|Home | About | Journals | Submit | Contact Us | Français|
Adipocyte differentiation is orchestrated by multiple signaling pathways and a temporally regulated transcriptional cascade. However, the mechanisms by which insulin signaling is linked to this cascade remain unclear. Here we show that the Med23 subunit of the Mediator Complex and its interacting transcription factor Elk1 are critical regulators of adipogenesis. Med23−/− embryonic fibroblast cells were refractory to hormone-induced adipogenesis. Knockdown of either Med23 or Elk1, or overexpression of dominant-negative Elk1, inhibited adipogenesis. In the absence of either Elk1 or Med23, Krox20, an immediate early gene stimulated by insulin during adipogenesis, was uninducible. Moreover, the adipogenic defect in Med23-deficient cells was rescued by ectopic expression of Krox20 or one of its downstream factors, C/EBPβ or PPARγ. Mechanistically, the insulin-stimulated, Med23-deficient preinitiation complex failed to initiate robust transcription of Krox20. Collectively, our results suggest that Med23 serves as a critical link transducing insulin signaling to the transcriptional cascade during adipocyte differentiation.
Understanding the molecular basis underlying adipocyte differentiation is crucial for better management of obesity and its associated diseases, the prevalence of which have escalated worldwide in recent years (Rosen and Spiegelman, 2006). Regulation of adipogenesis involves signaling by multiple extracellular factors and a temporally regulated transcription cascade (Rosen and MacDougald, 2006; Rosen and Spiegelman, 2006). Although these multiple signaling pathways can either positively or negatively regulate adipogenesis (Farmer, 2006), insulin signaling plays a critical role in promoting lipid accumulation and energy storage. It has been previously demonstrated that insulin transiently activates MAPK signaling (Bost et al., 2005; Porras et al., 1992; Prusty et al., 2002) and the expression of several rapid response genes during the early phase of adipogenesis (Farmer, 2006). However, the mechanism by which insulin-MAPK signaling transmits information to the adipogenic transcription cascade still remains unclear (Rosen and MacDougald, 2006).
The Mediator Complex is a multiprotein complex that bridges gene-specific transcription factors and the basal transcription machinery. Different transcription factors, in response to environmental and developmental cues, physically interact with different Mediator subunits to control specific genes and biological processes (Malik and Roeder, 2005). The Mediator Med23 (Sur2) subunit was originally identified as a genetic suppressor of a hyperactive ras phenotype in C. elegans (Singh and Han, 1995). In vivo and in vitro studies have demonstrated a specific interaction between mammalian Med23 and the ternary complex factor Elk1, which is activated by MAPK signaling (Stevens et al., 2002; Wang et al., 2005). Gel filtration chromatography and immunoprecipitation (IP) revealed that Med23−/− cells contain a stable Mediator Complex that supports the transcription of most genes (see Figure S1 available online) (Stevens et al., 2002; Wang et al., 2005). Furthermore, microarray analyses revealed a small set of genes that are downregulated in Med23−/− cells (Stevens et al., 2002; Wang et al., 2005), suggesting that Med23 may regulate specific cellular functions. Because the Ras-MAPK pathway has been shown to be involved in insulin signaling and adipogenesis, in this study we investigated the biological function of Med23 and the possible involvement of Med23 and Elk1 in the regulation of insulin signaling and adipogenesis. We found that both Med23 and its binding partner Elk1 are regulators of adipogenesis. By controlling the expression of Krox20 (Egr2), the first transcription factor induced during adipogenesis (Chen et al., 2005; Gonzalez, 2005), the MAPK-activated Elk1-Med23 interaction functions as a critical link that connects insulin signaling to the adipogenic transcriptional cascade.
