Regulation of Peroxisome Proliferator–Activated Receptor-γ Activity by Mammalian Target of Rapamycin and Amino Acids in Adipogenesis

3T3-L1 cells were obtained from American Type Culture Collection. Dexamethasone (DEX), isobutylmethylxanthine (MIX), insulin, Oil Red O, polybrene, and puromycin were purchased from Sigma. mTOR antibody was previously described (25). All other antibodies were obtained from commercial sources: anti–C/EBP-α, C/EBP-β, C/EBP-δ, PPAR-γ, adipsin, and p21^CIP from Santa Cruz Biotechnology; and anti-Foxo1, phospho-Ser256 Foxo1, and phospho-Thr24 Foxo1/Thr32 Foxo3 from Cell Signaling.

The retroviral constructs for mTOR and S6K1 were generated by inserting mutant mTOR cDNAs into pQCXIP (Clontech) via the NotI site or truncated S6K1 cDNA (corresponding to amino acids 24–398) via the EcoRI site. pBabe-puro-C/EBP-α (retroviral vector), luciferase reporters for PPAR-γ and C/EBP-α containing −603 to 62 bp of the PPAR-γ proximal promoter fragment in pXP2 (26), and −343 to 125 bp of the C/EBP-α proximal promoter in pGL3-BA (27) all were gifts from Dr. G.S. Hotamisligil’s laboratory (Harvard University School of Public Health). pMSV-C/EBP-α, -β, and -δ plasmids (28) were provided by Dr. A.D. Friedman (Johns Hopkins University). The PPAR-γ responsive element (PPRE) reporter was generated by inserting three copies of DR1 sequence (TTCTGACCTATGACCTGG) into pGL2-based basal reporter plasmid pTK-luc.

Both HEK293 and 3T3-L1 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) that contained 10% FBS at 37°C with 5% CO₂. For 3T3-L1 preadipocyte differentiation, 2 days after the cultures were confluent, they were induced to differentiate (defined as day 0) by incubation in DMEM with 10% FBS, 1 μmol/l DEX, 0.5 mmol/l MIX, and 1 μg/ml insulin. At day 2, the induction medium was replaced by DMEM with 10% FBS plus 1 μg/ml insulin only, and cells were then fed every 2 days with the same medium. At the time points indicated, the cells were either harvested for Northern or Western analysis or stained with Oil Red O.

BD EcoPack2-293 packaging cells (Clontech) were grown in 10-cm plates in DMEM with 10% FBS until they reached 70% confluence and then were transfected with 5 μg of pBabe-puro-C/EBP-α, pQCXIP-mTOR, or pQCXIP-RR-S6K1 plasmid using SuperFect transfection reagent (Qiagen). Six hours after transfection, the medium was replaced with fresh medium, and after another 48 h of incubation, the virus-containing medium was collected, filtrated, and either stored at −80°C or used immediately for viral infection as follows: 5 ml of the viral suspension was mixed with 5 ml of fresh DMEM that contained 10% FBS and 16 μg/ml polybrene and added to 3T3-L1 cells in a 10-cm plate at 50% confluence. After 24 h, the cells were further infected by replacing 5 ml of cell medium with 5 ml of viral suspension. After 48 h, the cells were split at 1:6 and seeded under selection of 200 μg/ml puromycin. Five to 7 days later, the selected cells were pooled and cultured for Western analysis or for differentiation experiments.

HEK293 cells grown in 24-well plates to 60–70% confluence were transfected with 0.25 μg of luciferase reporter plasmids using PolyFect (Qiagen). In each transfection, 0.25 μg of pMSV-C/EBP plasmid (28) or 0.05 μg of PPAR-γ/retinoid X receptor (RXR) plasmid was transfected. After 3 h, cells were incubated in fresh medium for an additional 24 h. Cells were treated with rapamycin for 3 h before lysis when indicated. Cells were lysed in 100 μl/well Passive Lysis Buffer and subjected to luciferase assay using Luciferase Assay System (Promega).

