Tuning melatonin receptor subtype selectivity in oxadiazolone-based analogues: Discovery of QR2 ligands and NRF2 activators with neurogenic properties

New multi-target indole and naphthalene derivatives containing the oxadiazolone scaffold as a bio-isostere of the melatonin acetamido group have been developed. The novel compounds were characterized at melatonin receptors MT 1 R and MT 2 R, quinone reductase 2 (QR2), lipoxygenase-5 (LOX-5)


a b s t r a c t
New multi-target indole and naphthalene derivatives containing the oxadiazolone scaffold as a bioisostere of the melatonin acetamido group have been developed.The novel compounds were characterized at melatonin receptors MT 1 R and MT 2 R, quinone reductase 2 (QR2), lipoxygenase-5 (LOX-5), and monoamine oxidases (MAO-A and MAO-B), and also as radical scavengers.We found that selectivity within the oxadiazolone series can be modulated by modifying the side chain functionality and coplanarity with the indole or naphthalene ring.In phenotypic assays, several oxadiazolone-based derivatives induced signalling mediated by the transcription factor NRF2 and promoted the maturation of neural stem-cells into a neuronal phenotype.Activation of NRF2 could be due to the binding of indole derivatives to KEAP1, as deduced from surface plasmon resonance (SPR) experiments.Molecular modelling studies using the crystal structures of QR2 and the KEAP1 Kelch-domain, as well as the Abbreviations: AD, Alzheimer's disease; AREs, antioxidant response elements; Ab, amyloid-b peptide; COSY, homonuclear correlation spectroscopy; GPCR, Gprotein-coupled receptor; HMBC, heteronuclear multiple bond correlation; HSQC, heteronuclear single quantum correlation; KEAP1, Kelch-like ECH associated protein 1; LOX-5, lipoxygenase-5; MAO-A and MAO-B, monoamine oxidases; MAP-2, microtubule-associated protein 2 marker; MMGBSA, molecular mechanics generalized born surface area approach; MT 1 R and MT 2 R, GPCRs of melatonin; MTDL, multi-target directed ligand; MW, microwaves; ND, neurodegenerative disease; NDGA, nordihydroguaiaretic acid; NRF2, nuclear erythroid 2-related factor; NSC, neural stem cells; ORAC, oxygen radical absorbance capacity assay; PAMPA-BBB, parallel artificial membrane permeability assay for the bloodÀbrain barrier; PDB, protein data bank; QR2, quinone reductase 2; ROS and RNS, reactive oxygen and nitrogen species; SGZ, subgranular zone; SPR, surface plasmon resonance assay;

