Phenolic Lipids Derived from Cashew Nut Shell Liquid to Treat Metabolic Diseases

,


■ INTRODUCTION
−9 The PPARs are transcription factors that act as heterodimers with the retinoid X receptor (RXR) and bind directly to DNA.Upon ligand activation, there is a conformational change in the receptor that promotes interaction with co-regulators, thereby modulating the recruitment of basal transcriptional machinery and influencing gene expression. 10,11Importantly, the PPARs are activated by endogenous ligands (fatty acids, lipid metabolites, and eicosanoids) at micromolar concentrations and synthetic ligands at nanomolar concentrations. 12,13There are three isoforms: PPARα, PPARγ, and PPARδ. 14PPARα is predominantly expressed in the liver and coordinately regulates the transcription of genes important in liver fatty acid uptake and fatty acid utilization, thereby decreasing plasma triglycerides. 15In contrast, PPARγ is most highly expressed in adipose tissue and serves to enhance insulin sensitivity and regulate adipogenesis. 16PPARδ is ubiquitously expressed including the small intestine, liver, brain, and skeletal muscle.Activation of PPARδ in skeletal muscle improves insulin sensitivity and enhances energy utilization as well as endurance exercise capacity. 17,18These properties make each PPAR receptor unique in its regulation of lipid metabolism and glucose homeostasis. 19ntihyperlipidemic agents, known as the fibrate class of drugs (gemfibrozil, clofibrate, fenofibrate, bezafibrate) (Figure 1), target PPARα and are effective at lowering hypertriglyceridemia and increasing high-density lipoprotein choles-terol (HDL-C). 20The insulin-sensitizing thiazolidinedione (TZDs) drugs (troglitazone, ciglitazone, rosiglitazone, pioglitazone) (Figure 1) target PPARγ and are used to improve insulin sensitivity in patients with type 2 diabetes. 8,16The glitazars (muraglitazar, tesaglitazar, aleglitazar, saroglitazar) (Figure 1) target PPARα and PPARγ and were developed to simultaneously treat hyperlipidemia and hyperglycemia.However, prolonged use of TZDs and glitazars lead to major PPARγ-related side effects including congestive heart failure, weight gain, bone loss, and edema, thus severely limiting their use and/or halting their further development. 16,19,21,22These side effects are specifically related to the high affinity, full agonistic activity, and potency of these compounds on the PPARγ receptor. 16,19,21 drug that combines the best features of both PPARα and PPARγ activation with an improved safety profile would be an attractive therapeutic modality for the treatment of metabolic syndrome, type 2 diabetes, and hyperlipidemia.23,24 Partial PPARγ agonists or weak PPARγ ligands have previously been shown in mice to have insulin-sensitizing effects without the side effect of weight gain.25−27 Compounds with these characteristics are termed selective PPAR modulators (SPPARMs) and are characterized by their tissue-selective and/or gene-selective activities.21,28 A key feature of SPPARMs is their ability to allow the separation of the negative effects of PPAR activation from the beneficial (therapeutic) effects.
−36 Herein, we developed novel PPAR agonists with properties similar to endogenous ligands in that they have balanced affinities and/or partial agonist activity for PPARα and PPARγ.This work was inspired by the structural similarity between fatty acids, which function as endogenous ligands of PPARs, and anacardic acid (1) and cardanol (2) mixtures the main phenolic compounds of the natural and technical-grade cashew nut shell liquid of Anacardium occidentale, respectively. 37Figure 2 shows the structural similarity of stearic acid with saturated anacardic acid.
■ RESULTS AND DISCUSSION Chemical Design.A new series of saturated anacardic acid derivatives were designed to identify structural features relevant for PPAR receptor recognition.To create molecular diversity and to delineate preliminary structure−activity relationships (SAR), cardanol was included as it is another major component of technical-grade CNSL.Our purpose was to determine whether these novel compounds derived from this natural resource could serve as SPPARMs with eventual application for the treatment of metabolic disease.Toward this goal, we turned our attention to scaffolds that share a C15 saturated alkyl chain as a basic feature of the fatty acid mimetics, in this case, sustainable and cost-effective for drug development. 38Also, the presence of an acid group completes the structural requirements for molecular recognition by PPAR.Thus, we designed derivatives of anacardic (6pentadecyl-2-hydroxybenzoic) and isoanacardic (4-pentadecyl-2-hydroxybenzoic) acids with O-substitutions in the phenol to obtain O-acetylated and O-methylated derivatives.For the cardanol series (3-pentadecylphenol and (Z)-(3-pentadec-8en-1-yl)phenol), the same O-variations designed for the acids were planned.To overcome the absence of the carboxylic group in cardanol, α-phenoxyalkyl acid derivatives were designed through the insertion of the carboxymethylene  subunit including its analogue with the geminal dimethyl group found in some fibrates (Figure 1).The design strategy of CNSL derivatives is shown in Figure 3.
Chemistry.Overall, 25 compounds (3−27) were synthesized following the routes outlined in Schemes 1−3.Isolations of the anacardic acids (1) and cardanol (2) mixtures, respectively, obtained from natural and technical CNSL, were performed according to Rossi et al. 31 To synthesize derivatives 4−8, we started with the hydrogenation of the unsaturated chains present in the mixture 1 using Pd/C 10% as catalyst in a Parr hydrogenation apparatus to obtain the saturated anacardic acid (3).Then, 3 was O-acetylated with acetic anhydride and H 3 PO 4 to give the acetoxy-derivative 4. Taking advantage of the differential reactivity of the phenol and carboxylic acid groups, obtaining acid 5 with methylated phenolic hydroxyl was carried out in two steps.First, 3 was converted to the derivative O,O-dimethylated 8 with methyl iodide in acetone at 102 °C using condenser cooling to −8 °C.
