Structures of PPARγ complexed with lobeglitazone and pioglitazone reveal key determinants for the recognition of antidiabetic drugs

The loop connecting H2b and H3, which is referred to as the Ω-loop, was poorly defined in the electron density maps in both lobeglitazone and pioglitazone-bound PPARγ LBDs. The Ω-loop is thought to be a very flexible region of LBD, serving as a gate to the ligand binding pocket²¹. The Ω-loop is more ordered in chain B, which results from stabilization of the loop by lattice contacts in the crystal.

The electron densities of lobeglitazone and pioglitazone were clearly visible in the ligand-binding pockets of chain A (Fig. 2A,C). The strong electron densities and the low B factors of lobeglitazone and pioglitazone indicate a high occupancy in the ligand-binding pocket (Table 1). The occupancy refinement using software Phenix showed that the lobeglitazone molecules in A and B chains, and the pioglitazone molecule in A chain, have 96%, 74%, and 82% occupancies, respectively. Lobeglitazone and pioglitazone are completely buried inside the hydrophobic pocket (Fig. 3B,C). Lobeglitazone occupies two sub-pockets (the AF-2 and Ω-pockets) of the Y-shaped ligand-binding pocket. Lobeglitazone, with its molecular volume of 577 ų, occupies 44% of the ligand-binding cavity, which is 10% more than the space occupied by pioglitazone (34%). The contact surface area of lobeglitazone to the ligand-binding pocket is 418 Ų, which is 1.33 times larger than the contact area of pioglitazone (314 Ų). The lobeglitazone bound to the B chain displays slightly weak electron densities and higher B-factors than the ligand in chain A, owing to the lack of hydrogen bonding to H12. Lobeglitazone and pioglitazone are bound to position their thiazolidinedione groups adjacent to H12, and adopt a U-shaped conformation with the hydrophobic chains wrapping around H3. The thiazolidinedione group makes multiple hydrogen bonds with the polar residues in the pocket. The nitrogen atom of the TZD head group makes a hydrogen bond with the hydroxyl group of Tyr473 in H12, stabilizing the active conformation of H12 (Fig. 3D,E). Two carbonyl groups of the TZD head group make hydrogen bonds with the side chains of His323, Ser289, and His449. The protein-ligand interactions observed in chains A and B were identical, except in the region around H12.

The apo form of the PPARγ LBD has a large internal cavity which can accommodate many water molecules. The tight TZD binding excludes water molecules in the AF-2 pocket and the central region of the hydrophobic cavity. Water molecules are present in the exposed region of the cavity and in the H1-pocket which is unoccupied by the ligand (Fig. 3B). The two nitrogen atoms of pyrimidine ring of lobeglitazone make hydrogen bonds with the two water molecules nearby (Fig. 3D). However, there is no direct water-mediated hydrogen bonds between the ligand and the protein. In the pioglitazone-bound form, there is no water molecules hydrogen bonding to the ligand. The electron densities of water molecules were not visible within the hydrogen-bonding distance to pioglitazone, suggesting that water molecules do not contribute to the specific interaction of TZD binding to the PPARγ LBD.

The binding mode of lobeglitazone is almost identical to rosiglitazone with good superposition of ligand structures in the binding pocket⁵ (Fig. 3F). However, lobeglitazone makes more extensive van der Waals contacts with the ligand-binding pocket than rosiglitazone and pioglitazone. The unique p-methoxyphenoxy group of lobeglitazone is positioned in the Ω pocket, making tight hydrophobic interactions with the residues of the pocket walls (Fig. 3D,E). The contacting surface area of the p-methoxyphenoxy group is 90 Ų, contributing 22% of the contact area of lobeglitazone to the binding pocket. However, there is no direct interaction with the disordered Ω-loop. The residues in the Ω-pocket (Ile249, Leu255, Ile281, Ile341, and Met348) interacting with the p-methoxylphenoxy group are relatively rigid, and undergo no conformational changes upon ligand binding. The binding of lobeglitazone to chain B, which displays an inactive conformation of H12, suggests that the extensive hydrophobic interactions significantly contribute to the high affinity of lobeglitazone to PPARγ.

