Differentiation of multipotent mesenchymal stem cells into lipid-accumulating adipocytes is a physiological process induced by transcription factors in combination with hormonal stimulation. We have used Affymetrix microarrays to compare the adipogenic differentiation pathways of NIH-3T3 fibroblasts induced to undergo in vitro differentiation by ectopic expression of early B cell factor (EBF)-1 or peroxisome proliferator-activated receptor (PPAR)γ2. These experiments revealed that commitment to the adipogenic pathway in the NIH-3T3 cells was not reflected in gene expression until 4 days after induction of differentiation. Furthermore, gene expression patterns at the earlier time points after stimulation indicated that EBF-1 and PPARγ2 induced different sets of genes, while the similarities increased upon differentiation, and that several genes linked to adipocyte differentiation were also transiently induced in the vector-transduced cells. These data suggest that the initial activation of genes associated with adipocyte development is independent of commitment to the adipogenic pathway and that EBF-1 and PPARγ2 induce adipocyte differentiation with comparable kinetics and efficiency.
- early B cell factor
- peroxisome proliferator-activated receptor-γ2
preadipocytes develop already in the early embryo and are the precursors of mature adipocytes, which are important for the metabolic homeostasis in both humans and mice. The molecular mechanisms that stimulate development of these cells from mesenchymal stem cells and fibroblastic cells have been under detailed investigation using in vitro differentiation systems as well as transgenic mouse models. This has resulted in the identification of a number of transcription factors with important roles in this process (3, 14). Among these are the peroxisome proliferator-activated receptor (PPAR)γ proteins (20, 21) as well as the C/EBP proteins-α (10), -β, and -δ (24, 26). On the basis of data from tissue culture experiments, these transcription factors act in an ordered hierarchy where C/EBPδ and -β precede C/EBPα and PPARγ (2, 24, 26, 27). C/EBPα is suggested to control the expression of PPARγ (5, 6, 28) and vice versa, creating a positive feedback loop that drives the cell into terminal adipocyte differentiation. The significance of the adipogenic transcription factors has also been investigated in vivo using genetically modified mice. C/EBPβ and -δ double-knockout mice have a decreased volume of white adipose tissue, although the adipocytes have a normal histological appearance and apparently normal expression of PPARγ and C/EBPα (19). Impairments in adipogenesis are also observed in mice lacking C/EBPα, and embryonic fibroblasts isolated from these animals do not differentiate to adipocytes when simulated with hormones in vitro (25). PPARγ-deficient mice die in utero, but experiments using a conditional knockout allele in vitro support the idea that this protein plays a crucial role in adipogenesis (12, 13). In addition, two helix-loop-helix proteins, adipocyte differentiation and determination factor (ADD)1/sterol regulatory element binding protein (SREBP) (8, 22) and early B cell factor (EBF)-1/O/E-1 (1, 7, 23) have been shown to stimulate adipogenesis. EBF-1 is expressed at the earliest stages of adipocyte differentiation and is in contrast to the C/EBPβ and -δ proteins, who themselves are rather poor stimulators of terminal adipogenesis in vitro (2, 3, 14), able to induce the adipogenic program in multipotent fibroblasts (1). Ectopic expression of EBF-1 led to enhanced adipogenesis in 3T3-L1 cells as well as mouse embryonic fibroblasts (1), suggesting this protein to be a potent mediator of fat cell differentiation. Real-time quantitative PCR analysis indicated that neither PPARγ2 nor C/EBPα were upregulated in immediate response to the expression of EBF-1 (1), leaving the mechanisms of action elusive.
To gain insight into the molecular mechanisms involved in adipogenesis, we have used Affymetrix microarrays to analyze the differences in mRNA levels in NIH-3T3 cells infected either with an empty retrovirus or with viruses carrying PPARγ2 or EBF-1 cDNAs. RNA was harvested 5–6 days after retroviral infection but before the addition of hormonal stimulators of adipogenesis as well as 2, 4, and 10 days after stimulation. These experiments revealed that, even though all the transduced cells initially responded in a similar fashion, only EBF-1- and PPARγ2-transduced cells differentiated into mature adipocytes. We were also able to observe specific differences in gene expression patterns in the EBF-1- and PPARγ2-transduced cells at the earlier time points of differentiation. However, these differences were gradually reduced as adipogenesis progressed. Thus we suggest that the initial induction of adipocyte marker genes is not linked to adipocyte lineage commitment and that EBF-1 and PPARγ2 are able to induce adipogenesis with similar kinetics and efficiency.
MATERIALS AND METHODS
Retrovirus production and infection.