Introduction
Alzheimer's disease (AD) is the most common age-associated neurodegenerative disorder (ND), whose main hallmarks are the extracellular deposits of the misfolded amyloid-b peptide (Ab), the intracellular accumulation of the hyperphosphorylated microtubule-associated tau protein and a massive neuronal loss [1].AD aetiology is not completely understood, although different causative factors have been postulated, such as oxidative stress and neuroinflammation [2].
Oxidative stress is the consequence of the progressive failure of the antioxidant defence systems, particularly noticeable in the brains of AD-patients, where the increased levels of oxidized biomolecules correlate to disease progression [3].Thereby, the control of transcription factors involved in the regulation of antioxidant genes and the inhibition of the main enzymes that produce reactive oxygen and nitrogen species (ROS and RNS) would be a valuable strategy to combat AD [4].
The nuclear erythroid 2-related factor (NRF2) is a transcriptional protein that regulates the expression of antioxidant and antiinflammatory enzymes [5].In non-stressed conditions, NRF2 is mainly located in the cytosol bound to the Kelch-like ECH associated protein 1 (KEAP1), which favours its proteasome degradation.Under pathological conditions (oxidative stress, toxic insults, etc.), KEAP1 changes its conformation and NRF2 is released and translocated to the nucleus, where it binds to the DNA antioxidant response elements (AREs), inducing the expression of antioxidant enzymes such as NAD(P)H quinone oxidoreductase and heme oxygenase-1 [6,7].
Monoamine oxidases (MAO-A and MAO-B) are outer mitochondrial membrane-bound flavoproteins, whose levels are increased in the brain of patients suffering NDs, such as AD and Parkinson's disease.Both enzymes catalyse the oxidative deamination of biogenic amines, including several neurotransmitters, what it produces hydrogen peroxide, ROS, and accordingly oxidative stress [8].Besides, overexpression of MAO-B increases the formation of aberrant Ab [9].In this regard, MAO inhibition could provide neuroprotection against AD, by regulating neurotransmitters' levels, ameliorating oxidative pathologies and stopping the deposition of amyloid plaques.
Neuroinflammation is another important player in AD that generates cytokines and neurotoxins, leading to neuronal damage and death [10].The iron-containing enzyme lipoxygenase-5 (LOX-5, EC 1.13.11.34) catalyses two early steps of the oxidation of the arachidonic acid, producing finally pro-inflammatory leukotrienes [11].LOX-5 is present in the CNS particularly in the hippocampus, where its levels appear to increase in AD [12].Several works have demonstrated that blockade of LOX-5 in transgenic mice diminishes cytokines levels and both Ab and tau pathologies [13,14].Therefore, LOX-5 inhibitors could be valuable therapeutic agents, as they could reduce neuroinflammation and aberrant protein aggregation.
In mammals, melatonin is secreted in nanomolar concentrations from the pineal gland to the blood, when it transmits the circadian rhythm to the peripheral organs by interacting with two G-proteincoupled receptors (GPCRs), called MT 1 and MT 2 (MT 1 R and MT 2 R) [15].Melatonin also interacts with a third binding site, MT 3 , which has been identified as the flavin adenine dinucleotide (FAD)-dependent enzyme quinone reductase-2 (QR2, EC 1.10.99.2) [16,17].Melatonin activates MT 1 R and MT 2 R at sub-nanomolar concentrations (K d ~0.1 nM), whereas its affinity as a QR2 inhibitor is about 100-fold lower [16].
QR2 catalyses the two-electron reduction of quinones into unstable hydroquinones, which are excreted through conjugation or reconverted to quinones in the presence of oxygen while producing ROS [18].Given that QR2 levels are increased in the hippocampus of AD patients, inhibitors of this enzyme are promising therapeutic agents to fight neurodegeneration by reducing the oxidative stress [19e21].
Adult neurogenesis is a matter of intense research, although its scope and function remain debatable [22,23].Restricted to small regions, namely the subventricular zone (SVZ) of the lateral ventricle and the subgranular zone (SGZ) of the hippocampus, the adult mammalian brain retains neural stem cells (NSC) niches that could develop and integrate new cells in the neuronal circuitry.Therefore, a neurogenic drug would induce the differentiation of NSCs into mature neurons capable of replacing those lost by neurodegeneration, allowing the brain to recover its own self-renewal capacity [24].
In recent years, numerous molecular targets and signalling cascades involved in neurogenesis have been identified and, as a consequence, different types of drugs have been evaluated in neuronal regeneration.For example, neurogenic properties have been found in MTRs' ligands, NRF2 activators, antioxidants and anti-inflammatory agents [25].
Melatonin modulates hippocampal neurogenesis by increasing both cellular precursor proliferation and survival in the hippocampi of aged mice [26].Given these outstanding properties, many research groups (including ours) have devoted great efforts aimed to the identification of new MTR ligands and their molecular interaction pathways [27e30].In this context, our melatonin analogue IQM316 was recently found to be capable of inducing hippocampal neurogenesis in mice at a healthy and sustainable rate, thus helping to preserve previous memories [31,32].We have also described the bioisosteric replacement of the acetamido group of melatonin by a series of reversed amides and azoles [29].Among these azole derivatives, 5-[2-(5-methoxy-1H-indol-3-yl)ethyl]-1,3,4-oxadiazol-2(3H)-one (1) was the most potent partial agonist at the human MTRs, displaying the highest affinity for both human MTRs (K i ¼ 35 and 4 nM for hMT 1 R and hMT 2 R, respectively).Moreover, this compound also showed potent neurogenic properties in vitro, better than melatonin itself, but it was predicted that it would not enter the CNS in view of the unfavourable results in in vitro blood-brain barrier (BBB) permeability assays.
Given the complex nature of AD, the multi-target directed ligand (MTDL) strategy has emerged as a valid alternative to fight different pathological cascades involved in this disease.To achieve maximum efficiency, MTDLs should promote endogenous defensive and/or regenerative pathways and be directed towards biomolecular targets located upstream in the neurotoxic cascades [2,33].
With these ideas in mind, in this work we planned to develop new melatonin-based MTDLs that could improve the CNS permeability of 1 and also incorporate additional activities on targets related to oxidative stress and neuroinflammation.For this purpose, we modified each part of the starting molecule by: (i) addition of different substituents to the oxadiazolone nitrogen; (ii) replacement of the oxadiazolone ring by an oxadiazolamine; (iii) insertion of a double bond in the linker; and (iv) replacement of the indole by a dihydronaphthalene or naphthalene ring (Figure 1).

