SCFA measurement.

SCFA concentrations were measured in the luminal content of different parts of the gastrointestinal (GI) tract of C57BL/6 mice fed a low-fat diet. After sacrifice, luminal content was collected in tubes containing phosphoric (H₃PO₄) and isocaproic acid, after which the tubes were thoroughly mixed and stored at −20°C. For analysis, samples were thawed, mixed on a vortex, and centrifuged at 14,000 rpm for 5 min. The supernatant was collected and SCFA concentrations were determined by gas chromatography (Fisons HRGC Mega 2; CE Instruments, Milan, Italy) at 190°C using a glass column fitted with Chromosorb 101. The carrier gas was N₂ saturated with methanoic acid, and isocaproic acid was used as an internal standard.

Cell culture.

T84 and HT29 human colonic carcinoma epithelial cells were cultured in Dulbecco's modified Eagle medium (DMEM) or DMEM containing Ham's F-12 medium (DMEM F-12) (Lonza, the Netherlands) containing 5 or 10% fetal calf serum (FCS), respectively, and 1% penicillin-streptomycin at 37°C in a humidified incubator under 95% air and 5% CO₂. Cells were grown until 80 to 90% confluence and treated for the indicated time periods with sodium acetate, sodium propionate, and sodium butyrate from a stock in phosphate-buffered saline (PBS) to a final concentration of 1 or 8 mM. Alternatively, cells were treated with synthetic agonists for PPARα (Wy14643; 5 μM), PPARδ (GW501516; 1 μM), PPARγ (rosiglitazone; 1 μM), synthetic PPARγ antagonist GW9662 (5 μM), or trichostatin A (100 nM) from a stock in dimethylsulfoxide (DMSO). In one experiment, cells were preincubated with α-amanitin (10 μg/ml) from a stock in water for 1 h, followed by 24 h of coincubation with butyrate at 1 mM. After the indicated time points, medium was collected and used for analysis of ANGPTL4 by enzyme-linked immunosorbent assay (ELISA), and/or cells were harvested for subsequent RNA isolation.

siRNA transfection.

Silencing of the PPARγ gene in T84 cells was done using ON-TARGETplus SMARTpool for human PPARγ (Thermo Scientific Dharmacon, Etten-Leur, the Netherlands), representing a mixture of four short interfering RNAs (siRNAs): CAAAUCACCAUUCGUUAUC, GACAUGAAUUCCUUAAUGA, GAUAUCAAGCCCUUCACUA, and GACAGCGACUUGGCAAUAU. As a control, we used an ON-TARGETplus nontargeting pool (Thermo Scientific Dharmacon). Transfection was done according to the manufacturer's protocol using DharmaFECT 1 transfection reagent. The cells were transfected with 100 nM siRNA and incubated for 48 h, followed by treatment with butyrate (1 mM) for a period of 24 h. Thereafter, medium was collected for ANGPTL4 protein analysis by ELISA and the cells were used for RNA isolation and subsequent quantitative PCR (qPCR).

PPARγ ChIP.

T84 cells were grown in 10-cm plates to a density of approximately 5 million cells/dish and treated with butyrate (8 mM) or rosiglitazone (1 μM). After 1 h, medium was removed and cells were fixed in 1% formaldehyde in PBS. After 10 min, cross-linking was stopped by the addition of glycine to a final concentration of 0.125 M. The cells were washed three times with ice-cold PBS and resuspended in 0.3 ml of lysis buffer (0.1% SDS, 1% Triton X-100, 0.15 M NaCl, 1 mM EDTA, 20 mM Tris, pH 8.0). The lysate was frozen at −80°C until further analysis. Chromatin immunoprecipitation (ChIP) was carried out as described by Siersbæk et al. (21), with the exceptions that disuccinimidyl glutarate was not used for cross-linking and sonication was performed in 2 rounds of 10 on/off cycles of 30 s each. The PPARγ antibody used was PPARγ H100 (sc-7196; Santa Cruz Biotechnologies).

PPARγ stable reporter assay.