The Med23−/− (KO) mice were embryonic lethal around 10.5 days of gestation (data not shown). Therefore, the embryonic fibroblasts (MEFs) were prepared from day 9.5 wild-type (WT) and KO mouse embryos, self-immortalized, and subjected to the standard hormone induction of adipogenesis. A three-cocktail (insulin, 3-isobutyl-1-methylxanthine [IBMX], and dexamethasone [Dex]) induced a small amount of lipid accumulation in WT MEFs, but not in KO MEFs (Figure 1A). Troglitazone, a diabetes drug that promotes glucose uptake and fat accumulation, efficiently potentiated hormone-induced adipogenesis in WT MEFs (over 20%), but failed to do so in KO MEFs (Figure 1A). As confirmed by real-time PCR, expression of the adipocyte markers αP2, adipsin, and PPARγ at different time points postinduction was increased in wild-type cells, but not in KO cells (Figure 1B). It has been previously shown that BMP4 promotes the differentiation of mesenchymal stem cells into adipocytes (Tang et al., 2004). Similarly, we observed that pretreatment of cells with BMP4, followed by the application of troglitazone and the hormone cocktail, was highly potent in converting WT MEFs into adipocytes (over 80%). However, this powerful treatment failed to induce adipogenesis in KO cells (Figure 1A), suggesting that Med23-deficient MEFs are refractory to hormone-induced adipogenesis.
To further verify that Med23 plays an essential role in adipogenesis, Med23 expression was reduced in 3T3-L1 preadipocytes by RNAi. A stable Med23 knockdown cell line was established by retrovirus-mediated siRNA delivery in 3T3-L1 cells. A control cell line expressing a negative control oligonucleotide was generated in a similar manner. Following selection with puromycin, the cells were cultured to confluence and treated with the hormone cocktail. The MED23 protein level in the si-Med23 cells was reduced by approximately 90% compared with that of the control cell line (Figure 1C). Oil Red O (ORO) staining demonstrated that si-Med23 cells accumulated far fewer lipids than the control cells (Figure 1D), and real-time PCR confirmed the greatly attenuated expression of adipocyte markers in si-Med23 cells during differentiation (Figure 1E). To exclude off-target effects of the siRNA, the human Med23 gene, which is resistant to the siRNA specific for the mouse Med23, was reintroduced into the si-Med23 cell line using retroviral transduction. Cells with the incorporated gene were selected with hygromycin and then induced to differentiate. Ectopic expression of human MED23 partially rescued the defective adipogenesis in the si-Med23 3T3-L1 cells, as assessed by ORO staining and the expression of αP2, PPARγ, and adipsin (Figures 1C, 1D, and 1E). Collectively, the experiments in both the MEFs and the 3T3-L1 cells demonstrate that Med23 is essential for hormone-induced adipogenesis.
Elk1 is a member of the ternary complex factor family. Upon phosphorylation by MAPK, ELK1 interacts with the Mediator Complex via the Med23 subunit (Cantin et al., 2003; Stevens et al., 2002; Wang et al., 2005) and therefore is considered upstream of Med23. Although Elk1 has been shown to be involved in proliferation and neuronal differentiation (Buchwalter et al., 2004), its role in adipogenesis has not yet been reported. We established a 3T3-L1 cell line stably expressing an Elk1-specific siRNA (Figure 2A) and observed attenuated hormone-induced adipogenesis, as detected by decreased ORO staining and adipocyte marker gene expression (Figures 2B and 2C). A similar observation was made in MEFs expressing the Elk1-specific siRNA (data not shown).
Next, we further characterized the ELK1 and MED23 interaction. The MED23 protein is highly hydrophobic, and soluble recombinant MED23 has been difficult to obtain. However, it is now possible to express and purify a soluble form of tagged MED23 protein from baculovirus-infected insect cells. In vitro, only the Erk-phosphorylated GST-ELK1 activation domain bound to the immobilized MED23 protein, whereas the nonphosphorylated GST-ELK1 did not (Figure 2D). In 293T cells, a weak interaction between ELK1 and MED23 was detected when Myc-Elk1 and Flag-Med23 expression plasmids were cotransfected. However, a much stronger interaction between ELK1 and MED23 was observed when a plasmid expressing active Mekk was also cotransfected (Figure 2E). These results suggest that phosphorylation of ELK1 facilitates its binding to MED23.
