Exploration of Long-Chain Vitamin E Metabolites for the Discovery of a Highly Potent, Orally Effective, and Metabolically Stable 5-LOX Inhibitor that Limits Inflammation

Endogenous long-chain metabolites of vitamin E (LCMs) mediate immune functions by targeting 5-lipoxygenase (5-LOX) and increasing the systemic concentrations of resolvin E3, a specialized proresolving lipid mediator. SAR studies on semisynthesized analogues highlight α-amplexichromanol (27a), which allosterically inhibits 5-LOX, being considerably more potent than endogenous LCMs in human primary immune cells and blood. Other enzymes within lipid mediator biosynthesis were not substantially inhibited, except for microsomal prostaglandin E2 synthase-1. Compound 27a is metabolized by sulfation and β-oxidation in human liver-on-chips and exhibits superior metabolic stability in mice over LCMs. Pharmacokinetic studies show distribution of 27a from plasma to the inflamed peritoneal cavity and lung. In parallel, 5-LOX-derived leukotriene levels decrease, and the inflammatory reaction is suppressed in reconstructed human epidermis, murine peritonitis, and experimental asthma in mice. Our study highlights 27a as an orally active, LCM-inspired drug candidate that limits inflammation with superior potency and metabolic stability to the endogenous lead.


■ INTRODUCTION
−8 LCMs are produced from αtocopherol (1a) and other vitamin E forms (1b−d, 6a−d) by hepatic ω-oxidation, yielding ω-alcohols and then ωcarboxylic acids, which are excreted via bile and feces, shortened by successive β-oxidations, or conjugated with sulfate or glucoronate for urinary elimination. 7,8The LCMs α-T-13′-CH 2 OH (9a) and α-T-13′-COOH (12a) were detected at low nanomolar concentrations in human plasma, albeit with strong variation between individuals. 6,8,9These differences in 1a metabolism may provide an explanation for the mixed outcomes of human vitamin E intervention studies 8,10,11 and open the door toward personalized pharmacotherapy.Notably, LCMs reach the highest concentration in the liver, which correlates with the recently confirmed clinical efficiency of 1a in nonalcoholic fatty liver disease (NAFLD). 12e have shown that 12a accumulates within immune cells at sites of inflammation, such as the inflamed peritoneal cavity of mice, limits the inflammatory reaction in murine peritonitis, and suppresses bronchial hyperreactivity in experimental asthma by targeting 5-lipoxygenase (5-LOX). 6LCMs bind to an allosteric site between the 5-LOX catalytic and regulatory domains and inhibit the enzyme at concentrations that are reached in plasma for 12a. 6-LOX initiates the biosynthesis of powerful immunomodulatory lipid mediators from polyunsaturated fatty acids that are released from membrane phospholipids by cytosolic phospholipase (cPL)A 2 . 135-Lipoxygenase-activating protein (FLAP) transfers arachidonic acid to 5-LOX at the nuclear membrane, where leukotriene (LT)A 4 is synthesized via 5hydro(pero)xyeicosatetraenoic acid (5-H(P)ETE) as an intermediate. 14−21 5-LOX is also involved in the biosynthesis of lipoxins, resolvins, and other specialized proresolving mediators that orchestrate resolution, pathogen clearance, and tissue regeneration, 22,23 but the impact of 5-LOX in this process differs between immune cell populations. 14,24Remarkably, 12a strongly increased systemic resolvin E3 levels in mice during the resolution phase, whereas the 5-LOX inhibitor zileuton, which is in clinical use for asthma therapy, did not show a comparable effect. 6These findings suggest that 12a, besides suppressing acute inflammation, promotes resolution, which would be a major advantage in treating chronic inflammation. 22dditional targets besides 5-LOX were proposed to contribute to the anti-inflammatory effectiveness of LCMs.Garcinoic acid (δ-TE-13′-COOH, 13d), a potential LCM with unknown physiological relevance in humans, has agonistic activity on specific nuclear receptors, such as pregnane X receptor and peroxisome proliferator-activated receptor γ, a mechanism that is partially shared by other LCMs. 25,26Of note, nuclear receptor activation by compound 13d has been proposed to diminish Alzheimer's disease progression by interfering with β-amyloid oligomerization and deposition. 27urther immunomodulatory targets were reported for LCMs at supraphysiological concentrations.−30 Moreover, the 13′-carboxylic acid of δtocopherol (12b), but not α-tocopherol (12a), reduces enzymatic COX-2 activity, 6,31 and tocotrienol-derived 13′carboxylic acids (13a−d) inhibit microsomal prostaglandin E 2 synthase (mPGES)-1, 32 an inducible enzyme that is functionally coupled to COX-2 and responsible for excessive prostaglandin (PG)E 2 formation during inflammation. 33,34he in vivo relevance of COX-2 and mPGES-1 inhibition is unclear, 35 but their moderate inhibition might be beneficial to buffer substrate redirection from LT to PG biosynthesis. 36he recent insights into bioactive vitamin E metabolites promise access to a new generation of 5-LOX-targeting drugs that suppress inflammation without impairing but instead triggering resolution.LCMs potently inhibit 5-LOX, beneficially adjust lipid mediator profiles, and are enriched in immune cells at inflammatory sites. 6As endogenous metabolites, they might be less afflicted with toxicity than zileuton and diverse clinical candidates targeting 5-LOX. 37hallenges for drug development were, however, the limited knowledge about pharmacophores, the elusive oral availability, and the rapid hepatic LCM metabolism.We here explored the structural requirements for 5-LOX inhibition, taking differences in cellular uptake into account, and identified αamplexichromanol (α-TE-12a′,13′-diCH 2 OH, 27a), a highly potent allosteric 5-LOX inhibitor, which combines a favorable pharmacological profile with oral availability and superior metabolic stability.Compound 27a accumulates in immune cells, is efficiently distributed to inflamed regions, and shows anti-inflammatory effectiveness in experimental models of atopic dermatitis in vitro and murine peritonitis and bronchial hyperreactivity in vivo.

■ RESULTS
Design and Semisynthesis of Chromanols Inspired from Bioactive Vitamin E Metabolites.Compound 13d was extracted and purified from Garcinia kola nuts according to a previously described method with an optimized yield compared to the literature. 39,40Chromanols from the amplexichromanol series (10e, 27c, 27d) were isolated from Garcinia amplexicaulis stem barks. 41Both αand β-forms (13a, Journal of Medicinal Chemistry 13b, 27a, 27b) were obtained from the corresponding δ-forms of the garcinoic acid (13d) and amplexichromanol (27d) series through a two-step strategy, which involves the preparation of the mono-or bis-Mannich bases and their reduction with sodium cyanoborohydride. 42,43δ-(Z)-Garcinoic acid 13e has been previously isolated from Clusia grandiflora, more specifically from fruits, and subsequently characterized. 44In the current study, it was prepared from the corresponding alcohol 10e extracted from the stem barks of this plant and also found in G. amplexicaulis (Scheme 1).Two oxidative steps, namely, 2-iodoxybenzoic acid-mediated and sodium chloritemediated, were applied to the tosyl-protected precursor 49.Final hydrolysis led to 13e with 51% overall yield over four steps.

