Scaffold-based discovery of indeglitazar, a PPAR pan-active anti-diabetic agent

Dean R. Artisa,1, Jack J. Lina,1, Chao Zhanga, Weiru Wanga, Upasana Mehraa, Mylene Perreaultb, David Erbeb, Heike I. Krupkaa, Bruce P. Englanda, James Arnolda, Alexander N. Plotnikova, Adhirai Marimuthua, Hoa Nguyena, Sarah Willb, Maxime Signaevskyc, John Kralc, John Cantwella, Calvin Settachatgulla, Douglas S. Yana, Daniel Fonga, Angela Oha, Shenghua Shia, Patrick Womacka, Benjamin Powella, Gaston Habetsa, Brian L. Westa, Kam Y. J. Zhanga, Michael V. Milburna, George P. Vlasukb, K. Peter Hirtha, Keith Nolopa, Gideon Bollaga, Prabha N. Ibrahima, and James F. Tobinb,1

aPlexxikon Inc., 91 Bolivar Drive, Berkeley, CA 94710; bDepartment of Cardiovascular and Metabolic Diseases, Wyeth Research, 200 Cambridgepark Drive, Cambridge, MA 02140; and cDepartment of Surgery, State University of New York Downstate Medical Center, Box 40, 450 Clarkson Avenue, Brooklyn, NY 11203

Communicated by Axel T. Brunger, Stanford University, Stanford, CA, November 19, 2008 (received for review September 12, 2008) In a search for more effective anti-diabetic treatment, we used a activate all 3 PPARs has been a challenge because of the subtle process coupling low-affinity biochemical screening with high- differences between residues lining the ligand binding pockets of throughput co-crystallography in the design of a series of com- the receptors (4). In addition, recent focus in this area has pounds that selectively modulate the activities of all three perox- become centered on the concept of Selective PPAR Modulators isome proliferator-activated receptors (PPARs), PPAR␣, PPAR␥, and (SPPARMs)—particularly for PPAR␥, following from the ob- PPAR␦. Transcriptional transactivation assays were used to select servation that the efficacies found with full agonists such as compounds from this chemical series with a bias toward partial and were complicated by receptor- agonism toward PPAR␥, to circumvent the clinically observed side mediated side effects of weight gain, edema, and potential effects of full PPAR␥ agonists. Co-crystallographic characterization cardiovascular complications. A partial agonist might retain of the lead molecule, indeglitazar, in complex with each of the 3 efficacy but reduce the transcriptional effects thought to be PPARs revealed the structural basis for its PPAR pan-activity and its responsible for the attendant side effects (5). partial agonistic response toward PPAR␥. Compared with full We saw an opportunity to combine 2 challenging design elements PPAR␥-agonists, indeglitazar is less potent in promoting adipocyte into a single biological profile, resulting in the discovery of a unique differentiation and only partially effective in stimulating adiponec- pan-agonist of the PPARs. We have recently described an approach tin gene expression. Evaluation of the compound in vivo confirmed where low-molecular-weight (150–350Da) compounds, weakly ac- the reduced adiponectin response in animal models of obesity and tive in an initial biochemical screen against a panel of structurally while revealing strong beneficial effects on glucose, related targets, are subject to a second filtering process based on , , body weight, and other metabolic pa- high-throughput co-crystallography (6–8). Of the compounds in rameters. Indeglitazar has now progressed to Phase II clinical which binding orientations in the active sites of target molecules can evaluations for Type 2 diabetes mellitus (T2DM). be unambiguously determined by X-ray diffraction analysis, we selected those offering the most efficient access to chemistry as the adiponectin ͉ diabetes ͉ partial agonist ͉ PPAR pan-agonist ͉ starting points (or scaffolds) for discovery programs. Structure- Scaffold-based drug discovery guided computational techniques were used to aid in the prospec- tive evaluation of different chemical avenues of pursuit. Here we herapeutic approaches to Type 2 diabetes mellitus (T2DM), present the results of our application of this methodology in the Twhich currently affects Ϸ6% of adults in the United States discovery of agonists targeting multiple members of a nuclear (US Department of Health and Human Services, Centers for receptor subfamily. Disease Control and Prevention, Atlanta, GA; 2005), are gen- erally polypharmaceutical in nature, targeting effects on Results sensitivity and elements of the coincident dyslipidemia and Scaffold Screening Process. The target-naive screening library was cardiovascular diseases (1). However, polypharmacy in these designed by using compounds from a relatively narrow and low treatment regimens has been cited as a potential additional risk molecular weight range (150–350D), selected for diversity at factor (2), with many patients on 4 or more concomitant both the putative ‘‘scaffold’’ core and at the whole molecule level . A more effective strategy would be to use a single (6). Compounds capable of modulating receptor signaling were agent that possesses combined benefits from simultaneous in- hibition or stimulation of several related targets, without the risks associated with combination therapy. However, optimizing Author contributions: D.R.A., J.J.L., C.Z., M.P., D.E., J.K., B.P., G.H., B.L.W., K.Y.J.Z., M.V.M., activities against several targets is a complex design problem that K.P.H., K.N., G.B., P.N.I., and J.F.T. designed research; J.J.L., C.Z., W.W., U.M., M.P., D.E., H.I.K., B.P.E., J.A., A.N.P., A.M., H.N., S.W., M.S., J.C., C.S., D.S.Y., D.F., A.O., S.S., P.W., B.P., necessitates judicious choice of targets and requires new ways in G.H., B.L.W., and K.Y.J.Z. performed research; D.R.A., J.J.L., C.Z., W.W., U.M., M.P., D.E., J.A., which therapeutic agents are generated. A.M., S.W., M.S., J.K., S.S., B.P., G.H., B.L.W., K.Y.J.Z., M.V.M., G.P.V., K.P.H., K.N., G.B., P.N.I., Two classes of marketed therapeutics, the (as lipid- and J.F.T. analyzed data; and D.R.A., J.J.L., C.Z., K.Y.J.Z., G.B., and J.F.T. wrote the paper. lowering agents) and the glitazones (as insulin-sensitizing drugs) Conflict of interest statement: The Sponsor is a member of the Scientific Advisory Board of target related receptors known as PPAR␣ and PPAR␥, respec- Plexxikon, Inc. The authors are either employees of Plexxikon, Inc. or Wyeth Research with tively, whereas a third member of the subfamily, PPAR␦, has the exception of M.S. and J.K. been the target of intense preclinical interest as an avenue for Data deposition: The atomic coordinates and structure factors have been deposited in the Protein Data Bank, www.pdb.org (PDB ID codes 3ET0, 3ET1, 3ET2 and 3ET3 for 5-methoxy- treatment of dyslipidemia (3). A pan-agonist, capable of stim- indole-3-propionic acid (1) bound in PPAR␥, indeglitazar bound in PPAR␣, PPAR␦, and ulating the 3 peroxisome proliferator-activated receptors PPAR␥, respectively). (PPARs) as a group, would be expected to be particularly useful 1To whom correspondence may be addressed. E-mail: [email protected], in the treatment of T2DM from the standpoints of both efficacy [email protected], or [email protected]. and reduction in the additional risk factors associated with This article contains supporting information online at www.pnas.org/cgi/content/full/ polypharmacy. Despite the close structural relationship between 0811325106/DCSupplemental. these 3 receptors, the search for compounds which competently © 2008 by The National Academy of Sciences of the USA