Phoenix retroviral packaging cells were transfected with pBabe-puro retrovirus vectors in 100-mm dishes at 70% confluence, and virus supernatants were harvested after 48 h. Target cells were infected in 100- or 60-mm dishes at 50% confluence, using one volume of virus supernatant diluted in one volume of DMEM (Invitrogen) with 10% (vol/vol) calf serum, after which polybrene (Sigma, Stockholm, Sweden) was added at a final concentration of 6 μg/ml. The medium was replenished 8–12 h after infection; 24 h after infection, the cells were split 1:3 to 100-mm dishes and selected with 2 μg/ml puromycin (Sigma) for 72–96 h. Dishes with selected cells were trypsinized, pooled, and replated on 100- or 60-mm dishes for differentiation.
Cell culture and differentiation.
NIH-3T3 cells were propagated in DMEM containing 10% (vol/vol) calf serum at 37°C in a humidified atmosphere of 10% CO2. Medium was changed every second day in all experiments. NIH-3T3 cells were induced to differentiate 2 days postconfluence (day 0), using DMEM with 10% (vol/vol) fetal calf serum supplemented with the following inducers of adipogenesis: 1 μM dexamethasone (Sigma), 0.5 mM 3-isobutyl-1-methylxanthine (IBMX; Sigma), 1 μg/ml insulin (Roche Diagnostics, Bromma, Sweden), and 0.5 μM darglitazone for 2 days and 1 μg/ml insulin and 0.5 μM darglitazone for 2 additional days. From day 4, NIH-3T3 cells were cultured in DMEM with 10% fetal calf serum supplemented with 0.5 μM darglitazone (Medicinal Chemistry, AstraZeneca R&D, Molndal, Sweden). 3T3-L1 cells were propagated and differentiated as in Åkerblad et al. (1), with the exception that darglitazone was substituted for by 0.5 mM IBMX.
Oil Red O staining.
Culture dishes were washed twice in PBS and fixed for 30 min in PBS containing 4% formaldehyde. After a single wash in water, cells were stained with Oil Red O for 30 min. After the staining, dishes were washed twice in water and photographed. Oil Red O was prepared by diluting a stock solution [0.5 g of Oil Red O (Sigma) in 100 ml of isopropanol] with water (3:2) followed by filtration. Spectrophotometric quantification of Oil Red O-stained cultures was performed by isopropanol extraction of retained Red Oil O followed by analysis of the optical density at 510 nm.
Gene expression analysis.
RNA was prepared using TRIzol (GIBCO), and 7.5 μg of total RNA were annealed to a T7-oligoT primer by denaturation at 70°C for 10 min followed by 10 min of incubation of the samples on ice. First-strand synthesis was performed for 2 h at 42°C using 20 U of Superscript RT (GIBCO) in buffers and nucleotide mixes according to the manufacturer's instructions. This was followed by a second-strand synthesis for 2 h at 16°C, using RNaseH, Escherichia coli DNA polymerase I, and E. coli DNA ligase (all from GIBCO), according to the manufacturer's instructions. The obtained double-stranded cDNA was then blunted by the addition of 20 U of T4 DNA polymerase and incubation for 5 min at 16°C. The material was then purified by phenol-chloroform-isoamyl alcohol extraction followed by precipitation with ammonium acetate and ethanol. The cDNA was then used in an in vitro transcription reaction for 6 h at 37°C, using a T7 IVT kit and biotin-labeled ribonucleotides. The obtained cRNA was purified from unincorporated nucleotides on an RNeasy column (Qiagen). The eluted cRNA was then fragmented by incubation of the products for 2 h in fragmentation buffer (40 mM Tris-acetate, pH 8.1, 100 mM potassium acetate, 150 mM magnesium acetate); 20 μg of the final fragmented cRNA were then hybridized to Affymetrix chip U74Av2 (Affymetrix) in 200 μl of hybridization buffer (100 mM MES buffer, pH 6.6, 1 M NaCl, 20 mM EDTA, 0.01% Tween-20) supplemented with Herring sperm DNA (100 μg/ml) and acetylated BSA (500 μg/ml) in an Affymetrix Gene Chip Hybridization oven 320. The chip was then developed by the addition of FITC-streptavidin followed by washing using an Affymetrix Gene Chip Fluidics Station 400. Scanning was performed using a Hewlett Packard Gene Array Scanner. Hierarchical tree clusters were generated using the dCHIP program (9). Expression values were calculated according to the PM-MM model, and genes were filtered according to 0.50 < SD/mean < 10.00 and %P call in the array used ≥10%. All data are accessible in the Gene Expression Omnibus (GEO; GSE2192). For comparison of 3T3-L1 data generated on MG-U74Av1 chips (15) and the MG-U74Av2 chips from the present study, incorrect probe sets in the v1 series were filtered out using Affymetrix msk files, and remaining genes were directly matched by probe set ID or (for a very small set of genes) GeneID. Genes differing between 3T3-L1 cells and day 14 adipocytes were selected with 0.5 < SD/mean < 10 and P ≥ 25%, resulting in the definition of 470 developmentally regulated genes.