Chemistry. Synthesis of new indole-and naphthalene-based compounds
The microwave (MW)-assisted reaction (120 C, 10 min) of 1 with the corresponding alkyl halide in basic media (K 2 CO 3 ) using acetone as solvent gave derivatives 2e15, in moderate to excellent yields (49e92%) (Scheme 1).
Modifications in the linker were performed using the commercially available 5-methoxy-1H-indole-3-carbaldehyde (18) as starting material.The introduction of a methyl group in the beta-  position with respect to the indole ring was carried out from 18 and (carbethoxyethylidene)triphenylphosphorane by a Wittig reaction, giving the corresponding a,b-unsaturated ester 19 in high yields (95%).Then, 19 was reduced by catalytic hydrogenation to the saturated ester 20, followed by its transformation to the corresponding hydrazide and further MW-assisted cyclization in the presence of 1,1-carbonildiimidazole (CDI) to give the 1,3,4oxadiazol-2(3H)-one derivative 21 in 48% yield.
The conjugated double bond in the linker was introduced by the treatment of 18 with malonic acid in basic media that generate the a,b-unsaturated acid 22, in almost quantitative yield by a Knoevenagel-Doebner reaction.Activation of 22 with hydroxybenzotriazole (HOBt) and N-(3-dimethylaminopropyl)-N 0 -ethylcarbodiimide hydrochloride (EDC$HCl) as coupling agents, and a catalytic amount of 4-dimethylaminopyridine (DMAP) at rt during 3 h, followed by the addition of hydrazine hydrate, yielded the corresponding hydrazide in quantitative yield.It is worth-mentioning that the order of addition of the reagents is crucial for obtaining high yields.Once the hydrazide is formed, its MW-assisted cyclization with CDI afforded the indoleeNH-oxadiazolone derivative 23 in good yield (85%).The treatment of 23 with propargyl bromide in basic media (K 2 CO 3 ) and acetone in an MW-oven at 120 C during 10 min afforded the corresponding N-propargyl-oxadiazolone 24 (Scheme 1).Bioisosteric replacement of the oxadiazolone ring by an amide group was carried out from the a,b-unsaturated acid 22 by condensation with propargyl or allyl amine in acetonitrile at rt during 3 h, in the presence of EDC$HCl, HOBt, and DMAP, giving the corresponding propargyl (25) or allyl derivative (26) in moderate to good yields (Scheme 1).
Indole scaffold replacement by either a dihydronaphthalene or a naphthalene ring was carried out from commercial 7-methoxy-1tetralone (27) by two different synthetic routes.Reaction of 27 with a solution of methyl magnesium iodide in ether at rt for 3 h afforded the corresponding alcohol (non-isolated intermediate), which was treated with HCl to yield the 1-methyl dihydronaphthalene 28 in quantitative yield [34]. Oxidation of the above methyl group with selenium dioxide gave aldehyde 29, in low yield due to the easy aromatization of the dihydronaphthalene ring.In spite of the mild oxidative conditions used, a mixture of dihydronaphthalene 29 and naphthalene 30 [35] was obtained, in 25% and 20% yield respectively (Scheme 2).
Then, we planned to transform the dihydronaphthalene derivative 29 into the unsaturated acid 31 by the treatment with malonic acid in Knoevenagel-Doebner conditions, obtaining unexpected 1 H and 13 C NMR data for the isolated product.A deeper NMR analysis (COSY, HSQC, and HMBC) demonstrated that this structure corresponded to lactone 32, formed under basic conditions by an intramolecular cyclization (Scheme 2).Probably, the carboxylate group attacked the intracyclic double bond forming the intermediate enol, which could develop to lactone 32.Nevertheless, given that in solution this cyclization is an equilibrium, acid activation was possible after treatment with coupling reagents (EDC$HCl, HOBt).Further treatment with N 2 H 4 $H 2 O and CDI, led to the desired dihydronaphthalene-based NH-oxadiazolone 33, although in low yield (15%), probably due to the easy aromatization that could also give the corresponding naphthalene (non-isolated).
Then, the introduction of an a,b-unsaturated ester was carried out by olefination of vinyl triflates under palladium-catalysed conditions (Heck reaction), using ethyl acrylate as reactant and TEA as base.Since first attempts to obtain 35 in good yields failed, different experimental conditions in a MW oven were probed, modifying the palladium-based catalyst [Pd(OAc) 2 þPPh 3 or Pd(PPh 3 ) 2 Cl 2 ], temperature (100e140 C) and reaction time (10e40 min).Optimized conditions for the synthesis of 35 resulted from the use of Pd(PPh 3 ) 2 Cl 2 as catalyst and MW irradiation at 105 C for 15 min, obtaining the desired ester derivative in 87% yield.
Then, the ester 35 was hydrolysed under basic condition with lithium hydroxide (LiOH) followed by acid treatment with HCl to obtain the corresponding acid.This acid was not isolated but transformed into the corresponding hydrazide, followed by cyclization with CDI to afford the desired dihydronaphthalene-based oxadiazolone 33 in low yields (21%) due to its easy aromatization.In fact, the aromatic analogue was detected by HPLC-MS, although Given that the best yields were achieved using the Heck reaction in triflate derivatives, this strategy was applied for obtaining naphthalene derivatives.The conditions previously used in the synthesis of the vinyl triflate 34 (Scheme 2) were extended to the dimethoxylated derivative (Scheme 3).However, derivative 37 was obtained in low yields due to the formation of several by-products, probably due to the lesser electrophilic character of the carbonyl group.Aromatization of triflates 34 and 37 with 2,3-dichloro-5,6dicyano-1,4-benzoquinone (DDQ) at rt, led naphthalenes 38 and 39, respectively (Scheme 3).Then, the ethyl acrylic moiety was introduced by a Heck reaction, treating the corresponding triflate derivative with ethyl acrylate in the presence of Pd(OAc) 2 , 1,10phenanthroline and TEA, to yield naphthalene esters 40 and 41 in high yields (90 and 81%, respectively).These esters were hydrolysed under basic conditions, obtaining acids 42 and 43 in quantitative yields.Finally, these acids were transformed into the corresponding hydrazides, followed by cyclization with CDI to afford naphthaleneeNH-oxadiazolones 44 and 45 in good yields (91 and 85%, respectively).
All new compounds were isolated and purified using flash column chromatography (automatic IsoleraOne-Biotage equip), preparative thin-layer plates (TLC) or semipreparative HPLC (Waters Autopurification system).Purity (>95%) was evaluated by HPLC-MS and chemical structure was characterized by their spectroscopic data ( 1 H NMR, 13 C NMR).Complete NMR assignments were made by two-dimensional NMR experiments, mainly COSY (homonuclear correlation spectroscopy), HSQC (heteronuclear single quantum correlation) and HMBC (heteronuclear multiple bond correlation).