Journal of Medicinal Chemistry
The third aforementioned alkene reagent, 2-methylene-1,3propanediacetate, helped prepare amplexichromanol analogues.Eventually, 14, 16−18, 28, and 29 were synthesized in our group through a cross-metathesis approach 56 similar to the one recently reported by Gujarathi et al. 57 Tosyl ester was used as a protecting group of the phenol function rather than a silylated ether.This choice initially aimed at achieving one final deprotection step of all protecting groups.Practically such a strategy was successfully applied to the syntheses of 28 and 29 with the parallel removal of acetates and tosylate groups of 62 and 63.On the other hand, a two-step deprotection sequence was employed to access 14, 16, and 17.
Mannich bases, a chemical class with a wide structural variety, exhibit a broad spectrum of biological activities, including antitumoral, anti-inflammatory, antimicrobial, and antiviral properties. 58,59Besides obtaining structural insights from docking studies, we considered optimizing the physicochemical properties of pharmacologically relevant candidates.Water solubility may be further enhanced by adding a protonable group.Therefore, aminomethylation through the Mannich reaction was explored.This strategy has already been applied to 6d to provide C5-aminomethylated analogues with an antitumoral potential. 46In the current study, several mono-and bis-Mannich bases have been prepared either in the garcinoic acid (23, 32 24 6 ) or in the amplexichromanol (35−40) series following classical experimental conditions (Scheme 5).
Alkylation of the phenol function was potentially associated with a loss of 5-LOX binding affinity for the corresponding tocotrienolic ethers. 6Based on reported synthesis methods for the development of redox-silent antitumoral vitamin E analogues, two ethers (32, 45) 60 and one ester (31) 61 bearing both a hydrogen-bond donor and acceptor with different linker lengths were semisynthesized and evaluated (Scheme 5).
We have shown that immune cells strongly accumulate LCMs, with the endogenous vitamin E metabolite 12a being more efficiently enriched than 13d. 6As a consequence, 12a suppressed 5-LOX product biosynthesis in PMNL better than 13d, which was the more potent enzymatic 5-LOX inhibitor (Table 1, Schemes S1 and S2).This gain of potency in PMNL was limited to side-chain saturated derivatives and not evident for the respective α-garcinoic acid 13a.Together, substantial differences in the SARs of 5-LOX inhibition exist between cellfree and cell-based assays.The challenge was to enhance 5-LOX inhibition while maintaining an efficient uptake of the vitamin E-inspired compounds into immune cells.
Starting from 13a−d as promising leads, we focused on the length and functionalization of the side chain and explored options for chromanol substitution.Both inhibition of cell-free 5-LOX activity and the biosynthesis of 5-LOX products in activated PMNL were addressed (Table 1, Schemes S1 and S2).In a first step, we varied the side-chain length (14, 15a,  15b, 16) and found that inhibition of 5-LOX activity consistently decreased with shorter chain length within the δ-series (14, 15b, 16), both in cell-free and in cell-based assays.Side-chain shortening of the α-derivative 13a (from C13 to C11) yielding 15a was instead tolerated.We then addressed side-chain methylation, saturation, and chain length and found that 5-LOX inhibition was impaired for the desmethyl analogue at the C12 position (14) as well as by saturation of the Δ11′,12′ double bond (19a, 19b) and absent when truncated to C5 side chains (17, 18).Together, the natural LCM 13d excels in 5-LOX inhibition within this series of sidechain scaffold modifications.

Journal of Medicinal Chemistry
IC 50 = 0.11 μM) better than ω-carboxylic acids (13a−c), except for the δ-derivative 27d, which was less potent than 13d (Tables 1 and 2, Schemes S1 and S2).We hypothesize that the methylation pattern of the chromanol core (as found in natural vitamin E forms) impacts cellular uptake and thus 5-LOX inhibition.In support of this hypothesis, the inhibitory activity of α-amplexichromanol 27a further increased in PMNL (IC 50 = 0.04 μM), thereby exceeding the potency of the endogenous metabolite 12a (IC 50 = 0.08 μM). 6The βand γ-analogues 27b and 27c maintained their activity in the cell-based assay, whereas δ-amplexichromanol 27d was less effective, which again highlights the strong influence of chromanol substitution on cell-based 5-LOX product formation.Shortening to C9 (28) or C5 (29) side chains was detrimental, especially in the cell-based assay, and also side-chain saturation of 27d yielding 26 was not tolerated.Ring closure of the 12a′,13′-diol to substituted 1,3-dioxans (44−48) was explored because we previously found that bulky substituents, even without the hydrogen bridge donor, are tolerated in the ω-position. 6,42ome of the obtained 1,3-dioxanes potently inhibited cell-free 5-LOX although they were less effective in suppressing 5-LOX product formation in PMNL than 27a.Together, compound 27a carries an optimized side chain that qualifies for potent 5-LOX inhibition in cell-free and cell-based assays, thereby fulfilling the structural requirements for efficient 5-LOX binding and cellular accumulation.
We concentrated our further efforts on the chromanol moiety.Molecular docking studies suggest that the chromanol hydroxyl group is critical for the interaction with 5-LOX. 6In fact, O-methylation (20, 22) abolished 5-LOX inhibition by the ω-carboxylic acids 19b and 21, and ring closure as coumarin (a scaffold found in potent 5-LOX inhibitors 62 ) under formation of an annulated 2-oxobenzopyran (41) was also disadvantageous (Tables 1 and 2, Schemes S1 and S2).Accordingly, replacement of the phenolic alcohol by an aminocarbonylmethoxy group (32) to increase the distance between the chromanol core and the hydrogen-bond donor/ acceptor function was associated with a loss of inhibitory potency.On the other hand, 5-LOX inhibition was slightly enhanced in the cell-free assay by introducing a free carboxylic acid group as the O-succinyl ester (31).In activated PMNL, 31 inhibited 5-LOX comparably to 27d, potentially because the ester in 31 is intracellularly cleaved to the free hydroxyl group of 27d.
We then investigated whether also oxidative modifications at the chromanol core lead to potent 5-LOX inhibitors, as yielded by ω-oxidation of the unsaturated side chain.However, neither oxidation of the C5 or C7 methyl groups to an aldehyde (3, 4,  7) or carboxylic acid (5) nor additional halogenation (2, 8)  had any improving effect (Table S1, Schemes S1 and S2).The replacement of the C5 or C7 methyl groups by an aldehyde (33, 34) or hydroxymethyl moiety (46, 48) was, on the other hand, compatible with potent 5-LOX inhibition when

