The PPARγ ligand binding domain (aa 204-477) was crystallized in the presence of a variety of fatty acid ligands (). In all cases the protein crystallized under similar conditions and in the same space group/cell dimensions as the previously reported unliganded homodimeric structure (ref 19
, 1prg). In this structure one PPARγ monomer closely resembles the liganded form of the receptor as seen in other structures (e.g. ref 19
, 2prg). In the other monomer, the C-terminal helix 12 is displaced and occupies an alternative position. Simulated annealing composite omit maps (CNS 21
) were calculated following molecular replacement and preliminary refinement. It was immediately clear that all the structures contained fatty acids in the ligand binding pockets of both monomers. In the monomer in which helix 12 is displaced from the active position, modeling of the ligand conformation was often less confident, or impossible, despite clear evidence for an occupied ligand binding pocket. This suggests that the positioning of the lipid ligand and the stabilization of helix 12 are coupled processes. Here we focus our description on the binding modes of the fatty acids in the monomer with the active conformation of helix 12 since these are generally better ordered and are likely to represent the bona fide mode of ligand binding to the activated receptor.
PPARγ has a versatile fatty acid binding pocket
The central region of the PPARγ ligand binding pocket is surrounded by largely non-polar amino acids. However, at both ends of the cavity there are clusters of polar residues. Many of the synthetic ligands position a carboxylate group, or similar group, adjacent to helix 12 making multiple polar contacts. shows an overlay of the eight fatty acids that we have crystallized in the ligand binding cavity of PPARγ. The combined ligands sample c. 1,004 Å3 of the ligand binding cavity. In the majority of structures, a single fatty acid chain is bound so as to position its carboxylate group adjacent to helix 12, with the aliphatic chain wrapping around helix 3 (). For the oxidized fatty acids, additional polar contacts are made with the hydroxyl or keto groups. However the C18 fatty acids 9-HODE and 13-HODE are exceptions to this general mode of binding.
PPARγ in complex with 9- and 13-HODE
In the structure of PPARγ with 9-HODE (at 2.05Å resolution) two molecules of 9-HODE are clearly observed simultaneously bound in the ligand binding pocket (). The shape of the two lipids is similar, with each fatty acid adopting a C-shaped conformation. The first molecule binds with its head group adjacent to helix 12, linking 4 helical regions of the protein with hydrogen bonds to Ser289 (H3), His323 (H6), His449 (H11) and Tyr473 (H12) (). This is the canonical interaction that is observed in several other PPARγ structures with carboxylate containing ligands. The tail of the 9-HODE fatty acids occupies a region of the cavity that is occupied by particularly potent agonists such as farglitazar. Intriguingly Phe363, which exhibits high temperature factors, and is presumably mobile, in the apo and several liganded structures, is repositioned in this structure so as to interact with both ends of the fatty acid (i.e. carbons 2, 3, 4 and 16, 17, 18). The 9-hydroxyl group of the first 9-HODE ligand makes a hydrogen bond to the carboxylate of the second 9-HODE ligand.
Recognition of 9- and 13-HODE by PPARγ
The second 9-HODE ligand is positioned between helix 3 and the beta-sheet and is distant from helix 12 (). In addition to hydrogen bonding with the hydroxyl of the first 9-HODE, the carboxylate of the second molecule of 9-HODE forms a salt bridge with Arg288 (). This arginine also makes a salt bridge with Glu295. Interestingly this side chain is repositioned from the apo conformation in which it makes a hydrogen bond to the hydroxyl group of Ser289. The hydroxyl of the second 9-HODE molecule is positioned such that it is approximately equidistant from the carboxylate groups of the glutamic acid residues: Glu291 and Glu295.
In the same way that two 9-HODE ligands bind to PPARγ, there is evidence that two 13-HODE molecules can also bind to the receptor. However, unlike 9-HODE, only one molecule of 13-HODE is well-ordered in the structure at 2.35Å resolution (). In the simulated annealing composite omit map there is density of sufficient volume for a molecule in the canonical position close to helix 12 (). However, it was not possible to convincingly model any dominant conformation for this molecule. We therefore conclude that it is present in the cavity, but its position is not tightly constrained and it may not stabilize helix 12 as efficiently as the more ordered 9-HODE.
There is clear evidence, however, for a well-ordered 13-HODE molecule bound toward the edge of the cavity between helix 3 and the beta-sheet (). The fatty acid adopts a “question mark” shaped conformation with the hydroxyl making a hydrogen bond to its own carboxylic acidic group. The hydroxyl also makes a rather long hydrogen bond with the backbone amide of Ser342. The carboxylic acidic group of 13-HODE makes a salt bridge with Arg288. To make this salt bridge with the ligand, the side chain of Arg288 is positioned quite differently from the complex with 9-HODE such that the salt bridge with Glu295 is broken.
Complexes with DHA, 4-HDHA and 5-HEPA
The simulated annealing composite omit maps clearly show that DHA, 4-HDHA and 5-HEPA all bind in a similar fashion to PPARγ. Their carboxylate groups interact in the canonical fashion in the vicinity of helix 12. Of the three, the 5-HEPA appears to be the best defined after refinement (). This is likely because the hydroxyl group makes a good hydrogen bond with the hydroxyl of Tyr327. The hydroxyl of 4-HDHA is positioned directly over the aromatic ring of Phe282 and is probably making a favorable interaction with the pi-electron ring. It could also be making a rather long hydrogen bond to Gln286 (). The aliphatic tail of DHA may have multiple conformations since breaks are observed in the electron density of the refined maps.