For insulin stimulation, cells were incubated with DMEM with 0.3% FBS for 24 h and then stimulated with 1 μg/ml insulin for 3 h. Rapamycin and/or wortmannin were added 30 min before insulin stimulation, when indicated. 3T3-L1 adipocytes at day 6 were transfected with 0.5 μg of reporter plasmid using Lipofectamine Plus (Invitrogen). Rapamycin treatment and luciferase assays were performed as described above. For amino acid withdrawal, cells were washed once with Dulbecco’s phosphate-buffered saline and incubated in the same buffer that contained 4.5 g/l glucose, 1 mmol/l sodium pyruvate, 0.375% sodium bicarbonate, and 1× MEM vitamin solution (Cambrex) for 3 h. Readdition of amino acids was achieved by changing to the same medium for amino acid starvation plus a mixture of amino acids, which contained 2 mmol/l glutamine, 2× MEM amino acids solution, and 2× MEM nonessential amino acids solution (Cambrex). Troglitazone (10 μmol/l) was added along with rapamycin or amino acid withdrawal when indicated.

3T3-L1 preadipocytes were induced to differentiate by addition of the differentiation cocktail (DEX, MIX, and insulin). Typically, changes in cell morphology and the accumulation of lipid droplets were evident 3–4 days after induction, and adipocytes were fully differentiated in 7–8 days. Rapamycin completely blocked the differentiation and lipid accumulation, as shown by the Oil Red O staining in Fig. 1A. Rapamycin also effectively attenuated terminal differentiation when added at day 4 and removed at day 6 (Fig. 1A), suggesting that a rapamycin-sensitive pathway is involved in later stages of differentiation after the clonal expansion phase, consistent with previous reports (22, 23). Furthermore, rapamycin treatment of day 8 adipocytes for 4 days led to a reduction of lipid droplets and smaller adipocytes (Fig. 1B), indicating that rapamycin affects the maintenance of adipogenic characteristics of mature adipocytes. This effect is apparently not due to increased cell death (data not shown). It is not known whether rapamycin induces de-differentiation or promotes lipolysis in adipocytes.

The protein level of mTOR in fully differentiated adipocytes was higher than that in preadipocytes, accompanied by an increase in phosphorylation at the putative Akt site Ser2448, the functional significance of which is yet to be determined (29, 30) (Fig. 1C). It is interesting that the expression of S6K1 and 4E-BP1, two well-known downstream targets of mTOR, underwent even more drastic increase upon adipogenesis (Fig. 1C), as seen by other groups (23, 31). Because S6K1 is a positive regulator of protein synthesis downstream of mTOR, whereas 4E-BP1 is a negative regulator, the upregulation of both proteins did not provide clear insights into the involvement of mTOR signaling in adipogenesis. Obviously, a direct examination of mTOR would be necessary.

We next probed the involvement of mTOR in adipogenesis directly. A point mutation, S2035T, confers rapamycin resistance to mTOR functions by abolishing rapamycin binding to mTOR (32, 33). This rapamycin-resistant mTOR (RR-mTOR) was introduced into 3T3-L1 cells via retroviral infection followed by puromycin selection. The expression of the FLAG-tagged recombinant proteins from the viral source was confirmed by Western blots shown in Fig. 2B. The infected cells were induced to differentiate with or without rapamycin. Cells expressing RR-mTOR differentiated fully in the presence of rapamycin, as indicated by the appearance of rounded adipocytes and lipid droplets (Fig. 2A), as well as expression of the adipogenic markers adipsin and p21^CIP (Fig. 2D). However, cells expressing a kinase-inactive mutant (D2357E) of mTOR containing the rapamycin-resistant mutation (RR/KI-mTOR) or infected with empty vector control did not differentiate in the presence of rapamycin (Fig. 2A), suggesting that the kinase activity of mTOR is required for adipocyte differentiation. We then asked whether S6K1 mediates the adipogenic signaling of mTOR. A rapamycin-resistant mutant of S6K1 (RR-S6K1; NH₂-terminal 23 amino acids and COOH-terminal 104 amino acids deleted [34]) was introduced into 3T3-L1 cells by viral infection, but expression of this mutant S6K1 did not rescue adipocyte differentiation from rapamycin inhibition (Fig. 2A and B), although the recombinant protein displayed full rapamycin-resistant S6 kinase activity (Fig. 2C). Thus, if S6K1 is involved, it is not the only effector downstream of mTOR in the adipogenic process.

