The PPARγ2 A/B-Domain Plays a Gene-Specific Role in Transactivation and Cofactor Recruitment

Adenoviral vectors for investigation of the role of the PPARγ A/B-domain

To determine how the AF-1 activity affects the transcriptional profile of PPARγ, we constructed adenoviral vectors expressing HA-tagged PPARγ1 (AdHA-PPARγ1) and A/B-domain truncated PPARγ (AdHA-PPARγCDE) (Fig. 1, A and C) for comparison with HA-PPARγ2. Notably, equal protein expression (Fig. 1B and supplemental Fig. 1A) corresponded to equal mRNA levels of the HA-PPARγ transgenes and of the adenoviral AdE4 gene, a marker of transduction efficiency (Fig. 1D and supplemental Fig. 1B). These observations suggest that the protein stabilities of HA-PPARγ1, HA-PPARγ2, and HA-PPARγCDE are similar and show that the cells were subjected to equal adenovirus exposure. Addition of the specific PPARγ agonist rosiglitazone decreased protein levels of both HA-PPARγ2 and truncated HA-PPARγCDE (Fig. 1B). This observation is in agreement with previous studies reporting that agonists enhance the turnover rate of PPARγ (44).

To first determine whether the two PPARγ isoforms displayed any functional difference in our system, we compared the ability of HA-PPARγ1 and -γ2 to activate a number of known target genes. Interestingly, there was no difference in the transcriptional activity between the isoforms, neither in the absence nor in the presence of agonist (supplemental Fig. 1C). We therefore concentrated our subsequent studies on the comparison of full-length HA-PPARγ2 and AB-domain truncated HA-PPARγCDE.

Deletion of the PPARγ A/B-domain affects neither DNA binding in vitro nor nuclear localization

Immunostaining combined with confocal microscopy showed that the majority of the NIH-CAR cells (∼90%) were targeted by the adenovirus and that transduction efficiency was similar between the different cells in the culture.

There have been several reports on proteins that bind and induce nuclear export of the PPARs (45, 46, 47), but as shown in Fig. 1E, both HA-PPARγ2 and HA-PPARγCDE were primarily localized to the nucleus. To investigate whether deletion of the A/B-domain affects the DNA-binding activity of the adenovirally expressed HA-PPARγ, we performed EMSAs using nuclear extracts from infected NIH-CAR cells (Fig. 1F). These data clearly showed that in vitro DNA binding and heterodimerization on several different PPREs are fully maintained in the HA-PPARγCDE protein.

Thus, using adenoviral vectors, we are able to express equal and moderate levels of full-length and truncated HA-PPARγ in NIH-CAR cells and thereby compare their ability to acutely activate target genes. Both proteins localize to the nucleus in NIH-CAR cells and display equal binding to naked DNA in vitro.

The role of the A/B-domain in PPARγ-mediated transactivation is gene specific

In the absence of exogenous agonist, deletion of the A/B-domain inhibited HA-PPARγ-mediated transactivation of most genes investigated to levels close to the background expression (Fig. 2). Furthermore, addition of the PPARγ-specific antagonist GW9662 (48) did not suppress transactivation either by full-length or by truncated PPARγ (Fig. 2). These results corroborate previous findings that the concentration of endogenous agonist is low in NIH-CAR fibroblasts (9) and indicate that, at least in these cells, ligand-independent transactivation by the HA-PPARγ2:RXR heterodimer is primarily mediated through the PPARγ2 A/B-domain, rather than the ligand-binding domain (LBD) or RXR. Interestingly, however, in the presence of the potent PPARγ-specific agonist rosiglitazone, deletion of the A/B-domain only led to a decrease in the transactivation of some but not all target genes.

