Epigenetic mechanisms and metabolic reprogramming in fibrogenesis: dual targeting of G9a and DNMT1 for the inhibition of liver fibrosis

Objective Hepatic stellate cells (HSC) transdifferentiation into myofibroblasts is central to fibrogenesis. Epigenetic mechanisms, including histone and DNA methylation, play a key role in this process. Concerted action between histone and DNA-mehyltransferases like G9a and DNMT1 is a common theme in gene expression regulation. We aimed to study the efficacy of CM272, a first-in-class dual and reversible G9a/DNMT1 inhibitor, in halting fibrogenesis. Design G9a and DNMT1 were analysed in cirrhotic human livers, mouse models of liver fibrosis and cultured mouse HSC. G9a and DNMT1 expression was knocked down or inhibited with CM272 in human HSC (hHSC), and transcriptomic responses to transforming growth factor-β1 (TGFβ1) were examined. Glycolytic metabolism and mitochondrial function were analysed with Seahorse-XF technology. Gene expression regulation was analysed by chromatin immunoprecipitation and methylation-specific PCR. Antifibrogenic activity and safety of CM272 were studied in mouse chronic CCl4 administration and bile duct ligation (BDL), and in human precision-cut liver slices (PCLSs) in a new bioreactor technology. Results G9a and DNMT1 were detected in stromal cells in areas of active fibrosis in human and mouse livers. G9a and DNMT1 expression was induced during mouse HSC activation, and TGFβ1 triggered their chromatin recruitment in hHSC. G9a/DNMT1 knockdown and CM272 inhibited TGFβ1 fibrogenic responses in hHSC. TGFβ1-mediated profibrogenic metabolic reprogramming was abrogated by CM272, which restored gluconeogenic gene expression and mitochondrial function through on-target epigenetic effects. CM272 inhibited fibrogenesis in mice and PCLSs without toxicity. Conclusions Dual G9a/DNMT1 inhibition by compounds like CM272 may be a novel therapeutic strategy for treating liver fibrosis.

What is already known on this subject?
 The progression of liver fibrosis depends on the activation and transdifferentiation of hepatic stellate cells into a myofibroblastic phenotype.  Epigenetic mechanisms have been shown to control many aspects of fibrogenesis in the liver.  Metabolic reprogramming is emerging as a key process in the activation of fibrogenic cells in different organs.
What are the new findings?
 Together with DNMT1, the histone methyltransferase G9a is expressed in fibrogenic cells in cirrhotic human liver, in chronically injured mouse liver and upon activation of cultured mouse HSC.  G9a and DNMT1 expression is required for fibrogenic activation of HSC by TGF1.  Pharmacological targeting of DNMT1 and G9a with the novel first-in-class dual G9a/DNMT1 inhibitor CM272 counteracts the pro-fibrogenic responses and metabolic reprogramming of HSC elicited by TGF1.  CM272 administration shows antifibrogenic activity in clinically relevant mouse models of liver fibrosis and in human precision-cut liver slices without causing toxic effects.

How might it impact on clinical practice in the foreseeable future?
 The development of effective antifibrotic therapies is much needed not only for chronic liver disease but also for other organs like the lung and kidney. Targeting the complex epigenetic mechanisms involved in fibrogenesis with innovative molecules like CM272 may pave the way for better therapies.