Characterization in melatonin receptors
Experiments on human MT 1 R and MT 2 R were performed in Chinese hamster ovary cells, where these receptors were stably transfected.The QR2 binding assays were carried out using membrane homogenates of hamster brains.In all cases, displacement of the radioligand 2-[ 125 I]iodomelatonin was measured in the absence or presence of the tested compound and nonspecific binding was determined with melatonin, following described protocols [37e39].
Firstly, radioligand displacements were measured at a fixed concentration of compound (100 or 10,000 nM) in each receptor subtype.Then, binding constants (K i ) were calculated only for compounds with a radioligand displacement above 80%, using a range of 5 different concentrations of tested compound in 3 independent experiments.Melatonin was assayed for comparative purposes and results are gathered in Table 1.
The indoleeNH-oxadiazolone 1 showed high binding affinities for hMT 1 R and hMT 2 R (K i ¼ 35 and 4 nM, respectively) [29].In contrast, it did not show a substantial affinity toward QR2, as it only displaced 5% of the radioligand at 100 nM (Table 1).The introduction of substituents in the NH group of the oxadiazolone ring to give derivatives 2e15 caused a great decrease in the affinity for all melatonin receptors, reaching in some cases the total loss of activity.Only compounds with longer chains namely, ), and 15 [R ¼ (CH 2 ) 3 phenyl] showed a slight selectivity for hMT 2 R over hMT 1 R, with binding constants around 10 À7 M in the case of hMT 2 R.This affinity drop could be due to a possible hydrogen bond between the NH group of the oxadiazolone ring of 1 and the receptors, which is lost when it is substituted.The evaluation of derivative 4 in QR2 gave only a 68% displacement at 10 mM, so its K i was not calculated and the rest of the indoleoxadiazolones were not evaluated in QR2.
The replacement of the oxadiazolone ring by a 1,3,4-oxadiazol-2-amine to give 16 provided binding constants in the hundrednanomolar range for the three melatonin targets (K i ¼ 550, 110, and 330 nM, for hMT 1 R, hMT 2 R and QR2, respectively).Thus, this change maintained the binding constants for hMT 1 R and hMT 2 R and improved the affinity for QR2, compared to some N-substituted oxadiazolone derivatives (e.g., 12 and 15).
The addition of a methyl group in the aliphatic linker of 1 to give the beta-methyl derivative 21, resulted in the complete loss of affinity for hMT 1 R, a similar binding for hMT 2 R (K i ¼ 12 nM) and better affinity for QR2 (K i ¼ 230 nM) than the starting molecule 1.
Introduction of a double bond in the linker of the indole series provided interesting effects on the behaviour of these unsaturated oxadiazolones (23 and 24) and amides (25 and 26).These four derivatives lost their affinities toward hMT 1 R and hMT 2 R (K i s > 100 nM), while drastically improved their binding to QR2.Unsaturated indole-oxadiazolone compounds 23 and 24, and the propargyl amide derivative 25 showed K i s between 3.2 and 9.3 nM in QR2, one order of magnitude better than melatonin itself (K i ¼ 77 nM).
According to previously published results [41], we also found that the NH-indole fragment was not indispensable for activity on melatonin targets.In fact, the 7-methoxy-3,4-dihydronaphthalene 33 showed affinities toward hMT 2 R and QR2 in the 10 À7 M range (K i ¼ 410 and 260 nM, respectively), although resulted poor active in hMT 1 R. Ring aromatization to give the naphthalene counterpart greatly modified its profile, as 44 recovered a moderated affinity for hMT 1 R (K i ¼ 260 nM), maintained a similar affinity for QR2 (K i ¼ 270 nM), and interestingly, displayed a potent binding constant in hMT 2 R in the low-nanomolar range (K i ¼ 5.1 nM).The introduction of a second methoxy group in position 6 of the naphthalene ring was partially detrimental for affinity, since 45 was inactive in hMT 1 R, not as active as 44 in hMT 2 R (K i ¼ 69 nM), although 45 was one order of magnitude more active in QR2 (K i ¼ 35 nM).