The major difference in pioglitazone compared to rosiglitazone and lobeglitazone is the shorter length of the linker between pyridine and phenyl groups owing to the lack of methylamino group (Fig. 1A). However, the ethyl pyridine group of pioglitazone occupies a similar position to the pyrimidine group of lobeglitazone in the binding pocket (Fig. 3F). The pioglitazone-bound PPARγ LBD displays almost an identical LBD conformation to the lobeglitazone-bound form. This observation suggests that TZD drugs have a conserved binding mode to PPARγ and share the same therapeutic effects via PPARγ activation.

We recognized that the high potency of lobeglitazone compared to other TZD drugs could be clearly demonstrated by direct determination of ligand binding affinities to the PPARγ LBD. The binding affinities of lobeglitazone and pioglitazone to the isolated PPARγ LBD were not known. Therefore, we attempted affinity measurements of lobeglitazone and pioglitazone by isothermal titration calorimetry (ITC). The titration of ligand to receptor using ITC requires solubilization of the ligand into a free form. However, lobeglitazone and pioglitazone were almost completely insoluble in the aqueous buffer, which interfered with fast titration of the ligands to the PPARγ LBD. The ligand binding to PPARγ seems a slow process requiring extraction of ligands from the lipid aggregate and a conformational change of the ligand-binding domain. Therefore, as an alternative approach, we estimated the relative affinities of TZDs using thermal shift assays of the PPARγ LBD. It is known that ligand binding by nuclear receptors results in the stabilization of the LBD against thermal denaturation, and that the degree of stabilization correlates with the affinity of the ligand²⁴.

To confirm the results obtained from the DFS experiments, we performed an additional thermal shift assay, which does not require a fluorescent dye. We measured the concentration of soluble protein after heat denaturation of the PPARγ LBD loaded with three different TZDs (Fig. 4C). The apo PPARγ LBD lost solubility by protein precipitation in a highly cooperative manner with a midpoint temperature (T_m) of 38 °C. When bound to the agonist pioglitazone, the T_m value increased to 44 °C. Rosiglitazone displayed a similar stabilization of the PPARγ LBD to pioglitazone with a T_m of 48 °C. Notably, the PPARγ LBD complexed with lobeglitazone was significantly stabilized with a T_m of 54 °C, which is 10 °C higher than the T_m of the pioglitazone-bound PPARγ LBD. There is a discrepancy of absolute T_m values between the two thermal shift techniques, which might be caused by the differences of detection methodology and concentration of protein and ligand. However, the patterns of thermal stabilization by the TZDs are consistent with the order of lobeglitazone, rosiglitazone, and pioglitazone, from the highest to lowest T_m values. Thus, lobeglitazone seems to be an excellent stabilizer of the PPARγ LBD, indicating a higher affinity to PPARγ than the other TZDs tested.

The full-length PPARγ clone in a pCMV-SPORT6 plasmid (clone ID: hMU000317) was obtained from Korea Human Gene Bank, Medical Genomics Research Center, KRIBB, Korea. Lobeglitazone sulfate (Duvie^R tablets) and pioglitazone hydrochloride (Glyact^R tablets) were purchased from Chong Kun Dang pharmaceutical co. and Myungmoon pharmaceutical co., respectively. Rosiglitazone and pioglitazone with purity higher than 98% were obtained from Tokyo Chemical Industry Co.

DNA encoding the ligand-binding domain (residue 207–477) of human PPARγ (GenBank accession number: NM_138711↗) was amplified from the full-length clone by polymerase chain reaction (PCR). The primers for the PPARγ LBD were 5′-GTATTA GGATCC GAGTCCGCTGACCTCCG-3′ (forward) and 5′-GATACA CTCGAG CTAGTACAAGTCCTTGTAGATC-3′ (reverse). The PCR products were sub-cloned into the BamHI and XhoI sites of a modified pET28b vector. The PPARγ LBD was tagged with an N-terminal hexahistidine followed by a thrombin protease cleavage site (LVPR/GS). The plasmid containing the PPARγ LBD was transformed to Escherichia coli strain BL21(DE3).