To understand the mechanism by which mTOR regulates adipocyte differentiation, we focused on the transcription factors C/EBPs and PPAR-γ, which are known to play major regulatory roles in adipogenesis (4, 7, 8). As demonstrated by Western analyses shown in Fig. 3A, the protein levels of C/EBP-β and C/EBP-δ were highly induced after 1 day of differentiation and gradually decreased from day 4 to day 8; rapamycin treatment did not affect the early expression of those two proteins but seemed to prevent their subsequent decrease. Contrary to the rapid induction and transient expression profile of C/EBP-β and -δ, the protein levels of PPAR-γ and C/EBP-α, which were significantly blocked by rapamycin treatment (Fig. 3A), rose later during differentiation. At the same time, rapamycin also inhibited the expression of the adipogenic markers adipsin and p21^CIP (Fig. 3A), which are thought to be the direct or indirect targets of PPAR-γ (35, 36). Thus, rapamycin seems to block a later stage of adipogenesis by potentially targeting transcription factors C/EBP-α and PPAR-γ. Importantly, these rapamycin effects on the protein expression were closely mirrored by those on the corresponding mRNA levels, as demonstrated by the Northern analyses shown in Fig. 3B, suggesting that the inhibition likely occurred at the transcriptional level.

It has been reported that C/EBP-β and -δ are essential for adipocyte differentiation (37) and at least partly responsible for the expression of C/EBP-α and PPAR-γ (5). It was thus possible that the activity of C/EBP-β and C/EBP-δ toward the transcription of C/EBP-α and PPAR-γ could be affected by rapamycin. To test this possibility, we transfected 3T3-L1 preadipocytes with a luciferase reporter driven by the C/EBP-α promoter (27), along with C/EBP-β or C/EBP-δ expression plasmid. As shown in Fig. 4, overexpression of C/EBP-β and -δ enhanced the reporter activity as expected, but rapamycin had no effect, suggesting that the activation of C/EBP-α promoter by C/EBP-β or C/EBP-δ is independent of mTOR function. Next, 3T3-L1 preadipocytes were transfected with a luciferase reporter for the PPAR-γ promoter (26) along with the C/EBP-β or -δ expression plasmid, and the reporter activity was measured with or without rapamycin treatment of the cells. Although rapamycin caused a moderate decline in PPAR-γ promoter activity (Fig. 4), it is unlikely that this effect accounts for the complete absence of adipogenesis in the presence of rapamycin.

Because rapamycin significantly blocked the expression of C/EBP-α and PPAR-γ transcription, we asked whether ectopic expression of C/EBP-α and PPAR-γ would rescue adipogenesis in the presence of rapamycin. To address this question, we generated 3T3-L1 cells stably expressing C/EBP-α by retroviral infection. The endogenous C/EBP-α mRNA yields two isoforms of the protein, p42 and p30, by utilizing alternative open reading frames (38). Here we used a C/EBP-α cDNA engineered to remove the start codon for the p30 isoform (38); as a result, only the p42 isoform was expressed from the retroviral vector, allowing us to distinguish the recombinant C/EBP-α from its endogenous counterpart. It was reported that constitutive expression of C/EBP-α p42 confers spontaneous adipocyte differentiation without hormonal induction (38). Indeed, we observed phenotypic differentiation in C/EBP-α p42-expressing cells even before hormonal addition (data not shown), and insulin further stimulated adipogenesis (Fig. 5A). Rapamycin completely blocked the insulin-induced accumulation of lipid droplets, as shown by Oil Red O staining in Fig. 5A. We analyzed the protein expression profiles and found that the expression of PPAR-γ was intact in C/EBP-α p42-expressing cells in the presence of rapamycin, but the adipogenic marker p21^CIP was not induced under the same conditions (Fig. 5B). Thus, rapamycin inhibits adipogenesis even when both C/EBP-α and PPAR-γ proteins are expressed.