The PPARγ CALUX cell line was obtained from BioDetection Systems B.V. (Amsterdam, the Netherlands). PPARγ CALUX cells are based on human osteoblastic (U2-OS) cells (American Type Culture Collection [ATCC], Manassas, VA) stably transfected with a human PPARγ expression plasmid and a luciferase reporter construct (22). The cells were cultured as described previously in a 1:1 mixture of DMEM and Ham's F-12 medium (Invitrogen, Breda, the Netherlands) supplemented with 7.5% fetal bovine serum (FBS) (Invitrogen), 1% nonessential amino acids (Invitrogen), and penicillin-streptomycin (Invitrogen) to final concentrations of 10 U/ml and 10 μg/ml, respectively (22, 23). Once per month 200 μg/ml G418-disulfate was added to the culture medium.

The PPARγ CALUX assay was performed as described previously (22, 24). In short, for the exposure experiments, PPARγ-CALUX cells were plated in the 96-well plates with phenolred-free 1:1 mixture of DMEM and Ham's F-12 medium supplemented with 5% dextran-coated charcoal-stripped FBS (Invitrogen) at a volume of 200 μl per well. The next day, medium was refreshed and cells were incubated in triplicates with test compounds added to the culture medium. After 24 h, the cells were checked visually for cytotoxicity, the medium was removed, and the cells were lysed in luciferase cell culture lysis reagent (Promega, Leiden, the Netherlands). Luciferase activity was measured in cellular extracts using a Synergy HT multidetection microplate reader (BioTek Instruments Inc.). For each test compound at last two independent experiments were performed.

GAL4-PPAR stable reporter assay.

The GAL4-PPAR stable reporter assay was carried out in HeLa cells stably expressing a chimeric protein containing the ligand binding domain (LBD) of human PPARα, PPARδ, or PPARγ fused to the yeast transactivator GAL4 DNA binding domain (DBD) and a luciferase reporter gene driven by a pentamer of the GAL4 recognition sequence in front of a β-globin promoter. The cells were seeded in 96-well plates and incubated the following day with tested compounds for 24 h. At the end of the incubation, the luciferase activity was measured using a BMG LUMIstar Galaxy luminometer. Results expressed as relative light units (RLU) were obtained from experiments performed in triplicate. Rosiglitazone, GW7647, and L165041 (all at 1 μM) were used as positive controls for activation of PPARα, PPARδ, and PPARγ, respectively. The measurements were performed as a commercial service by Tebu-bio Laboratories (Le Perray-en-Yvelines, France).

ANGPTL4 ELISA.

ANGPTL4 levels in the medium were measured by ELISA as detailed previously (17). Briefly, 96-well plates were coated with anti-human ANGPTL4 polyclonal goat IgG antibody (AF3485; R&D Systems) and incubated overnight at 4°C. Plates were washed extensively between each step. After blocking, 100 μl of medium of cells was applied, followed by 2 h of incubation at room temperature. A standard curve was prepared using recombinant human ANGPTL4 (3485-AN; R&D Systems) at 0.3 to 2.1 ng/well. One hundred microliters of diluted biotinylated anti-human ANGPTL4 polyclonal goat IgG antibody (BAF3485; R&D Systems) then was added for 2 h, followed by addition of streptavidin-conjugated horseradish peroxidase for 20 min and tetramethyl benzidine substrate reagent for 6 min. The reaction was stopped by the addition of 50 μl of 10% H₂SO₄, and the absorbance was measured at 450 nm.

Cofactor recruitment assay.

Nuclear receptor PamChip arrays (PamGene, s'Hertogenbosch, the Netherlands) were used as described previously (25). Upon binding a ligand, PPARs undergo a conformational change that promotes the formation of a cofactor binding pocket, subsequently allowing interaction with the so-called LXXLL motif within some coregulators. The PamChip arrays consist of 53 peptides encompassing the LXXLL motifs of 21 different coregulator proteins. The sequences are provided as supplementary data in Koppen et al. (25). Briefly, the arrays were incubated with glutathione S-transferase-tagged PPARγ LBD (Invitrogen, Breda, the Netherlands) in the presence and absence of the ligands, i.e., sodium acetate, sodium propionate, sodium butyrate (each at 40 mM), and rosiglitazone (1 μM). Quantification of the interaction between PPARγ and coregulators was made using Alexa 488-conjugated anti-glutathione S-transferase rabbit polyclonal antibody (Invitrogen).

