Molecular recognition of nitrated fatty acids by PPARγ

Abstract

Peroxisome proliferator activated receptor-γ (PPARγ) regulates metabolic homeostasis and adipocyte differentiation, and it is activated by oxidized and nitrated fatty acids. Here we report the crystal structure of the PPARγ ligand binding domain bound to nitrated linoleic acid, a potent endogenous ligand of PPARγ. Structural and functional studies of receptor-ligand interactions reveal the molecular basis of PPARγ discrimination of various naturally occurring fatty acid derivatives.

PPARγ is a nuclear receptor that regulates adipocyte differentiation and glucose homeostasis¹. The synthetic PPARγ ligands rosiglitazone (Avandia) and pioglitazone (Actos) are thiazolidinedione derivatives (TZDs) used for the treatment of type 2 diabetes. Despite the importance of PPARγ in human physiology and drug discovery, the identity of physiological ligands remains elusive. The first implicated natural ligand of PPARγ was 15-deoxy-Δ^12,14-prostaglandin J₂ (PGJ₂), an oxidized fatty acid^2,3. Later, 9-hydroxyoctadecadienoic acid (9-HODE) and 13-hydroxyoctadecadienoic acid (13-HODE), two oxidized fatty acids present in oxidized low-density lipoprotein, were shown to activate PPARγ in macrophages⁴. Despite the low receptor affinity of these oxidized fatty acids, they are more potent than native fatty acid counterparts. Recent studies provide functional evidence that nitrated linoleic acid (LNO₂) and nitrated oleic acid (OA-NO₂) potently activate PPARγ at nanomolar concentrations^5–7, with net concentrations of nitrated fatty acid species exceeding 1 μM in human plasma lipids, supporting the idea that nitrated fatty acids are high-affinity, endogenous PPARγ ligands. This is also relevant to clinical therapeutics, as PPARγ, lipid and nitric oxide signaling pathways have been implicated in diabetes, obesity and cardiovascular diseases^8,9. Moreover, nitric oxide–mediated oxidative and nitrative reactions induce fatty acid nitration, thus representing a convergence of lipid and nitric oxide–mediated signaling. The discovery of nitrated fatty acids as potent PPARγ ligands suggests further functional linkages between these seemingly disparate signaling pathways. The inactivity of linoleic acid as a PPARγ ligand⁴ raises questions as to why the oxidation and nitration of linoleic acid can yield much more potent PPAR activators.

Although TZDs improve insulin sensitivity and lower plasma glucose levels, they have adverse side effects, including weight gain, fluid retention and hepatotoxicity¹⁰. In addition, the recently identified cardiovascular risks of rosiglitazone¹¹ may jeopardize further development and clinical application of TZD-based PPARγ ligands. A drug-design strategy based on endogenous PPARγ ligands may yield more efficacious PPARγ-targeted drugs, but little is known about how PPARγ interacts with natural ligands such as LNO₂.

To unravel the biochemical mechanism of human PPARγ activation by nitrated fatty acids, we determined the ability of LNO₂ to promote recruitment of coactivator LXXLL motifs by PPARγ using the AlphaScreen biochemical assay¹² (Supplementary Methods online). Both LNO₂ and rosiglitazone strongly enhanced the interaction of PPARγ with various coactivator LXXLL motifs from the family of steroid receptor coactivators (SRC2−3 and SRC1−2), CREB binding protein (CBP), TRAP220 and PGC-1α, indicating that LNO₂ functions as a PPARγ agonist (Supplementary Fig. 1a online). Similarly to rosiglitazone, LNO₂ also potently promoted the interaction of PPARγ with coactivator motifs in a concentration-dependent manner (Supplementary Fig. 1b,c). PPARγ showed only weak interaction with coactivators upon addition of 20 μM of either 9-HODE or 13-HODE (Supplementary Fig. 1a), in agreement with their in vivo effects⁴. These results reaffirm that nitrated fatty acids are potent PPARγ ligands.