To further compare the ability of full-length and A/B-domain-deleted HA-PPARγ to acutely activate target genes in the presence of rosiglitazone, we performed transcriptome analyses of three independent biological replicas in parallel with the transcriptome analyses described above (Table 1, supplemental Table 3, and Fig. 2). For most PPARγ target genes, the A/B-domain appeared to have only very minor effects on gene expression in the presence of rosiglitazone. Notably, this group of A/B-domain-independent target genes included all the known and putative PPAR target genes involved in both mitochondrial and peroxisomal β-oxidation of fatty acids, which we found to be up-regulated in response to HA-PPARγ2 expression (Table 1 and Fig. 2). This was also the case for genes associated with lipolysis, insulin/IGF-I signaling, and glycogen/glucose metabolism (Table 1 and Fig. 2). In fact, only a small subset of 25 genes proved to be dependent on the A/B-domain to become fully activated. Interestingly, most of the PPARγ target genes involved in lipid storage are included in this group (Table 1 and Fig. 2). The only gene in this cluster that was induced equally well by full-length and truncated HA-PPARγCDE encodes ADRP. Like the products of the A/B-domain-dependent target genes CIDEC/FSP27 and PLIN, ADRP is localized to lipid droplets but is in contrast selectively associated with lipid droplets subjected to lipolysis (49). Whereas the expression of CIDEC/FSP27 and PLIN is limited to a few tissues that employ specific signaling pathways to regulate lipogenesis and lipolysis, ADRP is ubiquitously expressed (35, 50) and can be activated equally well by PPARα and PPARγ2 in contrast to other genes involved in lipid storage such as PLIN and A-FABP (9).

Activation of the PPARγ:RXR heterodimer with RXR agonist could not compensate for the lack of the A/B-domain activity in target gene regulation. However, in combination with rosiglitazone HA-PPARγCDE, transactivation of the highly A/B-domain-dependent genes CD36 and CIDEC/FSP27 was potentiated by the RXR agonist (supplemental Fig. 2).

It has previously been found that the A/B-domain confers subtype specificity to the PPARs in part by preventing nonselective target gene activation (13). In keeping with these results, RT-qPCR experiments demonstrated that in the presence of rosiglitazone, HA-PPARγCDE was approximately twice as potent as full-length HA-PPARγ2 in the transactivation of the gene encoding 3-hydroxy-3-methyl-glutaryl-coenzyme A synthase 2 (HMG-CoA S2) (Fig. 3C), which has previously been shown to be activated selectively by PPARα (9). This observation could not be explained by increased affinity of the A/B-domain-truncated receptor toward the HMG-CoA S2 PPRE in vitro (Fig. 1F, right). The induction of HMG-CoA S2 could be detected only by RT-qPCR analysis, and the array data did not reveal the existence of other genes where the A/B-domain significantly inhibited HA-PPARγ2 transactivation in this cell type.

Both PPARα and PPARγ have a strong N-terminal transactivation domain, whereas that of PPARδ is very weak (15). To investigate whether the A/B-domain of PPARα could substitute for the PPARγ2 A/B-domain, we constructed an adenoviral vector encoding an HA-tagged chimeric PPAR consisting of the PPARα A/B-domain fused to the CDE domains of PPARγ (HA-PPARαA/BγCDE) (Fig. 3, A and B). Notably, we found that the transcriptional activity and specificity of the chimeric PPAR was comparable to that of HA-PPARγCDE on all lipogenic target genes investigated (Fig. 3D), thereby clearly demonstrating that PPARα and -γ A/B-domains are functionally distinct and that PPARγ is unique in its ability to promote activation of the lipogenic subset of target genes. In contrast, the PPARα A/B-domain potentiated transactivation of the PPARα-selective target gene HMG-CoA S2 (Fig. 3C).

These results show that the PPARγ A/B-domain contributes considerably to the lipogenic subtype characteristics of PPARγ by enabling full transactivation of the subset of target genes involved in lipid storage. By contrast, although the PPARγ2 A/B-domain may account for some weak ligand-independent transactivation, it plays a minor role in the activation of the remaining PPARγ target genes. Finally, the A/B-domain renders PPARγ less efficient in activating the expression of the PPARα selective target gene HMG-CoA S2 and possibly other PPARα target genes yet to be identified.