Next, the methyl ester 8 was hydrolyzed with t-BuOK in dimethyl sulfoxide (DMSO) at 40 °C leading to derivative 5. To synthesize derivatives 6 and 7, the acid 3 was transformed into the corresponding methyl salicylate 6 by the Fischer reaction in methanol catalyzed with H 2 SO 4 and the ester was converted to the O-acetyl derivative 7 with acetyl chloride/ triethylamine (TEA) in dichloromethane (Scheme 1).
The derivatives 12−23 have as the precursor the saturated cardanol 9 obtained by the catalytic hydrogenation of mixture 2 with Pd/C 10% in ethanol.To synthesize compounds 12− 17 (isoanacardic acid series), 9 was transformed into salicylaldehyde 10 by regiospecific ortho-formylation with paraformaldehyde, MgBr 2 , and triethylamine in THF under reflux.Then, 10 was converted to the methoxy-derivative 11 with methyl iodide at 65 °C.Both aldehydes 10 and 11 were oxidized with NaClO 2 in the presence of NaH 2 PO 4 in a mixture DMSO/CH 2 Cl 2 (1:1), providing the respective acids 12 and 14.In turn, the acids were converted to the methyl esters 15 and 17 with methanol catalyzed by sulfuric acid.To synthesize derivatives 13 and 16, compounds 12 and 15 were acetylated with acetic anhydride catalyzed by phosphoric acid to give the O-acetyl derivatives (Scheme 2).To synthesize compounds 18 and 19 (cardanol series), 9 was converted to the acetoxy (18) and methoxy (19) derivatives in similar procedures applied in the synthesis of compounds 4 and 17.
To approximate the structural characteristics among the derivatives of anacardic acid, isoanacardic acid, and cardanol, 9 was transformed into α-phenoxyalkyl esters by reaction of 9 with ethyl bromoacetate in acetone at room temperature or ethyl α-bromoisobutyrate in the presence of KI in acetonitrile at 82 °C to give the respective compounds 20 and 21.Finally, the ethyl esters 20 and 21 were hydrolyzed with LiOH in THF/H 2 O in the presence of the phase transfer catalyst Aliquat 336, at room temperature for 20 and at 82 °C for 21, to provide the α-phenoxyalkyl acids 22 and 23 (Scheme 2).
After preliminary screening of the saturated cardanol derivatives showed selective activation of PPARs (shown below), we synthesized a series of monounsaturated cardanol derivatives (24−27) to directly compare the impact of the unsaturation in the C15 tail (Scheme 3) to the corresponding saturated derivatives (20−23).
We next determined the relative potency and efficacy of the active CNSL derivatives by performing a full dose−response curve for the three human PPAR isoforms.The activity profiles and half-maximal effective concentration (EC 50 's) of fatty acids and our compounds were compared to positive controls that are synthetic agonists with high specificity, full agonist activity, and high potency for their respective PPAR receptor in the luciferase assays.These positive controls were GW7647 (PPARα agonist), rosiglitazone (Rosi, PPARγ agonist), and GW0742 (PPARδ agonist), respectively.Individual E max values were calculated relative to positive controls (set to 100%) and are reported in Figure S1 and summarized in Table 1.For some compounds, solubility was limiting and saturation of the activity was not achieved.Against PPARα, C10:0, C14:0, and C18:0 displayed partial agonist activity compared to GW7647 (Figure S1A).Similarly, 3, 5, 12, 14, and 20−25 partially induced PPARα activity relative to GW7647 (Figure S1A).In contrast, C18:1n9, 4, 26 and 27 acted as full agonists of PPARα, with equal or greater RLU induction compared to GW7647 (Figure S1A).The EC 50 values of the CNSL derivatives ranged from 0.5 to 67 μM for PPARα (Table 1).Thus, many were more potent than the endogenous fatty acid ligands which activated PPARα between 13 and 40 μM.Against PPARγ, the CNSL derivatives 14, 20, 21, and 23 displayed partial agonist activity compared to Rosi (Figure S1B).Surprisingly, 3, 4, 22, and 27 exhibited similar or higher maximal activity relative to Rosi, though they were of lower potency, activating PPARγ at micromolar concentrations from 0.9 to 50 μM (Table 1).Interestingly, C10:0 was the only fatty acid tested that showed PPARγ activity (EC 50 54 μM, Table 1).No E max could be determined for 24−26 because they did not reach saturation against PPARγ.Activation of PPARδ was additionally explored with EC 50 values for the active compounds (Figure S1C).We found that C14:0, C18:1n9, 3, 4, and 22−27 partially induced transcriptional activation of human PPARδ compared to GW0742, a full PPARδ agonist with EC 50 of 3.5 nM (Figure S1C).PPARδ was weakly induced by fatty acids and the CNSL derivatives at micromolar concentrations, ranging from 10 to 100 μM (Table 1).Taken together, a subset of CNSL derivatives were successfully identified as a single, dual, and/or pan-PPAR agonists with partial or full agonistic activities for human PPAR isoforms.These data suggest that these CNSL derivatives represent a new chemical class of PPAR agonists that structurally mimic the endogenous fatty acid ligands of PPARs but generally with higher potency.