Given the importance of ELK1 phosphorylation, we mapped the ELK1 phosphorylation sites that are important for MAPK activation. Simultaneous mutation of two serine residues (S383 and S389) in the ELK1 activation domain to alanine significantly inhibited the activity of ELK1 in a luciferase assay (Figure 2F). We further analyzed the adipogenic phenotypes of 3T3-L1 cells overexpressing GFP, Elk, or Elk1 mutant (Figure 2G). Overexpression of either Elk1 or the S383A/S389A mutant (Elk1-mt) in 3T3-L1 cells slowed down cell growth and inhibited adipogenesis (Figures 2G and 2H). However, the generation of adipocytes from cells expressing Elk1-mt was more severely reduced (5%) compared with that from cells expressing wild-type Elk1 (30%; Figures 2G and 2H). The morphological result was confirmed by the real-time PCR-based quantification of adipocyte markers (Figure 2I). Thus, the S383A/S389A mutant displayed characteristics of a dominant-negative mutant. Taken together, these results suggest that Elk1 is an important regulator of adipogenesis.
In an effort to understand the mechanism by which Med23 regulates adipogenesis, we examined whether or not Med23 deficiency affects insulin signaling and cell cycle regulation. Although Med23 is an effector of growth signaling and is required for adipogenesis, siRNA knockdown of Med23 did not alter insulin-activated signaling, as indicated by the equal expression and phosphorylation levels of AKT, ERK, and ELK1 in the si-Ctrl and si-Med23 3T3-L1 cells following insulin stimulation (Figure S2A, compare lanes 2 and 4). Interestingly, however, the basal phosphorylation level of ERK and AKT appeared to change, possibly due to crosstalk or feedback regulation (Figure S2A, compare lanes 1 and 3). Importantly, the Med23 knockdown did not change the expression pattern of cell cycle regulators such as cyclin A, E, and D in a 24 hr time period after hormone induction (Figure S2B).
Postconfluent 3T3-L1 cells typically undergo one or two rounds of the cell cycle upon adipogenic hormone induction (mitotic clonal expansion), which is considered a prerequisite step for 3T3-L1 differentiation (Otto and Lane, 2005). It has been previously shown that MAPK signaling is required to initiate this mitotic clonal expansion during adipogenesis (Tang et al., 2003). However, knockdown of either Elk1 or Med23 in 3T3-L1 cells did not change cell doubling or cell cycle progression in response to the induction of adipogenesis by the three hormones (Figure S3). This suggests that Elk1 and Med23 most likely regulate adipogenesis by mechanisms other than an effect on insulin signaling, cell cycle progression, or mitotic clonal expansion.
In previous studies using microarrays and chromatin IP (ChIP) assays, it has been demonstrated that the Elk1-Med23 interaction controls the expression of the early growth response genes, including Krox20, in response to serum growth factor (Wang et al., 2005). Krox20 is highly expressed in adipose tissues (Soukas et al., 2001) and is the first transcription factor expressed upon insulin stimulation during adipocyte differentiation, although the mechanism of how insulin induces Krox20 expression is not clear (Chen et al., 2005; Gonzalez, 2005). Therefore, to better understand how Med23 controls adipogenesis and how insulin signaling affects Krox20 expression, we investigated whether Med23 also controls insulin-stimulated Krox20 expression in the early phase of adipocyte differentiation. RNA samples were collected from confluent WT and KO MEFs at different time points after insulin stimulation and analyzed by real-time PCR. Consistent with the previous report (Chen et al., 2005), in WT cells the expression of Krox20 was rapidly induced by insulin, reached a peak 30 min poststimulation, and decreased to the basal level 60 min poststimulation. However, this induction was abolished in KO cells (Figure 3A). Krox20 seems to be specifically induced by insulin, because insulin treatment with or without the other two hormones (IBMX and Dex) did not alter the intensity of Krox20 induction (data not shown). Consistent with the fact that Elk1 regulates adipogenesis, insulin-stimulated expression of Krox20 was also greatly reduced in cells following knockdown of Elk1 (Figure S4) or overexpression of a dominant-negative Elk1 mutant (data not shown). These data demonstrate that both Elk1 and Med23 control insulin-stimulated Krox20 expression.
