Activation of PPARγ Coactivator-1 Through Transcription Factor Docking

See allHide authors and affiliations

Science  12 Nov 1999:
Vol. 286, Issue 5443, pp. 1368-1371
DOI: 10.1126/science.286.5443.1368


Transcriptional coactivators have been viewed as constitutively active components, using transcription factors mainly to localize their functions. Here, it is shown that PPARγ coactivator–1 (PGC-1) promotes transcription through the assembly of a complex that includes the histone acetyltransferases steroid receptor coactivator–1 (SRC-1) and CREB binding protein (CBP)/p300. PGC-1 has a low inherent transcriptional activity when it is not bound to a transcription factor. The docking of PGC-1 to peroxisome proliferator-activated receptor γ (PPARγ) stimulates an apparent conformational change in PGC-1 that permits binding of SRC-1 and CBP/p300, resulting in a large increase in transcriptional activity. Thus, transcription factor docking switches on the activity of a coactivator protein.

Transcription factors exert their effects through docking of coactivator or corepressor proteins. Coactivators belong to two broad classes. In the first class, proteins like the SRC and CBP/p300 families contain histone acetyltransferase (HAT) activity that modifies chromatin structure (1). The second class, those in the vitamin D receptor–interacting proteins, thyroid receptor–associated proteins, and activator-recruited cofactor complexes, may mediate a direct connection between particular DNA binding proteins and RNA polymerase II (2–4). This model portrays the DNA binding transcription factor in a largely passive role, localizing the coactivator complex to genes that are marked for activation. Conversely, the coactivator complex has been thought to be constitutively active, requiring only proper positioning in the genome to initiate transcription. We have investigated the mechanism of transcriptional function and the role of transcription factor docking to coactivators. PGC-1 is a coactivator of nuclear respiratory factor–1 (NRF-1), PPARγ, and other nuclear receptors and is induced by exposure to cold temperatures. When expressed in cells, PGC-1 activates thermogenic gene expression, inducing uncoupling proteins, and stimulates mitochondrial biogenesis (5,6). We show that full-length PGC-1, despite these activities, is a relatively inactive molecule that requires docking to PPARγ or NRF-1 in order to undergo a conformational change and affect transcriptional activity.

Because PGC-1 has no HAT domains, we examined whether PGC-1 interacts with other HAT coactivators such as SRC-1, CBP/p300, and p300/CBP-associated factor (p/CAF). We fused full-length PGC-1 to the DNA binding domain (DBD) of yeast GAL4. Transcriptional output was measured with a reporter construct driven by GAL4 binding sites. As shown previously, PGC-1 had some transcriptional activity when fused to the GAL4 DBD (GAL4–PGC-1) (5) (Fig. 1A). However, this activity increased at least 10-fold when it was coexpressed with SRC-1, p300, or CBP, whereas p/CAF had no detectable effect (7). This suggests that this transcriptional enhancement could reflect a functional interaction with PGC-1. We tested for a physical interaction by performing coimmunoprecipitation experiments. When SRC-1 or CBP was coexpressed with PGC-1 in cells, immunoprecipitation of SRC-1 or CBP brought down GAL4–PGC-1 (Fig. 1B). Immunoprecipitated PGC-1 has a small amount of HAT activity (Fig. 1C); however, when SRC-1 and CBP were coexpressed either alone or in combination, a large increase in HAT activity was observed. Thus, PGC-1 can exist in transfected cells in a complex with SRC-1 and CBP/p300, and this complex has HAT activity. However, most of this activity derives from the CBP and SRC-1 components.

Figure 1

The transcriptional function of PGC-1 is activated by SRC-1 and CBP/p300. (A) COS cells were transfected with a 5XUAS/luciferase reporter gene (5XUAS, five copies of upstream activation sequence) and plasmids directing the expression of the GAL4–PGC-1, SRC-1, CBP, p300, and p/CAF (as indicated) and harvested for analysis of luciferase activity 48 hours later. Cytomegalovirus-driven β-galactosidase plasmid was used as transfection control. Error bars indicate SEM of five independent experiments made in duplicate. (B) PGC-1 interacts with SRC-1 and CBP in cells. COS cells were transfected with GAL4–PGC-1 or GAL4 with (+) or without (–) SRC-1 or CBP. Immunoprecipitates with monoclonal antibodies against SRC-1 or CBP were analyzed by protein immunoblot using specific monoclonal antibodies to GAL4 DBD (16). (C) PGC-1 is associated with SRC-1– and CBP-dependent HAT activity. COS cells were transfected with plasmids directing the expression of GAL4–PGC-1 or GAL4 and SRC-1 or CBP. After 48 hours, whole-cell extracts were immunoprecipitated with a polyclonal antibody to GAL4 DBD. HAT activity was measured in the immunoprecipitates with a liquid HAT assay (17); cpm, counts per minute. Error bars indicate SEM of two independent experiments made in duplicate. (D) PGC-1 physically interacts in vitro with SRC-1 and p300. GST–SRC-1 and GST-p300 fragments were incubated with full-length [35S]PGC-1 produced by translation in vitro. After extensively washing the glutathione agarose beads, the specifically bound [35S]PGC-1 was eluted and resolved by SDS-PAGE and detected by autoradiography (18). IVT, in vitro translation.