INTRODUCTION
The fibrogenic response is part of the natural reparative reaction in different tissues and organs. This process leads to the formation of a temporary extracellular matrix (ECM) which after wound repair is degraded and tissue architecture is restored. However, when damage persists, as occurs in liver chronic viral infection, alcohol abuse or in nonalcoholic fatty liver disease, the equilibrium between ECM production and removal is ultimately lost resulting in excessive accumulation of a dense ECM rich in fibrillar collagens. [1] This ECM is a physical barrier that perturbs organ's perfusion, contributes to loss of liver function, progression to cirrhosis and hepatocellular carcinoma development. [1,2] The pathological relevance of liver fibrogenesis has driven very active research over the past decades. One major finding was the realization of the highly dynamic nature of the process, including clinical findings showing fibrosis reversion upon removal of the causative agent. [1] The major cellular source of collagen are the liver myofibroblasts, mesenchymal cells mainly derived from hepatic stellate cells (HSC) and periportal fibroblasts. [3] In the normal liver HSC show a quiescent and differentiated phenotype which upon hepatic injury is substantially altered. A plethora of cytokines, small molecules and growth factors, with transforming growth factor- (TGF) playing a central role, contribute to HSC activation and conversion into proliferative and inflammatory collagen-secreting myofibroblasts. [3] A profound metabolic reprogramming, including a shift towards aerobic glycolysis, was recently identified as an essential mechanism in HSC activation. [4,5] Earlier evidence indicated that during fibrosis regression myofibroblasts were removed by apoptosis or entered a senescent profibrolytic state prone to immune-mediated clearance. [1] However, later studies demonstrated that upon cessation of injury a significant proportion of myofibroblasts also undergo reversion to a deactivated phenotype. [6,7] Together, these findings attest to the plasticity of HSC and provide valuable insights for the development of much needed antifibrogenic strategies. [8] Extensive changes in the HSC's transcriptome occur during their transition into hepatic myofibroblasts and upon cessation of injury their reversion to quiescence. [1,7] In this context epigenetic mechanisms are increasingly recognized to play a central role. [9] DNA hypomethylation has been associated with fibrogenic gene activation, while repression of genes that maintain HSC differentiation and quiescence was linked to increased methyl-CpGs abundance in their regulatory regions. [10] [11] Mechanistically, to control gene expression DNA methylation works in concert with other epigenetic modifications such acetylation and methylation of lysine residues in histones H3 and H4. [9] The methyl-CpG binding protein MeCP2 plays a key function in this process, orchestrating the activity of the histone methyltransferases (HMTs) enhancer of zeste homolog-2 (EZH2) and absent, small, or homeotic disc 1 (ASH1) during the reprogramming of HSC transcriptome to the myofibroblast phenotype. [12,13] From a translational perspective, it is important to consider that epigenetic modifications are reversible and amenable to pharmacological intervention. Indeed, the antifibrogenic effects of histone deacetylase inhibitors were already reported twenty years ago. [14] More recently it was demonstrated that targeting DNA-methyltransferases (DNMTs) with 5-azadeoxycytidine prevents HSC fibrogenic activation, while inhibition of HMTs halts hepatic fibrosis progression in mice. [15][16][17] Similarly, pharmacological inhibition of the H3K9 methyltransferase G9a, also known as euchromatic histone-lysine methyltransferase 2 (EHMT2), has been recently shown to reduce kidney and lung fibrosis, although the underlying mechanisms are not fully understood. [18,19]  Here we demonstrate the therapeutic potential of dual G9a/DNMT targeting in experimental liver fibrosis and show how this epigenetic mechanism can control TGFmediated pro-fibrogenic metabolic reprogramming and HSC activation. injection. Bile duct ligation (BDL) was performed as described. [12] [27] From day 2 postsurgery animals received daily injections of CM272 (2.5mg/Kg) or PBS (i.p.) and were humanely killed after 11 days. Animal care and procedures were approved by the Animal

Precision cut liver slices (PCLSs) experiments
Human liver tissue was obtained from normal resection margins surrounding colorectal metastases from adult patients undergoing surgical resection at the Freeman Hospital