It is generally believed that C/EBP-α and PPAR-γ induce each other’s expression in a positive feedback loop to promote and/or maintain adipocyte differentiation (1, 7, 8). Because rapamycin inhibited C/EBP-α and PPAR-γ expression without affecting C/EBP-β and C/EBP-δ activity, it seemed likely that rapamycin’s target would be the positive feedback loop between C/EBP-α and PPAR-γ. The data in Fig. 5B suggest that in the activation feedback loop between C/EBP-α and PPAR-γ, the induction of PPAR-γ by C/EBP-α remained intact in the presence of rapamycin, but apparently PPAR-γ activity was impaired as suggested by the lack of expression of PPAR-γ target p21^CIP and the endogenous C/EBP-α p30 isoform despite the presence of C/EBP-α and PPAR-γ. A simple model that would explain all of these observations is that mTOR specifically regulates PPAR-γ transactivation activity.

To examine this possibility, we transfected HEK293 cells with a luciferase reporter that contained triple repeats of the DR1 sequence, a potent PPRE (39), along with PPAR-γ and RXR expression plasmids. This reporter activity was stimulated by insulin in serum-starved cells, which was mostly eliminated by rapamycin (Fig. 6A), suggesting that mTOR may be required for the stimulation of PPAR-γ activity by insulin on PPRE-driven transcription. The activity of PPAR-γ monitored by this reporter was also inhibited by wortmannin, a specific inhibitor for phosphatidylinositol 3-kinase (PI3K) (Fig. 6A). Indeed, PI3K and its downstream effector Akt have been reported to play crucial roles in adipogenesis (40, 41). One of the downstream targets of Akt is the Foxo subfamily of Forkhead transcription factors. The phosphorylation and subsequent inhibition of Foxo proteins by Akt is critical for the insulin action on adipocyte differentiation (42, 43). Recently, it has been reported that Foxo1 antagonizes PPAR-γ activity through its direct interaction and disruption of DNA binding activity of PPAR-γ (44). Because mTOR is intimately linked to the PI3K pathway, we asked whether mTOR would also have an impact on PPAR-γ activity through Foxo. To this end, we analyzed the phosphorylation of various Foxo transcription factors during adipogenesis. As shown in Fig. 6B, phosphorylations of Foxo1 at Ser256, Foxo3 at Thr32, and Foxo1 at Thr24 all were inhibited by wortmannin treatment, but none of them was affected by rapamycin. Thus, mTOR and PI3K seem to regulate PPAR-γ transactivation via different mediators.

To further verify the regulation of PPAR-γ activity by mTOR in adipocytes, we transfected 3T3-L1 adipocytes (at day 6 differentiation) with the DR1 luciferase reporter. This PPRE-driven activity was significantly reduced by rapamycin treatment for 3 h (Fig. 7A). The short duration of rapamycin treatment ensured that the decrease in luciferase activity was not due to a general inhibition of protein synthesis by rapamycin, and, indeed, a cytomegalovirus (CMV) promoter-driven luciferase reporter was minimally affected under similar conditions (data not shown). The level of PPAR-γ protein was unchanged upon rapamycin treatment (Fig. 7A). Furthermore, the reporter activity was resistant to rapamycin treatment in RR-mTOR–expressing cells (Fig. 7B), confirming the role of mTOR in the regulation of PPAR-γ activity.