CNSL Derivatives Selectively Target PPARα-Responsive Genes for the Regulation of Lipid Metabolism in Primary Hepatocytes.To examine the ability of CNSL derivatives to activate PPARα target genes, mouse primary hepatocytes were treated with 50 μM of the CNSL derivatives for 16 h.Consistent with their ability to activate PPARα, the PPARα agonists, WY14643 (WY) and GW7647 (GW), significantly induced the mRNA expression of genes involved fatty acid uptake such as fatty acid binding protein 1 (Fabp1) and Cd36 (cluster of differentiation 36) (Figure 5A,B).Notably, dual PPARα/γ agonist, muraglitazar (Mura), also strongly upregulated the expression of Fabp1 and Cd36, whereas Rosi, the PPARγ agonist, did not.Fabp1 and Cd36 were significantly increased by 24 and 25.In contrast, 22 only induced gene expression of Fabp1 while 23 and 26 significantly upregulated Cd36 (Figure 5A,B).In addition, WY, GW, and Mura significantly increased the fatty acid oxidation genes fibroblast growth factor 21 (Fgf 21) and pyruvate dehydrogenase 4 (Pdk4), whereas induction by Rosi did not reach statistical significance (Figure 5C,D).PPARα-responsive genes Fgf 21 and Pdk4 were robustly increased in primary hepatocytes by 5, 12, 22, 23, and 27 compared to vehicle control (Figure 5C,D).Intriguingly, 3, 4, and 24 increased the mRNA expression of Pdk4 but not Fgf 21 (Figure 5C,D).Additionally, 12 upregulated only the expression of Fgf 21 and 14 failed to upregulate any of the target genes despite having an EC 50 for PPARα below that of 5 (17 vs 32 μM, respectively).These results suggest that depending on their structure, the CNSL derivatives can promote fatty acid uptake and oxidation in the liver by differential targeting of PPARα binding sites, leading to decreased circulating lipids and improved lipid metabolism.
CNSL Derivatives Differentially Activate PPARγ-Target Genes in the Process of 3T3-L1 Adipocyte Differentiation.PPARγ is considered a master regulator for the differentiation of pre-adipocytes into mature fat cells. 39ndeed, weight gain is a major side effect of the TZD class of insulin-sensitizing drugs. 40,41Insulin sensitization, however, requires the activation of PPARγ in adipose to promote the expression of adiponectin (encoded by Adipoq), an adipokine that signals via the liver and muscle to regulate glucose homeostasis. 42,43To gain an understanding of whether the In vitro screening of CNSL derivatives for PPARα, PPARγ, and PPARδ activity reveals a subset of selective pan-activators.HEK293 cells were transiently co-transfected with GAL4-hPPARα (A), GAL4-hPPARγ (B), or GAL4-hPPARδ (C) together with UAS-luciferase reporter and treated with positive controls (10 nM GW7647, 100 nM Rosi, and 10 nM GW0742) or 50 μM of indicated compounds for 16 h.Data represent mean ± standard deviation (SD) (N = 3).RLU, relative luciferase units = luciferase light units/β-galactosidase × time.Vehicle (DMSO) response was set to 1. C10:0, decanoic acid; C14:0, myristic acid; C18:0, stearic acid; C18:1n9, oleic acid.*P < 0.05 relative to corresponding vehicle, using one-way analysis of variance (ANOVA) with Holm−S ̌idaḱ correction.
CNSL derivatives would impact adipocyte differentiation and insulin sensitization, 3T3-L1 cells were differentiated in the presence of 25 μM of CNSL derivatives or 10 μM Rosi for 11 days.As expected, Rosi efficiently promoted the differentiation of 3T3-L1 cells into adipocytes, as indicated by the increase in Oil Red O staining relative to vehicle (Figure 6A, quantified in (B)).Interestingly, 3 and 24−26 were as efficient as Rosi at promoting adipocyte differentiation (Figure 6A,B).By contrast, 4, 14, 20, 21, 23, and 27 showed lower lipid accumulation in differentiated 3T3-L1 cells compared to Rosi (Figure 6A,B).In addition, Rosi significantly increased the expression of key early adipogenic genes, including PPARγ itself and CCAAT/enhancer binding protein α (Cebpα), as well as genes responsible for fatty acid uptake and storage such as adipocyte binding protein 2 (aP2), lipoprotein lipase (Lpl), and Cd36 (Figure 6C−G).Additionally, Rosi significantly upregulated the mRNA expression of the beneficial adipokine Adipoq and insulin-stimulated glucose uptake transporter glucose transporter type 4 (Glut4) (Figure 6H,I).We noted that CNSL derivatives were less effective at inducing genes that were involved in fatty acid update and adipogenesis compared to Rosi (Figure 6C−G), yet some CNSL derivatives still strongly induced the expression of Adipoq and Glut4, including 23 and 27, which had levels at or above those induced by Rosi (Figure 6H−I).These data indicate that CNSL derivatives may be beneficial for the treatment of insulin resistance without the severe consequence on body weight.No data could be obtained for 22 in 3T3-L1 cells as it was found to be toxic over the time period required for differentiation.It is unclear why 22 is selectively toxic in 3T3-L1 and HEK293 cells (Table 1), but not primary hepatocytes (Figure 5).An 3-(4,5dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay performed in HEK293 cells dosing CNSL derivatives at 25 μM found that only 22 showed significant toxicity (data not shown).In summary, 20, 21, 23, and 27 show selective PPARγ target gene activation in 3T3-L1 cells that could potentially separate their effects on adipocyte differentiation from its favorable glucose-lowering effects.