Expression of G9a and DNMT1 in activated HSC.
We performed immunohistochemical staining of liver tissue samples from patients with viral cirrhosis and ALD. We detected the presence of G9a and DNMT1 in activated myofibroblasts (-smooth muscle actin, -SMA-expressing cells) (Fig. 1A, Supporting Fig. S1 and S2). G9a and DNMT1 were also detected in mouse liver myofibroblasts after chronic CCl 4 injury or BDL (Fig. 1B, Supporting Fig. S3). Next, we examined the expression of G9a and DNMT1 in quiescent and culture-activated mouse HSC. We found that G9a and DNMT1 protein levels were significantly induced between day 1 and day 4 of culture in parallel with -SMA, a marker of HSC myofibroblastic transdifferentiation ( Fig. 2A). [3] The expression of ubiquitin-like with PHD and RING finger domains-1 (UHRF1), a key coordinator of DNA methylation during DNA replication and a functional adaptor between DNMT1 and G9a, [29] was also increased in culture-activated HSC ( Fig. 2A). The mRNA levels of these three genes also increased upon HSC activation ( Fig. 2A). Interestingly, in LX2 cells, a well-characterized model of human HSC,[30] TGF stimulation induced the rapid recruitment of the three proteins to the nuclear chromatin subfraction, without significantly changing their expression (Fig. 2B).
Combined, these observations suggested a role for G9a and DNMT1, together with UHRF1, in HSC activation. To directly address this point, we examined TGF responses in LX2 cells after siRNA-mediated knockdown of these genes (Supporting Fig. S4). We found an overall impairment of TGF-activated profibrogenic gene expression, an effect that was particularly strong upon G9a downregulation (Fig. 2C).  S5A and B). Next we observed a marked impairment of TGF effects on key fibrogenic genes expression, including COL11 and TGF1 itself ( Fig. 3A and Supporting Fig.   S5C), while glial fibrillary acidic protein (GFAP), a marker of quiescent HSC, [3,31] was upregulated (Fig. 3A). These effects were reproduced in primary human HSC (Supporting Fig. S5D). Interestingly, culture-activation of primary mouse HSC was also reduced by CM272 treatment, as indicated by the expression of Col1a1, Timp1 and lecithin-retinol acyltransferase (Lrat) (Supporting Fig. S5E). In agreement with the impaired response to TGF1 stimulation when G9a and DNMT1 were knocked-down in LX2 cells (Fig. 2C) we found that combined treatment with the DNMT1 inhibitor decitabine and the G9a inhibitor BIX01294 also dampened the pro-fibrogenic responses to this growth factor (Supporting Fig. S5F). Together with TGF, hypoxia is considered a major driver of liver fibrogenesis.[32,33] Consistently, we found that hypoxia stimulated LX2 cells growth and that CM272 inhibited this response as well as basal cell growth under normoxia (Fig. 3B). Moreover, fibrogenic gene expression induction by hypoxia was also blunted by CM272 (Fig. 3B) (Fig. 3C).
Gene ontology (GO) functional classification fundamentally identified categories related to cell growth, differentiation, signalling, metabolism, chromatin regulation and response to hypoxia (Fig. 3C). Accordingly, when we applied gene set enrichment analysis (GSEA), a significant positive enrichment in genes of the KEGG peroxisome proliferatoractivated receptor (PPAR) signaling pathway, as well as the reactome "metabolism of steroid hormones and vitamins A and D", was detected in cells treated with CM272 ( Fig.   3D). Also consistent with our GO analyses and with the effects of CM272 on TGF and hypoxia-mediated fibrogenic activation, we found significant negative enrichments in gene sets involved in TGF, platelet derived growth factor receptor- (PDGFR) and hypoxia-inducible factor (HIF) pathways (Fig. 3D). Interestingly, a negative enrichment was also observed in the KEGG glycolysis/gluconeogenesis gene set (Fig. 3D).
Collectively, these findings indicate that G9a/DNMT1 targeting with CM272 profoundly affects the fibrogenic activation of liver myofibroblasts and the involved metabolic adaptations.

