Repositioning PPARγ for Enhanced Covalent Inhibitor Efficacy

Peroxisome proliferator-activated receptor gamma (PPARγ) regulates gene expression programs that influence cellular differentiation and insulin sensitization in response to ligand binding. The discovery in 1995 that PPARγ is a molecular target for thiazolidinedione (TZD)-containing synthetic ligands (Lehmann et al., 1995), which were first reported in 1983 (ciglitazone from Takeda) from phenotypic screens as insulin sensitizers (Fujita et al., 1983), led to a significant interest in pharmacologically modulating PPARγ for therapeutic intervention in patients with type 2 diabetes. Several TZDs, including pioglitazone (Actos by Takeda), rosiglitazone (Avandia by GSK), and troglitazone (Rezulin by Daiichi Sankyo & Parke-Davis), were approved by the U.S. Food and Drug Administration (FDA) for use in patients with type 2 diabetes. However, patient reports of undesired side effects and adverse events in the 2000s associated with clinical use of TZDs that included edema, congestive heart failure, risk of bone fracture, and others led to FDA black box warnings or withdrawal from the market (Soccio et al., 2014). TZDs function as pharmacological agonists that activate PPARγ-mediated transcription by stabilizing an active ligand-binding domain (LBD) activation function-2 (AF-2) coregulator surface conformation (Shang et al., 2019), which promotes coactivator recruitment to PPARγ-bound regions of chromatin resulting in increased expression of PPARγ target genes (Haakonsson et al., 2013). Recent mechanism of action studies have also revealed a potential path for separating the insulin sensitizing effects of PPARγ-binding ligands from undesired side effects. The field now appreciates that ligand binding can influence the structure and function of two distinct surfaces in the PPARγ LBD (Frkic et al., 2021). Ligand binding can stabilize the AF-2 coregulator interaction surface in a conformation that enhances recruitment of transcriptional coactivator complexes that regulates an adipogenic gene program or, in a non-mutually exclusive manner, inhibit obesity-linked phosphorylation of Ser273 by Cdk5 to influence expression of an insulin sensitizing gene program (Choi et al., 2010). Subsequent studies have focused on developing so-called next-generation selective PPARγ modulators, which contained compound scaffolds different from TZDs and comprise a wide range of pharmacological activities: PPARγ partial agonists (Choi et al., 2010), transcriptionally neutral antagonists (Choi et al., 2011), and transcriptionally repressive inverse agonists (Marciano et al., 2015; Stechschulte et al., 2016) can inhibit Cdk5-mediated phosphorylation of Ser273. To determine if off-target effects may be responsible for the beneficial or side effects of TZDs and other PPARγ-binding compounds, researchers have used two compounds originally reported in 2002—GW9662 by GlaxoSmithKline (Leesnitzer et al., 2002) and T0070907 by Tularik (Lee et al., 2002)—as covalent inhibitor antagonists of ligand binding to the PPARγ ligand-binding domain (LBD). GW9662 and T0070907, which bind via a halogen exchange reaction to a reactive cysteine residue (Cys285 or Cys313 PPARγ isoform 1 or 2, respectively) that points into the orthosteric ligand-binding pocket of PPARγ, were shown to inhibit binding of [3H]-rosiglitazone in a radiolabeled ligand-binding assay and inhibit agonist-induced transcription and adipocyte differentiation in cells cotreated with the covalent inhibitor and rosiglitazone. Crystal structures show overlapping orthosteric ligand-binding modes for covalent inhibitors (Brust et al., 2018; Chandra et al., 2008) and non-covalent PPARγ ligands including full agonists such as rosiglitazone (Nolte et al., 1998) or partial agonists including MRL-24 and nTZDpa (Bruning et al., 2007), which provided additional support for the field to use GW9662 and T0070907 as covalent inhibitors. However, in 2014, we showed that GW9662 and T0070907 do not block all ligands from binding to the PPARγ LBD—a phenomenon we originally called ‘alternate site’ ligand binding (Hughes et al., 2014). Other studies have confirmed that non-covalent synthetic ligands and cellular metabolites can cobind to the PPARγ LBD in the presence of GW9662 or T0070907 when used as covalent inhibitors (Arifi et al., 2023; Brust et al., 2017; Hughes et al., 2016; Jang et al., 2017; Laghezza et al., 2018; Leijten-van de Gevel et al., 2022; Shang et al., 2018). To gain structural insight into the non-covalent and covalent ligand cobinding mechanism, we recently reported seven X-ray crystal structures of PPARγ LBD cobound to a covalent inhibitor (GW9662 or T0070907) and different synthetic non-TZD PPARγ modulators (Shang and Kojetin, 2024). We surprisingly found that the non-covalent ligand-binding event we previously described at the alternate/allosteric