In Vivo Screening of the CNSL Derivatives in Zebrafish Embryos Harboring Transgenic Human

Journal of Medicinal Chemistry
PPARs.To assess the in vivo activity and tissue distribution of the CNSL derivatives, we used the ligand trap zebrafish screening system.In this model, fusion proteins of the GAL4 DNA binding domain and the ligand-binding domain (LBD) of human PPARs are expressed under the control of a heat shock promoter. 44A green fluorescent protein (GFP) reporter responsive to GAL4 binding is included in the transgene so that after exposure to heat shock, the 2-day post-fertilization embryos express GFP in tissues where their endogenous respective PPAR ligand is present (Figure 7, vehicle).In the presence of exogenously added WY14643, full agonist for PPARα, GFP expression in embryos expands from its relatively restricted pattern to include strong expression in the CNS, epidermis, heart, retina, and muscle.Rosiglitazone treatment strongly increased PPARγ activity in the tail bud, CNS and heart.In contrast, GW0742 increased PPARδ activity in CNS and muscle.All CNSL derivatives were dosed at each receptor's EC 50 , unless otherwise indicated, to allow us to compare the signal intensity and tissue distribution between compounds and account for their varied affinities.4 and 23 strongly activated PPARα in the brain, yolk sac, heart, and muscle.In contrast, 21 and 27 demonstrated restricted activation of PPARα, predominantly in brain.Compared to activation by Rosi, CNSL derivatives were mostly active in tail bud (4, 23, 27), forebrain (23, 27) and hindbrain (23, 27) in the PPARγ fish line.In the PPARδ zebrafish line, there was substantial basal activity and none of the tested CNSL derivatives increased GFP signal beyond vehicle.Interestingly, 20 did not activate any of the PPARs suggesting there are in vivo barriers that preclude it from accessing the PPARs.These data support the value of an in vivo screening system since this differential response could not be predicted based on the EC 50 values or gene expression performed by in vitro screening alone.Overall, these data suggest that selected CNSL derivatives (4, 21, 23, 27) reveal tissue-specific activation of PPARα and/or PPARγ in an in vivo zebrafish embryo model.
Selected CNSL Derivatives Bind and Stabilize the hPPAR-LBDs.To determine whether the CNSL derivatives bound directly to PPARs, we performed a thermal shift assay with purified hPPARα-, hPPARγ-, and hPPARδ-ligand-binding domains (LBDs).Direct binding of a ligand to purified LBD protein should lead to stabilization against thermal denaturation.We found that 4, 23, and 27 induced large thermal shifts for the PPARα, PPARγ, and PPARδ LBDs with midpoint melting temperatures (T m ) that were +4 to +17 °C higher than the T m of DMSO control groups (Figure 8A  four strong and essential hydrogen (H)-bonds with the surrounding polar residues of PPARα (S280, Y314, H440, and Y464 45 ) and PPARγ (S289, H323, H449, and Y473 46−48 ) (Figure S2D,E).This is not the case with PPARδ, where CNSL form only one strong H-bond with Y437 and two weaker Hbonds with H413 and H287, 49 and no H-bonding interaction with T253 when we used 3U9Q structure of PPARδ for the docking calculation.These results agree with the thermal stabilization assay, indicating the structural key for the highest PPARα/γ activity/selectivity is the carboxylic group engagement.As a result, we hypothesized that the activity of the CNSL derivatives was highly dependent on the ability to form these four strong H-bonds with their carboxylic functional groups.The long 15 carbon aliphatic chains of the CNSL derivatives fold very well within the hydrophobic pockets.We found unique π−π interactions between the phenyl group of the CNSL derivatives with the histidine side chains located in the hydrophobic pocket of each receptor (H440 in PPARα, H449 in PPARγ and H413 in PPARδ).In addition, the gemdimethyl group of 23 fits well in the hydrophobic pocket, formed by Q277, V444, and L456 in PPARα; Q286, L453, and L465 in PPARγ; and Q250, M417, and L429 in PPARδ (not shown).In contrast, with the anacardic acid derivative 4, the phenyl moiety gains similar π−π stacking interactions with the histidine side chains of the PPAR isoforms.These π−π stabilizing stacking interactions are not possible with the fatty acid ligands that lack the phenyl ring.