262–267 ͉ PNAS ͉ January 6, 2009 ͉ vol. 106 ͉ no. 1 www.pnas.org͞cgi͞doi͞10.1073͞pnas.0811325106 Downloaded by guest on September 24, 2021 OH A Helix 3 B O

H323 H9 H1 O Y327 1 Y473 H10 H8 N H Pocket B H4 H2 OH H449 H3 H5 O H7 H12 O

H11 N H6 O Pocket A S O 2 H2'

Y314 C D H323 E OH H287 O F318 Y327 F291 Y473 Y437 Y464 H413 O H449 H440 N O S O Q286 Q250 3 Q277 O (Indeglitazar)

Fig. 1. Discovery and structural characterization of indeglitazar. (A) Crystal structure of scaffold 1 with PPAR␥. The blue mesh encompassing 1 delineates the productive hydrophobic interaction space common to PPAR␣, PPAR␥, and PPAR␦ (see SI Experimental Procedures). The 4 residues shown in sticks comprise the core signaling linkage to the ligand acidic moiety. (B) Overlap of indeglitazar in complex with PPAR␣ (red), PPAR␥ (green), and PPAR␦ (blue). (C–E) Close-up of the individual structures in the region highlighted from B. Note the water-mediated interactions in D and E. The chemical structures of the compounds 1, 2, and 3 (indeglitazar) are shown.

identified based on a minimal signal for significant response Pocket A in Fig. 1A, leaving Pocket B unoccupied. This feature of relative to the background by using proximity-based co-activator 1 was unique among the known PPAR ligands that had been recruitment assays with the ligand-binding domain (LBD) for observed in published co-crystal structures. Both pockets A and B