Affymetrix GeneChip microarray analysis.

Microarray analysis was performed on T84 cells treated with 0.1 μM rosiglitazone or 1 mM butyrate for 24 h. Total RNA was prepared using TRIzol reagent (Invitrogen) and purified on columns using the RNeasy Minikit (Qiagen, Venlo, the Netherlands). RNA quality and integrity were verified with the RNA 6000 Nano assay on an Agilent 2100 Bioanalyzer (Agilent Technologies, Amsterdam, the Netherlands). Hybridization, washing, and scanning of the Affymetrix Human Gene 1.1 ST array plate was performed on Affymetrix GeneTitan. Scans of the Affymetrix arrays were processed using packages from the Bioconductor project (26). Raw signal intensities were obtained by robust multiarray (RMA) normalization (27). Probe sets were defined according to Dai et al. using remapped chip definition file (CDF), version 15, based on data from the Entrez gene database (28).

Microarray analysis of colon of mice fed inulin was carried out as described above using the Affymetrix Mouse Gene 1.1 ST array plate.

3T3-L1 adipogenesis.

3T3-L1 fibroblasts were amplified in DMEM-10% FCS and subsequently seeded into six-well plates. Two days after the cells reached confluence, the medium was changed to DMEM-10% fetal calf serum containing 0.5 mM 3-isobutyl-1-methylxanthine, 2 μg/ml insulin (Actrapid), 0.5 μM dexamethasone, and either the short-chain fatty acids or rosiglitazone (1 μM). After 2 days, the medium was changed to DMEM-10% fetal calf serum containing 2 μg/ml insulin and the short-chain fatty acids or rosiglitazone. Cells were harvested after two more days for RNA isolation (day 4). This mild adipogenesis protocol permits assessment of PPARγ agonist activity of added compounds (29). Oil red O staining was performed at day 10 using a standard protocol.

In a second protocol, 2 days after the cell confluence (i.e., day 0), the medium was changed to DMEM-10% fetal calf serum containing 0.5 mM 3-isobutyl-1-methylxanthine, 5 μg/ml insulin, and 1 μM dexamethasone. After 3 days, the medium was changed to DMEM-10% fetal calf serum containing 5 μg/ml insulin. The medium was subsequently changed every 3 days, and no further insulin was added after 6 days. At day 10, 3T3-L1 adipocytes were treated with rosiglitazone (1 μM) and the short-chain fatty acids (8 mM) for 24 h.

Adipogenesis in SGBS.

Human Simpson-Golabi-Behmel syndrome (SGBS) adipocytes were cultured and grown to confluence in 0F medium (DMEM–F-12–biotin–panthotenate–penicillin-streptomycin) plus 10% FCS. One day postconfluence, the medium was changed to Quick-diff medium, which is 3FC medium (0F medium plus 0.01 mg/ml human apotransferrin, 2 × 10⁻⁸ M insulin, 0.1 μM cortisol, 0.2 nM T3) with 50 nM dexamethasone, 500 μM 3-isobutyl-1-methylxanthine, and 2 μM rosiglitazone or 8 mM SCFA. On day 4, medium was changed to 3FC medium and changed every 4 days until day 15. Expression of differentiation markers and PPARγ targets was determined by qPCR at day 15. DMSO was used as a control.

RNA isolation and qPCR.

RNA was isolated from T84 cells using RNeasy columns or TRIzol. One μg of total RNA was reverse transcribed with iScript (Bio-Rad, Veenendaal, the Netherlands) or a Fermentas cDNA synthesis kit according to the instructions from the manufacturer. cDNA was amplified on a Bio-Rad CFX384 real-time system using Sensimix (Bioline, GC Biotech, Alphen aan de Rijn, the Netherlands). Cyclophilin or 36B4 was used as a housekeeping gene. PCR primer sequences were taken from the PrimerBank and ordered from Eurogentec (Seraing, Belgium). Primer sequences are presented in Table 1.

Modeling of butyrate binding to PPARγ.