Evaluation of the oxygen radical absorbance capacity (ORAC)
As a measure of antioxidant properties, we tested new indole-and naphthalene-based compounds in the ORAC assay, following described protocols [43e46].
(±)-6-Hydroxy-2,5,7,8tetramethylchromane-2-carboxylic acid (trolox), the aromatic part of vitamin E responsible for its scavenging properties, was used as the internal standard with the arbitrary value of ORAC ¼ 1.0.Results are expressed as trolox equivalents (trolox mmol/tested compd.mmol) in a comparative scale that indicates if a compound is a better (ORAC > 1.0 trolox equiv) or a worse oxygen radical scavenger (ORAC < 1.0 trolox equiv) than vitamin E. Melatonin was also evaluated for comparative purposes, giving an ORAC value 2.3fold higher than trolox.This activity fully agrees with the ORAC value previously described by Sofic et al. (2.0 trolox equiv) [47], thus supporting the reliability of our experiments.
As shown in Table 1, all indole derivatives displayed ORAC values ranging from 1.3 to 3.0 trolox equiv., independently of the nature of the substituent attached to position 3 of the heterocycle.Consequently, they could be considered as excellent antioxidant agents.Surprisingly, the unsaturated derivative 23 did not show a substantial difference in the ORAC value (2.4 trolox equiv) compared to its saturated analogue 1 (2.7 trolox equiv), despite the view that an increase of the conjugation between the indole and oxadiazolone rings would bring about augmented radical scavenger properties.
Replacement of indole by a dihydronaphthalene or naphthalene core was detrimental to the antioxidant activity in the ORAC assay.Dihydronaphthalene 33 displayed a reduced ORAC value (0.4 trolox equiv) and naphthalenes 44 and 45 were inactive at the maximum concentration tested (10 mM).In summary, it can be inferred that in these compounds the radical absorbance capacity was mainly due to the presence of the indole ring, rather than other structural motifs.

Induction of the transcription factor NRF2 pathway in the AREc32 cell line
The ability to activate NRF2 was evaluated using an NRF2dependent luciferase reporter assay in the AREc32 cell line [40,48].Cells were cultured for 24 h and then treated with In general, saturated indole-oxadiazolone derivatives (e.g., 4) did not show activity at the maximum concentration tested (30 mM).On the contrary, the unsaturated indole-oxadiazolone 23 and its N-propargylic counterpart 24 were able to increase luciferase activity, with CD values of 15.1 and 1.8 mM, respectively (Table 1).Removal of the oxadiazolone ring in 24 to give the propargylic amide 25 caused a decrease by one order of magnitude in NRF2 activation (CD ¼ 27.4 mM).When the propargyl moiety was replaced by an allyl fragment (26) the activity vs. NRF2 was lost.Dihydronaphthalene and naphthalene analogues were inactive.The above results indicated the importance of the presence of the indole core and a double bond in the linker to achieve good induction of NRF2.

Binding studies by surface plasmon resonance (SPR) assay
As explained in the introduction, activation of NRF2 could be due to conformational changes in KEAP1 that preclude the union between both proteins and stimulate the expression of antioxidant enzymes by NRF2.To evaluate the ability of our molecules to interact with the Kealch domain of KEAP1, a surface plasmon resonance (SPR) experiment was carried out.The Kelch domain was obtained from Escherichia coli transformed with pET15b-6xHIS-Kelch plasmid and induced by adding isopropyl-1-thio-b-D-galactopyranoside (IPTG).6xHIS-Kelch was purified using the ProBond Purification System (Invitrogen).The Kelch domain of KEAP1 was immobilized on a carboxymethyl cellulose sensor chip (CM5) and a solution of the corresponding molecule (100 mM) was then fluxed over the sensor chip in order to detect a change in the SPR signal, resulting from the interaction between the compounds and the Kelch domain.The well-known KEAP1 binder ML334 was used as internal reference [49], which produced a change in the SPR signal of 10 resonance units (Fig. 2).
As shown in Fig. 2, SPR responses of compounds 23e25 were equal to, or greater than ML334, while the rest of the compounds tested (4, 21, 26, and 45) gave fewer values.Importantly, the most active compounds in the phenotypic assay based on AREc32 cells showed a better binding to KEAP1 than the less active derivatives.Therefore, we can tentatively think that the activation of NRF2 by compounds 23e25 in the cellular assay could be due to the binding of these molecules to KEAP1.

Inhibition of human MAO and LOX-5
All compounds were evaluated as inhibitors of human MAO-A, MAO-B and LOX-5 and the results of the most representative compounds are collected in Table 2.For the sake of clarity, data of inactive compounds in the three enzymes are omitted.The hMAO-A/B inhibition was determined by the production of oxygen peroxide from a common substrate for both isoenzymes (p-tyramine) and quantified by the Amplex Red MAO assay kit [50].(R)-Deprenyl, iproniazid and moclobemide were also tested for comparative purposes.The inhibition of hLOX-5 was performed following the fluorescence-based method described by Pufahl et al. [51], using the two well-known inhibitors zileuton and nordihydroguaiaretic acid (NDGA) as internal references.