We mapped the domains of SRC-1 and p300 responsible for interactions with PGC-1 by producing fusion proteins with glutathione S-transferase (GST). PGC-1 interacted with a domain of SRC-1, encompassing amino acids 782 to 1139 (Fig. 1D). An interaction was also observed between amino acids 1805 and 2441 of p300 and PGC-1. The region of SRC-1 that bound PGC-1 overlaps the region known to interact with p300 and p/CAF. Likewise, the region of p300 interacting with PGC-1 overlaps the region known to interact with p/CAF, SRC-1, and the viral oncoprotein E1A (8, 9).

To provide an independent test of whether SRC-1 was required for the transcriptional activity of PGC-1, we transfected fibroblasts derived from mice that lack SRC-1 (10). In wild-type cells, GAL4–PGC-1 activated transcription of the reporter that was three times as efficient as that in mutant (knockout) cells (Fig. 2A). This loss of efficiency was regained when SRC-1 was replenished. Expression of CBP could overcome this deficiency of SRC-1, but overexpression of p/CAF could not. Thus, SRC-1 is required for the full transcriptional activity of PGC-1, although some residual activity is present even in the absence of SRC-1. Although it is not completely specific for CBP/p300, the viral oncoprotein E1A is an inhibitor of these coactivators by suppressing HAT activity (11) and competing for coactivator binding (12). E1A inhibits the transcriptional activity of PGC-1, with or without coexpression of additional CBP (Fig. 2B).

Figure 2

Transcriptional activity of PGC-1 is dependent on SRC-1 and CBP/p300. (A) PGC-1–dependent transcriptional activity is reduced in cells lacking SRC-1. Fibroblasts were obtained from wild-type (wt) mice or from those lacking SRC-1 [knockout (KO)] (10) and were transiently transfected with 5XUAS reporter, GAL4–PGC-1 plasmid with (+) or without (–) vector (empty plasmid), SRC-1, CBP, and p/CAF. Luciferase assays were performed 48 hours after transfection. (B) E1A viral oncoprotein inhibits PGC-1–dependent transcriptional activity. Transfection experiments were performed as in (A) except for the E1A 12S expression plasmid. Error bars in (A) and (B) indicate SEM of four independent experiments made in duplicate.

Deletion analysis revealed that transcriptional activation was localized to amino acids 1 to 170 of PGC-1 (Fig. 3A). Full-length PGC-1 is much less active than constructs that delete amino acids 170 to 350. This putative negative regulatory domain of PGC-1 amino acids 170 to 350 overlaps with the domain of PGC-1 that is involved in docking PPARγ and NRF-1 (5, 6). To investigate whether transcription factor docking might be an activating event for PGC-1, we cotransfected GAL4–PGC-1 with various alleles of PPARγ and NRF-1 (Fig. 3B). Transcriptional activation was increased equivalently with wild-type PPARγ or with several alleles lacking or defective in the following transcriptional activation domains: a point mutation in the AF-2 activation domain (AF-2m) [Glu499 → Gln499 (E499Q)], deletion of both activation domains (ΔAF1-AF2m) or a fragment of PPARγ that contains only amino acids 128 to 229 (DNA binding/hinge region). This fragment of PPARγ is known to be involved in mediating the docking of this receptor to PGC-1 (5). We confirmed that this fragment of PPARγ does not interact with CBP/p300 or SRC-1, using proteins prepared in vitro (13). Likewise, NRF-1 wild type and a dominant negative of NRF-1 [deletion lacking amino acids 305 to 503 and is transcriptionally inactive (6)] also increased the transcriptional activity of PGC-1 (Fig. 3B). PPARγ and the transcriptionally inactive PPARγ fragment (amino acids 128 to 229) also activated GAL4-p/CAF, a coactivator that otherwise has very low intrinsic transcriptional activity in this assay (Fig. 3B). This suggests that the function of transcription factor docking as an activating event can be generalized to different coactivators.

Figure 3

PPARγ and NRF-1 augments the transcriptional activity of PGC-1. (A) Mapping of the activation domain of PGC-1 (19). Transient transfection experiments were performed as in Fig. 1A. (B) Stimulation of PGC-1– and p/CAF-dependent transcriptional activity by expression of transcription factor. COS cells were transfected with GAL4–PGC-1 or GAL4-p/CAF expression plasmids and with indicated alleles of PPARγ (AF-2m, point mutation E499Q; ΔAF1-AF2m, AF2m that lacks amino acids 1 to 128) and NRF-1 (DN, deletion of NRF-1 that lacks amino acids 305 to 503). Error bars in (A) and (B) indicate SEM of four independent experiments made in duplicate.