To determine the specific effect of the insulin-MAPK-Elk1-Med23 pathway on Krox20 expression, we treated the cells with various inhibitors of signal transduction during insulin induction (Figure 3B). Only a MEK1 inhibitor, U0126, selectively inhibited the insulin-mediated expression of Krox20 in WT MEFs; inhibitors specific to other signaling pathways, including SB203580 (to p38), H-89 (to PKA), and LY294002 (to PI3K) did not inhibit insulin-stimulated Krox20 expression. Interestingly, the inhibition of p38 further enhanced the insulin-induced expression of Krox20, possibly due to the crosstalk between p38 and Erk (Keeton et al., 2003). In KO cells, none of the inhibitors had an obvious effect on the low level of Krox20 expression, because deletion of Med23 already abolished the Krox20 response to insulin (Figure 3B). These data suggest that insulin-induced expression of Krox20 is MAPK dependent and that Med23 is required for connecting insulin-MAPK signaling to Krox20 gene activation.
It has been previously reported that following insulin-activated expression of Krox20, both C/EBPβ (Farmer, 2006) and KLF5 (Oishi et al., 2005) are expressed during the early phase of adipogenesis. Therefore, we tested whether C/EBPβ and KLF5 are also affected by Med23 in WT and KO MEFs. To eliminate the mixed effects of multiple hormones, the cells were treated with insulin alone (without IBMX and Dex). The mRNA expression level of both C/EBPβ and KLF5 in WT MEFs was modestly increased after insulin treatment. Expression reached the peak level (2- to 3-fold induction) 2 hr poststimulation and then returned to the basal level 4 hr poststimulation (Figures 3C and 3D). In contrast, in KO MEFs the expression level of both C/EBPβ and KLF5 did not change significantly throughout the time course, and a reduced basal level of expression was observed (Figures 3C and 3D). This suggests that loss of Med23 attenuates both the insulin-induced and basal level expression of C/EBPβ and KLF5. These results demonstrate that Med23 not only controls the MAPK-dependent expression of an early immediate gene, Krox20, but also modulates downstream adipogenic transcription factors. Furthermore, because PPARγ expression was also abolished in KO cells (Figure 1B), Med23 appears to be required for the entire adipogenic transcriptional program.
Our data suggest that Med23 acts upstream of Krox20 and the adjunct adipogenic transcription cascade. Therefore, we expect that overexpression of Krox20 should reverse the defect in adipogenesis resulting from Med23 deficiency. To test this, si-Med23 3T3-L1 cells were infected with a retrovirus encoding Krox20 and the hygromycin-resistance gene for selection (Figure 3E). As shown in Figures 3F and 3G, overexpression of Krox20 rescued the differentiation defect of si-Med23 3T3-L1 cells, as indicated by ORO staining and real-time PCR analysis of adipocyte marker gene expression. Additionally, forced expression of Krox20 in the KO MEFs also partially restored hormone-induced adipogenesis (data not shown).