Statistical analysis of expression data.
Before the ANOVA analysis, expression data for each treatment and time point were transformed by where Yg (treatment,time) is a lognormal measure of “gene” (probe set) g at a specific time point under a specific treatment; n denotes the number of technical repetitions, and wgk denotes the Affymetrix U74Av2 array signal for gene (probe set) g in the kth repetition. Data were normalized against ug, which denotes the average Affymetrix U74A array signal for gene (probe set) g in the Novartis gene atlas for mouse (17). This transform produced lognormal measures of Yg for genes that could be used for ANOVA analysis. The time curve for each gene over time was normalized further by baseline subtraction, before ANOVA analysis. Normality of Yg was confirmed by visual inspection of quantile-quantile plots against the normal distribution.
For 160 sets of functionally coupled and coexpressed genes, S1,…, S160, defined in the Supplemental Data of Ref. 11, one-way ANOVA (balanced ANOVA) was used to test for changes in the average Y for genes in each set across the experimental time points. ANOVA P values were corrected for multiple testing by the Bonferroni method (multiplied by 160).
Real-time quantitative and conventional PCR analysis.
Total RNA was isolated using RNA STAT-60 (BioSite, Täby, Sweden) according to the manufacturer's instructions. DNA was removed from RNA preparations (DNA-free kit; Intermedica, Stockholm, Sweden), and first-strand synthesis was performed using random primers (Superscript First-Strand Synthesis System for RT-PCR, Invitrogen). Quantitation of mRNA was performed with a quantitative, real-time PCR approach based either on Taqman technology (Applied Biosystems, Stockholm, Sweden) or the use of SYBR-Green. The threshold cycle (Ct) for the endogenous control 36B4 mRNA and the target signals was determined, and the relative RNA quantification was calculated, using the comparative Ct method where ΔCt is Ct(target) − Ct(36B4). The ΔCt values were used to calculate 2−ΔCt. Real-time quantitative PCR experiments were performed in triplicate.
The following oligonucleotides were used in real-time quantitative PCR analysis: mouse IGF-II (exon 1), forward 5′-CCCGTGTCCAGGAAAACG-3′; mouse IGF-II (exon 1), reverse 5′-AAGGGCAGTCCCCAGTGAAC-3′; mouse IGF-II (exon 1), probe 5′-FAM-TCCCCGGGTCTTCCAACGGAC-TAMRA-3′; mouse 36B4, forward 5′-GAGGAATCAGATGAGGATATGGGA-3′; mouse 36B4, reverse 5′-AAGCAGGCTGACTTGGTTGC-3′; mouse PPARγ2, forward 5′-AACTCTGGGAGATTCTCCTGTTGA-3′; mouse PPARγ2, reverse 5′-CCTATGAGCACTTCACAAGAAATTACC-3′; mouse PPARγ2, probe 5′-6-FAM-CCAGAGCATGGTGCCT-MGB-3′; mouse EBF-1, forward 5′-CCTGGTGTGGTGGAAGTCACA-3′; mouse EBF-1, reverse 5′-CTACACAGCACTCAATGAACCCAC-3′; mouse EBF-1, probe 5′-6-FAM-CTGTCGTACAAGTCCA-MGB-3′; human EBF-1, forward 5′-CCTGGTGTTGTGGAAGTCACA-3′; human EBF-1, reverse 5′-GCTCAACGAACCCACCATC-3′; and human EBF-1, probe 5′-6-FAM-TGTCCTACAAATCTAAGCAGT-MGB-3′.
Oligonucleotides for quantitative Taqman real-time PCR not listed were ordered as assay on demand (Applied Biosystems).
Conventional RT-PCR analysis was performed using cDNA prepared as above and a 25-μl PCR reaction composed of 2 U of Taq (Invitrogen) in 0.1 mM dNTP and the buffer recommended by the manufacturer. The primers were added at a final concentration of 1 μM, and the samples were cycled according to the following protocol: 95°C, 2 min, followed by the indicated number of cycles at 95°C, 45 s; 62°C, 45 s; and 72°C, 45 s.