Because transcription factor binding stimulates the activity of PGC-1, we examined whether there is a similar alteration in the binding of SRC-1 and CBP/p300 to PGC-1. Notably, the 1 to 400 amino acid portion of PGC-1, containing the transcription factor docking domain, is reduced in binding of SRC-1 and p300, in relation to the 1 to 180 amino acid portion of PGC-1 that lacks this domain (Fig. 4A). This binding of SRC-1 and p300 to the 1 to 400 amino acid fragment of PGC-1 is strongly stimulated by the 128 to 229 amino acid fragment of PPARγ. In contrast, this piece of PPARγ had no effect on the binding of SRC-1 and p300 to the 1 to 180 amino acid domain of PGC-1. A green fluorescent protein (GFP) control polypeptide had no effect on any of these binding events. We also investigated whether the binding of SRC-1 and p300 to PGC-1 could be stimulated by coexpression of PPARγ or NRF-1 in cells. Coexpression of either transcription factor with PGC-1 increased the amount of SRC-1 and p300 that coimmunoprecipitated with PGC-1 (Fig. 4B). Expression levels of SRC-1 and p300 were similar for each sample (14). These data illustrate that the docking of PPARγ or NRF-1 to PGC-1 augments the transcriptional activity of this coactivator by increasing the binding to SRC-1 and p300.

Figure 4

Docking of PPARγ to PGC-1 increases the binding of p300 and SRC-1. (A) GST–PGC-1 fragments (2 to 5 μg) were incubated with [35S]SRC-1 and [35S]p300. Ten micrograms of His-tagged PPARγ (amino acids 129 to 228) or His-tagged GFP were added to the indicated samples (20). Dashes indicate no addition of His-tagged proteins. (B) PPARγ and NRF-1 increase the binding of PGC-1 to SRC-1 and p300 in cells. Immunoprecipitates performed from transient transfected cells with an antibody to flag (to precipitate transfected flag–PGC-1 protein) were analyzed by protein immunoblot with specific monoclonal antibodies against SRC-1 and p300. (C) A PPARγ fragment induces a conformation change in the PGC-1 NH2-terminus. In vitro translated flag–PGC-1 was incubated with different amount of trypsin and with a PPARγ fragment (amino acids 128 to 229) or a GFP control. Digested bands were resolved by SDS-PAGE and analyzed by protein immunoblot analysis with an antibody to flag (21). C indicates control, without any trypsin added.

These data also suggest that the docking of PPARγ to PGC-1 induces a conformational change that reveals a cryptic binding capacity. To explore this further, we incubated in vitro translated PGC-1 with limiting concentrations of trypsin, in the presence or absence of the PGC-1 docking domain of PPARγ. The lowest amounts of trypsin used (1 μg) caused a cleavage of PGC-1 in the presence of PPARγ but not in the absence of PPARγ at any trypsin concentration used (Fig. 4C). Because PGC-1 was detected through an NH2-terminal immunotag, this induced cleavage site could be localized to amino acids 180 to 220. This is considerably distant from the PPARγ binding site located at amino acids 338 to 403 (5). Hence, transcription factor docking causes an activating conformational change in PGC-1.

On the basis of these results, it is necessary to modify the view that transcription factors function in gene activation primarily to recognize cognate DNA sequences and localize coactivator machines. Three independent lines of evidence show that transcription factor binding activates PGC-1. First, the transcriptional activity of GAL4–PGC-1 is enhanced by coexpression with NRF-1 and PPARγ or the fragment of PPARγ that mediates docking. Second, this same fragment of PPARγ enhances SRC-1 and p300 binding to PGC-1 in vitro and in vivo. Last, protease digestion reveals a new site of cleavage, distal to the PPARγ binding site, when this transcription factor docks. It is likely that three distinct states of PGC-1 exist (Fig. 5), in what we term a “spring-trap” model. When PGC-1 is not docked on a transcription factor, it has low transcriptional activity and minimal ability to assemble a complex involving SRC-1 and CBP/p300. Upon docking to an appropriate target, it undergoes a conformational change that is permissive for cofactor binding and thus sets the trap. This state of PGC-1 then captures SRC-1, CBP/p300, and perhaps other components, stimulating gene expression. Undocked but active PGC-1 could theoretically bind SRC-1 and p300 and thus interfere with their functions on various DNA-bound targets. It is also noteworthy that p/CAF shows a similar activation through docking of PPARγ, suggesting that the stimulatory role of transcription factor binding to coactivator is likely to be a more general effect.

Figure 5

Three-state model of PGC-1 activation. PGC-1 occurs in multiple states with different transcriptional activity. In a low-activity state (state I), PGC-1 is unbound to any transcription factor (TF). PGC-1 docking to a transcription factor causes a conformational change that results in a state permissive for cofactor binding (state II). This is in equilibrium with state III, where SRC-1, CBP/p300, and perhaps other proteins bind and activate transcription at this locus.

These data also represent an example of positive regulation of coactivator transcriptional activity. The E1A viral oncoprotein negatively regulates CBP/p300 (11, 12), whereas recent data suggest that calcium flux is an activating signal for the stimulatory transcriptional activity of CBP (15). It seems likely that coactivators and corepressors will have multiple modes of regulation, including temporal control of expression (like PGC-1) and regulation of inherent activity through transactivation factor docking and signal transduction systems.

  • * To whom correspondence should be addressed. E-mail: bruce_spiegelman{at}


Stay Connected to Science

Navigate This Article