For modeling the structure of the complex between PPARγ and butyrate, HADDOCK version 2.1 was used (30). HADDOCK is a highly successful modeling approach that makes use of structural knowledge when available to drive the docking procedure. In this case, the X-ray crystal structure of PPARγ with bound decanoic acid (DA; 3U9Q) was used to identify the likely binding site for short-chain fatty acids such as butyrate (31). The docking was performed using the web server version of HADDOCK (32), with the following modifications. The high-temperature rigid-body torsion angle dynamics and the first rigid-body cooling stage were not performed (i.e., the number of steps was set to 0), and the second cooling phase of torsion angle dynamics with flexible side chains at the interface was started with a reduced initial temperature of 500K instead of the default 1,000K.

All calculations were performed with CNS1.2 (33). Nonbonded interactions were calculated with the OPLS force field using a cutoff of 8.5 Å (34). The electrostatic potential (Eelec) was calculated by using a shift function, while a switching function (between 6.5 and 8.5Å) was used to define the Van der Waals potential (EvdW). The HADDOCK score (in arbitrary units [AU]) was used to rank the generated poses. It is a weighted sum of intermolecular electrostatic (Eelec), van der Waals (EvdW), desolvation (Edesolv), and ambiguous interaction restraint (AIR) energies with weight factors of 0.2, 1.0, 1.0, and 0.1, respectively.

The ambiguous interaction restraints used to drive the docking of butyrate to PPARγ were derived from the X-ray crystal structure of PPARγ with bound DA (3U9Q), where protein residues near the decanoic acid were identified as the butyrate binding site. All protein residues within 3.9 Å of the decanoic acid (residues 278, 281, 282, 285, 289, 323, 356, 360, 363, 449, 469, and 473) were defined as active restraints for the rigid-body docking phase and as passive restraints for the subsequent semiflexible refinement stage. The ligand was considered active for both docking phases. This strategy effectively pulls the butyrate ligand into the binding site during rigid-body docking while allowing a thorough exploration of the binding pocket during the refinement stage. With the protein active-site residues defined as passive during the semiflexible refinement stage, the ligand is not restrained to any particular orientation or location within the defined binding site. For comparison, a second docking trial was performed with active residues for the entire PPARγ binding site, defined as amino acids within 3.9Å of either the partial agonist decanoic acid (from 3U9Q, residues 278, 281, 282, 285, 289, 323, 356, 360, 363, 449, 469, and 473) or the full agonist-nitrosylated fatty acid (from 3CWD, residues 285, 286, 288, 289, 326, 327, 330, 340, 341, 364, 449, and 473). Docking with this wider binding pocket definition did not give additional solutions.

At the end of the docking protocol, clustering based on ligand pairwise root mean square deviation (RMSD) criteria was performed, and the best-scoring structure of the best-scoring cluster was taken as the best solution. The clustering distance cutoff was reduced from the default 7.5-Å cutoff for protein-protein complexes to 1.75 Å, which is more suitable for protein-ligand complexes.

The interactions between decanoic acid and PPARγ were analyzed by submitting the X-ray crystal structure 3U9Q to the HADDOCK refinement server (32), which only performs a gentle refinement in explicit solvent and returns statistics similar to those of a full docking run.

PPARγ mediates induction of ANGPTL4 by SCFA.

Because the medium-chain fatty acid (MCFA) decanoic acid was recently shown to be a direct ligand of PPARγ and ANGPTL4 is a sensitive target of PPARγ in colonocytes (31, 38), we hypothesized that butyrate upregulates ANGPTL4 expression in colonocytes by activating PPARγ. To test the responsiveness of ANGPTL4 to PPAR activation in T84 and HT29 cells, the cells were treated with synthetic agonists for PPARα, PPARδ, and PPARγ. The PPARγ agonist rosiglitazone was the most potent inducer of ANGPTL4 protein secretion (Fig. 2A) and mRNA (Fig. 2B) in both cell types, followed by the PPARδ agonist GW501516 and the PPARα agonist Wy14643. These results are consistent with the high expression of PPARG mRNA in T84 cells (Fig. 2C, top) and, to a lesser extent, in HT29 cells (Fig. 2C, lower) and the corresponding protein levels of the three PPARs in the two cell lines (39–43). Upon incubation with rosiglitazone, a time- and dose-dependent increase in ANGPTL4 protein was observed in T84 cells (Fig. 2D and ​andE),E), indicating that ANGPTL4 production is highly sensitive to PPARγ activation. Accordingly, we considered PPARγ as a possible candidate mediating induction of ANGPTL4 mRNA by butyrate.