PPARγ A/B-domain deletion results in decreased lipid accumulation 48 h after transduction

It has previously been demonstrated that NIH-3T3 fibroblasts constitutively expressing ectopic PPARγ2 can be induced to accumulate lipid and develop an adipocyte-like phenotype (17). Because the PPARγ2 A/B-domain proved important for complete induction of genes involved in lipid storage, we compared the ability of HA-PPARγ2 and HA-PPARγCDE to promote cellular lipid accumulation. By measuring the incorporation of [¹⁴C]acetate into triglyceride 48 h after adenoviral transduction, we established that even such brief exposure to AdHA-PPARγ was sufficient to induce lipid accumulation in the NIH-CAR cells. Notably, HA-PPARγ2 expression increased the incorporation of [¹⁴C]acetate by 62%, whereas the A/B-domain-truncated HA-PPARγCDE displayed a much lower lipogenic potential with an incorporation of only 28% above background (Fig. 4B).

Full-length and truncated PPARγ bind equally well to genomic target sites but differ in their ability to recruit RNAPII

To investigate the molecular mechanisms underlying the potentiation of transactivation by the PPARγ2 A/B-domain, we performed chromatin immunoprecipitation (ChIP) experiments. Similar to in vitro DNA binding (Fig. 1F), ChIP assays with antibodies against HA or RXR demonstrated that binding of the HA-PPARγ2:RXR and HA-PPARγCDE:RXR heterodimers to genomic PPREs is comparable (Fig. 5). The only exception is the A-FABP PPRE to which the truncated HA-PPARγCDE binds less well than full-length HA-PPARγ2. Thus, in general, the PPARγ2 A/B-domain does not appear to affect binding to genomic target sites, either for A/B- domain-dependent or for A/B-independent target genes.

We next investigated the ability of full-length and truncated HA-PPARγ to recruit RNAPII to the corresponding target promoters. For all promoters analyzed, RNAPII recruitment (Fig. 6) was found to reflect the mRNA expression pattern (Fig. 2). Thus, for A/B-domain-dependent target promoters, deletion of the A/B-domain decreased RNAPII recruitment by HA-PPARγCDE, whereas RNAPII was recruited equally well or even slightly better by the truncated PPARγ on the A/B-domain-independent target genes.

The PPARγ A/B-domain is important for the recruitment of CBP and p300 to the A/B-domain-dependent target genes

Because receptor binding to target sites in chromatin was unaffected by the A/B-domain deletion, we sought to establish whether the PPARγ A/B-domain is important for recruitment and/or stabilization of specific coactivators. Recruitment of thyroid hormone receptor-associated protein 220 (TRAP220), a PPAR-interacting subunit of the mediator complex, was increased on all PPREs investigated upon HA-PPARγ expression although to different extents (Fig. 7). Interestingly, for all target genes except A-FABP, TRAP220 recruitment was not markedly dependent on the PPARγ2 A/B-domain. This indicates that the difference in TRAP220 recruitment at the A-FABP PPRE may simply reflect the difference in receptor binding, and we conclude that the deletion of the PPARγ A/B-domain does not interfere with TRAP220 recruitment, which is in keeping with the previously reported interaction through the PPARγ LBD (51).

The coactivator Tip60 interacts with PPARγ through the A/B-domain and has been demonstrated to promote adipogenesis and lipid accumulation in 3T3-L1 cells (18). Nevertheless, we were unable to detect Tip60 recruitment to the PPREs of PPARγ target genes in NIH-CAR cells (results not shown). The acetyl transferase CBP and the highly homologous p300 have been shown to interact with both the A/B- and E-domains of PPARγ2 (19). Interestingly, we found that deletion of the A/B-domain severely decreased or abolished HA-PPARγ-mediated recruitment of CBP on A/B-domain-dependent but not independent target genes (Fig. 8). p300 was recruited to only a subset of targets, but recruitment followed the same pattern as for CBP, i.e. for A/B-domain-independent genes, recruitment of p300 was unaffected by deletion of the A/B-domain, whereas A/B-domain deletion completely abolished recruitment to the A/B-domain-dependent gene A-FABP (Fig. 8). These data indicate that a major function of the PPARγ A/B-domain is to potentiate recruitment or stabilization of coactivator complexes containing CBP and p300 to a subset of target genes.