Journal of Medicinal Chemistry
combined with ω-oxidation of the side chain.Motivated by this finding, we explored a variety of structurally diverse substituents.Introduction of a methyl ester residue at the C5 methyl group of 13d in combination with methylation of the 13′-carboxylic function was detrimental (25), whereas basic tertiary amines and N-heterocycles yielded inhibitors of cellfree 5-LOX that were comparably potent to the nonsubstituted derivative (for pyrrolidino, 35; for dimethylamino, 23, 36; for morpholino, 37; or for piperidinyl substitution, 38).An additional more distant nitrogen atom (N-methyl-piperazinyl, 39) was detrimental, and substitution of both C5 and C7 positions with dialkylated aminomethyl groups (24, 40) was not tolerated.While substituents at C5 and C7 of chromanol provide limited options to improve 5-LOX inhibition, all of these modifications substantially impaired the suppression of 5-LOX product biosynthesis in PMNL (Tables 1 and 2, Schemes S1 and S2), and we conclude that the natural chromanol moiety of vitamin E already represents an evolutionally optimized compromise between potent 5-LOX inhibition and access to the enzyme in innate immune cells.The 12a′,13′-dihydroxylated α-tocotrienol 27a combines these features and was selected as a promising drug candidate for further pharmacological studies.Figures 1 and S1 summarize the SARs for 5-LOX inhibition in a network that was calculated based on the structural similarity of the derivatives, with the size of the symbols indicating their potency in cell-free (Figure S1) and cell-based assays (Figure 1).
Mechanism of 5-LOX Inhibition and Binding Mode.Compound 27a inhibited human recombinant 5-LOX independent of the substrate concentration (Figure 2A) in a reversible manner (Figure 2B).Nonspecific inhibition through detergent-sensitive colloidlike aggregates was excluded by supplementing the detergent Triton X-100, which did not substantially impair 5-LOX inhibition by 27a (Figure 2C).Moreover, the concentrations of 27a that exhibit antioxidative properties exceed those that effectively inhibit 5-LOX, as measured by radical scavenging (Figure 2D).
We have recently shown for 13d that it binds 5-LOX close to Trp102, Trp13, Trp75, and Arg101 at the interface of the catalytic and regulatory C2-like domain. 6To investigate whether 27a targets the same site, we monitored the intrinsic Trp fluorescence of human 5-LOX.The spectral center of mass was substantially shifted to a shorter wavelength by both 13d and the endogenous vitamin E metabolite 12a (Figures 2E,  S3A, and S3B), which indicates an altered chemical environment of Trp following ligand binding. 63This trend was also evident for 27a, but the wavelength shift was less pronounced (Figure 2E).We conclude that the 5-LOX binding pose of 27a is similar but not identical to the ω-carboxylic acids 12a and 13d.Site-directed mutagenesis studies strengthen our hypothesis.The inhibitory activity of 13d is sensitive to Trp102Ala/ Trp13Ala/Trp75Ala triple mutation and Arg101Asp replacement, 6 whereas these mutations failed to affect 5-LOX inhibition by 27a (Figure 2F).
Molecular docking studies confirm results from mutation experiments and point out poses where 13d interacts with NH of Trp102 (H-bond with phenolic oxygen) and with Arg101 through an ionic interaction involving the ω-carboxylic acid function of 13d (Figure S2A).Binding of 27d within the 5-LOX allosteric site similarly involves an interaction of the phenol function with Trp102.However, at the tail of the side chain, the binding with Arg101 of the less polar allylic diol would be weaker than the one described above for the carboxylic acid group.Thus, the allylic diol exhibits multiple stabilizing interactions with various other amino acids, such as Val110 and Glu134 (Figure S2B).Site-directed mutagenesis studies showed that the 5-LOX binding affinity of 13d depends on interactions with Trp102 and Arg101. 6Docking poses for 27a suggest differences in the binding mode compared to the δ-form 27d (Figures 2G and S2B).Despite the substitution of the chromanol by three methyl groups, the heterocycle of 27a, as for 13d and 27d, still lies in the vicinity of Trp102 albeit with a shift downward.Consequently, the phenol function does not interact with Trp102 but it exhibits two hydrogen bonds with Asp170.This result supports the less pronounced wavelength shift observed in the fluorescence emission spectra (Figure 2E).Results from the docking studies tend to demonstrate that interactions of the phenol function are more efficient than the ones from the diol moiety to anchor amplexichromanols 27a and 27d in a close range of Trp102.Eventually, as highlighted by 5-LOX site-directed mutagenesis experiments (Figure 2F), a direct interaction with Trp102 is not fully required as long as the ligand strongly interacts with other spatially close amino acid residues, such as Asp170, through hydrogen bonds.
Acute cytotoxic effects were excluded at effective concentrations of 27a that inhibit 5-LOX (≤1 μM).Compound 27a impaired neither mitochondrial dehydrogenase activity in peripheral blood mononuclear cells (PBMCs) (Figure S4A), an indicator of cell viability, nor membrane intactness in monocytes within 2 or 24 h (Figure S4B).At higher concentrations, 27a induced a rapid release of lactate dehydrogenase (LDH), which was not evident for the endogenous vitamin E metabolite 12a (Figure S4B).
Our data point toward a direct inhibition of 5-LOX in immune cells.On the one hand, the enzyme was potently inhibited in PMNL by 27a in the presence of exogenous arachidonic acid (IC 50 = 0.04 μM), which uncouples LT formation from the cPLA 2 -triggered release of arachidonic acid and its transfer to 5-LOX by FLAP (Figure 3A).On the other hand, 27a neither reduced the availability of free arachidonic acid in E. coli-stimulated macrophages (Figure 3D) nor influenced intracellular levels of Ca 2+ , an essential cofactor for cPLA 2 α and 5-LOX activation, at submicromolar concentrations (≤1 μM) (Figure 3E).At higher concentrations of 27a (10 μM), Ca 2+ influx was substantially reduced, as previously reported for 13d 6 and other ω-carboxylic acids. 64ipophilic carboxylic acids like 12a are often afflicted with strong plasma protein binding, leading to a loss of inhibitory activity in tissues and blood. 65In fact, the dialcohol 27a suppressed 5-LOX product biosynthesis superior to 12a in activated human blood, when treated with either A23187 (Figure 3F) or the physiological stimuli LPS and N-formylmethionyl-leucyl-phenylalanine (fMLP) (Figure 3G).Together, compound 27a targets 5-LOX and potently inhibits LT biosynthesis in innate immune cells and blood, thereby surpassing the endogenous vitamin E metabolite 12a.
To investigate the consequences on the lipid mediator network, we stimulated human macrophages (M1 subtype) with pathogenic E. coli and performed a comprehensive metabololipidomics analysis.At concentrations that do not effectively inhibit mPGES-1 (1 μM), 27a preferentially suppressed the biosynthesis of 5-LOX-derived products, i.e., LTB 4 , its isomers, 5-HETE, 5,15-diHETE, 5-hydroxyeicosapentaenoic acid (HEPE), and 7-hydroxydocosahexaenoic acid (HDHA) (Figure 4H).On the other hand, PGE 2 levels slightly increased, which we ascribed to the 5-LOX substrate arachidonic acid being channeled into PGE 2 biosynthesis.Thus, our data indicate that the impact of 27a on the lipid mediator profile is mainly determined by 5-LOX inhibition, as shown for macrophages (Figure 4H) and confirmed in monocytes (Figure S5).
Superior Metabolic Stability against Side-Chain Truncation.We studied the hepatic metabolism of 12a and 27a using a human liver-on-chip. 6The biochip resembles a liver sinusoid and consists of a hepatic and vascular compartment, which are separated by two porous membranes that serve as cell scaffolds (Figure 5A).For the hepatic compartment, we selected HepaRG cells, which are considered to be the closest to human primary hepatocytes among liver cell lines, and differentiated them into hepatocyte-like and biliary-like cells.The endothelial layer mimics the endothelial lining of the liver sinusoid and consists of human umbilical vein endothelial cells (HUVECs).
The metabolism of 12a and 27a substantially differs.While 12a is truncated to metabolites with varying side-chain lengths (Figure 5B), 27a is preferentially conjugated with sulfate or oxidized at 12a′ or 13′ to the respective ω-carboxylic acid (Figure 5C).Truncation of the side chain was instead negligible.Despite being a substrate for hepatic ω-oxidation, 27a did not induce CYP 450 enzyme expression in HepaRG cells, as exemplarily studied for CYP3A4 (Figure 5D).
Next, we investigated the metabolism of 27a after oral administration in mice.Compound 27a was taken up within 1 h, reaching plasma concentrations (0.03−0.9 μM) that effectively inhibit 5-LOX in vitro (Figure 5E).We confirmed that the unsaturated chain of 27a is resistant to degradation  S2.