site, which is proximal to the orthosteric ligand-binding pocket, can instead correspond to the ligand adopting the original orthosteric binding mode where the covalent inhibitor adopts a binding mode that permits ligand cobinding (Figure 1a) with some variation depending on the specific covalent and non-covalent ligand pair. Furthermore, biochemical and protein NMR studies suggested a potential mechanism explaining why T0070907, a corepressor-selective pharmacological inverse agonist, is a more effective covalent inhibitor of ligand binding than GW9662, a transcriptionally neutral pharmacological antagonist. GW9662 stabilizes an active-like LBD conformation, whereas T0070907-bound LBD exchanges between two long-lived conformations corresponding active- and repressive-like states (Brust et al., 2018). Crystal structures of PPARγ LBD reveal that in the transcriptionally repressive state, when cobound to T0070907 and NCoR1 corepressor peptide, a critical regulatory element called helix 12 adopts a solvent-occluded conformation within the orthosteric ligand-binding pocket (Figure 1b; Shang et al., 2020). When T0070907 is used as a covalent inhibitor, ligand cobinding resulted in the disappearance of the repressive-like state and stabilization of an active-like state that is similar to the GW9662 cobound ligand state (Shang and Kojetin, 2024). These data suggested a mechanism where helix 12 and a non-covalent ligand compete to occupy the orthosteric ligand-binding pocket (Figure 1)—and ligand cobinding in the presence of the pharmacological inverse agonist T0070907, when used as a covalent inhibitor, selects or induces an active LBD conformation where helix 12 adopts a solvent-exposed active conformation. This mechanism is consistent with the two-step mechanism described for agonist binding to PPARγ (Shang and Kojetin, 2021). Figure 1 In the discussion of our previous study (Shang and Kojetin, 2024), we suggested that covalent inhibitors with improved pharmacological corepressor-selective inverse agonist functions—that better stabilize a repressive-like LBD conformation where helix 12 adopts a solvent-occluded confirmation within the orthosteric pocket—may more effectively inhibit ligand cobinding to the orthosteric pocket. Here, we test this hypothesis using two covalent inverse agonists, SR33065 and SR36708, which we recently reported with improved efficacy over T0070907 (MacTavish et al., 2025). Biochemical and NMR-based structural biology ligand cobinding assays show that although SR33065 and SR36708 have similar inverse agonist efficacy, they display ligand-specific differences when used as a covalent inhibitor—and they still do not completely block ligand binding to PPARγ. Finally, we also show that binding of a non-covalent ligand is not blocked by another previously described covalent inhibitor, SR16832 (Brust et al., 2017), which appears to function via a different mechanism that does not involve corepressor-selective inverse agonism and helix 12 occupancy of the orthosteric pocket. In the nuclear receptor field, small molecule antagonists that bind at the same site as the orthosteric endogenous ligand are used, when available, as pharmacological chemical tools to probe the functions of blocking ligand-induced nuclear receptor activities. However, aside from their use as inhibitors of ligand binding, antagonists can also display pharmacological properties on their own. Pharmacological nuclear receptor ligands are classified based on the molecular basis by which they influence transcription. Agonists recruit coactivator proteins to activate transcription, inverse agonists recruit corepressor proteins to repress transcription, and the mechanisms by which antagonists function are not fully understood but are thought to involve competition with an endogenous ligand resulting in no change in coregulator interaction and transcription or a decrease in transcription without recruiting a corepressor protein (Strutzenberg et al., 2019). GW9662 and T0070907 were first reported as antagonist inhibitors of ligand binding to PPARγ (Lee et al., 2002; Leesnitzer et al., 2002). However, more than a decade later, we showed these antagonist inhibitors do not in fact block binding of all PPARγ ligands (Hughes et al., 2014) and display unique pharmacological activities (Brust et al., 2018) linked to their ability to stabilize the PPARγ LBD in a transcriptionally repressive conformation with helix 12 docked in the orthosteric pocket (Shang et al., 2020). The PPARγ LBD is a dynamic conformational ensemble that exchanges between transcriptionally active and repressive conformations in the absence of ligand (Shang et al., 2020). Ligand binding can shift the LBD conformational ensemble between graded agonist and graded inverse agonist states (MacTavish et al., 2025; Shang et al., 2019). We recently proposed an allosteric mechanism that describes