It is known that proteins are very dynamic and X-ray crystal structures that represent a static snapshot of a protein, are packed in a crystal, and are frozen at a liquid nitrogen temperature.To investigate the stability of the key H-bonds formed by the polar warhead of 23 we conducted 10 ns molecular dynamics simulations using Desmond (D. E. Shaw Research) using NPT ensembles at 310 K constant temperature and 1 atm constant pressure using the OPLS3 force field.Average distances for the key interactions (H-bond, π−π, and cation−π interactions) as well as the percentage of time they were observed during the simulation were different for the three PPAR isoforms (Figure S3).We observed that all four key H-bonds were observed during 98−99% of the simulation time in PPARα.In the case of PPARγ, only three H-bonds remained very stable, while the H323 H-bond was only formed 88% of the time.The situation was even worse in the case of PPARδ.In this isoform, only the Y437 H-bond was very stable, all other H-bonds were formed and remained stable only 69, 79, and 80% of the time for H287, H413, and T253, respectively (Figure S3).It is worth noting that during the docking calculation compound 23 did not form a H-bond with T253 (Figure S2F).This is due to the fact that the conformation of the T253 side chain in the 3U9Q structure of PPARδ precludes the formation of this H-bond.During the 10 ns molecular dynamics run, the side chain of T253 was able to flip and form the H-bond with one of the carboxylate oxygens of 23 80% of the time.These observations are consistent with the relative potency of 23 against the α-, γand δ-PPAR isoforms of 0.5, 0.9 and 33 μM, respectively.Also, analysis of the dynamic structure of the proteins using fully solvated, long molecular dynamics simulations shed more light on the nature and stability of protein−ligand interactions.In vivo testing of CNSL derivatives using transgenic zebrafish that express human PPARα, PPARγ, or PPARδ reveal tissue-specific activation.Activation of human PPAR in the zebrafish embryo results in GFP expression.Basal activity of PPARα, PPARγ, and PPARδ is observed with vehicle (DMSO) treatment and is strongly increased in the presence of the full agonist for each receptor.Positive controls are WY (WY14643) for PPARα, Rosi (rosiglitazone) for PPARγ, and GW (GW0742) for PPARδ, respectively.Compounds were screened at their respective EC 50 's determined from their dose−response curves in HEK293 cells with a few exceptions: 4 was screened at 1.5 μM for all receptors because of toxicity at higher concentrations; for PPARδ, 20 and 27 were screened below their EC 50 's due to toxicity at higher concentrations.Each image depicts a representative embryo.Note that embryos incubated with 27 were imaged using a different microscope.
In Vitro Metabolic Stability of the CNSL Derivatives and Pharmacokinetic (PK) Profile of 23 in C57Bl/6 Mice.We investigated the metabolic stability of the most active CNSL derivatives in vitro using mouse liver microsomes by measuring the rate of disappearance of the parent compound over time (0, 120 min) (Table 2).The cardanol derivatives 20, 21, 23, and 27 all displayed high stability (t 1/2 > 2 h), whereas the anacardic acid derivative 4 had a short half-life of only 2 min (Table 2).
Next, we performed a pharmacokinetic study of 23 in mice.Compound 23 (LDT409) was selected for further study because of its promising in vitro profile that identified it as a relatively potent (EC 50 < 1 μM), stable (t 1/2(microsomes) > 2 h), dual partial agonist of PPARα and PPARγ and showed selective tissue activation in the in vivo zebrafish models.A single dose of 23 was administered to C57BL/6 male mice either intraperitoneally (40 mg/kg) or orally (100 mg/kg) and plasma was collected for analyte analysis by liquid chromatography/tandem mass spectrometry (LC/MS/MS) (Figure 9).The resulting pharmacokinetic parameters are presented in Table 3.The maximum concentration (C max ) and half-life were 102 ± 12 mg/L and 4.0 h, respectively, after intraperitoneal (IP) administration of 40 mg/kg.After oral administration, the C max was 76 ± 5 mg/L (100 mg/kg dose), indicating a good relative bioavailability (F rel ) against the IP dose of 38% (Table 3).The terminal half-life was calculated to be 2.3 h after oral dosing, and 23 was almost completely eliminated within 24 h (Figure 9B).These data support the potential for chronic daily dosing of 23 in mice as they suggest that such dosing regimens would result in plasma concentrations above the respective PPARα/PPARγ EC 50 's for at least 16 h.

■ CONCLUSIONS
Herein, we report the design and synthesis of novel compounds derived from anacardic acid and cardanol, phenolic lipids that are abundant in CNSL, the waste byproduct of the cashew nut industry.Importantly, these derivatives retained structural similarity to fatty acids that are known to endogenously activate PPARs.When tested against a panel of PPAR receptors in vitro, several compounds were found to be single-, dual-, or pan-PPAR agonists with partial agonist activity and low micromolar potency.Detailed characterization of adipocyte and hepatocyte responses, and in vivo biodistribution studies in zebrafish embryos led to the identification of the lead compound, 23 (LDT409), which is a novel partial pan-PPAR agonist with potent and balanced affinity for PPARα and PPARγ and weak binding affinity to PPARδ.Notably, markers of the beneficial glucose-lowering effects of PPARγ agonists (Rosi) in adipocytes were retained with 23.Overall, based on the desirable in vitro pharmacological activity results and the favorable in vivo pharmacokinetic profile, 23 may represent a sustainable resource from which to generate affordable agents to treat dyslipidemia and type 2 diabetes.
Primary Hepatocytes.Mouse primary hepatocytes were isolated by collagenase perfusion as previously described. 50Cells were plated onto type I collagen-coated plates at 0.5 × 10 6 cells per well for 2 h in attachment media (William's E Media, 10% charcoal-stripped FBS, 1× penicillin/streptomycin, and 10 nM insulin) and then switched to overnight media (M199 media, 5% charcoal-stripped FBS, 1× penicillin/streptomycin, and 1 nM insulin).Ligand treatments were carried out on the following day in M199 media without FBS.The cells were harvested 16 h later for RNA extraction.