Journal of Medicinal Chemistry
but did not detect sulfate conjugates (Figure 5E).We conclude that 27a is less susceptible to side-chain degradation than the endogenous metabolite 12a and speculate about speciesspecific differences in 27a sulfation between mice and humans.
Improved Epidermal Homeostasis in Experimental Atopic Dermatitis.LT levels are substantially increased in relapsing inflammatory skin lesions in atopic dermatitis, as are other lipid mediators and cytokines. 67To study the effect of 27a on the inflammatory processes leading to skin dysfunction, we used reconstructed human epidermis (RHE) and induced cytokine stress.Interleukin (IL)-8, an indicator of the severity of inflammation in atopic dermatitis, 68 was upregulated (Figure 6A), as was thymic stromal lymphopoietin (TSLP) (Figure S6A), which is an IL-7-like cytokine that propagates skin lesions by mediating the recruitment and polarization of Th2type CD4 + cells. 69Morphological and functional changes in the epidermal structure are characterized by spongiosis and the loss of keratohyalin granules (Figures 6B and S6B).The latter is associated with a diminished expression of filaggrin, one of the crucial proteins to maintain skin barrier function. 70Along these lines, cytokine stress triggers transepidermal water loss (Figure 6C) and allows the fluorescent dye Lucifer yellow to pass the stratum corneum (Figures 6D and S6C).
Moreover, the pH increased by trend (Figure 6E), which is believed to hamper skin barrier homeostasis, thereby interfering with the innate immune defense in atopic dermatitis. 71Dexamethasone, used as a control, suppressed these cytokine-induced responses, as expected.
Compound 27a reduced IL-8 levels (Figure 6A), attenuated the epidermal disorganization induced by cytokine treatment (Figures 6B and S6B), and substantially improved skin barrier functions (Figure 6C, D and S6C).The pH was decreased even below the basal level (Figure 6E).Thus, compound 27a limits the inflammatory reaction and supports epidermal homeostasis in experimental atopic dermatitis.On the other hand, 27a did not influence TSLP expression in the RHE (Figure S6A), rather excluding a major impact on Th2mediated tissue damage.Cytotoxic effects of 27a on human keratinocytes were excluded under our experimental conditions (Figure S6D).
Attenuated Inflammation in Murine Peritonitis.The anti-inflammatory efficacy of 27a was investigated for zymosan-induced murine peritonitis in vivo, which represents a model of acute inflammation that relies on LTs and other lipid mediators (Figure 7A). 72During the onset of inflammation, zymosan activates resident peritoneal macrophages that produce LTC 4 .This cysteinyl-LT increases vasopermeability in postcapillary venules.Compound 27a, given i.p., strongly decreased LTC 4 levels in the exudate (Figure 7B) and consequently reduced vascular permeability (Figure 7C).The progressive phase of inflammation is instead dominated by infiltrated neutrophils that generate the potent chemoattractant LTB 4 .Compound 27a was also active at this stage and substantially lowered LTB 4 levels, both in the exudate (Figure 7D) and by trend in plasma (Figure S7A).The influx of cells into the peritoneal cavity (Figure 7E), which is dominated by neutrophil infiltration, 72 was accordingly reduced.

Journal of Medicinal Chemistry
trations of 27a in the peritoneal exudate were considerably lower (0.5−4 nM) (Figure 7G) but likely still in a range that inhibits 5-LOX, given the strong intracellular accumulation of 27a in immune cells in vitro (Figure 3B).Cysteinyl-LTs were decreased to a basal level by 27a at a dose of 3 mg/kg (Figure 7H).The clinically used 5-LOX inhibitor zileuton (at 3 mg/ kg) did not evoke a stronger effect.Metabololipidomics profiling showed a preferential drop of arachidonic acidderived 5-LOX products in exudates from 27a-treated mice and a moderate reduction of various other LOX-derived metabolites (Figure 7I).The latter likely depends on the reduced availability of free polyunsaturated fatty acids.Conclusively, 27a is an orally active 5-LOX inhibitor that potently suppresses LT formation and associated peritoneal inflammation in vivo.
We have recently shown that the endogenous metabolite 12a augments the systemic concentration of resolvin E3, a specialized proresolving lipid mediator that actively terminates the resolution of inflammation. 6,22Although our study focused on acute inflammation and has only limited predictive power for the resolution phase, we found resolvin E3 levels already being moderately increased by 27a at 30 min post zymosan injection in individual mice (Figure S7B).It is tempting to speculate that 27a shares the putative proresolving potential of the endogenous lead compound.

Diminished Airway Hyperreactivity in Experimental
Asthma.Since LTs play a pivotal role in the pathogenesis of asthma by propagating lung inflammation, immune cell infiltration, and bronchoconstriction, 15,73 we investigated the effects of 27a in an experimental model of asthma.Mice were treated with 27a p.o. 60 min before being sensitized with ovalbumin on days 0 and 7 and monitored up to 21 days (Figure 8A).Systemic concentrations of 27a peaked at the days after administration (days 1 and 8) and then rapidly declined again in the plasma and lung (Figure 8B).Gavage of 27a lowered lung levels of LTB 4 and its isomers (Figure 8C), blocked ovalbumin-induced bronchial hyperreactivity to carbachol (Figure 8D), and fully restored the adrenergic bronchial relaxation induced by salbutamol (Figure 8E).Together, compound 27a is orally available and distributed to the lung, thereby effectively suppressing pulmonary LT levels and asthmatic airway contraction.

■ DISCUSSION AND CONCLUSIONS
Inspired by the endogenous vitamin E metabolite 12a, which may mediate immunomodulatory effects of vitamin E, 6,7 we designed a novel class of potent 5-LOX inhibitors that limit inflammation.We previously explored 13′-garcinamides, which impress through their potent 5-LOX inhibitory activity, 42 but were substantially less active in innate immune cells.Extended SAR studies revealed oxidative ω-modifications of the side Journal of Medicinal Chemistry chain as a superior strategy toward potent 5-LOX inhibitors that maintain their privileged access to innate immune cells.These two criteria were best realized by the C12a′-/C13′dihydroxylated α-tocotrienol 27a, which inhibits 5-LOX in a substrate-independent manner at concentrations that do not allow efficient radical scavenging.Compound 27a seems to bind 5-LOX slightly displaced from 12a and 13d in the vicinity of Trp102, as suggested by molecular docking studies, Trp fluorescence spectroscopy, and site-directed mutagenesis.In contrast to ω-oxidation, the chromanol core and the unsaturated side chain of 27a offer little space for structural optimization.Although we identified modifications that turn 27a in an even more potent 5-LOX inhibitor, the pharmacologically relevant inhibition of 5-LOX product formation in PMNL was not enhanced.
Compound 27a strongly accumulates in PMNL, reaching a comparable intracellular concentration as 12a.Accordingly, 27a superiorly inhibits 5-LOX product biosynthesis in activated PMNL and human blood but not in monocytes.
These cell-type-specific differences are in favor of our hypothesis that LCMs are efficiently taken up by cells through specific transport systems.Within cellular lipid mediator biosynthesis, 5-LOX is the direct and primary target of 27a, and its inhibition shapes the lipid mediator profile of activated human macrophages and monocytes.Although 27a is considerably more selective than 12a, 6 both compounds additionally inhibit mPGES-1 at high concentrations, which might be beneficial to damp the redirection of the 5-LOX substrate arachidonic acid toward proinflammatory PGE 2 biosynthesis.
Compound 27a is orally available and reaches plasma and lung concentrations that effectively inhibit 5-LOX in vitro.Rapid clearance of 27a from the plasma is accompanied by substantial retardation in tissues, including inflamed lungs.Along these lines, 27a is more stable against side-chain truncation than 12a.While we did not detect 27a metabolites in mice in vivo (within 90 min), 27a was efficiently sulfated or ω-oxidized to a carboxylic acid in a human liver-on-chip (within 48 h), which might be related to the kinetics or species-specific differences in 27a metabolism.It is tempting to speculate that both metabolites of 27a, the sulfate and ωcarboxylic acid, possess 5-LOX inhibitory activity or, in the case of the 27a sulfate, might be hydrolyzed to the active compound in tissues that highly express sulfatases, such as the lung and liver. 74e demonstrate anti-inflammatory efficacy of 27a in murine models of peritonitis and asthma in vivo, for both i.p. and p.o. administration, and in experimental atopic dermatitis in vitro using RHE.Compound 27a effectively suppresses LT levels along with LT-driven hallmarks of inflammation, i.e., immune cell infiltration, vasopermeability, and bronchial hyperreactivity.
In conclusion, we here present compound 27a, a potent and orally active LCM-inspired 5-LOX inhibitor that shares the favorable pharmacological profile of the endogenous vitamin E metabolite 12a but is even more potent, selective, and metabolically stable, thereby allowing an efficient suppression of inflammation in vitro and in vivo.Whether 27a shares the proposed proresolving activities of 12a needs further investigation.