Open in a separate window

Fig 2

PPARγ potently stimulates ANGPTL4 in colon adenocarcinoma cells. Synthetic agonists for PPARα (Wy14643 [WY]; 5 μM), PPARδ (GW501516 [GW]; 1 μM), and PPARγ (rosiglitazone [Rosi]; 1 μM) stimulate ANGPTL4 secretion (A) and mRNA expression (B) in T84 and HT29 cells. Cntl, control. (C) Amplification curve of PPARα, PPARδ, and PPARγ mRNA as determined by qPCR of T84 and HT29 cells. Sizes of amplicons varied less than 10%. (D) Time course of induction of ANGPTL4 protein in medium by rosiglitazone (1 μM). (E) Dose-dependent induction of ANGPTL4 protein in medium by rosiglitazone. Unless indicated otherwise, cells were treated for 24 h. Error bars represent SD. An asterisk indicates a result significantly different from that of the control according to Student's t test (P < 0.05).

To investigate whether the stimulatory effect of butyrate on ANGPTL4 secretion occurs via PPARγ, we used the PPARγ antagonist GW9662. Induction of ANGPTL4 secretion by butyrate in T84 cells was strongly diminished by GW9662 (Fig. 3A), suggesting that butyrate increases ANGPTL4 secretion mainly by activating PPARγ. Similarly, siRNA-mediated knockdown of PPARγ mRNA by about 60% led to a very significant 40% decrease in the butyrate-induced ANGPTL4 mRNA level (Fig. 3B) and butyrate-induced ANGPTL4 protein secretion (Fig. 3C), again suggesting that the effect of butyrate on ANGPTL4 is largely mediated by PPARγ. Stimulation of PPARγ mRNA by butyrate (Fig. 3B) likely reflects PPARγ autoregulation. To further pursue the role of PPARγ in ANGPTL4 regulation by butyrate, we performed chromatin immunoprecipitation to study binding of PPARγ to the ANGPTL4 gene. The results indicate that PPARγ occupies the PPAR binding site within intron 3 of the ANGPTL4 gene (Fig. 3D), which is known to mediate PPAR responsiveness (44, 45). Occupancy was modestly induced by both butyrate and rosiglitazone. Taken together, these data strongly suggest that induction of ANGPTL4 by butyrate in T84 cells is mediated by PPARγ.

Open in a separate window

Fig 3

Induction of ANGPTL4 by butyrate is mediated by PPARγ. (A) Inhibitory effect of PPARγ antagonist GW9662 (5 μM) on induction of ANGPTL4 secretion in medium by rosiglitazone (10 nM) and butyrate (1 mM) in T84 cells. Effect of siRNA-mediated PPARγ knockdown (B, left) in T84 cells on butyrate-induced upregulation of ANGPTL4 mRNA (B, right) and ANGPTL4 protein in medium (C). (D) PPARγ ChIP-qPCR on selected loci in T84 cells treated with butyrate (8 mM) or rosiglitazone (1 μM) for 24 h. Bars represent the mean recovery plus ranges from two independent experiments. Error bars represent SD except when indicated otherwise. An asterisk indicates a significantly different result from that of the control according to Student's t test (P < 0.05).

To investigate whether butyrate acts as a general inducer of PPARγ-dependent gene regulation, we performed microarray analysis on T84 cells treated either with butyrate (1 and 8 mM) or rosiglitazone. Remarkably, effects of butyrate on global gene expression were much more pronounced than those of rosiglitazone, which is likely linked to its property as a potent HDAC inhibitor (Fig. 4A; also see Fig. S1 in the supplemental material). Interestingly, more overlap was observed between 1 mM butyrate and rosiglitazone than between 8 mM butyrate and rosiglitazone (see Fig. S1). Cluster analysis identified clusters of genes induced by rosiglitazone and butyrate (Fig. 4A). Genes in one cluster, which, besides ANGPTL4, included PPAR targets HMOX1, PDK4, and UCP2 (46–48), were induced by 1 and 8 mM butyrate. Genes in another cluster, which included other PPAR targets, including AQP8 and PLIN2 (49, 50), were induced by 1 mM butyrate and surprisingly were not induced or even suppressed by 8 mM butyrate. The role of PPARγ in mediating the stimulatory effect of 1 mM butyrate on expression of PLIN2 and UCP2, representing two distinct clusters, was verified by siRNA (Fig. 4B).