Because Elk1 and Med23 may regulate the expression of other genes besides Krox20, it is possible that other Med23-regulated genes are also involved in adipogenesis. Therefore, a gene profiling experiment was performed in the si-Ctrl and si-Med23 3T3-L1 cells, and a small number of genes (34) were identified that are both Med23 regulated and insulin responsive (Table S1). Krox20 was the most insulin-responsive (11.9-fold induction) of the 34 genes identified. Although we cannot rule out the possibility that one of the other genes may also regulate adipogenesis, the facts that Krox20 overexpression rescues adipogenesis in Med23-deficient cells and that Krox20 knockdown blocks adipogenesis in both 3T3-L1 and NIH 3T3 cells (Chen et al., 2005) suggest that Krox20 is an essential factor connecting insulin signaling to adipogenesis. We also noticed that another transcription factor, Egr1, which is a closely related familymember of Krox20 that is also targeted by Med23 (Wang et al., 2005), is significantly induced by insulin (Table S1). However, in contrast to the effect of Krox20, ectopic expression of Egr1 failed to restore hormone-induced adipogenesis in the si-Med23 3T3-L1 cells (Figure S5), which further underscores the specific role of Krox20 in the adipocyte lineage.
To further verify the pivotal role of Med23 in regulating the adipogenic transcription cascade, we tested whether factors downstream of Krox20 (C/EBPβ and PPARγ) are able to reverse the adipogenic defect in Med23-deficient cells. To address this question, KO MEF cell lines stably expressing C/EBPβ or PPARγ were generated. A GFP-expressed cell line was also generated as a control. These cells were cultured to confluence and then subjected to hormone-induced differentiation. The ectopic expression of C/EBPβ or PPARγ in KO cells robustly promoted adipogenesis (>90%) and adipogenic marker gene expression (Figure S6). These results further support that Krox20, C/EBPβ, and PPARγ are downstream of Med23, because ectopic expression of any one of them is able to bypass the need for Med23 in adipogenesis.
To determine how Med23 controls insulin-MAPK-activated Krox20 transcription, ChIP experiments were performed to measure the binding of ELK1, the Mediator Complex, Pol II, and general transcription factors to the Krox20 locus in response to insulin in confluent si-Ctrl, si-Elk1, and si-Med23 3T3-L1 cells. Both ELK1 and TBP constitutively bind to the Krox20 promoter, but not to the coding region, at an equivalent level regardless of Med23 status and insulin application (Figures 4A and 4E). This most likely indicates that they function as the platform for the formation of the preinitiation complex (PIC). The knockdown of Elk1 reduced ELK1 binding at the Krox20 promoter by 50% (Figure S7A), which led to a 50% decrease in both the basal and the insulin-stimulated levels of the Mediator Complex binding at the Krox20 promoter (Figure S7B). Consistent with the in vivo and in vitro data demonstrating that phosphorylated ELK1 interacts with MED23 (Figures 2D, 2E, and 2F), the addition of insulin enhanced the co-occupancy of ELK1 and Mediator at the Krox20 promoter (6-fold increase), as detected in a sequential ChIP (re-ChIP) experiment (Figure S7C). These results suggest that recruitment of the Mediator Complex at the Krox20 promoter is dependent on ELK1 and insulin stimulation. Consistently, insulin stimulation increased the binding of the Mediator Complex, Pol II, and TFIIE at the Krox20 promoter (3- to 6-fold increase) in si-Ctrl 3T3-L1 cells (Figures 4B, 4C, and 4F). The level of Pol II binding at the Krox20 promoter in the presence of insulin and Med23 is comparable to that at the promoter of a constitutively expressed gene, EF2, but is significantly above that at the promoter of an inactive gene, IL12b (Figure S8). Interestingly, in the si-Med23 cells, there seems to be a Med23-independent enhancement of Pol II and TFIIE binding to the Krox20 promoter (approximately 1-fold increase) upon insulin stimulation (Figures 4C and 4F), and Mediator Complex binding is also slightly increased (Figure 4B). These enhancements are probably not due to the residual amount of Med23, because similar increases in Mediator Complex and Pol II binding were also observed in the insulin-stimulated KO MEFs (Figures S9B and S9C). However, only in the WT cells, and not in the Med23-deficient cells, Pol II at the Krox20 promoter was able to initiate transcription efficiently, as indicated by the enhanced Pol II binding to the Krox20 gene coding region in the presence of insulin (Figure 4D and S9D) and the greatly enhanced Krox20 mRNA level (Figure 3A). Taken together, these data suggest that insulin stimulation and Med23 are both required for significant recruitment of Pol II and active transcription from the Krox20 promoter. Upon insulin stimulation, the Elk1-Med23 interaction promotes a productive PIC that leads to robust transcription, whereas the Med23-deficient PIC fails to do so.