each of the 3 PPARs and a compound concentration of 100 ␮M. were found to be energetically favorable regions of the binding site MEDICAL SCIENCES Candidates with weak activity against all 3 receptors, against any in the pan-activity field analysis. Because compound 1 almost 2 of the pair-wise combinations, or containing chemotypes completely filled pocket A, we focused lead optimization chemistry related to those in the first 2 groups but active against only one on the indole N-1 position, from which substitution could be of the receptors, were selected for co-crystallization with at least directed to Pocket B through the narrow region restricted by helix 1 of the 3 receptors. Just over 25% of the 170 compounds thus 3 (H3). Analogs that could be obtained from various chemical selected were found to yield co-structural data in complex with modifications of 1 (e.g., alkylation, acylation, sulfonylation, etc.) at least 1 of the receptors. were subsequently evaluated computationally by using a molecular dynamics-based molecular mechanics/Poisson-Boltzmann solvent Scaffold Selection, Validation, and Optimization to a Pan-Agonist. A accessible surface area (MM/PBSA) approach to calculate binding ‘‘pan-activity field’’ was constructed using multiple structures for energy improvements (6, 9). The preference for an aryl sulfonamide each of the PPAR isoforms (see Experimental Procedures and substituent was clearly indicated by these studies, driven by both supporting information (SI) Experimental Procedures), enabling a geometric and electrostatic complementarity with the site. delineation of the energetically favorable regions of the binding site Validation of 1 as a scaffold for pan-active compounds was common to the 3 receptors. The crystallized scaffolds were then carried out via the synthesis of the N-phenyl sulfonamide derivative evaluated for suitability of chemistry targeting the regions of the 2, which demonstrated significantly improved activity when com- field predicted to give the most productive energies of pan-activity. pared with 1, with up to 100-fold increases in potency for the One of the first compounds to emerge from this analysis was PPARs observed [EC50s(␮M): PPAR␣,1.3;PPAR␥,1.3;PPAR␦, 5-methoxyindole-3-propionic acid (1). Compound 1 possessed 10]. Based on this success, additional sulfonamide analogs were weak to barely detectable agonist activity against the 3 PPARs evaluated computationally, and 20 were selected for synthesis. [EC50s(␮M): PPAR␣, Ϸ100; PPAR␥, Ϸ150; PPAR␦, Ͼ200]. The These compounds were then evaluated in biochemical assays, and structure of compound 1 in complex with the LBD of PPAR␥ is a subset from this group was selected for further characterization shown superimposed on the pan-activity field in Fig. 1A. The in cellular transactivation assays to measure both potency and the 5-methoxyindole core of 1 binds entirely in the region denoted degree of agonist response. Remarkably, whereas these compounds

Artis et al. PNAS ͉ January 6, 2009 ͉ vol. 106 ͉ no. 1 ͉ 263 Downloaded by guest on September 24, 2021 Fig. 2. Cellular activity of indeglitazar and its effect on the expression of adiponectin and in vivo adiponectin response. (A–C) Transactivation assay activity of indeglitazar (filled circles) and reference compound (open circles) in PPAR␣, PPAR␥, and PPAR␦, respectively. Note the partial response of indeglitazar toward PPAR␥. (D) Preadipocyte differentiation stimulated by rosiglitazone (open circles) and indeglitazar (filled circles). (E) Taqman analysis of the expression of adiponectin by mature adipocytes treated by indeglitazar and rosiglitazone. (F) Effect of indeglitazar and pioglitazone on adiponectin levels in the ob/ob mice after 14 days of treatment.

were full agonists against PPAR␣, the compounds as a group activities, clean target profiles, and excellent drug metabolism and showed a significantly reduced agonist response against PPAR␥ pharmacokinetics properties, Compound 3 was selected for devel- (averaging Ϸ40% of the full response of rosiglitazone). For several opment, and will be referred to hereafter by the United States compounds the response was reduced to Ϸ50% against PPAR␦ as Adopted Names (USAN) reference indeglitazar. well when the known PPAR␦ ligand L165041 was used as a reference. Structural Basis of the Partial Agonism for Indeglitazar. Structural From this group, we wished to select a compound with balanced characterization of indeglitazar in complex with the ligand-binding potency for the 3 receptors, partial response against PPAR␥, and domain for each of the 3 PPARs (Fig. 1 B–E) revealed that the more full response against PPAR␦. Of all of the compounds, scaffold indole core is bound in the center of pocket A and the compound 3 demonstrated the most balanced pan-agonist activity propionic acid and sulfonamide substituents are anchored in ori- profile in cells [EC50s(␮M): PPAR␣, 0.99; PPAR␥, 0.85; PPAR␦, entations consistent with our design. Interestingly, the propionic acid side-chain adapts to the specific microenvironment in each 1.3] as well as biochemically [EC50s(␮M): PPAR␣, 0.51; PPAR␥, 0.37; PPAR␦, 2.7]. Partial responses for 3, observed for both PPAR defined by a tetrad of aromatic residues, His-323 and ␥ Ϯ ␦ Tyr-327 on helix 4/5 (H4/5), His-449 on helix 11 (H11), and Tyr-473 PPAR (45% 10%, vs. the full agonist rosiglitazone) and PPAR ␥ (67% Ϯ 18%, vs. the full agonist L-165041) (Fig. 2 A–C) suggested on helix 12 (H12 or AF-2, PPAR residue numbering). As a result, the set of binding interactions made by the carboxylate group differs an interesting SPPARM profile consistent with the desired goals of ␥ ␦ this effort. Compound 3 showed no activation of other nuclear subtly in each context. Strikingly, for PPAR and PPAR com- receptors and no significant activity against the major cytochrome plexes, a water molecule mediates the carboxylate interaction to one of the surrounding residues, suggesting the possibility of water P450 enzymes as well as a diverse panel of other therapeutic targets involvement in the partial agonism of indeglitazar for the 2 recep- at a concentration of 10 ␮M (data not shown). Subsequent phar- tors. In the PPAR␥-indeglitazar complex in particular (Fig. 1D), the macokinetic assessment of 3 in Sprague–Dawley (S-D) rats and water is found between Tyr-327 and the carboxylate of indeglitazar rhesus monkeys indicated excellent systemic exposure and oral and within 2.8Å of one of the sulfonamide oxygens. Consequently, bioavailability in both species (Table 1). Because of the pan-agonist the side-chain hydroxyl group of Tyr-327 is displaced by Ͼ2Å when compared with its position in the PPAR␥-rosiglitazone complex structure (10). Because Tyr-327 and 2 other aromatic residues Table 1. Oral pharmacokinetics of indeglitazar in rat, monkey, ␥ and human (His-323 and His-449) in PPAR are optimally positioned to interact with Tyr-473 from helix 12, the readjustment of the Dose, Cmax, AUC0-ϱ, t1/2, F, hydrogen bond network caused by the water molecule results in less Species mg/kg Food ng/ml ng*hr/ml hr % stable AF-2 binding and a consequential significant decrease in the ␥ Rat 10 Fasted 20,250 59,830 1.4 98 observed degree of agonism for PPAR . Monkey 2 Fasted 12,319 217,100 23.9 72 Consequences of Partial Agonism in Cells: A Decreased Stimulation of Human 2 Fasted 6880 128,700 17.8 NA Adiponectin. Functional effects of indeglitazar on adipocytes Human 4 Fasted 8517 141,900 21.3 NA thought to be mediated through PPAR␥ were evaluated to explore