Binding of various oxidized fatty acid ligands to PPARγ
Oxo fatty acids couple covalently to PPARγ
The structures of PPARγ bound to the oxo fatty acids 4-oxoDHA and 6-oxoOTE revealed an unambiguous positioning of the ligands. In both cases there was very clear electron density indicating that the ligand is covalently bound at position C8 to the SH group of Cys285 (). The Cys285 side chain adopts alternative rotamers to accommodate slightly different positioning of the two ligands. The carboxylates of the two ligands are in very similar positions, surrounded by the polar sidechains of Tyr473, His449, His323, Ser289. The serine is positioned almost equidistance from the carboxylate and keto groups of 6-oxoOTE. Other than this the 6-oxoOTE keto group makes no polar contacts. As would be expected, the ketone of 4-oxoDHA is positioned very differently from that of 6-oxoOTE such that it is positioned between, and hydrogen bonds with, the sidechains of Tyr327 and Lys367.
Efficient covalent binding of oxo Fatty acids
To exclude the possibility that the covalent coupling of the oxo ligands seen in the crystal structures was an artifact of crystallization, or exposure to high intensity X-rays, we performed a mass spectroscopic (MALDI-ToF) analysis of the PPARγ LBD after a brief incubation with various ligands. shows that under these conditions (20 minutes at 20°C with a four-fold excess of ligand) about 60% of the receptor binds PGJ2 covalently. For those ligands that contain a conjugated ketone, essentially 100% of the receptor becomes covalently coupled with ligand. Conversely the hydroxy fatty acids show no evidence of covalently coupling to the receptor. These data support the biological relevance of the observations of covalent coupling in the crystal structures for 4-oxoDHA and 6-oxoOTE.
The covalent coupling is the result of a Michael addition (conjugate addition) for which the organic reaction is well established and illustrated in . The Michael acceptor contains an electron withdrawing group conjugated to an activated carbon that is then subject to attack by the nucleophile - in this case the SH group of Cys285. Intriguingly, in the case of 4-oxoDHA, there are two conserved sidechains, tyrosine Tyr327 and lysine Lys367, that are positioned such that they can form hydrogen bonds with the keto group of the ligand. It is likely that the tyrosine sidechain catalyses the conjugate addition through enhancing the electron withdrawing effect of the keto group22
. The position of these sidechains may play a role in determining the specificity of ligands that efficiently couple to the receptor.
Covalent ligands strongly stabilize the PPARγ LBD
It is well established that ligand binding by nuclear receptors results in the ligand binding domain becoming stabilized to thermal denaturation and furthermore that the degree of stabilization correlates, in many cases with the potency of the ligand. To explore the effect of covalent vs non-covalent ligand binding to PPARγ we used CD at 222 nm to monitor the thermal denaturation of PPARγ (). In the absence of ligand, PPARγ loses it folded structure in a highly cooperative fashion, with a melting temperature of 46°C. When bound to the synthetic agonist rosiglitazone the melting temperature increases to 51°C. In complex with the more potent agonist farglitazar, PPARγ remains folded at much higher temperatures, melting at 60°C.
Thermal denaturation studies of PPARγ LBD
PPARγ in complex with 4-HDHA is modestly stabilized and melts at 48°C. 4-oxoDHA is a more effective stabilizer of PPARγ such that the receptor has a melting temperature of 52°C (slightly higher than that for rosiglitazone).
The bottom panel in summarizes the melting temperatures of all the ligands tested. It is striking that all the complexes with oxo fatty acids denature at a higher temperature than those with hydroxy fatty acids (and also higher than the rosiglitazone bound receptor). Thus these ligands would appear to be excellent stabilizers of the PPARγ structure.
Covalent ligands are potent inducers of PPARγ targets
We used two assays to determine whether ligands that form covalent bonds with PPARγ are more effective activators of transcription than their non-covalent counterparts. In the first assay (left-hand panels in ) we measured the activity, in Cos1 cells, of a reporter gene activated by a Gal4 DBD:PPARγ LDB chimera. In the second assay we measured mRNA levels of the PPARγ target gene FABP4/aP2 in dendritic cells, derived from normal peripheral human monocytes23
. In both assays, it is clear that the oxo fatty acids that covalently couple to the receptor are up to an order of magnitude more active than the hydroxy fatty acids. This activity represents activation of PPARγ, rather than the RXR partner in the heterodimer, since the various oxidized fatty acids give negligible activity, even at 25μM, with a Gal4DBD:RXRLBD construct (Supplementary Fig. 1
). Furthermore RXR selective ligands do not induce FABP4/aP2 expression in dendritic cells (Supplementary Fig. 1
). Experiments to determine whether oxo fatty acids might result in a longer-term response than non-covalent ligands were largely inconclusive. However, when ligand was removed after 12 hours, there was some suggestion that covalent ligands might have prolonged activity (Supplementary Fig. 2
). Further investigation is required to clarify this issue.
Activity of oxidized fatty acids in cell based assays
Cys285 is essential for the activity of oxo fatty acids
To confirm that the greater activity of the oxo fatty acids is dependent upon the formation of a covalent bond with Cys285 we tested that ability of 4-oxoDHA and 4-HDHA to activate, in Cos7 cells, PPARγ in which Cys285 had been substituted by a serine or alanine side chain (). Interestingly both the mutants showed reduced activity with both ligands (and also pioglitazone - data not shown). However it was clear that the activity of 4-oxoDHA is less than that of 4-HDHA. This contrasts sharply with the wild-type receptor for which 4-oxoDHA is more potent than 4-HDHA.
Cys285 is essential for the activity of covalent ligands