The following primers were used for amplification: 36B4 primers as above; mouse acyl-CoA ligase (L), forward 5′-ATGGAAGTCCATGAATTGTTCC; mouse acyl-CoA L, reverse 5′-GATGAACTGCTCGGAGCAAG; mouse phosphoenolpyruvate carboxykinase (PEPCK), forward 5′-CTTCTCTGCCAAGGTCATCCAG; mouse PEPCK, reverse 5′-ATCCGAGTCATGATCCGCATGC; hormone-sensitive (Hs) lipase, forward 5′-GACACAGAGACACCAGCCAACG; and Hs lipase, reverse 5′-GATCAGAGGTGAGTGGCATCTC.
Protein extracts, electrophoretic mobility shift assay.
DNA probes were labeled with γ-[32P]ATP by incubation with T4 polynucleotide kinase (Roche), annealed, and purified on a 5% polyacrylamide Tris-borate-EDTA (TBE) gel. Nuclear extracts, prepared as in Ref. 1, were incubated with labeled probe [20,000 counts/min (cpm), 3 fmol] for 30 min at room temperature in binding buffer (10 mM HEPES, pH 7.9, 70 mM KCl, 1 mM dithiothreitol, 1 mM EDTA, 1 mM ZnCl2, 2.5 mM MgCl2, 4% glycerol) with 0.75 μg of poly(dI/dC) per reaction (Pharmacia). DNA competitors were added 10 min before the addition of the DNA probe. The samples were separated on a 6% polyacrylamide TBE gel, which was dried and subjected to autoradiography. Competitors were added at molar excesses, as indicated in the respective figure legends. Antibodies used for supershift experiments were purchased from Santa Cruz Biotechnology (Santa Cruz, CA); catalog numbers are actin, SC-1616; C/EBPα, SC-61; C/EBPβ, SC-150; C/EBPδ, SC-151; and C/EBPε SC-158.
Oligonucleotides used for electrophoretic mobility shift assays (EMSAs) were as follows: octamer (Oct), forward 5′-TTCATTGATTTGCATCGCATGAGACGCTAACATCGTACGTTC; Oct, reverse 5′-GAACGTACGATGTTAGCGTCTCATGCGATGCAAATCAATGAA; mouse obesity (Ob)-C/EBP site, forward 5′GCCGGACAGTTGCGCAAGTGGCACTG; and mouse Ob-C/EBP site, reverse 5′-CAGTGCCACTTGCGCAACTGTCCGGC.
RNA expression analysis suggests that EBF-1 and PPARγ2 induce adipocyte differentiation of NIH-3T3 cells with comparable kinetics and efficiency.
To increase our understanding of the molecular processes involved in transcription factor-stimulated adipogenesis, we decided to investigate the gene expression profiles in NIH-3T3 cells stimulated to undergo adipogenesis. To this end, we prepared RNA from vector-, human EBF-1-, and PPARγ2-transduced cells 5–6 days after infection but before hormonal induction of differentiation (day 0) as well as 2, 4, and 10 days poststimulation using dexamethasone, IBMX, insulin, and darglitazone. Cells were stained with Oil Red O 10 days after stimulation to assess lipid accumulation, and while EBF-1- as well as PPARγ2-infected cells displayed a high content of Oil Red O-stained lipids, the vector-transduced cells did not (data not shown). RNA was prepared from at least two differentiation experiments, converted into biotin-labeled cRNA, and hybridized to Affymetrix microarrays containing ∼12,000 sequence tags. The data were analyzed with dCHIP software to identify differences in gene expression pattern in the various NIH-3T3 cell cultures (Fig. 1A). One set of genes was upregulated transiently 2 and 4 days after stimulation in all three experimental groups (cluster I). These genes included several markers of cell activation, including C/EBPδ and c-myc (Fig. 1A and Supplemental Data S1; available at the Physiological Genomics web site).1 The addition of stimuli also resulted in reduced expression of a group of genes, independent of whether the cells had been transduced with vector or transcription factor-containing viruses (cluster IV). Thus several genes are regulated in a similar way, independent of the virus used for transduction, suggesting a strong general effect of the hormonal stimulators per se. Among the genes induced at day 2 by the hormonal stimulation, several were expressed at similar levels in the control and EBF-1-transduced cells but to lower levels in the PPARγ2-infected cells and vice versa (clusters I and IV). Hence, the terminal fate of the cellular differentiation process cannot be predicted from gene expression profiles at this stage. At day 4, the EBF-1-infected cells resembled the PPARγ2-transduced cells rather than the control cells, with higher expression of gene cluster III. Investigation of the genes located in this cluster suggested that they included several genes involved in fatty acid metabolism, such as lipoprotein lipase, fatty acid synthase, fatty acid binding protein-4 (aP2), and diacylglycerol acyltransferase (Supplemental Data S1). This similarity became even more obvious at day 10, when cluster III genes were expressed mainly in the EBF-1- and PPARγ2-transduced cells. To get a quantitative overview of differences in gene expression patterns between PPARγ2-, EBF-1-, and vector-infected cell cultures, we calculated the number of genes that were up- or downregulated more then twofold upon comparison of the samples, pairwise and at various time points (Fig. 1B). Before hormonal stimulation (day 0), there was a greater difference between EBF-1 and PPARγ2 (46 genes) than between EBF-1 and vector (10 genes) or between PPARγ2 and vector (24 genes). At day 2 after hormonal stimulation, there was still a larger difference between PPARγ2- and EBF-1-transduced cells (79 genes) compared with vector and PPARγ2 (16 genes) or vector