Mechanisms of the inhibitory effects of CM272 on hepatic myofibroblasts activation.
In view of the antagonism of CM272 on TGF1 cellular responses we first checked wether TGF1 signaling could be affected. We found that CM272 treatment attenuated SMAD3 phosphorylation in response to TGF1 in LX2 cells (Supporting Fig. S6A).  (Fig. 4A). Consistently, the relative contribution to ATP production of glycolysis vs oxidative phosphorylation, which was increased by TGF was mitigated by CM272 treatment (Fig. 4B). TGF-triggered lactate production, a hallmark of the glycolytic phenotype contributing to fibrogenesis, [4] was also attenuated by CM272 (Fig. 4C). Changes in the expression of key glycolytic and gluconeogenic genes have been mechanistically linked to metabolic reprogramming and activation of fibrogenic cells. [3,35,37,38] Consistently, when glycolysis was inhibited using the glucose analog 2-deoxy-D-glucose (2DG) (Supporting Fig. S7A) we found that TGF1-mediated fibrogenic gene expression in LX2 cells was impaired (Supporting Fig.   S7B). Next, we tested the expression of the glycolytic genes hexokinase-I (HK-I), 6-  S8B). Very interestingly, the expression of the rate-limiting gluconeogenic enzymes phosphoenolpyruvate-carboxykinase (PEPCK) and fructose-1,6-bisphosphatase-1 (FBP1), repressed during fibrogenic activation, [4] was also inhibited by TGF, but restored under CM272 treatment (Fig. 4D). Moreover, the expression of the transcription factor and metabolic regulator peroxisome proliferator activated receptor gamma coactivator-1 (PGC-1, recently identified as a key guardian of lung fibroblasts quiescence, [19,37] was also repressed by TGF and was potently reactivated upon CM272 treatment (Fig. 4D). These responses to CM272 were reproduced in human primary HSC (Supporting Fig. S9A). Importantly, the upregulation of FBP1 and PGC-1 expression by CM272 (Fig. 4E) was related to the on-target pharmacological actions  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   C  o  n  f  i  d  e  n  t  i  a  l  :  F  o  r  R  e  v  i  e  w  O  n  l  y 15 of this molecule. By qChIP analyses we found that CM272 reduced the levels of the repressive H3K9me2 mark in the proximal promoters of FBP1 and PGC-1 (Fig. 4F). At the DNA level, FBP1 promoter was found hypermethylated in a region previously associated with its transcriptional repression in cancer, [26] and DNA methylation was reduced upon CM272 or decitabine treatment (Fig. 4G). Regarding PGC-1, we did not find significant levels of DNA methylation (Fig. 4G), suggesting that its transcriptional repression could be mainly mediated by G9a-H3K9 dimethylation, which indeed was reversed by CM272 treatment (Fig. 4F). In support of these notions we observed that FBP1 expression was upregulated by decitabine or BIX01294, and together both agents had an additive effect, while PGC-1 expression was induced only by BIX01294 (Supporting Fig. S9B).

CM272 inhibits hepatic fibrogenesis in vivo.
Next, we examined the antifibrogenic potential of CM272 in different mouse models.
First, we tested the effects of CM272 on the acute activation of HSC upon single CCl 4 injection. We found that CM272 administration 24h after CCl 4 markedly inhibited HSC activation as indicated by -SMA expression (Supporting Fig. S10A). The antifibrogenic activity of CM272 was also evident in chronic liver injury. Mice received CCl 4 twice a week for 6 weeks, and for the last two weeks were treated with CM272 or its vehicle (Fig.   5A). -SMA and Sirius red staining for collagen deposition demonstrated reduced liver fibrosis in CM272-treated mice (Fig. 5A), corroborated by decreased expression of collagen-I1 (ColI1), -Sma and Tgf (Fig. 5B). Interestingly, expression of Pkm2, previously identified as a marker of glycolytic activation in liver myofibroblasts and a key regulator of glycolysis and the serine-glycine pathway, [4,43] was induced by CCl 4 administration. Noteworthy, Pkm2 expression was significantly attenuated by CM272 treatment, as was that of Phgdh ( Fig. 5A and 5B). The antifibrotic effects of CM272 were  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   C  o  n  f  i  d  e  n  t  i  a  l  :  F  o  r  R  e  v  i  e  w  O  n  l  y 16 reproduced in a model of cholestatic liver injury induced by BDL, as demonstrated by reduced -SMA immunostaining, collagen deposition and expression of fibrogenesis-and glycolysis-related genes ( Fig. 5C and D). As in the CCl 4 model, Pkm2 expression was also increased in areas of active fibrosis and was downregulated by CM272 (Fig. 5C). No significant differences in serum transaminases and creatinine levels nor body weight were found between vehicle and CM272 treated mice in either model, while a decrease in the hepatic expression of pro-inflammatory cytokines was noticed (Supporting Fig. S10B and C). Together these findings demonstrate that in vivo targeting of G9a/DNMT1 with CM272 during ongoing liver injury has antifibrotic potential and is exempt of overt toxicity.

