DNMT1 Is Regulated by ATP-Citrate Lyase and Maintains Methylation Patterns during Adipocyte Differentiation

INTRODUCTION

The process of adipocyte differentiation occurs upon the activation of a well-characterized transcriptional program that enables precursor cells to take on the characteristics of differentiated adipocytes, including the ability to take up glucose in an insulin-responsive manner and store triglyceride in lipid droplets (1–3). The transcription factor peroxisome proliferator-activated receptor gamma (PPARγ) is considered the master regulator of adipogenesis, and CCAAT/enhancer-binding proteins (CEBPα, -β, and -δ) also play key roles in this process. The differentiation process is associated with broad changes in the epigenome, as characterized in recent chromatin immunoprecipitation-sequencing (ChIP-Seq) studies (4, 5). Activating marks, including H3K9ac, H3K27ac, and H3K4me2/3, increase at promoters and enhancers of genes that are induced during adipocyte differentiation (4, 5). Chromatin remodeling is initiated within the first few hours of the differentiation process, with recruitment of multiple transcription factors to key “hot spots” (6). Additionally, an epigenomic transition state has been characterized at sites bound by the glucocorticoid receptor (GR) and CEBPβ, which display transient increases in histone acetylation at early time points in adipogenesis (5).

Epigenomic changes during adipocyte differentiation are also reflected in global levels of histone marks. In previous work, we showed that levels of histone H3 and H4 acetylation in 3T3-L1 adipocytes increase during differentiation and are regulated in a manner dependent on glucose availability (7). This nutrient-dependent acetylation requires the enzyme ATP-citrate lyase (ACL), which generates acetyl coenzyme A (acetyl-CoA) through the cleavage of citrate. ACL is present in both the cytoplasm and the nucleus, suggesting that acetyl-CoA is produced in both of these compartments by this enzyme and implying that nuclear production of acetyl-CoA might be important for the regulation of nuclear acetylation events (7). Increased histone acetylation during differentiation occurs in parallel with decreased levels of repressive histone H3K9 di- and trimethylation marks (8). Silencing of the H3K9 demethylase KDM4C impairs adipogenesis, indicating that downregulation of this mark is important for differentiation (8).

Coordination of activating and repressive histone modifications during adipocyte differentiation could potentially occur passively; for example, increases in acetylation could dilute out the effects of methyltransferases acting on the same lysine residues. On the other hand, it is possible that mechanisms exist to actively coordinate these marks. Several acetyl-proteomics studies have recently been published, which have revealed acetylation to be an extremely prevalent modification in cells, with thousands of cellular proteins acetylated (9–12). Among these, a central theme emerging from these studies is that many mitochondrial and cytoplasmic metabolic enzymes are acetylated (13). At least several of these enzymes are acetylated in a nutrient-responsive manner, and acetylation may play a key role in fine-tuning metabolic regulation (13). For example, the gluconeogenic enzyme phosphoenolpyruvate carboxykinase (PEPCK) is acetylated in a glucose-responsive manner, promoting its ubiquitination and proteosomal degradation; thus, under high-glucose conditions, gluconeogenesis might be suppressed in part through increased degradation of PEPCK (14). In addition to metabolic enzymes, numerous chromatin-modifying enzymes were also found to be acetyl-proteins (9, 10). We therefore hypothesized that ACL has a role in epigenetic regulation beyond histone acetylation, through the regulation of other types of chromatin-regulatory proteins.

The DNA methyltransferase 1 (DNMT1) has been shown to be acetylated on at least 16 lysines (10, 15, 16), and acetylation has been implicated in targeting DNMT1 for degradation (15). DNA methylation can occur either de novo or as a maintenance function to maintain methylation patterns on daughter DNA strands during replication. DNMT3a and DNMT3b function predominantly in de novo DNA methylation, whereas DNMT1 is thought to be the main maintenance methyltransferase (17, 18). In this study, we examined a potential role for ACL in the regulation of DNMT1. Although DNA methylation has been well studied in other contexts, such as cancer (19, 20), its role in the regulation of adipogenesis is poorly understood. In classic studies, Taylor and Jones showed that short-term treatment of C3H10T1/2 or 3T3 cells with DNMT inhibitors such as 5-azacytidine triggered spontaneous differentiation to multiple phenotypes, including adipocyte, that emerged several weeks later (21). On the other hand, treatment of 3T3-L1 preadipocytes with DNMT inhibitors for 2 days prior to induction of differentiation impaired the cells' capacity to differentiate, suggesting that DNA methylation may be important for licensing preadipocytes for adipogenesis (22). The roles of DNA methylation and DNA methyltransferases after the induction of adipogenesis remain to be elucidated.