and EBF-1 (17 genes). Interestingly, the difference in gene expression patterns between the EBF-1- and PPARγ2-transduced cells was markedly reduced to only 14 genes at day 4 after stimulation (Fig. 1B). This was in contrast to an increased difference between vector- and PPARγ (26 genes)- or EBF-1 (70 genes)-transduced cells. The differences between the control cells and PPARγ2 or EBF-1 cultures further increased at day 10, with differential expression of 407 genes (Fig. 1B) in PPARγ2- and 429 genes in EBF-1-transduced cells, while the number of genes differentially expressed in the PPARγ2- and EBF-transduced cells was only slightly increased compared with day 4 (14 to 45 genes). To investigate the expression of PPARγ2 and endogenous mouse EBF-1 as well as ectopically expressed human EBF-1 during the differentiation process, we performed quantitative PCR of the transduced cells (Fig. 1C). This suggested that PPARγ2 could be detected even before stimulation in the PPARγ2-transduced cells, while only minute amounts could be detected in the EBF-1-transduced cells. However, some PPARγ2 expression could be detected 2 days after stimulation, and, after 4 days, the level of PPARγ2 was as high in the EBF-1- as in the PPARγ2-transduced cells. The fact that we used a human EBF-1 cDNA in our retrovirus allowed us to selectively follow the expression of endogenous and ectopically expressed EBF-1. No human EBF-1 was detected in the PPARγ2-transduced cells, while such transcripts were detected in human EBF-1-transduced cells. The expression of human EBF-1 was accompanied by an increased level of endogenous transcripts; mouse EBF-1 is likely to be a result of autoregulation of EBF-1 via its own promoter (16).
The development of multipotent NIH-3T3 cells into adipocytes is critically dependent on strong stimulators and ectopic expression of transcription factors. To relate the expression patterns we observed to those in less complex models, we compared our gene expression pattern to that obtained when analyzing the differentiation of preadipocytic 3T3-L1 cells. This was achieved by comparing the expression pattern of overlapping probe sets in our data to that of nonstimulated and terminally differentiated (14 days after stimulation) 3T3-L1 cells (previously published by Ross et al., Ref. 15), allowing us to analyze expression of 470 developmentally regulated genes (Fig. 2A, Supplemental Data S2). This revealed that the two major clusters, cluster I with expression restricted to 3T3-L1 cells and cluster II with expression in differentiated adipocytes, were highly overlapping in the two model systems. This notion was also supported by establishment of a hierarchical tree (Fig. 2B) where the samples from PPARγ2- and EBF-1-transduced cells at day 10 after stimulation clustered with differentiated 3T3-L1 cells, while the nonstimulated cells clustered at another branch of the tree. It was also notable that the vector-transduced day 10 sample from the stimulated 3T3 cells clustered with the nonstimulated (day 0) samples, while the day 2 and 4 expression patterns rather followed those of the corresponding EBF-1- or PPARγ2-transduced cells.
To further analyze the obtained gene expression patterns, we investigated the induction of seven adipocyte-selective genes taken from a recently reported analysis of the Novartis mammalian gene expression atlas (Refs. 11 and 18, Table 1). The dynamic changes of these externally defined differentiation markers were analyzed with respect to coordinated induction (S. Nelander et al., unpublished observations). Induction of adipocyte markers, measured as the average expression for these seven marker genes, was detected in both the PPARγ2- and EBF-1-infected cells (Fig. 2C). This induction was statistically significant (ANOVA; P = 0.000036 and P = 0.000005, respectively, Table 1). Moreover, the time-dependent expression of adipocyte genes was strikingly concordant in the PPARγ2- and EBF-1-infected cells (Fig. 2C). In linear units, the mean level of induction (between day 1 and day 10) of adipocyte markers in the PPARγ2- and EBF-1-infected cells was 11- and 14-fold, respectively. To a lesser degree, changes were seen also in the vector-infected cells (P = 0.0075). However, the level of induction was much less (3-fold), and the expression of adipocyte genes did not follow the same time course as in the treated cells (Fig. 2C). Hence, the analysis of the expression data showed that adipocyte genes are induced selectively in the treated cells and that these genes are induced in a similar time-dependent fashion in both PPARγ2- and EBF-1-infected NIH-3T3 cells. Using the same ANOVA procedure, we evaluated other gene sets against our expression data. Trends were seen in gene sets of a housekeeping character such as protein synthesis and metabolism, suggesting gradual alterations in basic cellular functions (Table 1). In quantitative terms, however, the induction of adipocyte differentiation markers was by far the strongest (Table 1). These data suggest that, even though EBF-1 and PPARγ2 induce different genes at early stages, the overall gene expression patterns suggests that they induce adipocyte differentiation with a comparable kinetics and efficiency.