264 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0811325106 Artis et al. Downloaded by guest on September 24, 2021 Table 2. In vivo efficacy of indeglitazar in Zucker rat model of diabetes Vehicle Indeglitazar

Parameters Day 0 Day 21 P value* Day 0 Day 21 P value*

Glucose, mg/dl 465.8 Ϯ 41.9 455.7 Ϯ 97.7 497.5 Ϯ 87.4 256.3 Ϯ 73.9 0.04

HbA1C,% 2.7 Ϯ 0.05 3.1 Ϯ 0.2 0.05 2.8 Ϯ 0.1 1.5 Ϯ 0.3 0.002 TG, mg/dl 517.0 Ϯ 105.4 688.0 Ϯ 96.8 577.8 Ϯ 63.4 311.0 Ϯ 114.7 0.04 Tot. Cholesterol, mg/dl 106.3 Ϯ 5.4 97.0 Ϯ 22.7 104.8 Ϯ 5.4 62.8 Ϯ 19.9 0.04 HDL-c, mg/dl 45.5 Ϯ 9.7 51.3 Ϯ 5.2 45.0 Ϯ 8.0 35.0 Ϯ 13.4 Adiponectin, mcp/ml 7.8 Ϯ 0.7 4.9 Ϯ 0.8 0.02 7.8 Ϯ 0.2 4.8 Ϯ 1.2 0.02 Body weight, g 325.7 Ϯ 2.8 392.3 Ϯ 21.2 0.007 322.6 Ϯ 6.4 373.4 Ϯ 28.9 0.07