E. coli cells transformed with the plasmid encoding the PPARγ LBD were grown to an OD₆₀₀ of 0.6 at 310 K in Luria-Bertani medium. Cells were induced by adding isopropyl β-D-1-thiogalactopyranoside to a final concentration of 0.3 mM, and were incubated for 12 hours at 289 K prior to harvesting. The cells expressing the PPARγ LBD were re-suspended in 3X phosphate buffered saline (PBS) containing 30 mM imidazole and lysed by sonication. After centrifugation at 13,000 rpm for 45 min, the supernatant containing the PPARγ LBD was applied to a Ni-NTA affinity column. The protein was eluted from the column using 0.1 M Tris–HCl pH 7.0, 0.3 M NaCl, 0.3 M imidazole. To obtain PPARγ complexed with lobeglitazone and pioglitazone, three tablets of Duvie^R and Glyact^R were dissolved into the elution buffers containing the PPARγ LBD to final ligand concentrations of 60 μM and 2.9 mM, respectively. The protein was concentrated to 10 mg/ml, and the his-tag was removed by cleavage with thrombin protease with a 2-hour incubation at room temperature. The ligand-bound PPARγ LBDs were subjected to size exclusion chromatography (SEC) using a Superdex 200 column (GE Healthcare) which was pre-equilibrated with 20 mM Tris–HCl pH 7.5, 150 mM NaCl. To sustain the ligand-bound forms of the PPARγ LBDs during SEC, additional Duvie^R and Glyact^R tablets were dissolved into the equilibration buffers to final concentrations of 5.2 μM lobeglitazone and 229 μM pioglitazone, respectively. The peak fractions containing the ligand-bound PPARγ LBDs were concentrated to 30 mg/ml for crystallization. The PPARγ LBD loaded with rosiglitazone was prepared by the same procedures. Rosiglitazone was added into the affinity and SEC eluted fractions with a final concentration of 60 μM.

Preliminary crystallization experiments for the PPARγ LBD-ligand complex were carried out at 295 K in 96-well crystallization plates using a multichannel pipette and customized crystallization screening solutions by dispensing 0.8 μl protein solution and 0.8 μl precipitant solution. Initial crystals of the PPARγ LBD appeared after 5 days using a solution consisting of 0.1 M HEPES–NaOH pH 7.0, 20% PEG 8000. The crystallization conditions were further optimized using the hanging-drop technique in 15-well screw-cap plates. A drop consisting of 2 μl protein solution was mixed with 2 μl precipitation solution and equilibrated against a 1 ml reservoir solution. The PPARγ LBD crystals with typical dimensions of 0.1 × 0.1 × 0.15 mm appeared in one week. Crystals of PPARγ-pioglitazone were grown in 0.1 M HEPES–NaOH pH 7.5, 17.5% PEG 8000, 10% EG. Crystals of PPARγ-lobeglitazone were grown in 0.1 M HEPES–NaOH pH 7.5, 17.5% PEG 8000, 10% DMSO. The crystals of the TZD-bound PPARγ LBD were cryoprotected in the reservoir solution supplemented with 10% glycerol and flash-cooled by immersion in liquid nitrogen. Crystals were preserved in a cryogenic N₂-gas stream (~100 K) during diffraction experiments.

Diffraction data for PPARγ-pioglitazone and PPARγ-lobeglitazone crystals were collected at a wavelength of 0.98 Å using an ADSC Q315r CCD detector on the 5C beamline at Pohang Light Source (PLS), Pohang Accelerator Laboratory. All data were processed and scaled using HKL-2000 (HKL Research Inc.) and handled with the CCP4 program suite³⁹. The structures of the pioglitazone- and lobeglitazone-bound PPARγ were determined by molecular replacement using the structure of the PPARγ LBD (PDB code: 4EMA) excluding the bound-rosiglitazone as a search model. The F_o-F_c maps showed clear electron densities of PPARγ and the bound ligands. The nine residues at 264–272 (269–274 in B chain) in the Ω-loop and a few residues in the C-termini were not visible due to disorder and were not modeled.