EBF-1 and PPARγ2 induce distinct gene expression profiles in NIH-3T3 cells.
Despite the fact that the gene expression patterns in the PPARγ2- and EBF-1-transduced cells became similar at late stages of differentiation, there were some significant differences at earlier time points, indicating that EBF-1 and PPARγ were acting on different target genes. To investigate this further, we compared the genes induced by EBF-1 and PPARγ2 with the vector-transduced cells before stimulation of the cells (day 0) (Fig. 3A and Supplemental Data S3). This indicated that only three genes were activated and one gene repressed more than twofold in both the EBF-1- and PPARγ2-transduced cells. The upregulated genes proliferin and small inducible cytokine A2 (SCYA2, MCP1) are both secreted signaling molecules, while the repressed gene CTLA-2α encodes a cysteine protease. Several genes appeared to be specifically regulated in EBF-1-transduced cells, and quantitative real-time PCR analysis of a selection of these genes supported the microarray data (Fig. 3B). One specific feature of EBF-transduced cells was an upregulation of IGF-II expression. We could also detect an upregulation of the coregulated H19 gene (Fig. 3B), and quantitative real-time PCR analysis supported a 10- to 15-fold increase of both IGF-II and H19 mRNA in EBF-1-transduced cells at day 0 compared with vector- or PPARγ2-transduced cells (Fig. 3B). Thus gene expression analysis suggests that, despite the fact that both PPARγ2 and EBF-1 expression induced terminal adipogenesis and a similar gene expression pattern at day 10, the two transcription factors induce different sets of genes before the induction of adipogenesis with hormonal stimulation.
EBF-1 stimulates adipogenic conversion of 3T3-L1 cells in the absence of exogenous PPARγ ligand.
The gene expression patterns at the earliest stages of adipogenic differentiation of the NIH-3T3 cells argued against the full adipogenic effect of EBF-1 being mediated through the induction of PPARγ2 expression. However, to investigate this further, we explored the adipogenic potential of EBF-1 in the absence of exogenously added PPARγ ligand by transduction of 3T3-L1 preadipocytes with either PPARγ2 or EBF-1 and stimulation with dexamethasone and IBMX. The cellular accumulation of lipids was then quantified by spectrophotometric analysis of the Oil Red O content. This suggested that EBF-1-transduced cells accumulated three times as much lipids as the vector-transduced cells and twice as much as the PPARγ2-transduced cells 4 days after initiation of stimulation (Fig. 4A). A similar ratio of lipid content was found 6 days after stimulation, while the lipid content of PPARγ2-transduced cells approached that of the EBF-transduced cells 8 days after the initial stimulation (Fig. 4A). Analysis of the expression of aP2 by real-time PCR revealed an increased expression in the EBF-1- as well as PPARγ2-transduced cells 4 days after stimulation (Fig. 4B). This level was further enhanced at 6 and 8 days, with a higher expression of aP2 in EBF-1-transduced cells at all time points investigated. These data suggest that EBF can stimulate adipogenesis in 3T3-L1 cells in the absence of synthetic PPARγ ligand.
Induction of adipocyte-associated genes is not linked to commitment into terminal adipocyte differentiation.