3T3-L1 Adipocyte Differentiation Assays.3T3-L1 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum.Adipocyte differentiation was induced in 2day post-confluent cells by treating the cells with 100 μg/mL isobutylmethylxanthine, 1 μM dexamethasone, and 5 μg/mL insulin with 10% FBS in DMEM (day 0).Rosiglitazone and CNSL derivatives were added at the start of differentiation.Two days later, the cells were switched to fresh medium containing 5 μg/mL insulin with 10% FBS.After an additional 72 h, the cells were switched to maintenance media containing 10% FBS in DMEM.Thereafter, the media was changed every 2 days.The cells were harvested on day 11 for RNA expression, and Oil Red O staining was used to estimate lipid accumulation.
Oil Red O Staining.Following 11 days of differentiation, the cells were washed with PBS twice and fixed in 10% neutral buffered formalin (Sigma) for 1 hr at room temperature, followed by two double-distilled H 2 O (ddH 2 O) and two washes with 60% isopropanol.The cells were then stained with 0.6% (w/v) Oil Red O solution (60% isopropanol, 40% water) for 10 min at room temperature, followed by five washes with double-distilled H 2 O to remove unbound dye, and images were taken with a Leica M205 FA microscope.
RNA Isolation, cDNA Synthesis, and Real-Time Quantitative PCR (QPCR) Analysis.Total RNA was extracted from cells using RNA STAT-60 (Tel-Test Inc.), followed by DNase I treatment (RNase-free; Roche), and reverse-transcribed into cDNA with random hexamers using the High Capacity Reverse Transcription System (ABI; Applied Biosystems).QPCR analysis was performed on an ABI 7900 in 384-well plates using 2× SYBR Green PCR Master Mix (ABI).Relative mRNA levels were calculated using the comparative Ct method normalized to 36B4 mRNA for primary adipocytes and cyclophilin mRNA for differentiated 3T3-L1 adipocytes.
Compound Screening in Transgenic Zebrafish Line.Ligands were screened in human PPARα, PPARγ, and PPARδ-expressing zebrafish lines.One day post-fertilization PPARα, PPARγ, and PPARδ heterozygous embryos were heat-shocked at 37 °C for 30 min, dechorionated, and then arrayed.Embryos were maintained in embryo water, including 0.075 g/L NaHCO 3 , 0.018 g/L sea salt, and 0.0084 g/L CaSO 4 •2H 2 O. Before ligand incubation, the embryo water was replaced with fresh embryo water supplemented with CNSL derivatives.The CNSL derivatives were screened at EC 50 's.Embryos were incubated at 28 °C for 14 h in the presence of ligands, anesthetized with Tricaine (Sigma), and imaged for GFP fluorescence using an ImageXpress Velos laser scanning cytometer.
Liver Microsome Stability Assays.Metabolic stability assays were performed by incubating the CNSL derivatives with mouse liver microsomes.Briefly, 1 μL of 1 mM ligands was incubated with NADPH producing system (100 μL of 50 U/mL isocitrate dehydrogenase, 25 μL of 0.1 M MgCl 2 , and 25 μL of 0.1 M DLisocitrate trisodium salt), 324 μL of 0.1 M K + phosphate buffer, and 12.5 μL of 3.3 mg/100 μL NADPH.The reaction was incubated at 37 °C for 5 min before 12.5 μL of 20 mg/mL B6C3F1 mouse microsomes (Thermo Fisher) was added.CNSL derivatives (20, 21, 23, and 27) were incubated at 37 °C for desired time points (0 and 120 min) except 4 that was incubated for indicated time points (0, 2.5, and 5 min).At each time point, the reaction was quenched with 1 mL of methyl-tert-butyl ether (MTBE) and the mixture was vortexed for 3 min at room temperature.Once all of the time points were collected, the samples were centrifuged at 21 130g at room temperature for 3 min.The tubes were put in an ice bath (consisting of dry ice and isopropanol) to freeze the bottom layer.The top layer was poured into a 5 mL brown glass vial and dried under nitrogen gas.Samples were resuspended in 250 μL of HPLC-grade methanol prior to LC/MS/MS analysis.The disappearance of the parent compound (measured by changes in compound peak area/internal standard peak area) was plotted as the natural logarithm with respect to time to obtain the rate of disappearance, k (the slope of the line).Half-life (t 1/2 ) was calculated using the equation ln 2/k.Pharmacokinetic (PK) Study.Wild-type (WT) male C57Bl/6 mice (7−10 months old; n = 4 per two time points) were treated with 23 (LDT409) at 40 mg/kg, IP.The compound was administered in a solution of 5% DMSO, 5% Tween-80, and 90% saline.Four mice were sampled per time point and each animal was sampled twice, first by saphenous vein at 0.08, 0.16, 0.33, and 0.5 h, and next by terminal collection (trunk blood by decapitation) at 1, 2, 4, and 6 h.In a separate cohort of WT male C57Bl/6 mice, 23 was orally administered at 100 mg/kg in peanut butter pellets.Mice were acclimated to the peanut butter pellets (without drug) for 7 consecutive days, after which time the peanut butter pellets were consumed in less than 1 min.Twenty mice were used for the oral PK study with n = 4 per time point.Blood collection was performed by saphenous vein at 0.25, 0.5, 1, 1.5, and 2 h after drug administration, followed by the second terminal collection at 4, 8, 16, 24, and 38 h after administration.Plasma was obtained by centrifugation of blood at 500g at 4 °C for 20 min and stored at −80 °C.Plasma samples (50 μL) and standards were extracted with 250 μL of methanol containing the internal standard 4, vortexed, and centrifuged at 21 130g at 4 °C for 5 min.The supernatant was transferred to autosampler vials for LC-MS/MS analysis.