CM272 has antifibrotic activity in human precision-cut liver slices (PCLSs).
To further validate the antifibrotic effects of CM272 we used human PCLSs cultured in a newly designed bioreactor that allows modeling active fibrogenesis induced by pathophysiological stimuli; TGF and PDGF-BB.
[28] First, we observed that G9a, DNMT1 and UHRF1 expression was significantly increased after 96h in culture compared to freshly isolated tissues, and TGF+PDGF-BB enhanced this response (Fig. 6A).
Immunohistochemical analyses of PCLSs detected G9a and DNMT1 proteins in regions of the parenchyma enriched in -SMA positive cells (Fig. 6B). Next, we tested the effects of CM272 treatment on TGF+PDGF-BB-mediated fibrogenic activation of PCLSs ( Fig. 6A). PCLSs were incubated with TGF+PDGF-BB in the absence or presence of CM272 or the TGF1 receptor-1 inhibitor (Alk5i) SB-525334.
[28] As shown in Fig. 6C, the upregulation of fibrogenic gene expression elicited by TGF+PDGF-BB was significantly attenuated by CM272. Consistently, soluble collagen secretion into the culture media, its deposition in the fibrotic matrix, and -SMA staining were also markedly inhibited ( Fig. 6D  17 medium, indicative of metabolic glycolytic reprogramming, was inhibited not only by SB-525334 but also very efficiently by CM272 (Fig. 6F). Accordingly, FBP1 expression was downregulated by TGF+PDGF-BB treatment while that of PKM2 and PHGDH was induced (Fig. 6G). These changes were also reversed by SB-525334 and CM272 ( Fig. 6G). Immunohistochemical staining of PCLSs detected PKM2 expression in fibrogenic cells, validating the activation of glycolysis in human liver tissues by fibrogenic stimuli, and its inhibition by CM272 (Fig. 6H). Our PCLSs model may also provide valuable information on potential hepatotoxic effects of experimental therapies in a human liver tissue environment.
[28] We measured a series of parameters, including albumin and urea levels, and lactate dehydrogenase (LDH), AST and ALT activities in conditioned media from control and CM272 treated PCLSs. We found no significant differences on these markers of hepatocellular function and injury in comparison with controls ( Fig. 7A), and no major histological alterations upon H&E staining were observed either (Fig. 7B). In summary, we have identified novel epigenetic targets involved in liver fibrosis and demonstrated that their dual targeting with an innovative "epi-drug" can inhibit progression of liver fibrosis even in the absence of treating the underlying disease. We have also provided extended evidence on the role of metabolic reprogramming in liver fibrogenesis, and how this can be manipulated at the epigenetic level to halt or reverse the process. CM272 might be also considered for the treatment of fibrotic processes in other organs like the lung and kidneys, in which this condition has devastating effects.
What is already known on this subject?
 The progression of liver fibrosis depends on the activation and transdifferentiation of hepatic stellate cells into a myofibroblastic phenotype.  Epigenetic mechanisms have been shown to control many aspects of fibrogenesis in the liver.  Metabolic reprogramming is emerging as a key process in the activation of fibrogenic cells in different organs.

What are the new findings?
 Together with DNMT1, the histone methyltransferase G9a is expressed in fibrogenic cells in cirrhotic human liver, in chronically injured mouse liver and upon activation of cultured mouse HSC.  G9a and DNMT1 expression is required for fibrogenic activation of HSC by TGF1.  Pharmacological targeting of DNMT1 and G9a with the novel first-in-class dual G9a/DNMT1 inhibitor CM272 counteracts the pro-fibrogenic responses and metabolic reprogramming of HSC elicited by TGF1.  CM272 administration shows antifibrogenic activity in clinically relevant mouse models of liver fibrosis and in human precision-cut liver slices without causing toxic effects.

How might it impact on clinical practice in the foreseeable future?
 The development of effective antifibrotic therapies is much needed not only for chronic liver disease but also for other organs like the lung and kidney. Targeting the complex epigenetic mechanisms involved in fibrogenesis with innovative molecules like CM272 may pave the way for better therapies. reprogramming, including a shift towards aerobic glycolysis, was recently identified as an essential mechanism in HSC activation. [4,5] Earlier evidence indicated that during fibrosis regression myofibroblasts were removed by apoptosis or entered a senescent profibrolytic state prone to immune-mediated clearance. [1] However, later studies demonstrated that upon cessation of injury a significant proportion of myofibroblasts also undergo reversion to a deactivated phenotype. [6,7] Together, these findings attest to the Here we demonstrate the therapeutic potential of dual G9a/DNMT targeting in experimental liver fibrosis and show how this epigenetic mechanism can control TGFmediated pro-fibrogenic metabolic reprogramming and HSC activation.

