Histone H3 Lysine 4 Dimethylation Signals the Transcriptional Competence of the Adiponectin Promoter in Preadipocytes*

Adipogenesis is regulated by a coordinated cascade of sequence-specific transcription factors and coregulators with chromatin-modifying activities that are between them responsible for the establishment of the gene expression pattern of mature adipocytes. Here we examine the histone H3 post-translational modifications occurring at the promoters of key adipogenic genes during adipocyte differentiation. We show that the promoters of apM1, glut4, gpd1, and leptin are enriched in dimethylated histone H3 Lys4 (H3-K4) in 3T3-L1 fibroblasts, where none of these genes are yet expressed. A detailed study of the apM1 locus shows that H3-K4 dimethylation is restricted to the promoter region in undifferentiated cells and associates with RNA polymerase II (pol II) loading. The beginning of apM1 transcription at the early stages of adipogenesis coincides with promoter H3 hyperacetylation and H3-K4 trimethylation. At the coding region, H3 acetylation and dimethylation, as well as pol II binding, are found in cells at later stages of differentiation, when apM1 transcription reaches its maximal peak. This same pattern of histone modifications is detected in mouse primary preadipocytes and adipocytes but not in a related fibroblast cell line that is not committed to an adipocyte fate. Inhibition of H3-K4 methylation by treatment of 3T3-L1 cells with methylthioadenosine results in decreased apM1 gene expression as well as decreased adipogenesis. Taken together, our data indicate that H3-K4 dimethylation and pol II binding to the promoter of key adipogenic genes are distinguishing marks of cells that have undergone determination to a preadipocyte stage.

The influence exerted by the post-translational modifications of histones over the regulation of gene expression has been extensively studied in the past few years. Numerous studies have shown a clear link between the pattern of histone modifications found at promoter regions and gene transcription, thus leading to the statement of the histone code hypothesis (1), which postulates that the pattern of histone post-translational modifications in a locus considerably extends the amount of information conveyed by the genomic code. Histone H3 and H4 hyper-acetylation in promoter regions is closely correlated with gene activation in organisms ranging from yeast to mammals, and transcriptionally active euchromatin regions are highly enriched in acetylated histones (1)(2)(3)(4)(5). Unlike acetylation, histone H3 methylation can be equally associated with either transcriptional activation or repression. Methylation of the lysine residue Lys 4 of histone H3 (H3-K4) 3 correlates with activation of gene expression in most systems (2, 4 -7), whereas H3 Lys 9 (H3-K9) methylation is involved in the establishment and maintenance of silent heterochromatin regions (8). Moreover, lysine residues can be mono-, di-, or trimethylated in vivo, thus providing a further layer of complexity and exponentially increasing functional diversity (9,10). The recent identification of LSD1, the first histone demethylase to be characterized, which shows specificity for the mono-or dimethylated lysine 4 residue of histone H3 (11), suggests that methylation is subjected to dynamic regulation by a higher turnover rate than initially thought. Therefore, histone methylation may play a role in acute regulation of gene expression, just like histone acetylation does. Histone acetylation and methylation can also be detected at the coding regions of genes, in some cases correlating with elongation by RNA polymerase II (Pol II) (7,(12)(13)(14)(15)(16). Thus, epigenetics has become a factor that cannot be overlooked when unraveling the regulation of gene transcription in cells, either acutely during interphase or programmatically throughout complex processes of development and differentiation.
Adipogenesis, or the development of adipocytes from undifferentiated precursor cells, is a complex process that has been extensively studied at the hormonal and transcriptional levels. Mouse 3T3-L1 fibroblasts provide an in vitro model that recapitulates faithfully the main steps of the in vivo differentiation process (17). A well established hormonal stimulation initiates differentiation in these cells by activating a cascade of transcription factors that culminates with activation of CCAAT/enhancer-binding protein ␣ and PPAR␥ (17). These two factors are the main regulators of adipocyte gene expression and are between them responsible for the transcription in differentiated cells of many genes, including the adipokine apM1 (adiponectin) (18,19) and the glucose transporter glut4 (20,21).
Interestingly, a number of reports have also demonstrated a role during adipogenesis for coregulators with chromatin-modifying activities. Thus, induction of high level PPAR␥ expression in fibroblasts requires recruitment of SWI/SNF to its hyperacetylated promoter (22). PPAR␥, in its turn, recruits to its target promoters the histone acetyltransferases CREB-binding protein and p300, and this interaction is essential for adipocyte differentiation (23,24). On the other hand, the retinoblastoma protein RB is associated with the histone deacetylase HDAC3 and blocks adipogenesis in cycling fibroblasts by binding to PPAR␥, thus resulting in recruitment of deacetylase activity to the target promoters of the factor (25) and inhibition of adipogenic gene expression. In accordance with these results, histone deacetylase inhibitors have been recently shown to facilitate adipogenesis from undifferentiated fibroblasts (25,26), whereas HDAC1 overexpression blocks adipogenesis (26,27). Taken together, all of these data indicate that the post-translational modification of histones participates in the regulation of gene expression during the differentiation process of adipocytes.