*Only P Ͻ 0.05 are shown

the consequences of the partial response against this receptor. In an a consequence of synergy between the 3 PPAR activities or because assay of preadipocyte differentiation, measuring in part functional of the SPPARM profile of the compound, or a combination of these insulin sensitization capability of the cells, indeglitazar showed an factors. EC50 of 0.32 ␮M compared with rosiglitazone, which showed an The oral activity of indeglitazar was assessed in the ob/ob model EC50 of 0.013 ␮M, although the maximal response obtained from of diabetes and . Indeglitazar significantly de- the 2 compounds was comparable (Fig. 2D). Separately, a Taqman creased glucose, insulin, triglycerides, and free fatty acid levels analysis after stimulation of mature adipocytes with indeglitazar or (Table 3). These effects were comparable to pioglitazone on rosiglitazone showed up-regulation of the adipose hormone adi- reducing glucose levels, triglycerides, and free fatty acids, although ponectin (Fig. 2E). However, levels achieved with rosiglitazone a significantly greater reduction of insulin levels were observed. were twice that achieved with indeglitazar, suggesting that the Body weight differences in this study were not significant. As difference between full and partial agonism might be reflected by expected, pioglitazone increased adiponectin levels 3.5-fold, this measure. whereas indeglitazar raised adiponectin levels only 1.9-fold (Fig. 2F This finding was of particular interest because of the role of and Table 3). These data are consistent with the partial agonism adiponectin (11), which has been shown to modulate insulin sen- observed in cell-based studies and also suggest that the insulin sitivity and can be used to ameliorate insulin resistance in animal sensitizing activities of indeglitazar are at least partially indepen- models. Numerous studies have noted a strong increase in the levels of dent of adiponectin. adiponectin in response to treatment with PPAR␥ agonists in animal models and in humans (12). This suggested that indeglitazar, if effica- Lipomics Analysis Shows Engagement of All Three Receptors in Vivo. cious, might have an in vivo activity profile distinct from glitazone-like To further assess the role that each PPAR isoform plays in the PPAR␥ agonists that would be worthy of investigation. pharmacological activity of indeglitazar, we performed metabolo- Indeglitazar Treatment in Vivo Lowers Glucose with Decreased Effects mic analyses of mice treated with indeglitazar (Table S1). Five on Adiponectin and Weight. An initial assessment of in vivo activity hundred individual lipid metabolites were analyzed using a mass- was carried out using the Zucker rat model of diabetes. As shown spectrometry-based approach (Lipomics Technologies). As ex- pected for a compound with PPAR␣ and PPAR␥ activity, the in Table 2, significant lowering of glucose, HbA1C, triglycerides, and total cholesterol were observed after i.v. treatment with 10 mg/kg levels were significantly decreased (17). More impor- indeglitazar once per day for 3 weeks. Notably, the level of tantly, the composition of several lipid classes was significantly adiponectin (on day 21) was essentially unchanged in treated vs. altered. Cholesterol esters were depleted of essential fatty acids untreated animals (4.8 mcg/ml vs. 4.9 mcg/ml), thus the observed (18:2n6, 18:3n3) and major n3 fatty acids (20:5n3, 22:6n3) while MEDICAL SCIENCES reductions in glucose and HbA1C were achieved in an adiponectin- enriched in n7 fatty acids (18:1n7, 16:1n7). In addition, 20:3n9 levels independent fashion. This efficacy was also observed in the absence were elevated indicating increased ⌬-9 and ⌬-6 desaturase activity. of effects normally associated with PPAR␥ agonists. There was a The triglyceride pools were depleted of dietary fatty acids (18:2n6, decrease of 4.8% in total body weight relative to vehicle-treated 18:3n3) and the major n3 fatty acids (20:5n3, 22:5n3, and 22:6n3) animals (day 21, 373.4g treated vs. 392.3g untreated, Table 2), in while enriched with saturated fatty acids (14:0, 15:0, and 18:0). The contrast to other glitazone and non-glitazone PPAR␥ agonists, triglyceride pool was also substantially enriched with 20:4n6. The which have been shown to cause significant weight gain in this phosphotidylcholine pool was also depleted of dietary essential model at similar levels of glucose lowering (13–16). There were also fatty acids (18:2n6, and 18:0) while enriched with 16:0n7 fatty acids, no increases in heart or weight on necropsy at the end of this 18:1n9, and 20:3n9. The increases in both saturated and monoun- study. These differences in the effects of indeglitazar in vivo may be saturated fatty acids is indicative of increased de novo lipid synthesis

Table 3. In vivo efficacy of indeglitazar in ob/ob mouse model of diabetes and insulin resistance Parameters Vehicle Indeglitazar, 10 mg/kg Pioglitazone, 10 mg/kg

Glucose, mg/dl 365.7 (Ϯ 16.3) 121.9 (Ϯ 8.6) 102.7 (Ϯ 6.0) Insulin, ng/ml 12.3 (Ϯ 1.3) 5.1 (Ϯ 0.7) 7.8 (Ϯ 1.1) TG, mg/dl 307.3 (Ϯ 40.5) 58.4 (Ϯ 9.8) 46.7 (Ϯ 3.6) FFA, mM 1.1 (Ϯ 0.11) 0.71 (Ϯ 0.08) 0.75 (Ϯ 0.05) Adiponectin, mcg/ml 9.5 (Ϯ 0.46) 18.1 (Ϯ 2.6) 33 (Ϯ 5.8) Body weight, g 51.6 (Ϯ 1.1) 52.8 (Ϯ 0.9) 52.4 (Ϯ 0.9) Body weight change, g 3.9 5.8 5.2

Artis et al. PNAS ͉ January 6, 2009 ͉ vol. 106 ͉ no. 1 ͉ 265 Downloaded by guest on September 24, 2021 Fig. 3. In vivo effect on body weight. (A and B) Effects of indeglitazar and fenofibrate on body weight and triglyc- eride levels in a hamster DIO model after 14 days of oral dosing. (C) Effects on body weight of indeglitazar and pioglitazone dosed orally for 6 weeks in obese bonnet macaques.