LC-MS/MS Analysis.Samples from liver microsome stability assays and plasma samples from pharmacokinetics were analyzed by LC-MS/MS using a 6410 Triple Quadrupole instrument (Agilent Technologies, Santa Clara, CA) with electrospray ionization.Positiveion mode was used for compounds 20 and 21, and negative-ion mode for compounds 4, 23, and 27.Samples were separated on a C18 column (Zorbax XDB, 4.6 × 50 mm 2 , 3.5 μm) at a flow rate of 0.4 mL/min.Column temperature was set at 40 °C and injection volume was 10 μL.For positive-ion mode, the mobile phases consisted of HPLC-grade water (A) and MeOH (B), both containing 5 mM ammonium acetate.The following gradient was run: 0−1 min, 90% B; 1−3.For negative-ion mode, the mobile phases consisted of HPLCgrade water (A) and acetonitrile (B), both containing 1% formic acid.The following gradient was run: 0−1 min, 90% B; 1−3.Thermal Shift Assays.An Optim 1000 instrument (Unchained Labs) was used to record static light scattering signals during a temperature ramp using laser excited light at a wavelength of 473 nm.Changes in light scattering intensity reflect changes in aggregate size.Samples (9 μL) at a protein concentration of 1 mg/mL (∼25 μM) were heated in 0.5 °C increments from 25 to 70 °C.The heating rate between temperature intervals was set to 1 °C/min.A typical temperature scan took two and a half minute including 30 s for thermal equilibration.All measurements were done in duplicate.
The protein structures were prepared using the Protein Prep wizard in Maestro by assigning bond orders using CCD database, adding hydrogens and filling in missing side chains using Prime.For side chains with multiple occupancies, the highest occupancy state was chosen.Protonation states for the co-crystallized ligands were generated at pH = 7 ± 2. The orientation of the water molecules; side chains of glutamines and asparagines, tautomers, and protonation states of histidines were sampled; and the hydrogens of the altered species were minimized.The protonation states of the protein side chains were assigned with PropKa at pH 7. Waters with less than three H-bonds to nonwaters were deleted.At the end, the protein underwent a hydrogen-only energy minimization using the OPLS3 force field.
During the protein GRID preparation, keeping in mind the largely elongated binding site, the binding site box was elongated in one direction to accommodate ligands with length up to 25 Å.Terminal hydroxyls and thioalcohols of C275, C276, T279, S280, Y314, and Y464 in PPARα; C285, S289, Y327, and Y473 in PPARγ; and C249, T252, T253, and Y437 in PPARδ were kept rotatable during the docking process.Possible H-bond constraints to side chains of Y314, H440, and Y464 in PPARα; H323, H449, Y473, and S289 in PPARγ; and H287 and H413 in PPARδ were set up but not used during the docking stage.
All ligands were prepared with LigPrep procedure using default settings, ionized to pH 7 ± 2, and energy-minimized using the OPLS3 force field.
The ligand docking was done using both Glide SP and XP scoring functions with flexible ligand sampling, and no constraints used, with 100 poses included for the post-docking minimization.Ten poses for each ligand were saved and viewed with Pose Viewer to allow visualization of flexible hydroxyls and thioalcohols.
Molecular Dynamics Simulations.To investigate stability of the key H-bonds formed by the polar warhead of 23, we conducted molecular dynamics simulations using Desmond (D. E. Shaw Research, a part of the Schrodinger's Drug Discovery suite) using NPT ensembles at 310 K constant temperature and 1 atm constant pressure using the OPLS3 force field.A simulated box of 10 Å 3 was set up around the protein and filled with water molecules (SPC water model).Excessive charges were neutralized to make the total net formal charge of the system equal to 0, and 0.15 M NaCl was added to the buffer.The Nose−Hoover chain thermostat was used to keep the temperature constant, and the Martyna−Tobias−Klein barostat was used to keep the system pressure constant.A nonbonding cutoff of 9.0 Å was used for Coulombic interactions, which is the electrostatic interactions between electric.A 2 fs time step was used, and a trajectory frame was saved every 5 ps.The total simulation time was 10 ns for each of the PPAR isoforms.Distances for key Hbonding, π−π, and cation−π interactions were tracked during the simulations and analyzed afterward.
Statistical Analysis.Data handling, analysis, and graphical representations were performed using GraphPad 8.0 software (GraphPad, San Diego, CA).Statistical differences were determined by one-way analysis of variance followed by Holm−S ̌idaḱ; P < 0.05 was accepted as statistically significant.

Figure 2 .
Figure 2. Similarity of chemical structures of stearic acid and saturated anacardic acid.
−C).These shifts were similar or greater than what we observed with each receptor's positive control: GW7647, rosiglitazone, and CAY10512 had ΔT m values of +5.2, +5, and +7.7 °C, respectively (Figure 8A−C).In contrast, a negative thermal shift was observed when 20 was incubated with the LBDs of PPARα and PPARγ (Figure 8A,B), whereas, 21 had no effect on the T m values for all three PPARs.These data suggest 20 and 21 do not directly interact with the ligand-binding pockets.Given that both 20 and 21 are ethyl esters, they would require de-esterification prior to binding the PPAR LBDs, a process that could occur only with intact cells.These results suggest that 4, 23, and 27 directly interact and stabilize the ligandbinding domains of PPARα, PPARγ, and PPARδ.In Silico Docking of 23 to PPARα, PPARγ, and PPARδ Uncovers Key H-Bonding and Unique Hydrophobic π−π Interactions Not Present with Endogenous Fatty Acids.Based on our extensive docking calculations of compounds 3, 4, and 20−23 with Glide SP and XP scoring functions, and Induced-Fit Docking (IFD) protocol, which produced similar results, we expected the carboxylic group of CNSL to bind within the ligand-binding domain of PPAR receptors (Figure S2A−C, with compound 23 bound), forming

Figure 6 .