■ EXPERIMENTAL SECTION
Isolation and Semisynthesis of Vitamin E Derivatives. 1 H and 13 C NMR along with 2D NMR data were obtained on a Bruker Avance DRX 500 MHz spectrometer (500 and 125 MHz, respectively; BRUKER, Bremen, Germany) or a JEOL JNM-ECZS 400 MHz spectrometer (400 and 100 MHz, respectively; JEOL Ltd., Akishima, Tokyo, Japan) in deuterated chloroform, methanol, or acetone and calibrated using the residual undeuterated solvent resonance as an internal reference.IR spectra were recorded on a Thermo Scientific Nicolet iS5 FT-IR spectrometer (Thermo Scientific).Mass spectrometry analyses were performed on a JEOL JMS-700 (JEOL Ltd.) double-focusing mass spectrometer with reversed geometry, equipped with a pneumatically assisted EI or FAB source and on a BRUKER ESQUIRE 3000+ spectrometer (BRUKER) for ESI analyses in both positive and negative modes.Chromatographic analysis was performed on a Prominence-i LC-2030C (Shimadzu, Noisiel, France) equipped with a refrigerated autosampler and a column oven.The HPLC system was coupled to an evaporative light-scattering detector (ELSD SEDEX 90 LT, SEDERE).Then, 5 μL samples refrigerated at 10 °C were injected onto a Phenomenex Luna C18 column (150 mm × 4.6 mm, 5 μm) heated at 20 °C.A gradient of water with 0.1% formic acid (A) and acetonitrile with 0.1% formic acid (B) was applied (65% B for 1 min, 65−77% B within 4 min, 77% B for 8 min, 77−100% B within 5 min, 100% B for 6 min) at a flow rate of 1 mL/min.ELSD experiments were performed at 50 °C, and nitrogen was used as the nebulization gas (3.5 bar).Data were acquired and processed with LabSolutions Software (Shimadzu, Noisiel, France).Spectra for selected vitamin E analogues (13a, 13d, 27a, 27d) are shown in Figures S57−S60.Purity determined using ELSD experiments was ≥95%.Chromatographic

Journal of Medicinal Chemistry
separations, such as flash chromatography, were performed on IntelliFlash 310 (Analogix) using a silica gel column Chromabond flash RS column (Macherey-Nagel).Reactions under microwave irradiations were performed using an Anton Paar Monowave 300 microwave reactor (Anton Paar).All reactions under microwave irradiations were performed using the stirring option in borosilicate glass vials of 10 or 30 mL (G10 or G30) sealed with PTFE-coated silicone septa (at the end of the irradiation, cooling of reaction mixtures was realized by compressed air).The microwave instrument consists of a continuous focused microwave power output from 0 to 600 W. The target temperature was reached with a ramp of 3 min, and the chosen microwave power was maintained at a constant to hold the mixture at this temperature.The reaction temperature was monitored using a calibrated infrared sensor, and the reaction time included the ramp period.The microwave irradiation parameters (power, temperature, and time) were monitored by the Monowave software package.
Purified 5-LOX (0.5 μg in PBS pH 7.4 containing 1 mM EDTA and 1 mM ATP) was preincubated with the vehicle (DMSO) or test compounds for 10 min and prewarmed at 37 °C.5-LOX product formation was started by addition of arachidonic acid (Sigma-Aldrich; 20 μM or as indicated) and CaCl 2 (2 mM) and stopped by an equal volume of ice-cold methanol after 10 min at 37 °C.Major 5-LOX metabolites (all-trans isomers of LTB 4 and 5-HETE) were extracted on Sep-Pak C18 35 cc Vac Cartridges (Waters, Milford, MA), separated by RP-HPLC on a Nova-Pak C18 Radial-Pak column (4 μm, 5 mm × 100 mm, Waters) under isocratic conditions (73% methanol/27% water/0.007%trifluoroacetic acid) at a flow rate of 1.2 mL/min, and detected at 235 and 280 nm. 75Zileuton was used as the reference 5-LOX inhibitor.To investigate whether 27a reversibly inhibits 5-LOX, the purified enzyme was preincubated with 27a at 0.3 μM for 15 min.The reaction mix was then 10-fold diluted to a final compound concentration of 0.03 μM, and arachidonic acid (20 μM) and CaCl 2 (1 mM) were added after another 5 min incubation on ice.To recognize nuisance inhibition, Triton X-100 (Sigma-Aldrich; 0.01%) was added to the reaction buffer.
Human Whole Blood and Blood Cells.Human PMNL, PBMC, monocytes, and platelets were freshly isolated from leukocyte concentrates that we obtained together with human whole blood (containing 3.13% sodium citrate) from the Institute for Transfusion Medicine of the University Hospital Jena (Germany). 76Venous blood was collected in heparinized tubes (16 U heparin/mL blood), with informed consent of registered male and female healthy adult volunteers (18−65 years) who were fasted for at least 12 h.These volunteers regularly donated blood (every 8−12 weeks) and were physically inspected by a clinician.They had not taken antibiotics or anti-inflammatory drugs for more than 10 days before blood donation and were free of apparent infections, inflammatory disorders, or acute allergic reactions.
Leukocytes were concentrated by centrifugation (4000g/20 min/ 20 °C) of freshly withdrawn blood and subjected to density gradient centrifugation on a lymphocyte separation medium (LSM 1077, GE Healthcare, Freiburg, Germany).Erythrocytes were removed by dextran sedimentation and hypotonic lysis.PMNL were obtained from the cell pellet and platelets from the supernatant.The PBMC fraction was plated in cell culture flasks (Greiner, Nuertingen, Germany) for 1.5 h at 37 °C and 5% CO 2 in an RPMI 1640 medium (Sigma-Aldrich) with 5% FCS (Sigma-Aldrich), 2 mM L-glutamine (Sigma-Aldrich), and penicillin (100 U/mL)/streptomycin (100 μg/ mL, GE Healthcare) (monocyte medium) to isolate adherent monocytes.The purity of monocytes (>85%) was determined from forward and side scatter properties, and CD14 surface expression was determined by flow cytometry (BD FACS Calibur, Heidelberg, Germany).
5-LOX Product Formation by PMNL and Blood.Freshly isolated PMNL (5 × 10 6 ) suspended in PBS pH 7.4 with 1 mg/mL glucose or freshly withdrawn human whole blood were preincubated with the test compounds for 10 min (or as indicated) at 37 °C.5-LOX product formation in PMNL was triggered by addition of arachidonic acid (20 μM) and/or Ca 2+ -ionophore A23187 (2.5 μM; Sigma-Aldrich) followed by incubation for 10 min at 37 °C.The reaction was stopped with an equal volume of methanol.
Blood was treated with Ca 2+ -ionophore A23187 (30 μM) for 10 min at 37 °C or was first primed for 30 min with LPS (1 μg/mL) and then stimulated with fMLP (1 μM; Sigma-Aldrich) for 15 min.After the reaction was stopped on ice, plasma was prepared (600g, 10 min, 4 °C), and aliquots were mixed with an equal volume of methanol.Proteins were precipitated at −20 °C for 2 h and removed by centrifugation (600g, 15 min, 4 °C).
Fluorescence Spectroscopy of 5-LOX.Fluorescence spectroscopy measurements on human recombinant 5-LOX (Abcam, Cambridge, U.K.) were conducted in PBS pH 7.4 at 25 °C using an Edinburgh Instruments FLS920 Series fluorescence spectrophotometer (Livingston, U.K.).The effect of the ligand concentration was followed by means of tryptophan fluorescence, with excitation at 295 nm and emission collected between 310 and 450 nm, using excitation and emission slits of 5 and 10 nm, respectively.The concentration range of 5-LOX used through the experiments was 0.3−0.6 μM.5-LOX was titrated with a stock solution of the test compounds in dimethylformamide to yield an experimental compound concentration in the range of 0−10 μM.The final dimethylformamide concentration was kept under 5%.After each addition, the sample was incubated for 10 min before measurement.The fluorescence emission intensity values were corrected for dilution and background noise.All conditions were measured independently and in duplicate.
The fluorescence emission spectral shift was used to monitor ligand−protein binding since significant variations of intensity were not observed.The fluorescence spectral shifts were analyzed by monitoring the alterations in the spectral center of mass by estimating the intensity-averaged emission wavelength (⟨λ⟩), as calculated from the following equation, where F i is the fluorescence emission intensity at wavenumber ν i , and the summation was performed over the range of measured F values ( ) Molecular Docking Simulations.The molecular docking simulation was conducted with GOLD 2020.2.0 (CCDC, Cambridge, U.K.). 77The built-in CHEMPLP scoring function was used to rescore the outputted poses (10 best-scored poses were kept for each compound).The stable crystallographic tridimensional structure of 5-LOX was downloaded from the Protein Data Bank (PDB entry 3O8Y). 78Four in silico mutations were inserted to return to the 5-LOX wild-type sequence: E13W, H14F, G75W, and S76L.The residues were exchanged, and the structure was energetically minimized in Discovery Studio 3.5 (Biovia, San Diego, CA). 79ydrogen atoms were added with GOLD, using default settings.The binding site was constituted by 11 amino acid residues: Gln15, Arg101, Tyr81, Tyr100, Arg101, Trp102, Val110, Glu134, Asp170, Arg401, and Glu 622.Hydrogen-bond constraints were applied to Trp102-Hε1, Val110-HN, Glu134-Oε1, Asp170-Oδ1, Arg401-HH2, and Arg401-Hε with the constraint weight varying from 7 to 40 and the minimum H-bond geometry weight at 0.005.The ligands were allowed to detect internal H-bonds, flip pyramidal N, and flip amide bonds.Protein−ligand interactions of docking poses were analyzed using LigandScout 4.3 (Inteligand, Vienna, AT). 80ntracellular Concentrations of 12a and 27a.PMNL (1 × 10 7 /mL) were suspended in PBS pH 7.4 with 1 mg/mL glucose and incubated with a vehicle (DMSO, 0.1%), 27a or 12a (150 nM, each) for 20 min at 37 °C.Cells were centrifuged (1200g, 5 min, 4 °C) and