and is best explained by an increased flux of lipid toward adipose Discussion tissue, a hallmark of PPAR␥ agonists (18). We have shown that the scaffold-based methodology can be Treatment with indeglitazar also increased ⌬-6 desaturase activ- extended to the generation of compounds capable of targeted ity. This enzyme is the rate limiting step in the biosynthesis of receptor activation and beyond the initial scope focused on the polyunsaturated fatty acids, including 20:4n6 and 22:6n3. The discovery of enzyme inhibitors (6, 8). To our knowledge, this is 5Ј-flanking region of the ⌬-6 desaturase gene contains a DR-1 the first such report from any of the related fragment-based response element and its expression is increased by PPAR␣ ago- screening approaches (23, 24). By implementing a multireceptor nists (19). The depletion of n3 fatty acids is also consistent with screening system for biological and structural filtering of low- PPAR␦ activation. During prolonged exercise free fatty acids molecular-weight compounds through the PPARs, we identified released from adipose tissue are taken up by skeletal muscle and a scaffold with a unique binding mode (Fig. 1). Structure-guided used as fuel. It has been suggested that one of the physiological roles chemistry exploration of the scaffold led to the generation of a of PPAR␦ is to regulate this metabolic switch between glucose and series of selective pan-PPAR-activating compounds in 2 rounds fatty acid utilization and furthermore that the released fatty acids may of synthesis. provide the ligand for PPAR␦ to activate fatty acid oxidation (20). From this series of compounds, indeglitazar was identified as having balanced potency against the 3 PPARs with varied degree Treatment Results in Weight Loss in Obese Hamsters and Monkeys. We of agonist response in the decreasing order PPAR␣ϾPPAR␦Ͼ further investigated the differences in effects on weight between PPAR␥. Structural characterization of indeglitazar bound to the indeglitazar and single subtype agonists in 2 additional obesity 3 PPARs revealed that, for PPAR␦ and PPAR␥, a single water models. Indeglitazar was evaluated in hamsters fed a high fat diet, molecule was recruited into the signaling interface from ligand because these animals bear greater similarity to primates than mice to receptor. Interestingly, the 4 aromatic residues of the signaling ␥ or rats in their lipid metabolism (21). Interestingly, hamsters treated tetrad in the PPAR -indeglitazar co-structure overlap closely ␥ orally with 30 mg/kg indeglitazar for 2 weeks weighed significantly with the same set of residues observed in the complex of PPAR less than either control animals or animals treated with the PPAR␣ with another partial agonist, MRL20 (25). However, whereas Ϸ ligand, (Fig. 3A). This weight differential could be due MRL20 shows a response of 80% relative to rosiglitazone, Ϸ to increased lipid metabolism induced by this pan-agonist, as both indeglitazar is more substantively reduced to 45%. In the fed and fasted levels of triglycerides were significantly reduced in MRL20 complex, the incorporation of the structural water is indeglitazar-treated hamsters and not in fenofibrate-treated ani- precluded by what would be a steric clash with hydrophobic mals (Fig. 3B). functionality in the ligand. Our results showed that direct modulation of helix 12 and the surrounding key signaling resi- Indeglitazar also induced a dose-dependent decrease in weight dues by partial agonists with greatly reduced response can occur. after a 6-week treatment regimen in obese bonnet macaques (Fig. Furthermore, these results suggest that with proper design, it 3C). Interestingly, a recent study of PPAR agonists described a may be possible to fine tune the modulation of helix 12, synergistic effect from stimulating PPAR␣ and PPAR␦, which depending on the precise ligand structure and details of the resulted in a significantly enhanced weight loss effect over action on interactions within the complex. Similar observations regarding the individual receptors in mice (22). Our data suggests that this is the role of structural water in a partial agonist response have also not a mouse-specific observation, and that appropriate stimulation recently been made in the case of the ionotropic glutamate of the 3 PPARs can result in weight loss in both rodents and receptor (26). Our results suggest that this may be a more primates, in addition to beneficial effects on glucose, insulin, and generalizable phenomenon, and that dramatic changes in sig- lipid profiles in various models. Such a profile would be expected naling—and in this case transcription—may result from struc- to carry significant benefit if realized in the clinic, particularly if turally subtle distinctions in a signaling complex. from a single agent. Treatment of rodents with indeglitazar lowers blood glucose to the same extent as full PPAR␥ agonists (glitazones), but shows a Initial Clinical Pharmacokinetics and Safety. Based on this compelling diminished adiponectin response and reduced weight gain. Indeg- profile of activity, indeglitazar was selected for development. After litazar represents a class of new chemical agents that differentiate completion of appropriate preclinical safety studies, evaluation of the insulin-sensitizing and the adiponectin-stimulating effects as- the safety and pharmacokinetics of indeglitazar in humans was sociated with PPAR␥ agonists. Whereas the efficacy of indeglitazar carried out at doses of 2 and 4 mg/kg in a randomized, double-blind, in the absence of a strong adiponectin response might appear placebo-controlled, single-ascending dose trial in healthy volun- initially surprising, recent reports indicate that in adiponectin teers. The compound displayed excellent systemic exposure, with a knockout mice, the response to PPAR␥ agonists is diminished half-life under fasted conditions of Ͼ17 h (Table 1). Indeglitazar somewhat but not abolished and may be adiponectin-independent was safe and well tolerated at both doses. Based on these data, in muscle (27). A recent report has demonstrated that overexpres- indeglitazar was advanced to further clinical safety and efficacy sion of adiponectin in an obese mouse background can lead to an studies, and has progressed to Phase 2B efficacy and safety studies expansion of adipose tissue which causes dramatic additional weight in T2DM patients. gain along with normalization of glucose levels (28). Beyond partial