Human samples
Liver

Expression of G9a and DNMT1 in activated HSC.
We performed immunohistochemical staining of liver tissue samples from patients with viral cirrhosis and ALD. We detected the presence of G9a and DNMT1 in activated myofibroblasts (-smooth muscle actin, -SMA-expressing cells) (Fig. 1A, Supporting   Fig. S1 and S2). G9a and DNMT1 were also detected in mouse liver myofibroblasts after chronic CCl 4 injury or BDL (Fig. 1B, Supporting Fig. S3). Next, we examined the expression of G9a and DNMT1 in quiescent and culture-activated mouse HSC. We found that G9a and DNMT1 protein levels were significantly induced between day 1 and day 4 of culture in parallel with -SMA, a marker of HSC myofibroblastic transdifferentiation ( Fig. 2A). [3] The expression of ubiquitin-like with PHD and RING finger domains-1 (UHRF1), a key coordinator of DNA methylation during DNA replication and a functional adaptor between DNMT1 and G9a, [29] was also increased in culture-activated HSC ( Fig. 2A). The mRNA levels of these three genes also increased upon HSC activation ( Fig. 2A). Interestingly, in LX2 cells, a well-characterized model of human HSC,[30] TGF stimulation induced the rapid recruitment of the three proteins to the nuclear chromatin subfraction, without significantly changing their expression (Fig. 2B).
Combined, these observations suggested a role for G9a and DNMT1, together with UHRF1, in HSC activation. To directly address this point, we examined TGF responses in LX2 cells after siRNA-mediated knockdown of these genes (Supporting Fig. S4). We found an overall impairment of TGF-activated profibrogenic gene expression, an effect that was particularly strong upon G9a downregulation (Fig. 2C).  S5A and B). Next we observed a marked impairment of TGF effects on key fibrogenic genes expression, including COL11 and TGF1 itself ( Fig. 3A and Supporting Fig.   S5C), while glial fibrillary acidic protein (GFAP), a marker of quiescent HSC, [3,31] was upregulated (Fig. 3A). These effects were reproduced in primary human HSC (Supporting Fig. S5D). Interestingly, culture-activation of primary mouse HSC was also reduced by CM272 treatment, as indicated by the expression of Col1a1, Timp1 and lecithin-retinol acyltransferase (Lrat) (Supporting Fig. S5E). In agreement with the impaired response to TGF1 stimulation when G9a and DNMT1 were knocked-down in LX2 cells (Fig. 2C) we found that combined treatment with the DNMT1 inhibitor decitabine and the G9a inhibitor BIX01294 also dampened the pro-fibrogenic responses to this growth factor (Supporting Fig. S5F). Together with TGF, hypoxia is considered a major driver of liver fibrogenesis.[32,33] Consistently, we found that hypoxia stimulated LX2 cells growth and that CM272 inhibited this response as well as basal cell growth under normoxia (Fig. 3B). Moreover, fibrogenic gene expression induction by hypoxia was also blunted by CM272 (Fig. 3B) (Fig. 3C).
Gene ontology (GO) functional classification fundamentally identified categories related to cell growth, differentiation, signalling, metabolism, chromatin regulation and response to hypoxia (Fig. 3C). Accordingly, when we applied gene set enrichment analysis (GSEA), a significant positive enrichment in genes of the KEGG peroxisome proliferatoractivated receptor (PPAR) signaling pathway, as well as the reactome "metabolism of steroid hormones and vitamins A and D", was detected in cells treated with CM272 (Fig.   3D). Also consistent with our GO analyses and with the effects of CM272 on TGF and hypoxia-mediated fibrogenic activation, we found significant negative enrichments in gene sets involved in TGF, platelet derived growth factor receptor- (PDGFR) and hypoxia-inducible factor (HIF) pathways (Fig. 3D). Interestingly, a negative enrichment was also observed in the KEGG glycolysis/gluconeogenesis gene set (Fig. 3D).
Collectively, these findings indicate that G9a/DNMT1 targeting with CM272 profoundly affects the fibrogenic activation of liver myofibroblasts and the involved metabolic adaptations.