Adenovirus generation and purification

The generation of recombinant adenovirus encoding full-length HA-tagged mouse PPARγ2 and PPARα using the AdEasy cloning system (Stratagene, La Jolla CA) has previously been described (9). Similarly, PCR cloning was used to generate the construct encoding HA-tagged PPARγ1, the A/B-domain-deleted PPARγ (HA-PPARγCDE) and to create domain constructs, i.e. HA-PPARα-A/B and PPARγCDE, that were subsequently fused in a conserved region to avoid mutations in the chimeric HA-PPARαA/BγCDE. HA-PPARγ1, HA-PPARγCDE, and HA-PPARαA/BγCDE were cloned into pShuttle-CMV (Stratagene) (where CMV is cytomegalovirus) using the SalI/NotI, SalI/EcoRV, or SalI/NotI restriction sites, respectively. The adenoviruses were packed and amplified in 293-HEK cells (ATCC CRL-1573; American Type Culture Collection, Manassas, VA), and viral titers were estimated by the tissue culture infectious dose 50 (TCID₅₀) plaque assay-based approach (as described in the AdEasy protocol; Stratagene).

Adenoviral transduction

Adenoviral transduction was performed as previously described (9). Briefly, adenovirus was suspended in medium and added to NIH-CAR cells at 80% confluency. After 2 h transduction, the virus-containing medium was removed and new medium containing the vehicle dimethylsulfoxide (DMSO), 1 μm of the PPARγ-specific agonist rosiglitazone/BRL49653 (Novo Nordisk Copenhagen, Denmark) (25), or 1 μm of the PPARγ-specific antagonist GW9662 (Sigma-Aldrich, St. Louis, MO) (48) was added for an additional 6 h.

Nuclear extraction

Nuclei were purified from the transduced NIH-CAR cells by washing them twice in PBS followed by resuspension in lysis buffer containing 20 mm Tris-Cl (pH 7.5), 10 mm NaCl, 3 mm MgCl₂, 1 mm 1,4-dithioerythritol, 1 mm Na orthovanadate, 30 mm β-glycerol phosphate, and Complete. The cell membrane was disrupted by addition of 1% Nonidet P-40 followed by 12 strokes in a Dounce homogenizer. Nuclear extracts were prepared as previously described (61) and used for EMSA (62). The sequences of the annotated PPREs used as probes are listed in supplemental Table 2.

Western blotting

Adenovirally transduced NIH-CAR cells were harvested in hypotonic SDS sample buffer (63) and subjected to Western blotting as previously described (9). The membrane was probed with the following primary antibodies: anti-PPARγ (sc-7273; Santa Cruz Biotechnology, Santa Cruz, CA), anti-TFIIB (sc-225; Santa Cruz), anti-CBP (sc-369; Santa Cruz), anti-p300 (sc-584; Santa Cruz), anti-Sp1 (sc-59; Santa Cruz), and anti-HA (12CA5) (64). The secondary antibodies used for enhanced chemiluminescence detection (ECL; Amersham Phamacia Biotech, Piscataway, NJ) were horseradish peroxidase-conjugated goat antimouse IgG (P0447; Dako, Glostrup, Denmark) and swine antirabbit IgG (P0339; Dako).

Immunostaining and confocal microscopy

The NIH-CAR cells were seeded on glass coverslips and transduced with adenovirus. Subsequently the cells were washed in PBS supplemented with 2.5 mm Ca²⁺ and 4 mm Mg²⁺ and fixed with 4% formaldehyde. After incubation with 100 mm glycine, cells were permeabilized by incubation in 0.1% Triton X-100. Nonspecific binding sites were blocked by incubation in 3% BSA dissolved in PBS supplemented with Ca²⁺ and Mg²⁺. Cells were stained with αPPARγ (sc-7273; Santa Cruz) and subsequently with Alexa fluor 546-conjugated antimouse IgG (A11030; Molecular Probes, Eugene, OR) and 4′,6-diamidino-2-phenylindole (DAPI). Coverslips were mounted onto slides and observed through a Zeiss LSM 510 META confocal microscope. Two-photon excitation of the fluorophores was obtained using a MaiTai XF-W2S Ti:Sa laser (Broadband Mai Tai with 10-W Millennia pump laser, tunable excitation range 710–980 nm; Spectra Physics, Mountain View, CA). The excitation wavelength was 730 nm. Images were collected simultaneously in two different channels using band-pass filters of 460 ± 25 and 590 ± 25 nm for DAPI and Alexa fluor 546, respectively.