Upon induction of differentiation in 3T3-L1 adipocytes, cells undergo two synchronous rounds of cell division, termed mitotic clonal expansion (MCE), after which they undergo growth arrest followed by terminal differentiation (3). We hypothesized that DNMT1 might be important for maintaining patterns of DNA methylation during MCE and explored a role for ACL in the regulation of DNMT1 during adipocyte differentiation. We show here that DNMT1 plays a key role in regulating histone and DNA methylation during early adipocyte differentiation, that silencing of DNMT1 accelerates differentiation, and that ACL suppresses DNMT1 protein levels later in differentiation. Surprisingly, we failed to detect a difference in DNMT1 acetylation upon silencing of ACL. Instead, we found that ACL can suppress DNMT1 by promoting the expression of microRNA 148a (miR-148a), a microRNA known to target DNMT1. Ectopic expression of miR-148a recapitulates the early differentiation phenotype observed upon silencing of DNMT1 and can partially reverse a hypermethylation-mediated differentiation block. Overall, the results show that DNMT1 plays a role in maintaining DNA methylation patterns during early adipogenesis and that DNMT1 is suppressed later in adipogenesis in part through an ACL-miR-148a-dependent mechanism.

MATERIALS AND METHODS

RESULTS

We have previously shown that ACL-dependent production of acetyl-CoA can contribute to the regulation of nuclear histone acetylation (7). Protein levels of a key chromatin-modifying enzyme, DNMT1, were recently shown to be regulated by acetylation (15). To test whether ACL contributes to the regulation of DNMT1, we examined the effects of silencing ACL on DNMT1 during the adipocyte differentiation of 3T3-L1 and C3H10T1/2 cells. DNMT1 levels decreased by 4 days following induction of differentiation, and ACL silencing partially blocked this decrease in both cell lines (Fig. 1A and B). DNMT1 levels were similarly maintained during differentiation when cells were treated with an ACL inhibitor (Fig. 1C). ACL-dependent regulation of DNMT1 occurs at a posttranscriptional level, since DNMT1 mRNA levels were not elevated in the absence of ACL (Fig. 1A). Thus, ACL regulates DNMT1 protein levels during differentiation, suggesting that increased acetyl-CoA production during differentiation might mediate suppression of DNMT1.

To better understand the regulation of DNMT1 during adipocyte differentiation, we monitored DNMT1 mRNA and protein levels over a time course after induction of differentiation in 3T3-L1 cells. DNMT1 mRNA and protein levels increased by 24 h after induction of differentiation and decreased later in differentiation after growth arrest, consistent with a potential role for DNMT1 in maintaining DNA methylation during MCE in 3T3-L1 cells (Fig. 1D and E). DNMT3a and 3b mRNA levels, in contrast, are low during MCE and increase later in differentiation (Fig. 1D). To confirm that these patterns of DNMT1 regulation were reflective of chromatin-associated DNMT1, we fractionated 3T3-L1 cells at days 0, 1, and 6 of differentiation into cytoplasmic, nucleoplasmic, and chromatin-associated fractions. The majority of DNMT1 was present in chromatin, and chromatin-associated DNMT1 increased at day 1 and decreased later in differentiation (Fig. 1F). To assess whether similar regulation of DNMT1 also occurs in adipose tissue in vivo, we separated freshly harvested murine white adipose tissue into adipocyte and stromal-vascular fractions and examined DNMT1 levels. Consistent with the in vitro data, DNMT1 was present in the preadipocyte-containing stromal-vascular fraction but was almost undetectable in adipocytes (Fig. 1G).