To determine the molecular basis of the high-affinity binding of LNO₂ by PPARγ, we solved the crystal structure of PPARγ complexed with LNO₂ and the SRC1−2 LXXLL motif at 2.4 Å-resolution (Supplementary Table 1 online). The PPARγ ligand binding domain (LBD) forms a dimer, with helix 10 from each monomer forming a dimer interface that resembles that of the PPARγ–RXRα heterodimer complex¹³ (Fig. 1a). The binding mode of LNO₂ is similar to that of rosiglitazone, with one LNO₂ molecule occupying roughly 40% of the ligand binding pocket (Fig. 1b). The existence of two distinct C10 and C12 LNO₂ regioisomers in the PPARγ pocket was apparent from the highly revealing electron density map (Fig. 2). The acidic head group of both LNO₂ isomers forms several hydrogen bonds with the surrounding PPARγ residues, including Gln286 from helix 3 (Fig. 2a). These interactions are observed for both LNO₂ and rosiglitazone, and they support a crucial conserved mechanism for ligand-mediated activation of PPARγ. Indeed, the Q286A mutation that disrupts these interactions decreased the activation of PPARγ by both LNO₂ and rosiglitazone in cell-based assays using a PPARγ-response reporter (Fig. 3).

Figure 1.

The structure of the PPARγ LBD and LNO₂ complex. (a) Overall structure of the PPARγ–LNO₂–SRC1−2 complex in ribbon representation. (b) A closer view of the binding of LNO₂ to PPARγ. PPARγ is in green, the SRC1−2 peptide is in yellow and the bound LNO₂ is shown in stick representation with carbon, nitrogen and oxygen atoms depicted in pink, blue and red, respectively.

Figure 2.

The structural determinants of the PPARγ LBD and LNO₂ complex. (a) 2F_o – F_c electron density map (1.0σ) showing two bound LNO₂ isomers and the surrounding PPARγ residues. LNO₂ is shown in stick representation with carbon, nitrogen and oxygen atoms depicted in pink, blue and red, respectively. The key residues that determine PPARγ selectivity are noted, and hydrogen bonds are indicated by arrows. (b) Charged interactions of PPARγ residues with specific nitro groups in the LNO₂ ligand as determinants of LNO₂ selectivity. Hydrogen bonds are indicated by arrows. (c,d) The conformational changes of PPARγ induced by LNO₂. Overlays of the PPARγ–LNO₂ structure with the PPARγ–rosiglitazone structure, where LNO₂-bound PPARγ is in green and rosiglitazone-bound PPARγ is in gold. The conformational shifts of Glu343 toward the nitro group and the shift of Phe287 toward the LNO₂ backbone are indicated. The hydrophobic interaction between Phe287 and the LNO₂ backbone is shown with a dashed line.

Figure 3.

Functional correlation of the LNO₂–PPARγ interactions. (a) Schematic representation of PPARγ–LNO₂ interactions. The two isomers of LNO₂ found in crystals with the PPARγ LBD are shown in red and blue, respectively, whereas the black indicates the identical conformation and structure shared by those LNO₂ isomers. Hydrophobic interactions are indicated by dashed lines, and hydrogen bonds are indicated by arrows from proton donors to acceptors. Key hydrocarbon positions are indicated. (b) Effects of mutations of key PPARγ residues on LNO₂-mediated PPARγ transcriptional activity in cell-based assays. The cells were co-transfected with a 3×-PPRE luciferase reporter together with plasmids encoding either full-length wild-type PPARγ or the mutants, as indicated in the figure. The cells were treated with 1 μM rosiglitazone or LNO₂, respectively. The red dashed line indicates the activation level of wild-type PPARγ by LNO₂. Error bars indicate s.d.

The selectivity of PPARγ for binding LNO₂ is explained by the specific interactions of the two nitro groups with the PPARγ pocket residues (Fig. 3a). The C10 nitro group forms a hydrogen bond with Arg288, and the C12 nitro group interacts with Glu343 (Fig. 2b). Mutation of either residue decreased PPARγ activation by LNO₂ but had no effect on PPARγ activation by rosiglitazone (Fig. 3b), suggesting that nitro group interactions with Arg288 and Glu343 are required for stable binding of LNO₂. Notably, modeling studies indicate that the hydroxyl group in 9-HODE and 13-HODE occupies the same position as the nitro groups of the two LNO₂ isomers (Supplementary Fig. 2a,b online), suggesting that these hydrogen-bond interactions should be preserved in the complex between PPARγ and these oxidized fatty acid regioisomers. Notably, the R288H mutation detected in colon cancers induced decreased binding of PGJ₂ (ref. ¹⁴). Furthermore, the Arg288 residue has been implicated in the binding of the endogenous PPARγ ligand alkylglycerophosphate (AGP)¹⁵. Together, these data help to explain why PPARγ binds nitrated and oxidized linoleic acid but not unmodified linoleic acid^4,5.