In the present study, we examine the pattern of histone H3 acetylation and methylation found across the adiponectin locus throughout adipocyte differentiation. We describe a role for histone H3-K4 dimethylation in recruiting Pol II and marking the adiponectin gene as "poised" for transcription in undifferentiated fibroblasts that do not yet express it. Decrease of the H3-K4 methylation mark at the apM1 promoter by incubation of the cells in the presence of an inhibitor of methyltransferase activity results in decreased expression of the gene and reduced adipogenesis. The same pattern can be extended to other adipogenic genes such as glut4 or gpd1 (glycerol-3-phosphate dehydrogenase).

EXPERIMENTAL PROCEDURES
Cell Culture and Differentiation-Mouse 3T3-L1 and 10T1/2 fibroblasts were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (Sigma). Differentiation of 3T3-L1 cells to adipocytes was induced by treatment of 48 h postconfluent cells (designated D0) with an adipogenic mixture consisting of 1 g/ml insulin, 1 M dexamethasone, and 0.5 mM IBMX (all reagents from Sigma) in the presence of 10% fetal calf serum. The differentiation medium was withdrawn 2 days later (D2) and replaced with medium supplemented with 10% fetal calf serum and 1 g/ml insulin. After 3 more days in insulin-containing medium (D5), the cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal calf serum.
In order to inhibit AdoMet-dependent methyltransferase activity during differentiation, 1 mM 5Ј-methylthioadenosine (MTA) was added to 3T3-L1 fibroblasts 24 h after reaching confluence (D1) and maintained for 24 h more (thus reaching D0). The medium was then changed to fresh medium containing insulin, dexamethasone, and IBMX as described above but supplemented in this case with 0.5 mM MTA. For the rest of the differentiation process, the medium was replaced every 24 h in order to add fresh MTA.
Isolation of Mouse Preadipocytes and Adipocytes-Isolated epididymal fat tissue obtained from 6-week-old male C57BL/6J mice was rinsed briefly with PBS and minced with a razor blade in collagenase solution (0.2 mg/ml collagenase A in 100 mM HEPES, 120 mM NaCl, 4.8 mM KCl, 1 mM CaCl 2 , and 4.9 mM glucose, pH 7.4). The mixture was then allowed to digest for 30 min at 37°C with gentle shaking. The resulting cell suspension was allowed to settle for 5 min to separate into a supernatant containing adipocytes and an inferior layer composed mainly of preadipocytes. The adipocyte-containing supernatant was recovered by pipetting, and the infranatant was filtered through a 60-m cell strainer (BD Biosciences) to obtain the preadipocytes.
Oil Red O Staining-To assess the progression of cytoplasmic fat accumulation, intracellular triglyceride was stained by Oil Red O (Sigma). Briefly, growing cells were washed gently with PBS and stained with Oil Red O solution (0.36% Oil Red O in 60% isopropyl alcohol) for 1 h at 37°C. Excess stain was removed with 60% isopropyl alcohol, and cells were washed with water before being photographed under a light microscope.
RNA Isolation, Semiquantitative RT-PCR and Real Time RT-PCR-Total RNA was extracted using TRIzol (Invitrogen) according to the instructions of the manufacturer. Random-primed cDNA synthesis was performed at 37°C starting with 1 g of RNA, using the high capacity cDNA archive kit (Applied Biosystems, Foster City, CA). Adiponectin gene expression was measured by real time PCR in an ABI Prism 7900HT real time PCR system using the DNA double strand-specific SYBR green I dye for detection (Applied Biosystems). Results were normalized to abl1 expression levels. Expression levels of the other genes studied were measured by semiquantitative RT-PCR. Gene products of interest were co-amplified with an internal control gene (␤-actin or tbp (TATA-box-binding protein)) at a low cycle number to ensure that the two products were in the exponential phase of amplification. Amplification products were resolved on ethidium bromide-stained acrylamide gels, and low exposure images were analyzed using Image J software (28). Primers were designed to span an intron using Primerselect 4.0 (DNASTAR Inc., Madison, WI) software. The primer sequences used are listed on supplemental Table 1.