This report demonstrates that the Mediator Complex subunit Med23 andElk1 are pivotal regulators of adipogenesis. By controlling insulin-MAPK-stimulated expression of Krox20, Med23 connects the insulin signaling pathway to the adipogenic transcription cascade. Our study reveals a function of the Elk1-Med23-Krox20 axis in regulating adipocyte formation, adding Elk1 and Med23 to the growing list of key early regulators of adipogenesis, such as Krox20 (Chen et al., 2005), Klf5 (Oishi et al., 2005), Klf4 (Birsoy et al., 2008), and CREB (Reusch et al., 2000; Zhang et al., 2004). Collectively, these studies suggest the existence of an early transcriptional factor network that initiates the adipogenesis program, and Elk1/Med23 may represent the earliest point of regulation.
The effect of Med23 in adipogenesis is highly specific, because Med23 deficiency does not prevent the differentiation of embryonic stem cells into many other cell types of the three germ layers during embryoid body and teratoma formation (data not shown). Furthermore, knockdown of Med23 in C2C12 myoblasts did not inhibit induced skeletal myogenesis (Figure S10), emphasizing the specific role for Med23 in adipogenesis. Because the deletion of Med23 is embryonic lethal, the in vivo function of Med23 during adipogenesis awaits future study using a conditional knockout approach.
Growing evidence has demonstrated that the Mediator Complex functions as a control center in directing diverse biological processes. For example, another Mediator subunit, Med1, controls adipogenesis by binding to PPARγ, a transcription factor that is critical for the terminal stage of adipocyte differentiation (Ge et al., 2002). Other Mediator subunits are required for different biological functions, such as Med12 for regulating neuronal development (Wang et al., 2006) or Med15 for controlling TGFβ signaling during development (Kato et al., 2002) and SREBP-controlled cholesterol and fatty acid homeostasis (Yang et al., 2006). Modulation of the various interactions between different activators and Mediator subunits may provide novel strategies to manipulate these diverse biological functions. Importantly, the Elk1-Med23 interaction may represent a potential therapeutic target for obesity and its associated diseases.
Differentiation of MEFs and 3T3-L1 into adipocytes was described previously (Chen et al., 2005; Ge et al., 2002). For more detail, including methods for insulin stimulation and signaling inhibition, see Supplemental Experimental Procedures.
Establishing stable cell lines to knock down or overexpress a gene of interest were based on the manufacturer’s recommendation (Clontech). For more detail, see Supplemental Experimental Procedures.
Multiple serine residues within the Elk1 activation domain of Gal4-Elk1 (307–428) were mutated by site-directed mutagenesis to alanine residues using the KOD Hot Start DNA Polymerase kit (TOYOBO). Cotransfection and luciferase assay are described in the Supplemental Experimental Procedures.
We thank Guanzhen Yang for technical assistance, Kan Liao and Baoliang Song for reagents, Jun Yan and Haifang Wang for help with microarray analysis, and Jun Ma (University of Cincinnati) and Dangsheng Li for helpful comments and suggestions. This work was supported in part by grants from China MOST (2006AA02Z156, 2006CB943900, 2007CB947902, and 2009CB941100), the Shanghai MCST (06PJ14106), and CAS (KSCX2-YW-R-107). G.W. is a scholar of the “Hundred Talent Program.”
Supplemental Data include Supplemental Experimental Procedures, ten figures, and two tables and can be found with this article online at http://www.cell.com/developmental-cell/supplemental/S1534-5807(09)00168-3.