[¹⁴C]Acetate incorporation into cellular lipid

NIH-CAR cells were transduced with adenovirus and incubated with DMSO or rosiglitazone. Twenty-four hours after transduction, 35 nmol [¹⁴C]acetate (specific activity 57 mCi/mmol) was added to each well and incubation continued for another 24 h (this incubation period was chosen based on the results from a preliminary study, data not shown). Uptake was terminated by removal of the medium, followed by a wash with Krebs-Ringer buffer and ice-cold PBS. The cells were harvested in PBS and sonicated for four times for 30 sec each with 30-sec intervals at level 5 in the Bioruptor (Diagenode, Liege, Belgium). Protein concentration was determined on a Qbit (Invitrogen, Carlsbad, CA), and total lipid was extracted and purified using the Bligh and Dyer method (65). After drying in a stream of N₂, samples were redissolved in 1:2 chloroform/methanol and lipids were separated on silica thin layer chromatography plates (HPTLC silica gel 60; Merck, Darmstadt, Germany) developed in hexane-diethylether-acetic acid (70:30:1, vol/vol). The incorporated radioactivity was detected by a Typhoon TRIO scanner (Amersham Biosciences) and quantified using the ImageQuant TL software.

RNA extraction and cDNA synthesis

The adenovirally transduced NIH-CAR cells were harvested in 500 μl TRIzol. Total RNA was extracted by addition of chloroform and precipitated with isopropanol and centrifugation. The pellet was washed with 75% ethanol and redissolved in diethyl pyrocarbonate-treated water. From each preparation, 1 μg RNA was subjected to deoxyribonuclease I (Invitrogen) treatment, and cDNA was synthesized using random deoxynucleic acid hexamers and reverse transcriptase (RT) (First-Strand Kit; Invitrogen) as previously described (63). All experiments were performed in duplicate.

Microarray analysis

RNA samples from three individual experiments in which NIH-CAR cells were transduced with AdEmpty followed by addition of vehicle (DMSO) or transduced with either AdHA-PPARγ2 or AdHA-PPARγCDE and treated with rosiglitazone were purified further on RNAeasy columns (QIAGEN, Valencia, CA). Two micrograms of total RNA was used to synthesize double-stranded cDNA with the one-cycle cDNA synthesis kit (Affymetix Inc., Santa Clara, CA) using an oligo(dT) primer containing a T7 RNA polymerase promoter (Affymetrix). The cDNA was purified using the GeneChip clean-up module (Affymetrix) and used as the template for an in vitro transcription reaction to synthesize biotin-labeled antisense cRNA (IVT labeling kit; Affymetrix). After fragmentation at 94 C for 35 min in fragmentation buffer (40 mm Tris, 30 mm magnesium acetate, and 10 mm potassium acetate), 15 μg labeled cRNA was hybridized for 16 h to the Mouse 430 2.0 array (Affymetrix) that contains 45,000 probe sets, which analyze the expression level of over 39,000 transcripts and variants from over 34,000 well characterized mouse genes. The raw intensity data was normalized using the RMA algorithm implemented in the Bioconductor package affy running under R version 2.8.1. The normalized expression data were exported to the DNA Chip Analyzer program (dChip, November 2007, www.dchip.org). Three group comparison analyses were performed, HA-PPARγ2-vs.-AdEmpty, HA-PPARγ2-vs.-HA-PPARγCDE, and HA-PPARγCDE-vs.-AdEmpty, respectively. A gene was defined as being differentially expressed if it had a P value below 0.05 in a paired Welch t test (variance not equal), an absolute fold change greater than 1.5, and a difference of group means greater than 50. Functional clustering of the up-regulated genes was performed using Ingenuity Pathway Analysis (Ingenuity Systems Inc., Redwood City, CA).