In vitro bloodÀbrain barrier permeation assay (PAMPA-BBB)
In order to determine the capability of the new compounds to cross the BBB and reach their CNS targets, we evaluated them in the in vitro parallel artificial membrane permeability assay for the BBB (PAMPA-BBB) described by Di et al. [52], and modified by our group for testing molecules with limited water solubility [44e46, 53,54].The passive CNS permeation of new compounds through a lipid  extract of porcine brain was measured at room temperature.In each experiment, 11 commercial drugs of known brain permeability were also tested, and their permeability values were normalized to the reported PAMPA-BBB data.As previously established in the literature [52], compounds with P e > 4.0$10 À6 cm s À1 would be able to cross the BBB (cnsþ), whereas those displaying P e < 2.0$10 À6 cm s À1 would not reach the CNS (cns-).Between these values, the predicted CNS permeability remains uncertain (cns þ/ À).
In the melatonin-based family, the indoleeNH-oxadiazolone 1 and its beta-methyl analogue 21 (both bearing a saturated linker) were predicted as not CNS-permeable (P e < 2$10 À6 cm s À1 ).In contrast, all indoleeN-substituted oxadiazolone derivatives (2e15), showed permeability values exceeding 4.0 10 À6 cm s À1 and thus, they were predicted to be CNS-permeable.The indoleeoxadiazolamine 16 displayed negative CNS permeation, whereas the indoleeNH-oxadiazolone 23, which is the unsaturated analogue of 1, displayed a permeability value in the ambiguous range (P e ¼ 3.2$10 À6 cm s À1 ).The attachment of a propargylic moiety onto the oxadiazolone ring nitrogen of 23 gave 24, which -as expected-was found to be CNS-permeable.Unsaturated amides 25 and 26 showed negative CNS permeability.
The exchange of the indole heterocycle by a dihydronaphthalene (33) or a naphthalene (44 and 45) generated compounds with good CNS permeability values (P e ¼ 9.0, 9.8 and 4.5 10 À6 cm s À1 , respectively), even when the NH-unsubstituted oxadiazolone ring was retained.From these experiments we can conclude that, in general, the new indole and naphthalene derivatives with a single NH-unsubstituted motif are expected to be CNS-penetrating, whereas compounds with two NH groups would experience more difficulties to reach the CNS.

Drug-like calculations
To choose the best candidates for neurogenic assays, we studied the medicinal chemistry alerts of new indole-or naphthaleneoxadiazolones in two free databases, namely ZINC15 (http:// zinc15.docking.org/)[55] and Aggregator Advisor (http://advisor.bkslab.org/)[56] (Table S2 in the Supplementary Information).None of the new compounds was highlighted as a pan assay interference compound (PAINS), according to the ZINC15 database.However, the Aggregator Advisor website identified compounds 13 and 15 with a fairly high calculated LogP in the range reported for other aggregators.Thus, these compounds were discarded for further biological assays.
Moreover, we calculated several physicochemical and ADME properties (log P, solubility, pK a and Lipinski rule) of selected compounds (23, 24 and 44) in comparison with melatonin (Table S3 in the Supplementary Information).According to the Swiss Institute of Bioinformatics (http://www.swissadme.ch/)[57], these compounds showed a good drug-like profile and thus, they are good candidates for further pharmacological studies.

Neurogenic studies
We investigated the capacity of a selection of the new compounds, covering different structural features, to promote cell differentiation into neurons.To this end, hippocampal-derived neurospheres adhered to a substrate and cultured under differentiation conditions in the presence of compounds were used [27,58].The neurogenic potential of each compound was determined using fluorescence confocal microscopy, by direct observation of the expression of two immunofluorescent dyes linked to two well-known neuronal markers, b III -tubulin (TuJ-1 clone) shown in green and microtubule-associated protein 2 (MAP-2) in red.TuJ-1 is expressed in new born neurons in early stages of their differentiation, whereas the expression of MAP-2 indicates a consolidated neuronal stage [59].Vehicle-treated cultures (considered as basal) showed only a few scattered positive TuJ-1 or MAP-2 cells.In contrast, treatment with the compounds resulted in an increase in the number of TuJ-1 and MAP-2-positive cells compared to basal levels.
In the melatonin-based family several compounds promoted the expression of TuJ-1 and MAP-2, indicating positive neurogenic activities (Fig. 3).Comparing derivatives with a dimethylene liker, the indoleeoxadiazolone 4 displayed good neurogenic properties, whereas its indoleeoxadiazolamine counterpart 16 was inactive.This result points out the importance of the presence of an oxadiazolone ring for successful neurogenesis.
Among indole compounds bearing a double bond in the linker, the indoleeN-propargyloxadiazolone 24 was one of the most active derivatives, as it greatly increased the expression of both TuJ-1 and MAP-2.Conversely, its indoleeNH-oxadiazolone analogue 23 was inactive, suggesting that the N-propargyl group is beneficial for activity.
Both naphthaleneeNH-oxadiazolone derivatives 44 and 45 showed positive effects in the expression of TuJ-1 and MAP-2.In contrast, the dihydronaphthalene analogue 33 was inactive, suggesting that the naphthalene core is more useful for biasing the compounds towards neurogenesis.