Chromatin Immunoprecipitation Assays (ChIP)-ChIPs were performed and analyzed as described (29,30) with a few modifications. Briefly, in the case of 3T3-L1 cells and primary preadipocytes, cells fixed for 10 min with 1% formaldehyde were lysed with SDS (1% SDS, 50 mM Tris, 10 mM EDTA, pH 8.0). In the case of primary adipocytes, nuclei were prepared prior to SDS lysis by incubating the fixed cells in an adipocyte lysis buffer (5 mM PIPES, 80 mM KCl, and 0.5% Igepal, pH 7.9) for 1 h at room temperature. DNA was sonicated to obtain 500 -1000-bp fragments with a Branson 150 sonifier at 50% maximal power for six pulses of 30 s. The antibodies used were anti-acetyl-H3-K9/K14, anti-dimethyl-H3-K4, anti-trimethyl-H3-K4 (Upstate Biotechnology, Inc., Lake Placid, NY), and anti-Pol II C-terminal domain (Abcam, Cambridge, UK). Protein A-Sepharose blocked with bovine serum albumin and salmon sperm DNA was used to immunoprecipitate the complexes, except in the case of the Pol II ChIPs that were recovered using pan-mouse IgG-Dynabeads (Dynal Biotech).
ChIPs were analyzed by co-amplification of the PCR product of interest with both a positive control (the housekeeping ␤-actin or pgk1 (phosphoglycerate kinase 1) promoters) and a negative control (ins2 (insulin II), glut2, or myod1, all of them genes that are expressed in neither fibroblasts nor adipocytes). Multiplex PCR conditions were adjusted to ensure nonsaturation kinetics and similar amplification efficiencies for all amplicons within a reaction. Primers were designed to amplify segments located in a region of ϳ100 bp from the transcription initiation site of selected genes for analysis of promoter regions. In the case of apM1, primers were also designed within an exon and in the 3Ј-untranslated region zone for analysis of the coding and terminal regions, respectively. Primer sequences are listed in supplemental Table  1. Amplification products were run in a 12% acrylamide gel and stained with ethidium bromide, and low exposure images were analyzed using Image J software. Each PCR was performed at least twice with samples resulting from three independent ChIP experiments.
Acid Extraction of Histones and Western Blotting-3T3-L1 cells at different days of the differentiation process were scraped into ice-cold PBS, pelleted, and resuspended in 1 ml of lysis buffer (10 mM HEPES, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM dithiothreitol, and 1.5 mM phenylmethylsulfonyl fluoride, pH 7.9). Hydrochloric acid was added to a final concentration of 0.2 N, and samples were incubated on ice for 30 min. The clarified lysates containing acid-soluble proteins were dialyzed sequentially against 0.1 N acetic acid and H 2 O. Protein concentration was determined with the BCA protein assay (Pierce) using bovine serum FIGURE 1. H3-K4 dimethylation is the first modification detected in the promoter regions of adipogenic genes during adipocyte differentiation. A, expression of apM1, glut4, gpd1, adipor1, adipor2, and sod2 increases during adipogenesis, as detected by RT-PCR. The genes of interest were co-amplified with either tbp or actb as housekeeping controls, except for adipor1 and adipor2. B, to assess the histone modifications present at the apM1 promoter, ChIP assays were performed using 3T3-L1 cells at different stages of adipogenesis. Samples were immunoprecipitated with anti-acetylated (K9/K14) (top), anti-trimethylated (K4) (middle) and anti-dimethylated (K4) (bottom) histone H3 antibodies. A representative experiment for each antibody immunoprecipitation is shown at the left. The right panels show the densitometric analysis of the data of three independent experiments performed in duplicate. The intensity of the apM1 band (in arbitrary units) was corrected by the intensity of its corresponding actb band for each individual amplification, and the resulting ratio was multiplied by 100 to obtain percentage values. C, histone modifications detected at the promoter regions of glut4 and gpd1 display the same pattern as in the apM1 promoter, whereas the promoters of adipor2 and sod2 are already acetylated at D0. In each case, the gene of interest was co-amplified with both a positive (actb) and a negative (myod1, glut2, or ins2) control to assess the specificity of the amplification. The nonimmune (NI) lane shows the result of a control immunoprecipitation using nonimmune rabbit serum. The input lane illustrates the amplification rates of the three products when they are present at equimolar concentrations in a DNA sample prior to immunoprecipitation. *, p Ͻ 0.005; **, p Ͻ 0.0005. Diff days, differentiation days; NI, nonimmune.
albumin as a standard. Equal amounts of protein were resolved by 16% SDS-PAGE using the Tris-Tricine buffer system and transferred to a 0.2-m nitrocellulose membrane (Schleicher & Schuell). The primary antibodies were diluted 1:2,000 in PBS supplemented with 1% nonfat milk and visualized by blotting with horseradish peroxidase-conjugated secondary antibodies. Chemiluminescence was detected using the ECL reagents (GE Healthcare Bio-Sciences, Uppsala, Sweden), in a LAS3000 Lumi-Imager (Fuji Photo Film Inc., Valhalla, NY).