Journal of Medicinal Chemistry
washed thrice with PBS pH 7.4 with 0.1% fatty-acid-free BSA (Sigma-Aldrich).Compounds were extracted and analyzed by UPLC-MS/MS as described below for LCMs and derivatives.The intracellular concentrations of 27a and 12a were calculated assuming a spherical cell shape with a diameter of 13 μm (as measured using a Vi-CELL Series Cell Counter, Beckman Coulter, Krefeld, Germany) and an equal intracellular distribution.13d (9 pmol) was used as the internal standard.
LDH release was measured in monocytes using a CytoTox 96 nonradioactive cytotoxicity assay (Promega, Madison, WI) according to the manufacturer's instruction using Triton X-100 (0.8%) as the maximum LDH release control.
Purified sEH (60 ng) in 25 mM Tris HCl pH 7 with 0.1 mg/mL bovine serum albumin (BSA) was preincubated with the vehicle (DMSO) or 27a for 10 min at room temperature.The sEH substrate PHOME (50 μM, Cayman Chemicals) was added to start the enzymatic reaction, which was stopped after 60 min in darkness by addition of ZnSO 4 (200 mM).The formation of the fluorescent product 6-methoxynaphtaldehyde was measured using a NOVOstar fluorescence microplate reader (BMG Labtech, Ortenberg, Germany), with excitation at 330 and emission at 465 nm.The selective sEH inhibitor AUDA (100 nM, Cayman Chemicals) was used as the control.
Fatty acids and lipid mediators were extracted from macrophage supernatants or murine peritoneal exudates (1 mL) after addition of two aliquots of ice-cold methanol. 23 The columns were extensively washed with H 2 O and hexane before lipid mediators were eluted with methyl formate, evaporated to dryness, dissolved in methanol/H 2 O (50:50), and analyzed by UPLC-MS/MS. 23xtraction of LCMs and Derivatives.LCMs, 27a, and 27a metabolites were extracted from PMNL, medium of the liver-on-chip, murine plasma, peritoneal exudate, or homogenized murine tissue as described. 6In brief, PBS pH 7.4, methanol, chloroform, and saline (final ratio: 14:34:35:17) were successively added together with 13d Journal of Medicinal Chemistry (9 pmol) as the internal standard.The organic layer was evaporated, and the extracted LCMs were dissolved in methanol.
To analyze lipid mediators produced by monocytes, a flow rate of 0.8 mL/min and a column temperature of 45 °C were applied.Mobile phase A consisted of acetonitrile with 0.07% formic acid, and mobile phase B was acetonitrile/H 2 O (10:90) with 0.07% formic acid.Isocratic elution at A/B (30:70) for 2 min was followed by a linear gradient to A/B:70/30 within 5 min.Data were normalized to the internal standard PGB 1 , which allows the comparison of signal intensities between samples but not absolute quantification.
To quantify lipid mediators and fatty acids in macrophages, murine plasma, and murine peritoneal exudates, the mobile phase methanol/ H 2 O (acidified with 0.01% acetic acid) was ramped from 42:58 to 86:14 over 12.5 min, followed by isocratic elution with methanol/ H 2 O (98:2) for 3 min.Absolute quantification was based on 6 internal and 45 external standards. 23CMs, 27a, and 27a metabolites were separated at a flow rate of 0.8 mL/min and a column temperature of 45 °C, with acetonitrile/ H 2 O (10:90) as mobile phase A and acetonitrile as mobile phase B, both acidified with 0.07% formic acid.The linear gradient from A/B = 50:50 to 0:100 within 4.5 min was followed by isocratic elution with A/B = 0:100 for 1 min.MS parameters were adjusted according to Pein et al. 6 and as detailed in Table S3.Signal intensities were normalized to the internal standard 13d, and concentrations were calculated using external calibration curves as specified in Table S3.
Automatic peak integration was performed with Analyst 1.6 software (Sciex) using IntelliQuan default settings.
Correlation Network Analysis.The structure-based correlation networks of LCMs and derivatives were generated with Cytoscape 3.7.2software (Cytoscape Consortium). 84Tanimoto coefficients were calculated between the SMILES structures of compounds and represent the connecting edges in the correlation network, which is presented in an edge-weighted spring embedded layout.Nodes in close proximity represent structurally related compounds.The potency of the compounds to inhibit cell-free or cell-based 5-LOX activity is visualized by the node size, where bigger nodes represent compounds with higher inhibitory activity (lower IC 50 value).The node shape illustrates the compound series, and the node color indicates the series of garcinoic acids (13a-e), amplexichromanols (27a−d), tocopherols, and tocotrienols (1a−d, 6a−d).
Multiorgan-Tissue-Flow (MOTiF) Biochips.Biochips were made from polystyrene by injection molding and were equipped with a poly(ethylene terephthalate) (PET) membrane (TRAKETCH, thickness: 12 μm; pore diameter: 8 μm; pore density: 1 × 10 5 pores/ cm 2 ; Sabeu, Radeberg, Germany) that was integrated in the upper and lower parts of the biochip by heat-sealing with the bulk material.The top and bottom sides of the biochips and channels were sealed with an extruded PS bonding foil (thickness: 125 μm) by a lowtemperature bonding method.The upper and lower parts of the biochips were assembled using a double-sided adhesive film, and the chip surface was hydrophilized by oxygen plasma treatment to facilitate cell adhesion and prevent air bubble formation within the chambers and channels. 85,86UVECs were isolated from human umbilical cord veins and cultivated in an endothelial cell growth medium MV (Promocell, Heidelberg, Germany) with penicillin (100 U/mL)/streptomycin (100 μg/mL, GE Healthcare) up to passage 4. 87 On reaching 95% confluence, cells were subcultured at a density of 1.5 × 10 4 cells/cm 2 .
The liver-on-chip was prepared by seeding HUVECs (top: 3 × 10 5 , bottom: 1.5 × 10 5 ) in an endothelial cell growth medium MV on top of the membrane in the upper chamber and on the bottom of the membrane in the lower chamber, giving the cells 3−4 h to adhere before flipping the chip upside and seeding the bottom layer.After 2 days, differentiated HepaRG cells (3 × 10 5 ) were seeded on top of the lower and on the bottom of the upper membrane in a hepatocyte culture medium with hydrocortisone-hemisuccinate adjusted to 5 μM for 24 h at 37 °C and 5% CO 2 .The medium between the two membranes was renewed, and the vehicle (DMSO), 12a, or 27a were added.After incubation for 48 h at 37 °C and 5% CO 2, the medium was collected and the system was washed with 500 μL of methanol.12a, 27a, and their metabolites were extracted from the combined fractions and analyzed by UPLC-MS/MS. 6YP3A4 Expression in Hepatocytes.HepaRG cells were seeded at a density of 2.6 × 10 4 cells/cm 2 in Williams' E medium supplemented with 2 mM GlutaMAX (Thermo Fisher Scientific, Waltham, MA), 100 U/mL penicillin, 100 μg/mL streptomycin, 10% HyClone FCS (Thermo Fisher Scientific), 5 μg/mL insulin, and 50 mM hydrocortisone-hemisuccinate.At confluence after two weeks, HepaRG cells were shifted to a medium supplemented with 1.7% DMSO for two additional weeks to obtain confluent differentiated cultures containing equal proportions of hepatocyte-like and progenitors/primitive biliary-like cells.These differentiated hepatic cell cultures were exposed to the compounds in Williams' E medium supplemented with 2% FCS plus 0.2% DMSO for 48 h.