Theoretical studies. Conformational analysis of indole and naphthalene derivatives 23 and 44
As explained above, the indoleeNH-oxadiazolone derivative 23 showed marked selectivity towards QR2 (K i ¼ 6.6 nM), in comparison with hMT 1 R and hMT 2 R (K i > 100 nM).In contrast, substitution of a naphthalene ring for the indole heterocycle gave 44, which was a potent and selective agonist at hMT 2 R (EC 50 ¼ 1.1 nM), compared to hMT 1 R and QR2 (K i ¼ 260 and 270 nM, respectively) (Table 1).
With the aim of explaining this distinct behaviour in NH-oxadiazolone derivatives that only differ in the nature of the fused heterocycle, we performed a conformational analysis of 23 and 44 using quantum mechanics (Fig. 4).
The conformational analysis revealed higher energy barriers (~2 kcal mol À1 ) for rotation about the C2eC3 bond and a preferred flat conformation in the indoleeNH-oxadiazolone 23 (Fig. 4).This co-planar arrangement of the ring systems is prevented in the naphthalene derivative 44 due to steric clash as a consequence of the decreased C1eC2eC3 angle.These results seemed to indicate that the flat conformation of 23 might favour the stacking interaction with the isoalloxazine ring of FAD in QR2, whereas the slight non-planarity of 44 (dihedral angle of ~40 ) would be preferred for its agonist activity on hMT 2 R. On the other hand, the energy barriers for rotation about the C4eC5 bond are of similar magnitude in both molecules.

Molecular dynamics simulations on hQR2, hMT 1 R and hMT 2 R
To test the previous hypotheses, we undertook some molecular dynamics studies.In agreement with the previous statement, our modelled complexes of hQR2 (Fig. 5) and hMT 2 R (Fig. 6) with 23 and 44, respectively, show that the proposed stacking of the flat 23 molecule on FAD is reminiscent of that described for resveratrol [60], with the indole nitrogen establishing a hydrogen bond to the carbonyl oxygen of Gly174 that is similar to that observed for one of the resorcinol hydroxyls of this naturally occurring polyphenol.It can also be seen that the oxadiazolone nitrogen is engaged in hydrogen bonds to water molecules, which would be displaced when the attached hydrogen is replaced with a short substituent, as is the case for the propargyl group present in 24 (Fig. S1, Supporting Information).
In contrast, for the less coplanar naphthalene derivatives, the variety of binding orientations we found is commensurate with that observed in X-ray crystal structures of QR2 in complex with melatonin and derivatives [61].Nonetheless, the out-of-plane conformation of 44 is ideally suited to mimic the bound conformation of melatonin that has been observed in the XFEL structures of chimeric MT 1 R and MT 2 R constructs, very recently described by Stauch et al. [62] and Johansson et al. [63], respectively.At this location in our modelled MT 2 R, binding of 44 is largely stabilized by the phenyl ring of Phe192 ECL2 , additional hydrophobic interactions with neighbouring residues (including the highly conserved Trp264 6.48 that was replaced by Phe in the experimental constructs) and a number of direct and water-mediated hydrogen bonds with the polar side chains of Gln194 ECL2 , Asn162 4.60 , and Asn268 6.52 .The most important difference relative to MT 1 R is the orientation of Tyr200 5.38 , which is close to the side chains of Val178 ECL2 and Ser180 ECL2 and whose phenolic group hydrogen bonds to the carbonyl oxygen of Ala171 4.56 .In MT 1 R, the equivalent residues are Ala165 ECL2 and Thr167 ECL2 so that the side chain of Tyr187 5.38   points towards the membrane [62,63] and slightly different interactions are observed (Fig. S2, Supporting Information).

Molecular docking studies and molecular dynamics simulations on KEAP1
Molecular docking studies of 23 and 24 were performed on four different crystal structures of the KEAP1 Kelch domain (PDB entries 4XMB, 5CGJ, 5FNU and 4IFL) to consider, at least partially, protein flexibility.At physiologic pH, compound 23 exists as two different species, as deduced from its experimental pK a of 7.45 ± 0.09 (Fig. S3, Supporting Information).No substantial differences in the average binding energy values were observed in both species, named 23-0 and 23-1.On the other hand, the higher affinity score obtained for 24 correlated with the experimental NRF2 induction value in the AREc32 cell line (Table S1, Supporting Information).
The most stable molecular docking poses found for 23 and 24 on the four 3D-structures of the KEAP1 Kelch domain (Figs.S4 and S5, Supporting Information) were used as the starting structures for uMD simulations, which after 200 ns of production run showed some drifting of the bound ligand towards a new ligand-protein interface (Fig. 7).The binding site residues involved in the NRF2-KEAP1 interaction are very well known and have been reported by several authors [64].It has been shown that residues Arg380,415,483; Ser363,508,555,602; and Tyr525,572,334 on KEAP1 are essential to interact with NRF2.Our uMD simulations show that 23 and 24 do interact with some of these residues (Fig. 7) and also 23 experiences greater fluctuations within the binding pocket in KEAP1.This difference mostly arises from the fact that 24 interacts with hydrophobic residues Phe577 and Ala556 whereas 23 does not.
Finally, thermodynamic calculations over the last 10 ns of the production run, using the Molecular Mechanics Generalized Born Surface Area approach (MMGBSA) [65], revealed a noticeably higher affinity for 24 relative to 23 that is in consonance with the experimental finding that 24 leads to a larger activation and expression of NRF2 in AREc32 cells (Table 1).