Open in a separate window

Fig 4

Global effects of butyrate in T84 cells. T84 cells were treated with rosiglitazone (1 μM) or butyrate (1 and 8 mM) for 24 h and subjected to Affymetrix microarray analysis. (A) Hierarchical clustering based on Pearson's correlation with average linkage. (B) Effect of siRNA-mediated PPARγ knockdown in T84 cells on induction of PLIN2 and UCP2 mRNA by butyrate. AQP8 mRNA (C) and PLIN2 mRNA (D) levels were determined in T84 cells treated with SCFA for 24 h at the indicated concentrations. Error bars represent SD. An asterisk indicates a result significantly different from that of the control according to Student's t test (P < 0.05).

The biphasic regulation of AQP8 and PLIN2 by butyrate but not by propionate and acetate was confirmed by qPCR (Fig. 4C). Such a biphasic regulation is possible only if a gene is regulated by at least two distinct mechanisms: a stimulatory effect, presumably via PPARγ, which is already active at lower butyrate concentrations, and a suppressive effect impacting a subset of PPAR targets at higher butyrate concentrations. Overall, these data indicate that SCFA acts as a general inducer of PPARγ-dependent gene regulation in T84 cells. However, especially at higher concentrations, the main effects of SCFA in T84 cells are independent of PPARγ.

SCFA function as PPARγ agonists.

To further study activation of PPARγ by SCFA, we employed a stable PPARγ reporter assay. HeLa cells stably transfected with a fusion construct between the DNA binding domain of Gal4 and the ligand binding domain of PPARγ, PPARα, or PPARδ were incubated with increasing concentrations of SCFA. Remarkably, PPARγ was strongly activated by butyrate and, to a lesser extent, by propionate (Fig. 5A). In contrast, PPARα and PPARδ were not activated by any of the SCFA, even at higher concentrations.

Open in a separate window

Fig 5

SCFA are agonists of PPARγ. (A) Stable GAL4-PPAR chimera reporter assay showing activation of PPARγ but not PPARα and PPARδ by SCFA at concentrations of ≥1 mM. The dotted lines represent the levels of luciferase activity reached upon incubation with synthetic PPAR agonists. Cells were treated for 24 h. Error bars represent SD. (B) Stable PPARγ reporter assay showing activation of PPARγ by SCFA at concentrations of ≥500 μM. Cells were treated for 24 h. Note the different x axis for rosiglitazone and the SCFA. Error bars represent SD. (C) A nuclear receptor PamChip assay was used to measure the interaction between PPARγ and immobilized peptides corresponding to specific coregulator-nuclear receptor binding regions in the presence and absence of rosiglitazone (1 μM), butyrate (40 mM), or acetate (40 mM). Representative images are shown. (D) Quantitation of the PamChip assay results for rosiglitazone and butyrate compared to the control. Arrows point to the same peptides as those described for panel C.

To corroborate activation of PPARγ by SCFA, we performed an alternative reporter assay in U2OS cells stably transfected with a human PPARγ expression construct and a PPRE-luciferase reporter construct. Similar to the results described above, propionate, but especially butyrate, markedly activated PPARγ reporter activity, while acetate had minimal effect (Fig. 5B). The threshold for stimulating reporter activity was 0.5 mM for butyrate and 10 nM for rosiglitazone, indicating that butyrate has relatively weak PPARγ agonist activity.