Immunocytochemistry-Undifferentiated 3T3-L1 cells were plated on 4-well LabTek chambers (Nunc, Roskilde, Denmark) the day before staining. Differentiated 3T3-L1 adipocytes growing in 6-well dishes were trypsinized, and 5,000 cells were placed onto glass slides using a cytospin centrifuge. Cells were fixed in 9:1 ethanol/acetic acid for 1 min, rinsed with PBS, and blocked with 3% bovine serum albumin in PBS for 1 h at room temperature. Primary antibodies (diluted 1:1,000) were incubated overnight at 4°C in a wet chamber. Slides were then washed with PBS and incubated with the fluorescence-conjugated secondary antibody (Cy3-conjugated anti-rabbit from Jackson Immunoresearch, West Grove, PA) at a final dilution of 1:500 in PBS for 1 h at room temperature. After being washed in PBS, the preparations were mounted and coverslipped using FA mounting fluid (Difco). Images were collected using a Leica DC-300F fluorescence microscope and processed using Adobe Photoshop 7.0 (Adobe Systems Inc., San José, CA).
Statistical Analyses-Statistical significance of the changes of histone modifications detected at the apM1 promoter throughout adipogenesis was assessed by two-tailed Student's t test using the mean Ϯ S.D. calculated from six independent determinations performed for each differentiation day and obtained from three separate experiments performed in duplicate. Correlation coefficients that assess the linear association between transcriptional activity and promoter acetylation or methylation were computed between the log-transformed real time RT-PCR values for the apM1 RNA at different days of the differentiation process and the log-transformed densitometric analysis of each modification at the same day, referred to the level of that same modification found at the actb (␤-actin) promoter, which was co-amplified and used as an endogenous control for immunoprecipitation efficiency.

Dimethylation of Histone H3-K4 Is the First Modification Observed at the Promoters of Adipogenic Genes during Adipocyte Differentiation-
The mouse 3T3-L1 cell line is frequently used as a model to study adipocyte differentiation. The striking morphological and transcriptomic changes taking place during this process have been extensively detailed (31,32), but the role of histone post-translational modifications has not yet been thoroughly established. We focused our attention on a set of genes that are activated throughout adipogenesis. The adipokines apM1 and leptin, the glucose transporter glut4, and the enzyme gpd1 were not expressed in undifferentiated 3T3-L1 fibroblasts; transcription of these genes started during the early stages of adipogenesis, slowly rising and reaching a peak in mature adipocytes (Fig. 1A). The adiponectin receptors adipor1 and adipor2 and the enzyme sod2 (superoxide dismutase, mitochondrial), on the other hand, were already expressed in 3T3-L1 fibroblasts, although at low levels; transcription of these genes increased steadily throughout adipogenesis (Fig. 1A). We studied the presence of histone H3-K9/K14 acetylation, usually considered to be a mark for transcriptional activation (1,33), at the promoter region of the apM1 gene in 3T3-L1 cells at different days of the differentiation process by means of ChIP assays. Interestingly, H3-K9/K14 acetylation was not detected in undifferentiated fibroblasts, slowly rising and reaching a peak in mature adipocytes (Fig. 1B, top). This pattern closely mirrors the expression profile of the apM1 gene.
Another histone modification linked to transcriptional activation, H3-K4 trimethylation, followed a pattern similar to that of acetylation, with trimethylation slightly preceding the appearance of acetylation on the same region (Fig. 1B, middle). In contrast, H3-K4 dimethylation was already detected in undifferentiated 3T3-L1 cells and increased throughout adipogenesis, following a biphasic pattern (Fig. 1B, bottom). A similar study of the promoters of glut4, gpd1, and leptin shows that this temporal sequence of events is conserved in those other genes, with H3-K4 dimethylation the only mark detected in 3T3-L1 fibroblasts. (Fig.  1C and supplemental Fig. 1). On the other hand, the promoter regions of adipor1, adipor2, and sod2, which were already expressed in 3T3-L1 fibroblasts, were associated with hyperacetylated histone H3 in addition to dimethylated H3-K4 in undifferentiated cells. Moreover, H3-K4 trimethylation was also detected at the promoters of sod2 and adipor1 but not adipor2 in 3T3-L1 fibroblasts. However, the levels of histone acetylation and di-and trimethylation of these promoters increased slightly in mature adipocytes, coinciding with the maximal transcription of the genes, particularly in the case of adipor1 and adipor2 (Fig. 1C and supplemental Fig. 1).