266 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0811325106 Artis et al. Downloaded by guest on September 24, 2021 agonism toward PPAR␥, the favorable in vivo profile of indegli- Compound Synthesis. Compounds 2 and 3 (indeglitazar) were synthesized in 4 tazar presented here is also likely to derive from its concurrent steps from commercially available starting material, 5-methoxy-indolyl-3- activation of PPAR␣ and PPAR␦. PPAR␦ (29) and PPAR␣ (30) carboxaldehyde with different benzenesulfonyl chlorides. See SI Experimental agonists have been shown to cause an increase in energy flux Procedures and Scheme S1 for details of synthesis. through muscle and liver and the lipid analysis indicates that these effects are seen with indeglitazar treatment. Consistent with the Biochemical and Cellular Assays. The in vitro PPAR agonist activity was deter- weight loss seen in studies in obese rodents and monkeys, our mined by measuring compound-dependent interaction between recombinant observations suggest the possibility that treatment with indeglitazar PPAR-LBDs and biotinylated coactivator peptides by using the Alpha Screen avoids the unwanted effect of PPAR␥ full agonists on the adipose Technology (Perkin-Elmer). Transcriptional transactivation activities of the com- compartment and shifts the major metabolic impact on glucose and pounds were measured by the luciferase reporter assay using 293T cells cotrans- lipids from fat toward muscle. fected with pGal4-DBD-PPAR-LBD fusion. The preadipocyte differentiation assay Beyond indeglitazar, compound 1 has proven to be a rich and Taqman analysis were carried out following the protocol in SI Experimental Procedures. progenitor of analogs with widely differing profiles against the 3 PPARs, indicating the generality of the approach to the problem of subfamily specificity. By using the PPAR-ligand binding domain Cloning, Expression, Protein Purification, Crystallization, and Structure Deter- (PPAR-LBD) structural information in combination with the com- mination. See SI Experimental Procedures and Table S2 for details. putational techniques described herein, subtype specific regions of each of the 3 PPARs have been identified and used in the design Scaffold-Based Drug Design. Compounds were designed to maximize productive of analogs with enhanced selectivity for 1 or 2 of the receptors. interactions delineated by the PPAR pan-active fields (see SI Experimental Pro- Shifts from a pan-activity profile through different dual-activity cedures). The designed compounds were first modeled into the ligand binding profiles to compounds with 200- to 400-fold selectivity for each of pocket by scaffold-anchored placement. The protein–ligand complex was sub- the individual receptors have been achieved with various substitu- jected to molecular dynamics (MD) simulation using AMBER7 and PARM94 force tions on this scaffold (Fig. S1). Furthermore, when 46 compounds fields. The binding free energy of the complex was estimated using the MM/PBSA methodology (9). from this set were evaluated for pharmacokinetics in the rat, mean oral bioavailability was 66% at a dose of 2 mg/kg. Consequently, the Animal Studies. All experimental work was conducted in accordance with the approach described herein and the resulting group of compounds humane guidelines for ethical and sensitive care of the Institutional Animal Care represents both a powerful resource for exploration into the and Use Committee of the U.S. National Institutes of Health. Experimental details biological consequences of PPAR activation and a potential source are described in the SI Experimental Procedures. of new therapeutics as subpopulations of patients with metabolic disease are increasingly defined. ACKNOWLEDGMENTS. Diffraction data were collected at the Advanced Light Source and the Stanford Synchrotron Radiation Laboratory, which are supported Experimental Procedures by the U.S. Department of Energy, Office of Basic Energy Sciences under contract Note that detailed methods are included in SI Experimental Procedures. DE-AC03–76SF00098 and DE-AC03–76SF00515, respectively.