Mechanisms of the inhibitory effects of CM272 on hepatic myofibroblasts activation.
In view of the antagonism of CM272 on TGF1 cellular responses we first checked wether TGF1 signaling could be affected. We found that CM272 treatment attenuated SMAD3 phosphorylation in response to TGF1 in LX2 cells (Supporting Fig. S6A). SB-525334 regardless of TGF1 stimulation (Supporting Fig. S6C).
Metabolic reprogramming is emerging as a critical event in fibrogenic activation across different tissue types. [4,5,35,36] Therefore, we examined the effects of CM272 on oxygen consumption rate (OCR; a representation of mitochondrial activity) and the extracellular acidification rate (ECAR; a surrogate for glycolytic rate) in LX2 cells treated with TGF. As recently reported we found that TGF reduced OCR and increased ECAR,[36] however these effects were attenuated by CM272 (Fig. 4A). Consistently, the relative contribution to ATP production of glycolysis vs oxidative phosphorylation, which was increased by TGF was mitigated by CM272 treatment (Fig. 4B). TGF-triggered lactate production, a hallmark of the glycolytic phenotype contributing to fibrogenesis, [4] was also attenuated by CM272 (Fig. 4C). (FBP1), repressed during fibrogenic activation, [4] was also inhibited by TGF, but restored under CM272 treatment (Fig. 4D). Moreover, the expression of the transcription factor and metabolic regulator peroxisome proliferator activated receptor gamma coactivator-1 (PGC-1, recently identified as a key guardian of lung fibroblasts quiescence, [19,37] was also repressed by TGF and was potently reactivated upon CM272 treatment (Fig. 4D). These responses to CM272 were reproduced in human primary HSC (Supporting Fig. S9A). Importantly, the upregulation of FBP1 and PGC-1 expression by CM272 (Fig. 4E)  15 of this molecule. By qChIP analyses we found that CM272 reduced the levels of the repressive H3K9me2 mark in the proximal promoters of FBP1 and PGC-1 (Fig. 4F). At the DNA level, FBP1 promoter was found hypermethylated in a region previously associated with its transcriptional repression in cancer,[26] and DNA methylation was reduced upon CM272 or decitabine treatment (Fig. 4G). Regarding PGC-1, we did not find significant levels of DNA methylation (Fig. 4G), suggesting that its transcriptional repression could be mainly mediated by G9a-H3K9 dimethylation, which indeed was reversed by CM272 treatment (Fig. 4F). In support of these notions we observed that  Fig. S9B).

CM272 inhibits hepatic fibrogenesis in vivo.
Next, we examined the antifibrogenic potential of CM272 in different mouse models.
First, we tested the effects of CM272 on the acute activation of HSC upon single CCl 4 injection. We found that CM272 administration 24h after CCl 4 markedly inhibited HSC activation as indicated by -SMA expression (Supporting Fig. S10A). The antifibrogenic activity of CM272 was also evident in chronic liver injury. Mice received CCl 4 twice a week for 6 weeks, and for the last two weeks were treated with CM272 or its vehicle (Fig.   5A). -SMA and Sirius red staining for collagen deposition demonstrated reduced liver fibrosis in CM272-treated mice (Fig. 5A), corroborated by decreased expression of collagen-I1 (ColI1), -Sma and Tgf (Fig. 5B). Interestingly, expression of Pkm2, previously identified as a marker of glycolytic activation in liver myofibroblasts and a key regulator of glycolysis and the serine-glycine pathway, [4,43] was induced by CCl 4 administration. Noteworthy, Pkm2 expression was significantly attenuated by CM272 treatment, as was that of Phgdh ( Fig. 5A and 5B). The antifibrotic effects of CM272 were reduced -SMA immunostaining, collagen deposition and expression of fibrogenesis-and glycolysis-related genes ( Fig. 5C and D). As in the CCl 4 model, Pkm2 expression was also increased in areas of active fibrosis and was downregulated by CM272 (Fig. 5C). No significant differences in serum transaminases and creatinine levels nor body weight were found between vehicle and CM272 treated mice in either model, while a decrease in the hepatic expression of pro-inflammatory cytokines was noticed (Supporting Fig. S10B and C). Together these findings demonstrate that in vivo targeting of G9a/DNMT1 with CM272 during ongoing liver injury has antifibrotic potential and is exempt of overt toxicity.