To strengthen the link between active transcription and histone acetylation and methylation, we examined the promoter of pref1 (preadipocyte factor-1), a gene that is expressed in undifferentiated 3T3-L1 cells but whose expression decreases throughout adipogenesis and is finally silenced in mature adipocytes (34). At the promoter of pref1, all of these histone modifications associated with transcription were detected in undifferentiated fibroblasts but steadily decreased as adipogenesis progressed and transcription of the gene was silenced. In mature adipocytes, only low levels of H3-K4 dimethylation could be detected at the silenced pref1 promoter, as a lingering mark of past transcriptional events on that locus (supplemental Fig. 1).
To check if this recurring pattern is a peculiarity of this cell line or if it is also found in vivo, we isolated primary mouse preadipocytes and adipocytes. apM1 was not expressed in preadipocytes but was highly expressed in mature mouse adipocytes ( Fig. 2A). Supporting our findings with 3T3-L1 cells, the promoter region of the gene was already dimethylated in preadipocytes, whereas the three histone modifications studied were detected at the same region in differentiated adipocytes (Fig. 2B).
As another control, we used the fibroblast cell line 10T1/2, which can also be differentiated into adipocytes by hormonal stimulation but only after prior treatment of the cells with the DNA methyltransferase inhibitor 5-azacytidine (35). Surprisingly, in parental 10T1/2 cells not treated with 5-azacytidine, none of the histone modifications studied were detected in the apM1 promoter (see Fig. 5B), and hormonal stimulation with insulin, dexamethasone, and IBMX did not alter this situation (data not shown). Taken together, these data suggest that histone H3-K4 dimethylation at the promoter regions of adipogenic genes in 3T3-L1 fibroblasts or primary preadipocytes acts as an epigenetic mark that indicates that genes that are not yet expressed are transcriptionally competent, ready to be expressed later on during the differentiation process. 10T1/2 fibroblasts, which are not primed for adipocyte differentiation, do not present this epigenetic mark in those same genes.
Histone Modifications Spread through the apM1 Locus, Correlating with the Beginning of Gene Transcription-A host of previous studies have shown elevated levels of histone H3 hyperacetylation and K4 methylation localized to the 5Ј-proximal regions of transcriptionally active genes (2,7,36). The presence of the same modifications downstream of the transcription initiation site, on the other hand, is not as preeminent (12,37,38). We designed primers located in the coding region (exon 2) and the 3Ј-untranslated region of the apM1 locus to examine those regions for the presence of histone modifications (Fig.  3A). The first modification found at the coding region was again H3-K4 dimethylation (at D3 ; Fig. 3B, top), followed by H3 acetylation (at D5; Fig. 3B, middle). H3-K4 trimethylation was delayed in this region, being only barely detectable in fully differentiated adipocytes at D7 (Fig. 3B,  bottom), concurring with the maximal peak of gene expression. At the 3Ј-terminal region of the apM1 locus, we were only able to detect low levels of H3-K4 dimethylation transiently at D5 (Fig. 3C).
We correlated the expression level of apM1 as measured by real time RT-PCR with the densitometric analysis of histone modifications detected at the apM1 promoter and referred to the levels of the same modification detected at the ␤-actin promoter, which we used as an endogenous control for immunoprecipitation efficiency. A high correlation was found between gene expression and both H3-K9/K14 acetylation and H3-K4 trimethylation at the promoter region (Fig. 3D). In fact, these two modifications are also closely correlated between them. H3-K4 dimethylation at the promoter was detected well in advance of the beginning of transcription, and thus it is not surprising that it does not correlate well with gene expression (Fig. 3D).
The global levels of the three histone modifications studied were examined by Western blotting using acid extracts of 3T3-L1 cells at different stages of the differentiation process and by immunocytochemical staining of undifferentiated fibroblasts and differentiated cells (D5). H3 acetylation and H3-K4 trimethylation increased from fibroblasts to adipocytes, whereas H3-K4 dimethylation slightly decreased (Fig. 4A). The increase of H3 acetylation and H3-K4 trimethylation is also evident by immunocytochemical staining (Fig.  4B), but the most striking feature is the change of the distribution pattern of H3-K4 dimethylation during differentiation. Dimethylated H3-K4 was found strongly enriched along the nuclear border in undifferentiated 3T3-L1 fibroblasts but was dramatically rearranged in mature adipocytes (Fig. 4B, lower panels).