1. Grant RW, Devita NG, Singer DE, Meigs JB (2003) Polypharmacy and adherence 17. Harrity T, et al. (2006) , a novel dual (alpha/gamma) peroxisome proliferator- in patients with type 2 diabetes. Diabetes Care 26:1408–1412. activated receptor activator, improves diabetes and other metabolic abnormalities and 2. Austin RP (2006) Polypharmacy as a risk factor in the treatment of type 2 diabetes. Diabetes preserves beta-cell function in db/db mice. Diabetes 55:240–248. Spectrum 19:13–16. 18. Watkins SM, Reifsnyder PR, Pan HJ, German JB, Leiter EH (2002) Lipid metabolome-wide 3. Evans JL, Lin JL, Goldfine ID (2005) Novel approach to treat insulin resistance, type 2 effects of the PPARgamma agonist rosiglitazone. J Lipid Res 43:1809–1817. diabetes, and the : Simultaneous activation of PPARalpha, PPAR- 19. Tang C, Cho HP, Nakamura MT, Clarke SD (2003) Regulation of human delta-6 desaturase gamma, and PPARdelta. Current Diabetes Reviews 1:299–307. gene transcription: Identification of a functional direct repeat-1 element. J Lipid Res 4. Willson TM, Brown PJ, Sternbach DD, Henke BR (2000) The PPARs: From orphan receptors 44:686–695. to drug discovery. J Med Chem 43:527–550. 20. Tanaka T, et al. (2003) Activation of peroxisome proliferator-activated receptor delta 5. Rocchi S, et al. (2001) A unique PPARgamma ligand with potent insulin-sensitizing yet induces fatty acid beta-oxidation in skeletal muscle and attenuates metabolic syndrome. weak adipogenic activity. Mol Cell 8:737–747. Proc Natl Acad Sci USA 100:15924–15929. 6. Card GL, et al. (2005) A family of phosphodiesterase inhibitors discovered by cocrystallog- raphy and scaffold-based drug design. Nat Biotechnol 23:201–207. 21. McIntosh CHS, Pederson RA (1999) Nonisulin-Dependent Animal Models of Diabetes Mellitus. Experimental Models of Diabetes, ed McNeill JH (CRC Press, Boca Raton, FL), pp 7. Kumar A, et al. (2005) Crystal structures of proto-oncogene kinase Pim1: A target of MEDICAL SCIENCES aberrant somatic hypermutations in diffuse large cell lymphoma. J Mol Biol 348:183–193. 337–398. 8. Tsai J, et al. (2008) Discovery of a novel selective inhibitor of oncogenic B-Raf kinase with 22. Harrington WW, et al. (2007) The effect of PPARalpha, PPARdelta, PPARgamma, and potent anti-melanoma activity. Proc Natl Acad Sci USA 105:3041–3046. PPARpan agonists on body weight, body mass, and serum lipid profiles in diet-induced 9. Kollman PA, et al. (2000) Calculating structures and free energies of complex molecules: obese AKR/J mice. PPAR Res 2007:97125. Combining molecular mechanics and continuum models. Accounts of Chemical Research 23. Congreve M, Chessari G, Tisi D, Woodhead AJ (2008) Recent developments in fragment- 33:889–897. based drug discovery. J Med Chem 51:3661–3680. 10. Nolte RT, et al. (1998) Ligand binding and co-activator assembly of the peroxisome 24. Hajduk PJ, Greer J (2007) A decade of fragment-based drug design: Strategic advances and proliferator-activated receptor-gamma. Nature 395:137–143. lessons learned. Nat Rev Drug Discov 6:211–219. 11. Goldfine AB, Kahn CR (2003) Adiponectin: Linking the fat cell to insulin sensitivity. Lancet 25. Bruning JB, et al. (2007) Partial agonists activate PPARgamma using a helix 12 independent 362:1431–1432. mechanism. Structure 15:1258–1271. 12. Maeda N, et al. (2001) PPARgamma ligands increase expression and plasma concentrations 26. Hald H, et al. (2007) Partial agonism and antagonism of the ionotropic glutamate receptor of adiponectin, an adipose-derived protein. Diabetes 50:2094–2099. iGLuR5: Structures of the ligand-binding core in complex with domoic acid and 2-amino- 13. Brown KK, et al. (1999) A novel N-aryl tyrosine activator of peroxisome proliferator- 3-[5-tert-butyl-3-(phosphonomethoxy)-4-isoxazolyl]propionic acid. J Biol Chem activated receptor-gamma reverses the diabetic phenotype of the Zucker diabetic fatty 282:25726–25736. rat. Diabetes 48:1415–1424. 27. Nawrocki AR, et al. (2006) Mice lacking adiponectin show decreased hepatic insulin 14. Oakes ND, et al. (2005) , a dual PPARalpha/gamma agonist, ameliorates sensitivity and reduced responsiveness to peroxisome proliferator-activated receptor glucose and lipid intolerance in obese Zucker rats. Am J Physiol Regul Integr Comp Physiol 289:R938–946. gamma agonists. J Biol Chem 281:2654–2660. 15. Pickavance LC, Brand CL, Wassermann K, Wilding JP (2005) The dual PPARalpha/gamma 28. Kim JY, et al. (2007) Obesity-associated improvements in metabolic profile through agonist, ragaglitazar, improves insulin sensitivity and metabolic profile equally with expansion of adipose tissue. J Clin Invest 117:2621–2637. pioglitazone in diabetic and dietary obese ZDF rats. Br J Pharmacol 144:308–316. 29. Wang YX, et al. (2003) Peroxisome-proliferator-activated receptor delta activates fat 16. Wargent E, et al. (2005) Improvement of glucose tolerance in Zucker diabetic fatty rats by metabolism to prevent obesity. Cell 113:159–170. long-term treatment with the dipeptidyl peptidase inhibitor P32/98: Comparison with and 30. Barger PM, Kelly DP (2000) PPAR signaling in the control of cardiac energy metabolism. combination with rosiglitazone. Diabetes Obes Metab 7:170–181. Trends Cardiovasc Med 10:238–245.

Artis et al. PNAS ͉ January 6, 2009 ͉ vol. 106 ͉ no. 1 ͉ 267 Downloaded by guest on September 24, 2021