ChIP

The adenovirally transduced NIH-CAR cells were fixed with 1% formaldehyde for 10 min followed by incubation with 12.5 mm glycine. Cells were washed twice in PBS and harvested in lysis buffer containing 0.1% SDS, 1% Triton X-100, 0.15 m NaCl, 1 mm EDTA, and 20 mm Tris (pH 8). Cells were sonicated eight times for 30 sec each with 30-sec intervals at level 5 in the Bioruptor (Diagenode). The DNA concentrations were determined by measuring OD₂₆₀ and subsequently adjusted to equal levels. Protein A beads (Amersham Pharmacia Biotech) were prepared by three washes in IP buffer containing 0.15% SDS, 1% Triton X-100, 0.15 m NaCl, 1 mm EDTA, and 20 mm HEPES (pH 7.6) followed by 2 h incubation in IP buffer supplemented with 1 μg/μl BSA. One hundred micrograms of chromatin were diluted three times in IP buffer and incubated overnight with 1–2 μg antibody and 100 μl prepared beads rotating at 4 C. Beads were washed at 4 C, twice with buffer 1 containing 0.1% SDS, 0.1% deoxycholic acid (DOC), 1% Triton X-100, 0.15 m NaCl, 1 mm EDTA, and 20 mm HEPES (pH 7.6); once in buffer 2 containing 0.1% SDS, 0.1% DOC, 1% Triton X-100, 0.5 m NaCl, 1 mm EDTA, and 20 mm HEPES (pH 7.6); once in buffer 3 containing 0.25 m LiCl, 0.5% DOC, 0.5% Nonidet P-40, 1 mm EDTA, and 20 mm HEPES (pH 7.6); and twice in buffer 4 containing 1 mm EDTA, and 20 mm HEPES (pH 7.6). DNA-protein complexes were eluted with elution buffer (1% SDS and 0.1 m NaHCO₃) and de-cross-linked by adding 0.2 m NaCl followed by shaking overnight at 65 C. DNA was isolated by chloroform-phenol extraction. Immunoprecipitated DNA and 10% input DNA were analyzed by qPCR. Antibodies used for IP were anti-RXRα (sc-774; Santa Cruz), anti-CBP (sc-369; Santa Cruz), anti-p300 (sc-584; Santa Cruz), anti-TRAP220 (sc-8998; Santa Cruz), anti-polII (8WG16; Covance Research Products, Princeton, NJ), and anti-HA (12CA5) (64). For each ChIP experiment, the enrichment of target site and background control (β-globin promoter), respectively, was normalized to input DNA (i.e. recovery). To relate independent experiments, we subsequently determined the relative occupancy by normalizing the recovery of the target site and β-globin promoter in the individual samples to the recovery of the β-globin promoter in the control cells transduced with AdEmpty.

siRNA-mediated knockdown

Twenty-four hours before adenoviral transduction, NIH-CAR cells were switched to OptiMEM (Invitrogen) and transfected with combinations of predesigned siRNA against firefly luciferase [CGUACGCGGAAUACUUCGA(dT)(dT); Sigma-Aldrich, St. Louis, MO], CBP [GUAACUCUGGCCAUAGCUU(dT)(dT) Sigma-Aldrich], or p300 [GAAUGACUUUCUGAGGCGA(dT)(dT) Sigma-Aldrich] in a final concentration of 50 nm using RNAiMAX (Invitrogen). The transfection medium was replaced with normal culture medium 4 h before the adenoviral transduction.

We thank P. Sauerberg (Novo Nordisk A/S) for the rosiglitazone/BRL49653 ligand, L. A. Bagatolli for expert assistance with the confocal microscope, and T. Obsen and N. J. Færgeman for instructions on lipid analyses.