To explore the possibility that SCFA serve as direct PPARγ agonists, we used nuclear receptor PamChip arrays. In this system, the interaction between soluble nuclear receptors and 53 immobilized peptides corresponding to specific coregulator-nuclear receptor binding regions is studied. Using this system for PPARγ, we have previously generated ligand-specific coregulator interaction profiles (25). Both rosiglitazone and butyrate promoted the interaction between PPARγ and numerous coactivator peptides (e.g., CBP) (Fig. 5C). However, in contrast to rosiglitazone, butyrate did not relieve the interaction between PPARγ and the NCoR1 and NCoR2 corepressor peptides. Consistent with the PPARγ reporter data, acetate had little effect. Quantitative analysis of the PamChip arrays is presented in Fig. 5D. These data suggest that butyrate functions as a selective PPARγ modulator (SPPARM) (51).

Modeling of butyrate bound to PPARγ.

To find support for SCFA behaving as selective PPARγ modulators, we performed structural modeling of the binding of butyrate to PPARγ using HADDOCK (30, 32) (Fig. 6A). Docking of butyrate into PPARγ resulted in three different clusters of solutions, each showing butyrate bound in the ligand-binding pocket of the protein. The best cluster in terms of HADDOCK score was also the largest one (179/200 calculated structures) (Table 2) and contained the best scoring models. It reveals a binding mode very similar to that of decanoic acid within the crystal structure of PPARγ (3U9Q) (31). Specifically, the average RMSD between the butyrate and decanoic acid common atoms upon aligning the backbone atoms of the PPARγ (ligand RMSD) was 1.55 ± 0.29 Å for the best four cluster members. The remaining clusters have less favorable interaction energy.

Open in a separate window

Fig 6

Modeling of butyrate into the PPARγ binding pocket. (A) The model reveals the complex between butyrate and PPARγ with the best HADDOCK score (butyrate is shown as cyan sticks), overlaid with the crystal structure of the decanoic acid complex with PPARγ (3U9Q, a decanoic acid, is shown as green sticks) by aligning the protein backbone atoms of the two structures (ribbon displayed for the HADDOCK model). The displayed protein side chains are shown as thin cyan or green sticks and the side chains making contacts with the docked butyrate or decanoic acid, respectively. Hydrogen bond contacts between the butyrate and the protein are shown as yellow dashed lines. (B) Comparison of the binding location for coactivator peptide PGC-1α (blue ribbon) bound to PPARγ (HADDOCK model of the PPARγ-butyrate complex, left) with the binding site for the SMRT corepressor peptide (purple ribbon) binding to PPARα (1KKQ; right) (69). In both structures, the C-terminal portion of the PPAR molecule that forms helix AF-2 is colored pink for comparison. The structures were aligned using the backbone atoms of the receptors.

Interactions between butyrate and PPARγ are dominated by electrostatic interactions and hydrogen bonds between the carboxylate group and several hydrogen bond-donating protein side chains, which are very similar to those found in the structure with decanoic acid (Fig. 6A). In the latter can be found hydrogen bonds to four different side chains near the carboxylate group, namely, His323 HE2, Ser289 OG, His449 HE2, and Tyr473 OH. The top four structures of the best cluster (in terms of HADDOCK score) reveal specific hydrogen bonds from the butyrate oxygens to His323, Tyr473, and His449. The hydrogen bond to Ser289 is never observed, while in two of the structures a fourth hydrogen bond is made to Tyr327 OH. The average buried surface area (BSA) for the best four docking solutions is 286 ± 14 Ų, and the average HADDOCK score for the best four structures is −7 ± 3. In comparison, the buried surface area and HADDOCK score calculated for the decanoic acid-PPARγ structure are 478 Ų and −22.1, respectively.

We next explored the possible influence of butyrate on coactivator and corepressor binding (Fig. 6B). Fortunately, the structure of PPARγ with decanoic acid also contained a bound PGC1α coactivator peptide, allowing us to model butyrate docked within the crystal structure of PPARγ complexed with the PGC-1α peptide. However, no structures were available for corepressor (peptides) binding to PPARγ, which limited us to the crystal structure of PPARα complexed with an antagonist and the SMRT corepressor peptide (1KKQ). As shown in Fig. 6B, the binding sites for the coactivator (left, shown in blue) and corepressor (right, shown in purple) peptides occupy structurally analogous positions on PPARγ and PPARα and are likely to be mutually exclusive binding events. The C-terminal AF2 helix of PPARγ (shown in pink) forms part of the binding site for the coactivator peptide PGC-1α, and stabilization of this helix is known to promote binding of coactivator peptides (31). As can be seen from the structural comparison, the AF2 helix of PPARγ also occludes part of the structurally analogous binding site for the corepressor peptide binding to PPARα. The butyrate molecule in the HADDOCK model makes several contacts with the AF2 helix (such as the hydrogen bond to Tyr473 OH) that are very similar to the contacts that were observed between decanoic acid and the AF2 helix in the X-ray crystal structure of the decanoic acid-PPARγ complex (3U9Q), which is consistent with our finding using a coactivator peptide binding assay that butyrate can promote at least partial agonism of the PPARγ receptor but does not promote dissociation of corepressor peptides.