The final model of PPARγ-lobeglitazone including two molecules of PPARγ LBD, two molecules of lobeglitazone, and water molecules was refined to R_work/R_free values of 19.2%/22.3% at 1.7 Å resolution. Pioglitazone was visible only in the A chain of the PPARγ LBD homodimer. Electron densities of ligands in chain B were very weakly visible. Therefore, the ligand was not modeled in the B molecule of the PPARγ-pioglitazone complex. The structure of PPARγ LBD – pioglitazone was refined to R_work/R_free values of 20.2%/23.2% at 1.8 Å resolution.

The stabilization of the PPARγ LBD by ligand binding was monitored by increase of melting temperatures using differential scanning fluorimetry (DSF)^25,40. SYPRO orange is an environmentally sensitive dye. The unfolding process exposes the hydrophobic region of proteins and results in a large increase in fluorescence, which is used to monitor the protein-unfolding transition. DSF was performed using the CFX96 Real-Time PCR system (Bio-Rad) with a C1000 Thermal Cycler using FRET mode. The recombinant PPARγ LBD was pre-loaded with rosiglitazone, pioglitazone, or lobeglitazone during protein purification by applying the protein to size exclusion chromatography equilibrated with the buffer containing 60 μM of each TZDs. In a single well of a 96-well PCR plate, a 20 μL reaction solution contained 4 μM of PPARγ LBD, 6 μM of each TZD ligand, and 2 μL of 20 × SYPRO orange (diluted from 5000 × stock in DMSO). Owing to the high affinities of TZD ligands, the PPARγ LBD seemed to be almost fully occupied by each ligand, which allowed comparison of stabilization by different TZD drugs. Well plates were sealed with a transparent adhesive sheet. The real-time PCR machine was programmed to equilibrate samples at 25 °C for 5 minutes and then increase temperature to 80 °C at a rate of 1 °C/30 sec. The melting temperature of the protein was obtained as the lowest point of first derivative plot (dF/dT), as calculated by the software included with the RT-PCR system.

The apo or TZD-bound form of the PPARγ LBD was added to the SEC buffer containing 0 μM or 60 μM of each TZD ligand, respectively. In each PCR tube, 20 μl of reaction solution contained 0.1 mM of the purified PPARγ LBD and 60 μM of ligand. The samples were heat-treated using a PCR machine for 3 min at 15 temperature data points. The protein samples were then centrifuged at 13,000 rpm for 30 min to remove protein precipitate. The concentration of the soluble PPARγ LBD in the supernatant of each tube was measured by UV absorbance at the wavelength of 280 nm. The global curve fitting of the data points and the calculation of T_m values were done using software SigmaPlot.

The structures of the PPARγ LBD complexed with lobeglitazone, pioglitazone and rosiglitazone (PDB id: 3EMA) excluding water molecules were used for calculations of the binding energies of the TZDs, using the software AutoDock Vina²⁸. AutoDock Vina is implemented with an efficient scoring function for the estimation of protein-ligand affinity and a search algorithm for binding mode predictions. Since the high-resolution structures of the TZD-bound forms are available, the conformations of ligands and receptors were fixed and the program was run with “score only” option with an exhaustiveness value of 20. The dimensions of the search space centered in the ligand-binding pocket were 21 × 23 × 30 Å. The free binding energy values of the TZDs were converted to relative K_d values using the equation, ΔG = −RT ln K.

The coordinates and structure factors for the PPARγ LBD complexed with lobeglitazone and pioglitazone have been deposited in the RCSB Protein Data Bank with the accession codes, 5Y2T and 5Y2O, respectively.