We next sought to determine whether DNMT1 functionally regulates DNA methylation during early adipogenesis. DNMT1 was silenced by transfecting siRNA into preadipocytes 3 days prior to inducing differentiation. Bisulfite conversion and sequencing were performed to examine DNA methylation during differentiation. Glut4 and Pparg are both highly induced during terminal adipocyte differentiation. We examined methylation of 52 CpG dinucleotides within the Glut4 promoter and of 37 CpGs in the Pparg locus at days 0, 2, and 4 after induction of differentiation (Fig. 2A). Methylation of Glut4 and Pparg did not significantly change during the first 4 days of differentiation in control cells; however, upon silencing of DNMT1 and between days 0 and 2, during the MCE period, a pronounced decrease in DNA methylation was observed (Fig. 2A). Specifically, for Glut4, CpGs were on average 53% methylated at day 0 in control cells and this remained unchanged to day 4. In siDNMT1 cells, CpGs were on average 45% methylated at day 0, 33% at day 2, and 32% at day 4. For Pparg, CpGs assessed were on average 80% methylated at day 0 in control cells and 78% methylated at days 2 and 4. In siDNMT1 cells, CpGs were on average 75% methylated at day 0, 55% at day 2, and 53% at day 4. These data indicate that DNMT1 is critical for maintaining DNA methylation patterns during early adipogenesis.

In addition to regulating DNA methylation, DNMT1 performs scaffolding functions. The H3K9 methyltransferase G9a associates with DNMT1 at the replication fork and acts to maintain H3K9me2 during replication (28). Thus, it is possible that maintenance of histone methylation patterns during MCE is also dependent on DNMT1. We analyzed the promoters of Glut4 and Fabp4 (aP2) by chromatin immunoprecipitation (ChIP) 2 days after induction of differentiation, and H3K9me2 levels near the transcription start sites were indeed reduced in DNMT1-silenced compared to control cells (Fig. 2B).

To assess whether DNMT1-dependent regulation of these chromatin marks has a functional role in the regulation of adipogenesis in 3T3-L1 cells, we compared gene expression in the presence or absence of DNMT1. In 3T3-L1 cells, DNMT1 mRNA levels were suppressed by 94% at day 0 (day of induction, 3 days after transfection) and by 88% at day 2 of differentiation, and protein levels of DNMT1 remained nearly undetectable at day 2 (Fig. 3C). Silencing of DNMT1 led to early induction of adipocyte-specific genes, such as those for Glut4, Fabp4, adiponectin, and adipsin, as well as early upregulation of the master regulator of adipogenesis PPARγ (Fig. 3A). Modestly higher levels of FASN and LDHA were also observed at day 2 of differentiation (Fig. 3A). Silencing of DNMT1 in C3H10T1/2 cells promoted increased expression of adipocyte genes similar to that observed in 3T3-L1 cells (Fig. 3B). Some differences were observed in the sets of genes that are regulated by DNMT1 in 3T3-L1 cells and 10T1/2 cells, and effects of DNMT1 silencing in 10T1/2 cells were overall somewhat less pronounced than those in 3T3-L1 cells. These differences may reflect a differential need for DNMT1 activity in the two cell lines, since 10T1/2 cells do not undergo MCE and hence may not require DNMT1's maintenance methyltransferase activity but likely still utilize its scaffolding and repressor functions.

Phenotypically, lipid droplets were also visible earlier upon silencing of DNMT1, with small droplets apparent by 48 h after induction of differentiation (Fig. 3C). At this time point, few if any lipid droplets could be discerned in control cells (Fig. 3C). MCE still occurred in 3T3-L1 cells after silencing of DNMT1, though a small reduction in the final cell number was observed (Fig. 3D), consistent with DNMT1's role in proliferation (29). DNMT1 knockdown cells also consumed glucose at a rate comparable to or slightly higher than that of control cells upon induction of differentiation, indicating that DNMT1-silenced cells remain metabolically functional (Fig. 3E). Indeed, DNMT1-silenced adipocytes exhibited a similar capacity to take up glucose in response to insulin as control cells (Fig. 3F). These data suggest that DNMT1 functions during early adipogenesis to modulate the timing of induction of lineage-specific genes.