To investigate potential links between EBF-1 and PPARγ2 and other transcription factors known to have important roles in adipogenesis, we extracted the expression data of PPARγ, SREBP1, C/EBPβ, and C/EBPδ, genes suggested to be markers and inducers of adipogenesis, and displayed these as line charts (Fig. 5A). PPARγ was present in the PPARγ2-transduced cells at days 0 and 2, while the EBF-1-transduced cells expressed comparable amounts of this mRNA at days 4 and 10 after stimulation (Fig. 5A and Fig. 1C). The expression of PPARγ was also transiently induced in the vector-transduced cells, but at day 4 the expression levels were significantly lower than in the transcription factor-transduced cells, and at day 10 the expression appeared to be reduced to the levels observed at day 0. The mRNA levels of C/EBPδ and SREBP1 were increased to similar extents after hormonal stimulation in all three experimental groups (Fig. 5A). The expression of C/EBPβ mRNA was somewhat higher in the EBF-1- and PPARγ2-transduced cells compared with the control cells at day 10, while no such difference was observed at the earlier time points (Fig. 4A). To explore whether the observed mRNA expression pattern was also reflected in C/EBP DNA binding activity, we performed EMSA using nuclear extracts from either control or EBF-1-transduced cells (Fig. 5, B–C). As a control for extract amount and quality, we used a consensus decamer interacting with Oct-1 protein in all the nuclear extracts (Fig. 5B). The amount of C/EBP DNA binding activity was then determined by an EMSA using a C/EBP binding site from the ob-promoter (Fig. 5C). This indicated that both the vector- and EBF-1-transduced cells responded in a similar way to the stimuli with an increase in C/EBP binding activity at day 2. However, there was a more pronounced DNA binding activity in the EBF-transduced cells at day 10 after stimulation (Fig. 5C). Supershift analysis, using antibodies directed against various C/EBP proteins, supported the microarray data, suggesting that the major complexes formed were composed of C/EBPβ and to some extent -δ at day 2 after stimulation, while C/EBPβ was the dominant protein 10 days after stimulation. We were unable to detect any significant DNA binding activity of C/EBPα, even though this protein complex could be identified in the myeloid cell line WEHI-3 (data not shown). These findings indicate that EBF-1 is not acting directly by inducing DNA binding of early acting factors such as C/EBPδ and -β and that induction of these transcription factors is not directly linked to commitment into adipocyte differentiation.
Knowing that C/EBP proteins were transiently activated in vector-transduced cells, we wanted to investigate how this was reflected in the expression pattern of genes associated with adipocyte differentiation. Two distinct expression patterns were observed when we compared the expression of genes associated with adipocyte differentiation. The first can be exemplified by the expression patterns of PEPCK and stearoyl-CoA desaturase (Fig. 6A). These genes were only detected at day 10 after hormonal stimulation and only in cells that had undergone terminal adipocyte differentiation. The second group, involving more genes, including acetyl-CoA dehydrogenase, hormone sensitive lipase, adipose differentiation-related protein, fatty acid CoA-ligase, and fat-specific gene 27 (Fig. 6A), was also induced in the vector-transduced cells at day 4. The expression of these marker genes of adipogenesis was further increased at day 10 in both the EBF-1- and PPARγ2-transduced cells, whereas their expression declined in the vector-transduced cells. To verify the transient induction of adipocyte marker genes in the absence of terminal differentiation, we performed real-time PCR analysis of adipose differentiation-related protein, fatty acid CoA-ligase, and fat-specific gene 27 in the transduced NIH-3T3 cells (Fig. 6B). This analysis supported the idea that adipocyte marker genes are also transiently induced in cells that will not undergo terminal adipocyte differentiation. To investigate this further, we analyzed the expression patterns of PEPCK, hormone-sensitive lipase, and fatty acid CoA-ligase before stimulation and 4 as well as 10 days after stimulation of either vector- or EBF-1-transduced cells by RT-PCR (Fig. 6C). This indicated that the PEPCK message could only be detected in the EBF-1-transduced cells at day 4 and day 10 after stimulation and not in the vector-transduced cells at any time point. In contrast, we were able to detect the hormone-sensitive lipase and fatty acyl-CoA ligase message in all samples, including the stimulated vector-transduced cells. The expression levels did increase transiently the vector-infected cells, while the upregulation was maintained throughout the experiment in the EBF-1-transduced cells. Thus we conclude that the hormonal stimulation by dexamethasone, IBMX, insulin, and darglitazone by itself leads to increased expression of genes associated with the adipogenic program in NIH-3T3 cells but that this may be a result of induction of already weakly transcribed genes rather then activation of silent genes such as PEPCK.