SCFA do not induce PPARγ-mediated adipogenesis.

Given that SCFA can activate PPARγ-mediated transcription in reporter assays but only partially overlap a full PPARγ agonist in their coregulator binding profile, SCFA may not display the same capacity as full agonists in cell-based assays. Since PPARγ is a critical regulator of adipogenesis and mediates stimulation of adipogenesis by rosiglitazone, we subsequently used stimulation of 3T3-L1 adipogenesis as an in vitro readout for PPARγ activation by SCFA. It has been shown that NCoR1 represses 3T3-L1 adipogenesis and that dismissal of NCoR1 is essential for induction of adipogenesis by PPARγ agonists (52, 53). Since SCFA/butyrate did not release the association between NCoR1 and PPARγ as observed using the PamChip array, it may be expected that SCFA are unable to mimic the adipogenic effects of PPARγ agonists (29). Consistent with this notion and in contrast to rosiglitazone, SCFA did not induce 3T3-L1 adipogenesis, as revealed by oil red O staining (Fig. 7A) and gene expression analysis of adipogenesis markers, which also represent PPARγ targets (Fig. 7B). In fact, at higher concentrations butyrate and propionate suppressed adipogenesis (Fig. 7C). Furthermore, in contrast to rosiglitazone, butyrate did not upregulate PPARγ targets in differentiated 3T3-L1 adipocytes (Fig. 7D). Similarly, in contrast to rosiglitazone, SCFA were unable to induce adipogenesis in the human SGBS adipocyte model, as revealed by lack of induction of adipogenesis markers (Fig. 7E). Together with the data from colon cell lines and PamChip arrays, these data support the classification of SCFA as selective PPARγ modulators.

Open in a separate window

Fig 7

Butyrate and propionate inhibit 3T3-L1 adipogenesis. (A) Oil red O staining of 3T3-L1 adipocytes at day 10 treated with SCFA (8 mM) from day 0. The mix contained 2.67 mM of each of the SCFA. (B) Expression of differentiation markers and PPARγ targets was determined by qPCR at day 4. (C) Concentration-dependent effect of butyrate on 3T3-L1 differentiation when added at day 0, as determined by expression of differentiation markers at day 4. (D) Effect of SCFA (8 mM) and rosiglitazone (1 μM) on expression of PPARγ targets in fully differentiated 3T3-L1 adipocytes. Cells were treated for 24 h. (E) Effect of SCFA (8 mM) and rosiglitazone (1 μM) added on day 1 on human SGBS adipocyte differentiation and expression of adipogenesis marker genes at day 15. Error bars represent SD. An asterisk indicates a result significantly different from that of the control according to Student's t test (P < 0.05).

This study was supported by the Netherlands Nutrigenomics Centre, the Netherlands Organization for Scientific Research (NWO) (VICI grant 700.56.442; to A.B., visitor travel grant 040.11.299; to J.G., and TOP grant 40-00812-98-08030 to S.K.), and the Netherlands Consortium for Systems Biology.

S.A., H.R., S.M., E.K., M.M., G.J.H., and S.K. designed research; S.A., K.L., T.A., J.S.G., E.S., A.K., A.K.H., K.M., D.H., and S.A.-L. performed research; R.H. and B.V.D.B. contributed analytic tools; S.A., K.L., T.A., J.S.G., E.S., A.K., A.K.H., H.R., A.M.J.J.B., E.K., G.J.H., and S.K. analyzed data; S.K. wrote the initial draft of the paper. All authors read and edited the paper.