RNA Polymerase II Contacts the apM1 Promoter in Undifferentiated 3T3-L1 Cells-To investigate the recruitment of Pol II to the apM1 promoter region, we performed chromatin immunoprecipitations using an anti-Pol II antibody directed against the unmodified C-terminal domain. We observed that Pol II was already bound to the apM1 promoter in fibroblasts at D0 (Fig. 5A), although adiponectin was not yet expressed. The extent of the binding increased throughout adipogenesis, being easily detected in fully differentiated adipocytes (from D7 onward), when apM1 transcription reached its maximal level.
In 10T1/2 fibroblasts on the other hand, Pol II was not detected at the apM1 promoter region (Fig. 5B), and incubation of these cells with insulin, dexamethasone, and IBMX did not alter this situation (data not shown). These results indicate that the presence of dimethylated histone H3-K4 at the promoter region of the apM1 gene is associated with binding to the same region of an as yet inactive Pol II enzyme, prior to the start of transcription. These two events were observed in 3T3-L1 cells, which were already primed for differentiation, but not in a related fibroblast cell line which was not able to differentiate in response to hormonal stimulation. In the coding region of adiponectin, Pol II was detectable from D3/D4, coinciding with the start of high level transcription and the appearance of histone H3-K4 dimethylation in this region (Fig. 5C).
The Presence of H3-K4 Dimethylation at the Promoter Region Is Necessary for the Expression of apM1 in 3T3-L1 Cells-To check if the presence of H3-K4 dimethylation at the promoter regions of adipogenic genes is necessary for their later expression, we treated differentiating 3T3-L1 cells with MTA, an inhibitor of AdoMet-dependent methyltransferases that has been shown to inhibit histone H3-K4 methylation (36). We treated postconfluent 3T3-L1 fibroblasts 24 h before the stimuli of differentiation with 1 mM MTA to inhibit histone methylation prior to the start of adipogenesis. From D0 onward, 0.5 mM MTA was added to the cells along with the adipogenic mixture, with medium being replaced every 24 h in order to add fresh inhibitor. Immunocytochemical analysis of 3T3-L1 cells differentiated in the absence or presence of MTA shows that H3-K4 dimethylation is decreased by D5 of the differentiation process (Fig. 6A).
ChIP analysis of the apM1 promoter region at D4 shows a partial decrease of H3-K4 dimethylation (Fig. 6B). In order to investigate how inhibition of histone methylation affects apM1 expression, we performed semiquantitative RT-PCR. apM1 RNA was significantly decreased by D3, D4, and D5 of the differentiation process (Fig. 6C). The RNA levels of other adipogenic marker genes such as glut4 and aP2 were also decreased (data not shown). We could not detect a significant decrease in the RNA and protein level of transcription factor PPAR␥ (data not shown), which has been shown to regulate expression of apM1 (18), thus suggesting that reduced apM1 expression is not secondary to decreased PPAR␥.
Oil red staining showed a significant decrease of fat accumulation in the cytoplasm of cells differentiated in the presence of MTA both by D5 and D7 of the differentiation process (Fig. 6D), thus suggesting an important role for histone methylation in the regulation of adipogenesis.  . RNA polymerase II is already bound to the apM1 promoter in undifferentiated fibroblasts 3T3-L1. A, occupancy of Pol II was assessed by ChIP in 3T3-L1 cells at different stages of adipogenesis. apM1 promoter was co-amplified with the promoter regions of the highly expressed actb gene as a positive control and the silent myod1 gene as a negative control. Pol II contacts the apM1 promoter already in undifferentiated 3T3-L1, but the rate of occupancy increases in later stages of adipogenesis, coinciding with elevated transcription levels. B, ChIP assays performed using 10T1/2 fibroblast show that the apM1 promoter in those cells is neither acetylated nor di-or trimethylated. Pol II is not detected in the apM1 promoter in 10T1/2 cells. C, ChIP assay showing recruitment of Pol II to the coding region of the apM1 gene at different days during the differentiation process. In this region, Pol II was not detected at D0, D3 being the earliest time point tried at which we were able to detect it. myod1 is co-amplified as a negative control for nonspecifically precipitated DNA. Diff days, differentiation days; NI, nonimmune.