EBF-1 and PPARγ2 induce adipocyte differentiation of NIH-3T3 cells with comparable efficiency. Surprisingly, compared with vector-transduced cells, rather few genes changed in immediate gene expression after transduction of either EBF-1 or PPARγ2 before hormonal induction of differentiation, and none of these early changes could directly explain the effect on adipogenesis. There could be several explanations as to why these changes in gene expression do not explain the adipogenic effect of PPARγ2 or EBF-1, for example, that the specific genes were not present on these microarrays or that the proadipogenic effects involve posttranscriptional alterations. However, the expression pattern of fat cell genes in PPARγ2- and EBF-1-expressing NIH-3T3 cells does not support a crucial role for these transcription factors at the earliest stages of differentiation. Our results are in agreement with this idea, since all the cells initially respond to hormonal stimulation in a similar manner, although the outcome at day 10 is different. In addition, we find similar expression of early acting adipogenic transcription factors, such as C/EBPβ and -δ and SREBP1, in all cell cultures. Also, when the RNA expression patterns from the different day 4 cell cultures were compared, there was a global activation of genes associated with terminal adipogenesis in all cell cultures, which implies that expression of adipogenic markers cannot predict the outcome of the differentiation pathway. However, the expression of these genes appears to involve an upregulation of genes already actively transcribed in the nonstimulated cells, while another category of genes, such as PEPCK, is silent until day 4 after stimulation and also totally dependent on the presence of either EBF-1 or PPARγ2 for expression. The general expression of genes associated with adipogenesis, induced by the hormonal stimulators, is possibly mediated through C/EBPβ and -δ, the expression of which precedes that of PPARγ2. Hormone-induced C/EBPβ and -δ expression is, however, not linked to the outcome of the differentiation process, since terminal adipocyte differentiation is not supported in the absence of PPARγ2 or EBF-1. Another striking feature is the increased similarity in gene expression patterns and kinetics in EBF-1- and PPARγ2-infected cultures from day 4 to day 10. This is well illustrated by the array data obtained 4 days after stimulation. These were collected from four independent transduction/stimulation experiments and seven microarray hybridizations, displaying that these transcription factors induce highly similar expression patterns. This is the case, even though these two transcription factors are from different transcription factor families and are expressed distinctly during adipocyte development (1, 21, 24). Thus, on the basis of our data, we suggest that the initiation of the adipogenic program is independent of ectopic expression of EBF-1 or PPARγ2 but that expression of these proteins is essential for maintained expression of adipocyte genes and progression into terminal adipocyte differentiation under the conditions investigated. Hence, lineage commitment appears to be uncoupled from activation of a large number of adipocyte-related genes, providing another level of complexity to this process, possibly involving a genetic switch to allow for maintained expression adipocyte marker genes.
Although PPARγ2 and EBF-1 promoted the terminal differentiation process and induced similar expression patterns, the differential expression of 45 genes at day 10 indicates that there might be some differences in the nature and/or function of these NIH-3T3-derived adipocytes. Two differentially expressed genes that may be of special interest are adipsin and Glut4. The mature adipocytes obtained after PPARγ2-mediated differentiation express high levels of adipsin, while cells obtained using EBF-1 express higher levels of Glut4 (Supplemental Data S1). Higher expression of Glut4 could be explained by the finding that control elements regulating the Glut4 gene contain an EBF binding site, even though the primary role for this site is suggested to be transcriptional repression (4). However, the differences in expression of adipsin could not be attributed directly to differences in PPARγ2 expression, since the EBF-1- and PPARγ2-transduced cells express essentially the same PPARγ2 mRNA levels at day 10. Possibly, EBF-1 and PPARγ2 induce distinct types of adipocytes, and EBF-1 also exerts its action in parallel to the observed upregulation of PPARγ2. The relationship between EBF-1 and PPARγ2 needs to be established, but the fact that EBF-1-transduced NIH-3T3 cells are dependent on a PPARγ agonist to efficiently convert into adipocytes indicates a dependence on active PPARγ in this process.
Our results confirm the ability of EBF-1 to induce adipocyte differentiation of noncommitted fibroblastic cells, and at later stages the process induced by EBF-1 displays many similarities to that observed with PPARγ2. There are, however, discrete differences that may be crucial to control the metabolic profile of mature adipocytes. Fine-tuning of adipocyte development and regulation of the function of mature fat cells could be of importance for treating metabolic diseases like obesity, diabetes, and dyslipidemia.
This work was funded by AstraZeneca, the Swedish Medical Research council (VRM), the Swedish Foundation for Strategic Research (SSF) via Lund Strategic Center for Stem Cell Biology and Cell Therapy and the Medical Faculty at Lund University.
We are grateful to Dr. O. MacDougald for supplying gene expression data from 3T3-L1 cells.
↵1 The Supplemental Material for this article is available online at http://physiolgenomics.physiology.org/cgi/content/full/00015.2005/DC1.
Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).
Address for reprint requests and other correspondence: P. Åkerblad, Dept. of Molecular Pharmacology, AstraZeneca R&D Mölndal, S-431 83 Mölndal, Sweden (e-mail:).
- Copyright © 2005 the American Physiological Society