We next sought to determine the mechanism through which ACL regulates DNMT1 levels during differentiation. DNMT1 is an acetyl-protein, and acetylation of DNMT1 by the acetyltransferase Tip60 was shown to promote its association with the E3 ubiquitin ligase UHRF1, promoting DNMT1 ubiquitination and proteosomal degradation (15). Hence, higher levels of DNMT1 in the absence of ACL would be consistent with reduced acetylation (Fig. 1A). We first assessed whether Tip60 functionally regulates DNMT1 levels in 3T3-L1 cells. Indeed, similar to silencing of ACL (Fig. 1A), Tip60 silencing impaired suppression of DNMT1 protein levels during differentiation, without impacting DNMT1 mRNA levels (Fig. 4A). P/CAF has also been reported to be able to acetylate DNMT1 (16), although it has not been implicated in regulating DNMT1 stability. Consistently, DNMT1 levels could still be suppressed during differentiation after silencing of P/CAF (Fig. 4A). These data suggest the possibility that ACL regulates DNMT1 levels by modulating Tip60-dependent acetylation; however, it is also possible that ACL and Tip60 independently regulate DNMT1 levels. We have previously observed that some, but not all, acetylation events are subject to control by acetyl-CoA availability (7). To test whether ACL regulates the acetylation levels of DNMT1, we silenced ACL in 293T cells and examined acetylation of myc-tagged DNMT1. Surprisingly, we failed to detect a difference in DNMT1 acetylation upon silencing of ACL or upon acetate supplementation, suggesting that DNMT1 acetylation may not be strongly regulated by acetyl-CoA availability (Fig. 4B). Since we did not observe ACL-dependent regulation of DNMT1 acetylation, we sought an alternative mechanism to explain the effects of ACL on DNMT1 protein levels.

DNMT1 has also been reported in the literature to be directly or indirectly targeted by microRNAs (30–36). One of these, miR-148a, has also been shown to be induced during adipocyte differentiation and has been identified as a potential PPARγ target (37, 38). miR-148a has been previously shown to directly repress DNMT1 by binding to its 3′ untranslated region (UTR) (30). While this was shown in human cells, the sequence targeted by miR-148a is conserved in the murine DNMT1 3′ UTR (Fig. 5A). Interestingly, a second site targeted by miR-148a was also identified in the protein-coding region of human DNMT1 (33), although this site is not conserved in mice. We found that miR-148a is induced during differentiation between days 2 and 4, correlating with downregulation of DNMT1 levels (Fig. 5B). In previous work, we showed that ACL modulates expression of certain genes, such as Glut4, during adipocyte differentiation (7). We therefore tested whether ACL might similarly regulate miR-148a expression during adipocyte differentiation, and we observed that, indeed, silencing of ACL blocked induction of the miR-148a primary transcript levels during differentiation (Fig. 5C). An ACL inhibitor also impaired induction of miR-148a expression (data not shown). The role of ACL in regulating miR-148a is not general among microRNAs, since another adipocyte differentiation-induced microRNA, miR-103-1 (38), was induced equally well in the presence or absence of ACL (Fig. 5C). Expression of miR-148a was partially rescued by treating cells with acetate, demonstrating that it is regulated in an acetyl-CoA-dependent manner (Fig. 5D).

The Mir148a gene is not located within an intron and thus may be regulated independently of other genes. We examined histone acetylation at the Mir148a gene in the presence or absence of ACL and found that levels of histone H4 acetylation are reduced upon silencing ACL and can be rescued with acetate treatment (Fig. 5E). To test whether miR-148a can regulate DNMT1 in adipocytes, we ectopically expressed miR-148a in preadipocytes and induced differentiation. Expression of miR-148a blocked the induction of DNMT1 during MCE, resulting in lower levels of DNMT1 protein at days 1 and 2 of differentiation (Fig. 5F).

We next investigated whether miR-148a overexpression is sufficient to mimic the effects of DNMT1 silencing on gene expression during differentiation. Indeed, similar to silencing of DNMT1, overexpression of miR-148a resulted in an early induction of gene expression in differentiating adipocytes, with clear effects observed at day 2 (Fig. 6A). Later in differentiation, adipocyte-specific gene expression was comparable for control and miR-148a-expressing genes (data not shown). Lipid accumulation was also markedly enhanced upon expression of miR-148a (Fig. 6B). Normally, adipocytes undergo growth arrest prior to the onset of lipid droplet formation; however, the early time point at which we observed lipid droplets forming seemed to overlap the MCE period. To determine whether lipid droplet-containing cells were able to undergo cell division upon expression of miR-148a, we treated cells with Bodipy lipid dye at 44 h after induction of differentiation and imaged them live for an additional 20 h. While many of the miR-148a-transfected cells containing lipid droplets were already growth arrested, lipid droplet-containing cells could be observed to undergo cell division (Fig. 6C). This was quantitated in an independent experiment by scoring approximately 1,000 cells within 6 fields for both miR-Ctrl- and miR-148a-transfected cells between 40 and 60 h after inducing differentiation. Cells were scored positive or negative for observed cell division and for visible lipid droplets. While the number of cells undergoing division during this time frame was not higher upon overexpression of miR-148a, lipid could be detected in more cells (∼50% of cells versus ∼20% of control cells) by 60 h after inducing differentiation (Fig. 6D). Cells with and without lipid droplets were equally likely to divide (Fig. 6D). Thus, lipid accumulation and cell division are not mutually exclusive in 3T3-L1 cells, and the results suggest a role for DNMT1 in segregating these processes in time by restraining adipocyte-specific gene expression during early adipogenesis.