The CYP3A4 expression was quantified using an arbitrary unit of fluorescent mass representing the combination of intensity and area adjusted to cell number calculated by nuclei counting (Hoechst).
Experimental Atopic Dermatitis.Compound 27a was evaluated in an inflammatory model of RHE by Synelvia S.A.S. (Labege, France) according to the company's protocol. 88In brief, primary female human keratinocytes were used to generate RHE on polycarbonate filters as described. 88RHE was allowed to acclimate for 6 days (5% CO 2 , 37 °C) prior to incubation with 27a (5, 10, 25 μM), dexamethasone (1 μM), or vehicle (0.05% DMSO), and a proinflammatory cytokine cocktail for 4 days at 5% CO 2 and 37 °C.The culture medium including compounds and cytokines was changed on days 6, 8, and 10.Each culture condition was repeated three times.Acute cytotoxicity of 27a was excluded at the concentrations studied within 24 h by measuring the mitochondrial dehydrogenase activity (MTT assay).
Morphological changes (i.e., spongiosis and the formation of keratohyalin granules) were evaluated for sections of paraffinembedded RHEs after hematoxylin-eosin staining. 89The permeability of the skin barrier was estimated from the transepidermal water loss at Journal of Medicinal Chemistry the RHE surface. 90The surface pH was determined with a pH electrode as reported. 89The permeability of the stratum corneum was addressed by applying Lucifer yellow onto the RHE surface and inspecting the penetration by fluorescence microscopy. 88TSLP and IL-8 levels were analyzed by ELISA (according to the specifications of Synelvia S.A.S.) in the culture medium on day 2 and day 4 after addition of cytokines, respectively.
Animals.Male CD-1 mice (33−39 g, 6−8 weeks, Charles River Laboratories; Calco, Italy) and female BALB/c mice (20 g, 8 weeks, Charles River Laboratories) were fed with standard rodent chow and water and acclimated for 4 days at a 12 h light and 12 h dark schedule in a constant air-conditioned environment (21 ± 2 °C).Mice were randomly assigned to groups, and experiments were carried out during the light phase.Experimental procedures were conducted in conformity with Italian (DL 26/2014) and European (directive 2010/ 63/EU) regulations on the protection of animals used for scientific purposes and approved by the Italian Ministry.
Zymosan-Induced Peritonitis in Mice.CD-1 mice received 27a or zileuton, with DMSO (2%) in saline (0.5 mL) as a vehicle for i.p. administration and carboxymethylcellulose (0.5%) in 10% Tween 20 (0.5 mL) as a vehicle for p.o. administration.Zymosan (2 mg/mL in saline, i.p., 0.5 mL, Sigma-Aldrich) was injected at 30 min (i.p.) or 60 min (p.o.) post compound administration.Mice were sacrificed by inhalation of CO 2 after another 30 min to determine LTC 4 levels, lipid mediator profiles, metabolites, and vascular permeability and after 4 h to analyze LTB 4 levels and cell infiltration. 72Plasma and peritoneal exudates were collected, and cells were counted in exudates after trypan blue staining.Vascular permeability was measured by injection of Evans blue dye (40 mg/kg in saline, 0.3 mL, Sigma-Aldrich) into the tail vain directly before peritonitis induction. 72The exudate was centrifuged (3000g, 5 min), and the absorbance was measured at 610 nm (DU730 spectrophotometer, Beckman Coulter, Krefeld, Germany).Levels of LTB 4 and cysteinyl-LTs (dominated by LTC 4 ) were quantified in the exudate by ELISA (Enzo Life Sciences, Lorrach, Germany) according to the manufacturer's instructions.Compound 27a, its metabolites, and lipid mediators were extracted and analyzed by UPLC-MS/MS.
Experimental Model of Murine Asthma.BALB/c mice received 27a (p.o.) or vehicle (0.5% carboxymethylcellulose in 10% Tween 20; 0.5 mL) on days 0 and 7, 1 h (p.o) before being sensitized to ovalbumin (100 μg adsorbed to 3.3 mg of aluminum hydroxide gel, s.c., Sigma-Aldrich).Mice were sacrificed on days 1, 3, 8, 10, or 21 to collect lung and plasma.Bronchi were cut in rings of 1 to 2 mm length, placed in organ baths, and fixed to an isometric force transducer 7006 connected to a Powerlab 800 (AD Instruments, Ugo Basile, Comerio, Italy).After stretching the rings to a resting tension of 0.5g and equilibration for at least 30 min, the rings were challenged with carbachol (1 μM) until a reproducible response was observed.To assess bronchial reactivity, the cumulative response to carbachol (0.001 to 3.16 μM) was measured.Bronchial relaxation was determined from the cumulative response of precontracted bronchial tissues to salbutamol.Compound 27a, LTB 4 , and LTB 4 isomers were analyzed in plasma and lung homogenates by UPLC-MS/MS.Lung tissue (100 mg/mL) was homogenized in PBS pH 7.4 at 4 °C for 1−2 min using an Omni tissue homogenizer (Omni, Kennesaw, GA).
Statistics.Data are presented as mean with single values or mean ± SEM of n observations, where n represents the number of experiments or the number of animals, as indicated.The compound library was blinded for screening purposes but not for further biological evaluation, and samples from mouse peritoneum were blinded for counting infiltrated cells.Outliers were identified using a Grubb's test.Different groups were compared by one-way or two-way ANOVA for independent or correlated samples followed by Tukey's or Bonferroni's HSD post hoc test or by the two-tailed Student's t-test for paired or unpaired samples.Tests were conducted using a twosided α level of 0.05.P values <0.05 were considered statistically significant.Statistical calculations were performed using GraphPad Prism 9.1 (GraphPad Software, La Jolla, CA).IC 50 values were determined by graphical analysis using SigmaPlot 14.0 (Systat Software Inc., San Jose, CA).