how T0070907 is a more effective inhibitor of weakening orthosteric ligand binding than GW9662, not by direct clashing of these covalent inhibitors with an orthosteric ligand but via increased occupancy of helix 12 within the orthosteric pocket (Shang and Kojetin, 2024). Because T0070907 partially stabilizes the LBD in a repressive state that can exchange back to the active state, we hypothesized that covalent compounds such as SR33065 and SR36708 that more fully stabilize the repressive LBD state where helix 12 occupies the orthosteric pocket (MacTavish et al., 2025) may completely block ligand binding to the orthosteric pocket (Shang and Kojetin, 2024; Figure 8). Although our data here support this hypothesis, our data here show that SR33065 and SR36708 weaken but do not completely inhibit ligand binding. This observation indicates that the allosteric mechanism of stabilizing the LBD in a repressive conformation with helix 12 located in the orthosteric pocket is not enough to prevent a ligand from binding and, in doing so, likely pushes helix 12 out of the repressive conformation into an active conformation—similar to the two-step binding mechanism we reported for agonist binding (Shang and Kojetin, 2021). Figure 8 The allosteric ligand inhibition mechanism afforded by the pharmacological ligand series of GW9662, T0070907, SR33065, and SR36708 appears to differ from a different covalent PPARγ ligand, SR16832, which by NMR does not stabilize a repressive helix 12 conformation and would therefore likely inhibit binding due to direct clashing with other orthosteric ligands (Brust et al., 2017). However, our NMR data show that MRL-24 and nTZDpa can still bind to the PPARγ LBD that is pretreated with the SR16832 as a covalent inhibitor. Notably, our TR-FRET data show that nTZDpa cobinding is weakened when PPARγ LBD is pretreated with SR16832; however, MRL-24 shows no change in TR-FRET signal even though the NMR data shows that MRL-24 cobinds to SR16832-bound PPARγ LBD. This finding highlights an important warning for the field: functional assays that are used as a proxy for ligand binding, including TR-FRET coregulator interaction biochemical assays or luciferase reporter assays, may not be well suited to detect actual and direct ligand binding such as methods like protein NMR. This is of importance because in the PPARγ field, it is well known that a high-affinity ligand can bind and produce no change in luciferase reporter activity yet still affect gene expression (Choi et al., 2011). It is possible that by using peptides derived from different coactivators or corepressors, or different segments of the same coregulator—as certain segments of the same coregulator protein can bind to an NR LBD with high affinity and others bind with low affinity or not at all—subtle changes caused by ligand cobinding may be detected. Furthermore, biochemical assays using fluorescent or radiolabeled tracer ligands as a proxy for measuring ligand-binding affinity may also miss ligand cobinding events—once the tracer ligand has been displaced, or blocked by an antagonist, the assay is incapable of detecting any additional ligand-binding events such as ligand cobinding. It is not yet clear if other nuclear receptors share the same ligand-induced activity-dependent conformational ensemble to postulate if our findings here on PPARγ can translate to other nuclear receptors. However, our findings are likely applicable to other non-covalent orthosteric PPARγ ligands regardless of agonist, antagonist, or inverse agonist pharmacology—and generally suggest covalent ligand that stabilizes helix 12 within orthosteric pocket may function in some degree to inhibit binding of other orthosteric ligands. Our observations also raise the question as to how to design a completely effective inhibitor of ligand binding to PPARγ. It is possible that covalent analogs of SR33065 and SR36708 that not only stabilize a repressive conformation but also contain an additional covalent warhead to lock the repressive helix 12 in place may provide a route to design improved covalent inhibitors. It is also possible that SR16832 could be used as a parent compound to discover a covalent inhibitor that likely occupies more of the orthosteric pocket and the so-called alternate site region corresponding to the entrance cavity to the orthosteric pocket (Hughes et al., 2016; Hughes et al., 2014; Shang et al., 2018). Thus, although our data here provide further support to the allosteric ligand inhibition model, our findings also reveal a significant unmet need in the field and a word of caution. Three of the covalent ligands described here are commercially available and used by the field (GW9662, T0070907, and SR16832), but are not effective inhibitors of ligand binding to PPARγ—and they have pharmacological functions distinct from the inability to effectively block ligand binding.