To further validate the specific impact of charged residues on PPARγ–LNO₂ binding, we mutated several hydrophobic residues to arginine in the ligand binding pocket surrounding the nitro groups. These mutations should favor hydrogen bonding between the PPARγ pocket residues and the bound LNO₂. Accordingly, all these mutations substantially increased PPARγ activation by LNO₂ but not by rosiglitazone (Fig. 3b).

The binding of LNO₂ is further facilitated by the two cis double bonds of LNO_2, which introduce a 90° bend around the C9 to C13 position that allows the rest of the hydrocarbon chain to fit into a hydrophobic pocket of PPARγ (Fig. 2a). Conformational changes in two pocket residues (Glu343 and Phe287) are evidenced when the LNO₂–PPARγ complex is overlaid on the rosiglitazone–PPARγ structure (Fig. 2c,d). In response to LNO₂ binding, the charged side chain of Glu343 adopts a second conformation, allowing the receptor to form a hydrogen bond with the nitro group. The hydrophobic side chain of Phe287 also shifts from its rosiglitazone-bound conformation toward the hydrophobic tail (C18) of LNO₂, thus stabilizing LNO₂ binding by making additional hydrophobic interactions with the LNO₂ backbone. To validate the differential roles of Phe287 and Glu343 in binding to rosiglitazone versus LNO₂, we determined the in vitro ligand binding properties of PPARγ mutants using Alpha-Screen assays (Supplementary Fig. 3 online). The binding potency of LNO₂ was decreased by both mutations, whereas the binding potency of rosiglitazone was not affected. Mutations in Phe287 or Glu343 also reduced the activation of PPARγ by LNO₂ but not by rosiglitazone in cell-based assays, further affirming that these interactions are specific for LNO₂ (Fig. 3b).

The three PPAR subtypes (α, β/δ and γ) are characterized as fatty acid receptors with distinct selectivity for ligands^2,3,16. PPARγ has a preference for unsaturated or hydrophilic fatty acids, whereas PPARα can also bind saturated fatty acids¹⁶. Together with the structure of PPARβ/δ-bound eicosapentaenoic acid, the structure of LNO₂ bound to PPARγ presented herein reveals a molecular basis for PPAR selectivity toward various fatty acids. Despite the high degree of homology among the three PPARs, the charged residues Arg288 and Glu343 of PPARγ, which make specific contacts with the nitro groups, are not conserved (Supplementary Fig. 2c), indicating that the PPARγ ligand binding pocket has unique hydrophilic properties that dictate the discrimination of native fatty acids and those modified by oxidation and nitration reactions.

The specific interactions between crucial LBD residues of PPARγ and LNO₂ thus provide an important perspective regarding the recognition of nitrated and oxidized fatty acids by PPARγ. On the basis of the above structural observations, PPARγ contains two electrostatic binding epitopes: one epitope, in part comprising the C-terminal activation helix (AF2, Fig. 1b), is common for binding of the acidic group of LNO₂ and TZDs; the other epitope, comprising Arg288 and Glu343, is specific for the nitrated or oxidized groups of modified fatty acids. The differential binding modes of endogenous ligand versus the therapeutic drug rosiglitazone may allow the differential modulation of PPARγ ligand binding selectivity and affinity, thereby affecting the physiological outcome of different PPARγ ligands. As such, the structural mechanism may provide a more rational template for designing PPARγ ligands that would better mimic nitro group–receptor interactions by the endogenous PPARγ ligand LNO₂. On the basis of the structural insight into LNO₂–PPARγ interactions gained herein, we expect that new compounds can be developed that have a more favorable pharmacological impact than current TZD-based synthetic PPARγ ligands.