■ ASSOCIATED CONTENT
* sı Supporting Information to any Author Accepted Manuscript version arising from this submission.

Figure 1 .
Figure 1.Correlation network of the compound library for inhibition of 5-LOX in PMNL.The network visualizes structural similarity between compounds calculated using Tanimoto similarity.Nodes represent individual compounds, and connecting edges represent Tanimoto coefficients > 0.9.The node shape differentiates between derivatives derived from amplexichromanols (AC), garcinoic acids (GA), or other leads, and the filling highlights the parental series, i.e., amplexichromanol (red), garcinoic acid (blue), tocopherol, and tocotrienol (green).The node size reflects the potency (IC 50 values) of the compound to inhibit 5-LOX product formation in PMNL.For nodes with dotted lines, IC 50 values were not determined.

Figure 3 .
Figure 3. Compound 27a accumulates in innate immune cells and inhibits 5-LOX product formation.(A) Effect of 27a on 5-LOX product formation (LTB 4 , its isomers, and 5-H(P)ETE) in PMNL treated with Ca 2+ -ionophore A23187 or A23187 and arachidonic acid (AA).(B) Intracellular uptake of 27a or 12a by PMNL treated with 150 nM of the respective compound for 20 min.Average intracellular concentrations were calculated for spherical PMNL with a diameter of 13 μm.(C) Effect of 27a or 12a on 5-LOX product formation initiated by A23187 and AA in LPS-prestimulated monocytes.(D, E) Effect of 27a on AA release from Escherichia coli-stimulated macrophages (D) and fMLP-induced Ca 2+ influx in PMNL (E).(F, G) Effects of 27a or 12a on 5-LOX product formation in human whole blood stimulated with A23187 (F) or LPS and fMLP (G) upon preincubation with the compounds for 10 min (F, G) or 3 h (F).Data are expressed as mean ± SEM (A, C−G) or mean with single values (B) from n = 3 (A−E, F, left panel) and n = 4 (F, right panel, G) independent experiments.*p < 0.05, **p < 0.01, ***p < 0.001 vs control (A, C−G); RM one-way ANOVA + Tukey's post hoc test (A, C−G).

Figure 5 .
Figure 5. Metabolism of 27a in a human liver-on-chip and in mice.(A) Scheme of the human liver-on-chip model.(B, C) The liver-on-chip was loaded with compound (1 μM) and incubated for 48 h.Extracted chromatograms (representative of three independent experiments) and concentrations of 12a and metabolites (B) or 27a and metabolites (C).(D) Effect of 27a and zileuton (zil) on CYP3A4 expression in differentiated HepaRG cells after treatment for 48 h.Fluorescence images (20× magnification) visualize CYP3A4-expressing (CYP3A4 + ) and CYP3A4nonexpressing cells (CYP3A4 − ) and are representative of nine (w/o) or three independent experiments.(E) Extracted chromatograms (representative of eight mice) and concentrations of 27a and metabolites in plasma of mice with acute peritonitis 90 min post 27a administration (10 mg/kg, p.o.).Data are expressed as mean ± SEM (D) and single values (B, C, E) from n = 3 (B, C, D, 27a, and zil, E, w/o), n = 8 (E, 27a), and n = 9 (D, w/o) independent experiments.*p < 0.05, **p < 0.01, ***p < 0.001 vs control; two-tailed unpaired t-test of log data (B, C, E), ordinary one-way ANOVA + Tukey's post hoc test (D).The nomenclature used for LCMs is explained in TableS2.

Figure 6 .
Figure 6.Compound 27a relieves inflammation in reconstructed human epidermis (RHE).RHE exposed to 27a or dexamethasone (dex) was treated with a cytokine cocktail for 4 days to trigger the inflammatory reaction.(A) Concentration of IL-8 in the growth medium.The dotted line indicates basal levels without cytokine stress.(B) Morphological changes visualized by hematoxylin and eosin staining (scale bar is 50 μm; representative of three independent experiments).(C) Transepidermal water loss (TEWL) at the RHE surface.(D) Impermeability of the stratum corneum.The stratum corneum of cytokine-stressed RHE becomes permeable for Lucifer yellow (green) that diffuses into the viable cell layers, as shown in the insets in higher magnification (scale bar for the outer box is 20 μm, and scale bar for the inset is 10 μm; representative of three independent experiments).Nuclei were stained with 4′,6′-diamidino-2-phenylindole (DAPI, blue).(E) pH value at the RHE surface.Data are expressed as mean ± SEM (A, 27a) with single values (A, dex, C) or mean + floating bars from minimum to maximum values (E) from n = 2 (A) and n = 3 (C, E) independent experiments.Ordinary one-way ANOVA + Tukey's post hoc test.

Table 1 .
Inhibition of Human Isolated 5-LOX and 5-LOX Product Formation in Activated PMNL by Garcinoic Acid-Derived Compounds (12a−25) e a IC 50 values (μM).b Residual activities (% control) at 1 μM compound concentration.c Residual activities (% control) at 3 μM compound concentration.d Highlighted data (gray) are from Pein et al. 6 n.d., not determined.e All values are given as mean ± standard error of the mean (SEM) of single determinations obtained in 3−6 independent experiments.

Table 2 . continued
a IC 50 values (μM).b Residual activities (% control) at 1 μM compound concentration.c Residual activities (% control) at 3 μM compound concentration.d Highlighted data (gray) are from Pein et al. 6 n.d., not determined.e All values are given as mean ± SEM of single determinations obtained in 3−4 independent experiments.