Previous studies have established that the mTOR pathway transduces nutrient availability signals, especially amino acid sufficiency. Therefore, we examined whether amino acid sensing is involved in the regulation of PPAR-γ activity by mTOR. As shown in Fig. 7A, day 6 3T3-L1 adipocytes transfected with DR1 luciferase reporter were incubated in amino acid–deficient medium for 3 h. The reporter activity was significantly reduced by amino acid starvation, which was mostly restored when amino acids were replenished for 3 h in these cells. The restoration of PPAR-γ activity by amino acid readdition was abolished by rapamycin treatment. As a control, the CMV-driven luciferase activity was not significantly affected by such short-term amino acid starvation (data not shown). Taken together, these data suggest that PPAR-γ transactivation activity may be directly controlled by a nutrient-sensing signaling pathway, likely involving mTOR.

TZD drugs, as specific ligands for PPAR-γ (10), have been used to improve insulin sensitivity in various models of diabetes and obesity (11). It is interesting that when day 6 3T3-L1 adipocytes were treated with 10 μmol/l troglitazone, a TZD drug, the inhibitory effect of rapamycin on PPAR-γ activity (as measured by the DR1 reporter) was completely reversed (Fig. 8A). Likewise, the reduction of PPAR-γ activity upon amino acid starvation was reversed by troglitazone (Fig. 8A). Once again, the CMV-driven luciferase reporter was mostly unaffected under the same conditions. Accordingly, the expression of various adipogenic markers, including C/EBP-α, adipsin, and p21^CIP, was completely rescued by troglitazone treatment in the presence of rapamycin (Fig. 8B), and so was phenotypic differentiation including lipid formation and morphological changes (data not shown). These findings suggest that troglitazone neutralizes the inhibitory effect of rapamycin and amino acid deficiency by enhancing PPAR-γ activity. Troglitazone seems to act independent of the S6K1 pathway, because S6K1 phosphorylation at Thr398 (a putative mTOR site and an indication of S6K1 activity) was not rescued by troglitazone in rapamycin-treated cells (Fig. 8B). An obvious implication of the troglitazone effect on reversing rapamycin-inhibited PPAR-γ activity was that mTOR controls the production of a PPAR-γ ligand. However, conditioned media from normally differentiating adipocytes did not rescue differentiation in rapamycin-treated 3T3-L1 cells (data not shown). Thus, the major function of mTOR in the regulation of PPAR-γ activity is unlikely to be through the production of a ligand but rather through its direct impact on PPAR-γ transactivation activity, as suggested by the rapid inhibitory effect of rapamycin (Figs. 6 and 7). It is possible that troglitazone, as an unnatural ligand, overrides the normal requirements for PPAR-γ activation and thus counters rapamycin inhibition.

A number of studies were published reporting the inhibitory effect of rapamycin on adipogenesis (22–24). However, the role of mTOR was not directly examined, and the molecular mechanism of rapamycin inhibition was not explored in this important biological process. In our current study, we provide the first direct demonstration that mTOR is required for adipogenesis in the system of 3T3-L1 cells and that the catalytic activity of mTOR is indispensable for its adipogenic functions. To unravel the mechanisms of mTOR signaling in adipogenesis, we focused on the transcriptional regulatory network in adipocyte differentiation. Our findings have led us to propose a novel model (Fig. 9) in which the mTOR pathway specifically regulates the transactivation activity of PPAR-γ, which is essential for a positive feedback control of C/EBP-α expression as well as the adipogenic gene expression program and thus critical for both initiating and maintaining adipogenesis. Furthermore, mTOR signaling may serve to transduce nutrient availability signals to control the activity of PPAR-γ.