Not surprisingly, miR-148a-mediated repression of DNMT1 was far less potent in reducing DNMT1 levels than siRNA-mediated silencing of DNMT1. Yet, similar effects on gene expression and lipid accumulation were observed with either DNMT1 silencing or miR-148a overexpression. MicroRNAs typically target multiple mRNAs in order to implement their effects, and it is likely that miR-148a has additional targets that promote differentiation. Accordingly, silencing of DNMT1 and expression of miR-148a produced additive effects on promoting early expression of adipocyte-specific genes (Fig. 7A). Nevertheless inhibition of miR-148a impaired the decrease in DNMT1 levels during differentiation, indicating that miR-148a is indeed important in regulating DNMT1 levels during normal differentiation (Fig. 7B). miR-148a-dependent suppression of DNMT1 does not appear to be required for differentiation, however, since expression of differentiation markers was not different in control LNA and miR-148a LNA inhibitor-transfected cells, consistent with the model that miR-148a serves to suppress DNMT1 after it is no longer functionally required (Fig. 7B). Similarly, modest overexpression of DNMT1, accomplished by retroviral transduction of preadipocytes, did not suppress the cells' capacity to differentiate, indicating that the normal amounts of DNMT1 are sufficient to accomplish DNMT1's functions, and the presence of somewhat more DNMT1 does not impair differentiation (Fig. 7C). Overexpression of DNMT1 at high levels or stable overexpression of DNMT1 was not consistent with preadipocyte viability, and hence differentiation could not be tested under these conditions.

Although modest DNMT1 overexpression did not suppress differentiation, inhibition of histone demethylation causes histone hypermethylation, which does impair adipogenesis (8). Isocitrate dehydrogenase (IDH1 and IDH2) mutations occur in certain types of cancer and cause hypermethylation of histones and DNA. These epigenetic changes occur because the mutant IDH enzymes produce high levels of the metabolite 2-hydroxyglutarate (2-HG), which inhibits Jmjc histone demethylases and TET proteins, which are implicated in DNA demethylation (8, 39–41). Expression of mutant IDH proteins in 3T3-L1 adipocytes results in a hypermethylation phenotype and differentiation block (8). We therefore reasoned that if miR-148a can promote differentiation through regulation of DNA and histone methylation, it might be able to reverse the hypermethylation-mediated differentiation block caused by mutant IDH (Fig. 8A). Strikingly, in 3T3-L1 cells with IDH2-R140Q or IDH2-R172K mutants, overexpression of miR-148a partially restored differentiation, as assessed by lipid accumulation and gene expression (Fig. 8B and C). Additionally, histone methylation levels were normalized upon miR-148a expression (Fig. 8D). Hence, miR-148a can partially rescue an epigenetically mediated differentiation block.

DISCUSSION

In previous work we have shown a key role for ACL in the regulation of histone acetylation levels during adipocyte differentiation (7). In this study, we show that ACL activity also modulates levels of DNMT1 during adipocyte differentiation. Investigation of the role of DNMT1 during adipocyte differentiation demonstrated that DNMT1 expression is important for maintaining DNA and H3K9 methylation during the mitotic clonal expansion phase of adipocyte differentiation. Maintaining DNA methylation during MCE restrains inappropriate activation of adipocyte-specific gene expression and lipid accumulation while cells are dividing, and DNMT1 silencing results in accelerated adipocyte differentiation.