Figure 6.CNSL derivatives differentially regulate the expression of PPARγ target genes and adipocyte differentiation in 3T3-L1 cells.3T3-L1 fibroblasts were differentiated for 11 days in the presence of vehicle (DMSO), 25 μM of indicated CNSL derivatives or 10 μM rosiglitazone (Rosi).Cells were harvested for Oil Red O staining and mRNA on day 11.(A) Cells were imaged under 10× magnification (n = 2/per group).(B) Lipid accumulation was quantitated by spectrophotometric analysis of extracted Oil Red O. mRNA expression was analyzed by QPCR for two key regulators of adipogenesis, (C) Pparγ and (D) Cebpα; fatty acid uptake genes (E) aP2 (Fabp4), (F) Lpl, and (G) Cd36; adipose-specific adipokine gene (H) AdipoQ and glucose uptake gene (I) Glut4.Vehicle mRNA expression was set to 1 and Rosi value was set to 100%.Data represent mean ± SD (N = 3).*P < 0.05 vs vehicle and # P < 0.05 vs Rosi; using one-way ANOVA with Holm−S ̌idaḱ correction.

Figure 7 .
Figure 7.In vivo testing of CNSL derivatives using transgenic zebrafish that express human PPARα, PPARγ, or PPARδ reveal tissue-specific activation.Activation of human PPAR in the zebrafish embryo results in GFP expression.Basal activity of PPARα, PPARγ, and PPARδ is observed with vehicle (DMSO) treatment and is strongly increased in the presence of the full agonist for each receptor.Positive controls are WY (WY14643) for PPARα, Rosi (rosiglitazone) for PPARγ, and GW (GW0742) for PPARδ, respectively.Compounds were screened at their respective EC 50 's determined from their dose−response curves in HEK293 cells with a few exceptions: 4 was screened at 1.5 μM for all receptors because of toxicity at higher concentrations; for PPARδ, 20 and 27 were screened below their EC 50 's due to toxicity at higher concentrations.Each image depicts a representative embryo.Note that embryos incubated with 27 were imaged using a different microscope.

Figure 9 .
Figure 9.In vivo pharmacokinetic profile of 23 (LDT409) in C57BL/6 mice.The plasma concentration of 23 (LDT409) in mice after (A) a single intraperitoneal injection (IP) at 40 mg/kg and (B) oral administration in peanut butter treat at 100 mg/kg.Data represent mean ± SEM (N = 4 per time point).

Table 1 .
Activity of CNSL Derivatives for Human PPARs In Vitro a,b a EC 50 values were obtained using reporter gene assays.b E max 5−29% of the corresponding positive control.c E max 30−89% of the corresponding positive control.d E max ≥ 90% of the corresponding positive control.e Did not reach saturation at the highest concentration tested.f −: Not active.g Cellular toxicity.

Table 2 .
Mouse Liver Stability of Selected CNSL Derivatives In Vitro a Results are expressed as mean ± SD (N = 3).b To assess the rate of disappearance of 4, time points of 2.5 and 5 min were used.
a an autosampler (SIL-10AF), and a controller communication bus module (CBM-20A)].The column was a reversed-phase octadecylsilyl column (4.6 μm, 4.6 × 250 mm 2 , Shim-pack VP-ODS).The mobile phase was HPLC-grade acetonitrile containing 0.1% (v/v) trifluoroacetic acid (eluent A) at a flow rate of 1 mL/min.Samples were diluted in eluent A at a concentration of 1 mg/mL.Using the autosampler, 80 μL of each sample was injected and the absorbance at the 200−400 range was monitored for 40 min.

Table 3 .
Pharmacokinetic Parameters of 23 (LDT409) in Mice after a Single IP and Oral Administration SEM.AUC 0−∞ , area under the concentration−time curve from zero to infinity; C max , maximal concentration; C min , minimum concentration; CL, clearance; F, bioavailability; F rel , relative bioavailability (oral vs IP); k e , elimination rate constant; t 1/2 , half-life; t max , time of maximum concentration observed.

■
ASSOCIATED CONTENT * sı Supporting Information The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jmedchem.1c01542.Molecular formula strings (CSV) QPCR primer sequences; dose−response curves of PPAR-active CNSL derivatives demonstrate many dual PPARα/PPARγ agonists with low micromolar affinities; docking of 23 with Glide SP scoring function in PPAR isoforms: general view; statistical analysis of key interactions of compound 23 with the three PPAR isoforms; 1 H NMR, and 13 C NMR spectra; HRMS spectra; compound purity and HPLC traces (PDF) Docking models for 23 (LDT409) with PPARα (PDB) Docking models for 23 (LDT409) with PPARδ (PDB) Docking models for 23 (LDT409) with PPARγ (PDB) Corresponding AuthorsCarolyn L. Cummins − Department of Pharmaceutical Sciences, Leslie Dan Faculty of Pharmacy, University of Toronto, Toronto, Ontario M5S 3M2, Canada;