mTOR signaling is best known for its role in translational regulation via two major downstream effectors, S6K1 and 4E-BP1. The vast majority of cellular mTOR is found in the cytoplasm at steady state, but we previously identified a cytoplasmic-nuclear shuttling behavior for mTOR (45), which may suggest a nuclear function for mTOR. Several lines of evidence have implicated the direct involvement of mTOR in the regulation of transcription (for examples, see references 46–48), but the mechanisms are not well understood. So far, the only transcriptional regulator identified as a target of mTOR is STAT3, which is reportedly phosphorylated by mTOR at Ser727 and activated (48). In this study, we have revealed another transcriptional regulator as a potential target of mTOR (PPAR-γ) further extending the complexity of mTOR signaling. How mTOR functionally interacts with PPAR-γ is currently unknown and warrants future investigations. mTOR did not seem to affect the DNA binding affinity of PPAR-γ for the DR1 sequence (data not shown). The direct target of mTOR could be a nuclear cofactor of PPAR-γ, or as a potential scaffold protein mTOR could participate in the recruitment of a coactivator complex for PPAR-γ.

The PI3K/Akt pathway, downstream of insulin signaling, has previously been shown to be critically involved in preadipocyte differentiation. The activities of PI3K and Akt kinase rise during differentiation (40), and constitutively active Akt causes spontaneous differentiation of 3T3-L1 without hormone stimulation (41). Most recently, Akt has been demonstrated to regulate adipogenesis via phosphorylating and thus inactivating Foxo1 (42, 43), and Foxo1 is known to regulate directly PPAR-γ activity (44). The mTOR and PI3K pathways are intricately linked, and at least two distinct models have been proposed for their relationship (discussed in ref 49). In one model, the two pathways adopt a parallel relationship, and they converge on common downstream targets such as S6K1 and 4E-BP1. The alternative model places mTOR downstream of PI3K/Akt in a linear manner. Our observations suggest that in adipogenesis, these two pathways most likely work in parallel and converge on the activation of PPAR-γ: whereas the PI3K/Akt pathway seems to mediate insulin signals to downregulate Foxo1, the mTOR pathway most likely transduces nutrient-availability signals to regulate PPAR-γ via a target distinct from Foxo1 (Fig. 9).

The functions of the adipose tissue, including metabolism, energy homeostasis, and endocrine actions, are dependent on cellular energy status and nutrients (1). The mTOR pathway is known to sense amino acid availability (14) and more recently was shown to respond to cellular energy levels (50, 51). Our data suggest that adipocytes may utilize the mTOR pathway to sense nutrient availability and modulate the activity of PPAR-γ, revealing the first molecular link between nutrients and adipogenesis/adipogenic functions. It is interesting to note that skeletal myogenesis also requires mTOR signaling. Nutrient sensing through mTOR controls the autocrine production of insulin-like growth factor-II that instructs the initiation of the differentiation program in skeletal myoblasts (47). Hence, the mTOR pathway seems to be a master mediator of nutrient signals, governing a wide range of biology from cell growth to various types of cellular differentiation.

Rapamycin (also known as Rapamune or sirolimus) has been widely used as an immunosuppressant to prevent graft rejection in transplant recipients (52, 53). Recent clinical studies have demonstrated that one of the major side effects associated with rapamycin treatment is dyslipidemia, characterized by increased serum levels of triglyceride and cholesterol (54, 55). This toxicity of rapamycin is potentially the result of deregulated energy and lipid homeostasis in peripheral tissues including adipose and hepatic tissues. Our findings that rapamycin specifically inhibits PPAR-γ activity and subsequently disrupts the adipogenic functions provide a possible molecular explanation for this toxic effect of rapamycin. Furthermore, it would be reasonable to explore the possibility of mTOR regulation of other PPAR isoforms such as PPAR-α and PPAR-β/δ, because these factors are also critical for lipid metabolism and control of serum lipid levels (56). A significant implication of our observation that troglitazone reversed the inhibitory effect of rapamycin on PPAR-γ transactivation activity is the rationale for testing the possibility of using TZD drugs to counter the dyslipidemia side effect of rapamycin in transplant recipients.

This work was supported by grants from the National Institutes of Health (GM58064 and AR48914) and American Cancer Society (RSG-03-250-01-TBE) to J.C.

We are grateful to Dr. G.S. Hotamisligil and Dr. A.D. Friedman for providing various reagents (see research design and methods for details).