DISCUSSION
Recent studies have shown that the post-translational modifications of histones play a key role in regulating the programmatically determined gene expression patterns arising during differentiation of several cell types. In the ␤-globin locus, which has become a paradigm for the study of chromatin modifications during development, H3 acetylation and H3-K4 methylation closely correlate with gene activity throughout development (39). Increased histone hyperacetylation has been detected in the promoters of several cytokines during T-cell differentiation (40) and in the promoters of myogenic genes during muscle differentiation (41). We study here the pattern of two key histone post-translational modifications, H3-K9/K14 acetylation and H3-K4 di-and trimethylation, in several adipogenic genes throughout adipocyte differentiation. We focused most of our work on the apM1 gene, which can be considered a prototype for adipogenic genes, since it is expressed exclusively in adipocytes (42) and its expression is controlled by master adipocyte regulators PPAR␥ and CCAAT/enhancerbinding protein ␣ (18,19). Adiponectin transcription is turned on 2 days after the initiation of adipocyte differentiation and maintained at a high level in mature adipocytes. Adiponectin gene expression is decreased in a variety of disorders associated with insulin resistance, such as cardiovascular disease, obesity, and diabetes mellitus (42,43). We show here that the promoters of apM1 and other key adipogenic genes, such as glut4, leptin, or gpd1, are not associated with histone H3 hyperacetylation but are enriched in histone H3-K4 dimethylation in primary preadipocytes and 3T3-L1 undifferentiated fibroblasts, which do not yet express these genes. This signal is exclusive to adipogenic promoters, since it was not detected in other silent genes, such as glut2, ins2, or myod1. This epigenetic mark labels the adipogenic genes as competent for transcription and is associated with loading of Pol II to the same DNA regions. This scenario is similar to that described for the ␤-globin locus, where significant levels of H3-K4 methylation were found in the promoters of inactive genes, signaling the open state of the chromatin at those regions (7). Interestingly, H3-K4 dimethylation exhibited a characteristic distribution along the nuclear border when examined by immunocytochemical staining in 3T3-L1 fibroblasts. The region along the nuclear envelope is occupied by middle to late replicating chromatin in 3T3-L1 cells, mostly including silent genes (44). A recent report has shown that the neurogenic gene mash1 is located in the peripheral nuclear region in ES cells, where it is not expressed but relocates toward the interior of the nucleoplasm in differentiated neurons, at the time that its transcription is up-regulated (45). We observed a striking rearrangement of the dimethyl-H3-K4-enriched chromatin in differentiated 3T3-L1 adipocytes. The enrichment along the nuclear border disappeared, and H3-K4 dimethylation was then mainly found in the central region of the nucleus. Our data suggest that adipogenic genes are regulated by changes in chromatin structure and location, just as has been shown for mash1, but further experiments are needed to ascertain that.
It is noteworthy that 10T1/2 fibroblasts do not display H3-K4 dimethylation at the promoter region of the adipogenic genes studied. Pluripotential 10T1/2 cells can give rise to several specialized cell types, including adipocytes, but only after prior treatment with the DNA methyltransferase inhibitor 5-azacitydine (35). Unipotential 3T3-L1 fibroblasts, on the other hand, have undergone determination and can only be maintained as preadipocytes or be differentiated into adipocytes. Our data suggest that H3-K4 dimethylation is a distinguishing mark of cells that have undergone determination to a preadipocyte stage. Interestingly, mouse primary preadipocytes display high levels of H3-K4 dimethylation at the apM1 promoter, thus indicating that these cells have already been committed to an adipogenic fate.
Histone H3 acetylation is frequently associated with gene activity (2,4,5). Several studies have demonstrated a positive effect of inhibiting histone deacetylase activity on adipogenesis (25)(26)(27). A recent report contradicted those results by describing a inhibitory rather than stimulatory effect of the histone deacetylase inhibitor valproic acid on adipogenesis, thus arguing for a positive role of Class I histone deacetylases on adipocyte differentiation (46). In our study, the promoters of apM1, leptin, glut4, and gpd1 are not acetylated in 3T3-L1 fibroblasts, which do not express these genes. However, sod2, adipor1, and adipor2, which are already expressed in undifferentiated cells, display detectable levels of H3-K9/K14 acetylation. H3 hyperacetylation increased throughout adipogenesis in the apM1 promoter and spread to the coding region, and global H3-K9/K14 acetylation also increased. On the other hand, the promoter of pref1, a gene that follows an inverse pattern of transcriptional activity, being highly expressed in undifferentiated cells but progressively silenced throughout adipogenesis (34), was strongly acetylated and methylated in fibroblasts, but the modifications decreased in parallel to gene silencing. These results indicate that histone deacetylase activity at particular loci is also necessary for the correct regulation of transcription during adipogenesis.
