as an endogenous control.

Dram1 siRNA generation
Template for esiRNA [72,73] production against specific genes was prepared by PCR amplification from mESC or MEF cDNA. In vitro transcription was performed using T7 RNA polymerase, followed by digestion of the doublestranded RNA using RNase III. RNA was then transfected into mESCs using the DharmaFECT 1 transfection reagent (T-2001-02; Dharmacon, Lafayette, CO, USA). After 72 h, cells were lysed in TRIzol for RNA or RIPA lysis buffer to prepare whole-cell protein extract. Non-targeting esiRNA was prepared using GFP as a template in all experiments.

Chromatin immunoprecipitation (ChIP)
mESCs were grown in 10 cm plates at 70% confluency. After trypsinization, cells were fixed for 10 min at 37°C using a final concentration of 1% formaldehyde. Glycine was added to a final concentration of 125 mM to quench the crosslinking reaction. Cells were washed twice with ice-cold 19 PBS. One milliliter SDS lysis buffer was added to each sample and incubated at 4°for 30 min. This was followed by sonication using a BioruptorTM (UCD200) at a high setting for 30 s ON and 60 s OFF for 25 cycles. Subsequent steps were performed as described in [74]. The purified DNA was analysed by qPCR using specific primers as described in