CM272 has antifibrotic activity in human precision-cut liver slices (PCLSs).
To further validate the antifibrotic effects of CM272 we used human PCLSs cultured in a newly designed bioreactor that allows modeling active fibrogenesis induced by pathophysiological stimuli; TGF and PDGF-BB.
[28] First, we observed that G9a, DNMT1 and UHRF1 expression was significantly increased after 96h in culture compared to freshly isolated tissues, and TGF+PDGF-BB enhanced this response (Fig. 6A).
Immunohistochemical analyses of PCLSs detected G9a and DNMT1 proteins in regions of the parenchyma enriched in -SMA positive cells (Fig. 6B). Next, we tested the effects of CM272 treatment on TGF+PDGF-BB-mediated fibrogenic activation of PCLSs (Fig. 6A). PCLSs were incubated with TGF+PDGF-BB in the absence or presence of CM272 or the TGF1 receptor-1 inhibitor (Alk5i) SB-525334.
[28] As shown in Fig. 6C, the upregulation of fibrogenic gene expression elicited by TGF+PDGF-BB was significantly attenuated by CM272. Consistently, soluble collagen secretion into the culture media, its deposition in the fibrotic matrix, and -SMA staining were also markedly inhibited (Fig. 6D and E) 17 medium, indicative of metabolic glycolytic reprogramming, was inhibited not only by SB-525334 but also very efficiently by CM272 (Fig. 6F). Accordingly, FBP1 expression was downregulated by TGF+PDGF-BB treatment while that of PKM2 and PHGDH was induced (Fig. 6G). These changes were also reversed by SB-525334 and CM272 ( Fig. 6G). Immunohistochemical staining of PCLSs detected PKM2 expression in fibrogenic cells, validating the activation of glycolysis in human liver tissues by fibrogenic stimuli, and its inhibition by CM272 (Fig. 6H). Our PCLSs model may also provide valuable information on potential hepatotoxic effects of experimental therapies in a human liver tissue environment.
The PHGDH inhibitor NCT503 and the glucose analog 2-deoxy-D-glucose (2DG) were both from Sigma Aldrich. In vitro treatments were performed at indicated times and doses, and controls received the same concentrations of DMSO (always <0.1% of final volume).

Cells transfection with siRNAs
Human G9a, DNMT1 and UHRF1-specific siRNAs and control siRNA (siC) were from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Transfections were performed with 75 nM of each siRNA using Lipofectamine RNAiMAX reagent (Invitrogen, Grand Island, NY, USA) as we previously described, [8] and following the manufacturer's instructions. Cells were harvested 48h after transfection. Gene expression was confirmed by qPCR and western blotting after transfections.

Western blotting
Cells and tissues were lysed in RIPA buffer. Histones were extracted as described below.

Gene expression and microarray analyses
RNA was extracted using the automated Maxwell system from Promega (Madison, WI, USA) according to the manufacturer's instructions. For retro-transcription RNA samples were exposed for 1min at 90°C for denaturalization followed by 1h at 37°C using a mix  of constitutive expression H3F3A as we described, [9]. and Bioconductor, [11] were used for preprocessing and statistical analysis. LIMMA (Linear Models for Microarray Data) was used to find out the probe sets that showed significant differential expression between experimental conditions. [12] Genes were selected as significant using a criterion of B>0 and |logFC|>1. Functional enrichment analysis of Gene Ontology (GO) categories, [13] was carried out using standard hypergeometric test and the gene list ranked by logFC was also analyzed with Gene Set Enrichment Analysis (GSEA), [14]. Microarray data can be downloaded from Gene Expression Omnibus (GEO) public functional genomics data repository under the accession number GSE139504.