We found a significant correlation between both acetylation and H3-K4 trimethylation at the apM1 promoter region and gene expression. In fact, both post-translational modifications were closely correlated between them. These results agree with previous data showing that all "positive" histone modification marks are likely to be present at the same chromatin regions (4,38). Promoter H3-K4 dimethylation, on the other hand, did not correlate with expression. These results are in agreement with previous data showing a global correlation between gene expression and promoter H3 and H4 acetylation, but not H3-K4 dimethylation, in yeasts (47). However, in our case, we are referring to different transcriptional states of the same gene rather than to the total transcripts of a cell at a given time. We could not find significant levels of H3-K4 trimethylation outside the promoter region of the apM1 gene, although high levels of acetylation and dimethylation were easily detected. Previous studies have shown that H3-K4 trimethylation is associated with promoter clearance by Pol II (14 -16). Untargeted action of Set1 is proposed to account for genome-wide H3-K4 dimethylation levels in yeast (15). However, we could not find H3-K4 methylation at the promoters of a number of silent genes, such as ins2, or for that matter, on the coding region of apM1 in 3T3-L1 fibroblasts. Thus, if untargeted histone methyltransferase action is in fact the reason behind H3-K4 dimethylation at the coding region of apM1, some structural modification must have taken place at that chromatin region upon induction of differentiation to account for its accessibility to methyltransferases in differentiated adipocytes but not fibroblasts.
To study the role of H3-K4 dimethylation on apM1 expression, we differentiated 3T3-L1 cells in the presence of MTA, a known inhibitor of AdoMet-dependent methyltransferases, that has previously been shown to inhibit H3-K4 methylation (36). Accordingly, by D5 of the differentiation process, we observed a marked decrease of global H3-K4 dimethylation in cells treated with the inhibitor. At the apM1 promoter, we were only able to detect a partial decrease of dimethylation, which may reflect the low turnover rate of this epigenetic mark. The expres- Shortly, in pluripotential cells such as 10T1/2 that have not yet compromised to an adipogenic fate, neither H3-K9/K14 acetylation nor H3-K4 di-or trimethylation is detected at the apM1 promoter. In 3T3-L1 fibroblasts, which have undergone determination, the still silent promoter is associated with H3-K4 dimethylation and bound Pol II. In fully differentiated 3T3-L1 adipocytes, finally, the apM1 promoter becomes acetylated and trimethylated at the same time that Pol II and histone acetylation and methylation spread to the coding region. TIS, transcription initiation site.
sion of the gene was also partially decreased in cells differentiated in the presence of the inhibitor. Moreover, we observed a significant inhibition of differentiation in cells maintained in the presence of MTA. We could not detect decreased levels of PPAR␥ RNA or protein (data not shown), thus indicating that reduced apM1 expression and reduced adipogenesis are not secondary to decreased expression of the factor, considered to be the master regulator of adipogenesis. However, more experiments are needed to discern whether binding of PPAR␥ to the apM1 promoter is impaired in the absence of H3-K4 dimethylation.
A previous study found a 50% decrease of H3-K4 methylation at the nos3 promoter after 48-h incubation of endothelial cells in the presence of 3 mM MTA (36), which correlated with decreased expression of the nos3 gene. Similarly, at the inducible collagenase promoter, which has been shown to be rapidly dimethylated at H3-K4 following mitogen stimulation (48), preincubation with 2 mM MTA blocked the induction of the gene. Our results are in accordance with these data and argue for a positive role of H3-K4 dimethylation in regulating transcription of adipogenic genes and ultimately, the differentiation process. However, we must take into account that MTA is a generic inhibitor of AdoMet-dependent methylation and has been shown to partially decrease DNA methylation (49) and affect arginine methylation (50). Thus, although our data suggest that AdoMet-dependent methylation is required for proper adipocyte differentiation, further studies are needed to determine which other targets aside from H3-K4 methylation may be involved in the process.
In conclusion (Fig. 7), our results show that H3-K4 dimethylation and Pol II recruitment at the promoters of key adipogenic genes are distinguishing marks of cells that have undergone determination and have compromised with an adipocyte cell fate. These marks are absent from the same promoter regions in pluripotential cells, such as the 10T1/2 cell line, which represent a previous step in the differentiation process from totipotential cells to differentiated adipocytes. In differentiated adipocytes, on the other hand, active transcription is accompanied by promoter H3 acetylation and H3-K4 trimethylation as well as the spreading of these modifications toward the coding region. More studies are needed to determine which transcription factors and coregulators are involved in the establishment and maintenance of this modified chromatin state throughout adipogenesis. Unraveling the molecular mechanisms underlying the establishment of the transcriptional pattern of mature adipocytes will undoubtedly help us to develop novel therapeutic approaches for the treatment of obesity-related diseases.