It is not fully clear why it would be advantageous to inhibit induction of adipocyte-specific genes early in differentiation. High glucose consumption in adipocytes has been shown to result in elevated production of reactive oxygen species (ROS) (42). It is possible that by preventing cells from taking on their metabolic functions during the mitotic clonal expansion period, ROS-induced damage to DNA during MCE might be minimized. Perhaps more likely is that methylation may serve to ensure that adipocyte-specific genes are repressed until cells receive appropriate signals to differentiate. This interpretation is consistent with a study by Taylor and Jones (21), which showed spontaneous adipocyte differentiation after treatment with DNMT inhibitors. We observed that about 50% of CpGs within the Glut4 promoter and 80% of those at Pparg were methylated (Fig. 2A), and relatively high (at least 50%) methylation was previously observed at the adiponectin and CEBPα promoters in adipocytes as well (8). Hence, maintaining relatively high levels of methylation at promoters of adipocyte-specific genes could serve as a mechanism to suppress induction of these genes until certain signaling conditions are met. For example, in vivo, precursor cells contained in the stromal-vascular compartment of adipose tissue express the adipogenic transcription factor PPARγ (43), which is necessary and sufficient for adipogenesis (1), yet these progenitor cells presumably remain undifferentiated until receipt of cues to differentiate. Consistent with the possibility that DNA methylation restrains inappropriate differentiation, gene ontology analysis of highly methylated genes showed enrichment for genes involved in physiological responses to stimuli (44). Hence, we speculate that DNMT1-dependent DNA methylation may serve to silence these adipocyte-specific genes, even if PPARγ is expressed, until cells receive the proper signaling cues to differentiate.

Mitotic clonal expansion is critical for differentiation of 3T3-L1 cells and MEFs into adipocytes (1, 3). In humans, new adipocyte formation occurs throughout childhood and adolescence, with fat cell numbers plateauing around age 20, regardless of body fat mass (45). Adipocytes are replaced in adults at a rate of about 10% each year, indicating that some new differentiation occurs throughout the life span (45). Whether MCE is also important for adipogenesis in vivo is not fully clear, although some evidence suggests that it may play a role. For example, mice lacking E2F1, a transcription factor important for MCE, exhibit reduced body weight and fat weight gain when fed a high-fat diet (46). Regardless, PPARγ-expressing preadipocytes do divide in vivo (43), and DNMT1 is expected to be important in these cells for sustaining DNA methylation patterns. Consistently, DNMT1 levels are higher in the preadipocyte-containing stromal-vascular fraction of adipose tissue than in adipocytes (Fig. 1G).

The results of this study indicate that ACL regulates DNMT1 protein levels in addition to histone acetylation levels during adipocyte differentiation (Fig. 9). In professional metabolic cells such as adipocytes, regulating chromatin modifications in a nutrient-responsive manner may allow environmental modulation of gene expression. In support of this possibility, we have previously shown that ACL modulates the expression of a subset of genes involved in glucose metabolism, allowing nutrient-sensitive acetyl-CoA production to influence metabolic gene expression (7). The extent to which ACL-dependent regulation of DNMT1 regulates histone and DNA methylation levels is not yet clear, since ACL seems to function in downregulating DNMT1 levels later in differentiation when it is no longer required for the maintenance of methylation patterns during DNA replication. Inhibition of miR-148a does not suppress differentiation, suggesting that DNMT1's primary role is in the early part of differentiation. On the other hand, expression of mutant IDH proteins in preadipocytes results in hypermethylation and inhibition of capacity to differentiate. miR-148a partially rescues differentiation and restores normal histone methylation levels in these cells, likely through regulation of DNMT1 and additional targets. Interestingly, silencing of DNMT1 was not sufficient to restore differentiation in this context, highlighting the notion that microRNAs mediate their effects through regulation of networks of genes. More work is needed to understand the mechanisms through which this occurs, as well as whether miR-148a or other microRNAs could provide any therapeutic benefit in the context of IDH mutant cancers. Notably, miR-148a has been reported to promote both myogenesis and osteoclastogenesis (47, 48), suggesting that it can promote differentiation in multiple contexts. miR-148a is also suppressed in several types of cancer (49–55). Reexpression of miR-148a in cancer cells may promote a more differentiated state, since the epithelial-to-mesenchymal transition was suppressed upon expression of miR-148a in both hepatocellular carcinoma and non-small-cell lung cancer cells (55, 56).

In summary, this study demonstrates that DNMT1 is critical for maintaining DNA and histone H3K9 methylation levels during MCE and that it plays a role in restraining adipocyte-specific gene expression and lipid accumulation during early adipogenesis. Hence, DNMT1 may be a key player in separating proliferation and lipid accumulation temporally during 3T3-L1 differentiation. We also show that ACL contributes to the downregulation of DNMT1 protein levels after MCE and that this occurs at least in part through regulation of miR-148a, which targets DNMT1. miR-148a also promotes differentiation, likely through regulation of DNMT1 and other targets.