Conclusions
By structural modifications of our described hMT 1 R and (1), new indolee and naphthaleneeoxadiazolone MTDLs have been obtained and evaluated on targets related to oxidative stress and neuroinflammation.Introduction of lipophilic substituents on the nitrogen of the oxadiazolone ring improved their in vitro CNS permeation but was detrimental for their affinity for the GPCRs hMT 1 R and hMT 2 R. The addition of a methyl group in the aliphatic linker of 1 or the replacement of the oxadiazolone ring by a 1,3,4-oxadiazol-2-amine scaffold yielded compounds with K i s in the hundred-nanomolar range in melatonin targets, but they were predicted as not CNSpermeable.
Introduction of a double bond in the linker provided interesting effects on the behaviour of these less flexible indole-and naphthalene-oxadiazolone compounds.The 5-methoxyindole derivatives 23 and 24 displayed remarkable selectivity towards QR2 (K i ¼ 6.6 and 3.2 nM, respectively) in reference to hMT 1 R and hMT 2 R (K i s > 100 nM).In contrast, when the 5-methoxyindole of 23 was replaced by a 7-methoxynaphthalene the selectivity changed drastically, since 44 was preferably bound to hMT 2 R (K i ¼ 5.1 nM) compared to hMT 1 R and QR2 (K i ¼ 260 and 270 nM, respectively).Moreover, 44 was functionally characterized as a potent hMT 2 R agonist with a maximum activity of 85% with respect to melatonin and an EC 50 of 1.1 nM.
Conformational analyses using quantum mechanics revealed that the indole-NH-oxadiazolone 23 presents a co-planar arrangement of the ring systems, whereas the naphthalene-NH-oxadiazolone 44 shows a dihedral angle of ~40 between both heterocycles.These differences could explain the observed selectivity of 23 and 44 towards QR2 and hMT 2 R, respectively.Molecular modelling and computer simulations, using the crystal structure of QR2 and the recently described XFEL structures of chimeric MT 1 R and MT 2 R, have provided a rationale for the experimental selectivity and afford valuable insight that can aid in future drug design endeavours.
In the ORAC assays, indole derivatives were found to be as good radical scavengers as melatonin, whereas naphthalene-based compounds showed worse ORAC values.Thus, the radical absorbance capacity was mainly linked to the indole ring.
In relation to NRF2 induction, best results were obtained with the indoleeNH-oxadiazolone derivative with a double bond in the linker 23 and its N-propargylic counterpart 24 (CD ¼ 15.1 and 1.8 mM, respectively).When comparing with other derivatives, we can infer the importance of the presence of the indole core and a double bond in the connector to achieve a good induction of NRF2.From the SPR experiments using the KEAP1 Kelch domain, we could infer that the cellular NRF2 induction by compounds 23 and 24 could be due to the binding of these molecules to KEAP1.Moreover, molecular dynamics simulations on the KEAP1 Kelch domain showed that 23 and 24 interact with several essential residues involved in the binding KEAP1-NRF2.Thermodynamic calculations revealed higher affinity for 24 than for 23, according to the experimental finding that 24 leads to a larger activation and expression of NRF2 in cellular experiments.
Several saturated indoleeoxadiazolone derivatives were modest hLOX-5 inhibitors and inactive in hMAOs.However, the introduction of a double bond in the linker again reversed selectivity, as the indoleeNH-oxadiazolone 23 inhibited hMAOs but it is inactive in hLOX-5.Only the dihydronaphthalene-oxadiazolone 33 was found to inhibit the three enzymes, with IC 50 values in the two-digit micromolar range.
In neurogenic assays, several of the new melatoninoxadiazolone derivatives stimulated the differentiation of NSC to a consolidated neuronal stage, as they could contribute to the autorepair processes of the CNS.
Finally, it can be emphasized that it is possible to modulate the selectivity of these new oxadiazolone-based MTDLs by modifying the side chain functionality and co-planarity with the indole or naphthalene ring.

Chemistry. General methods
High-grade reagents and solvents were purchased from Fig. 7. Dynamic conformation of KEAP1-ligand complexes.A) KEAP1-23 complex.B) KEAP1-24 complex.Water molecules and ions are not depicted for clarity.KEAP1 is represented as a brown molecular surface [66].Residues involved into protein-ligand interaction are represented as sticks by colour type: green are polar residues, blue are positive charged residues and white are hydrophobic residues.Ligands are represented as sticks with grey carbon atoms, blue nitrogen atoms, red oxygen atoms and white hydrogen atoms.Thinner white sticks correspond to the ligand conformations during the last 10ns of the MD production.These images were made with VMD software support [67].The most stable molecular docking poses found for 23 and 24 on the 3D-structures of the KEAP1 Kelch domain with pdbcode 4XMB were used as the starting reference structure.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)commercial suppliers and were used without further purification.Reactions were followed by thin-layer chromatography (TLC) or high-performance liquid chromatography coupled to mass spectrometry (HPLC-MS).TLC was performed using Merck silica gel 60 F254 plates and compounds were detected under UV-light (l ¼ 254 or 365 nm) and/or stained with 10% wt.phosphomolybdic acid, ninhydrin, or vanillin solutions in EtOH.HPLC-MS data were acquired in an Alliance Watters 2695 coupled to a quadrupole mass spectrometer (Micromass ZQ).HPLC was equipped with a SunFire C 18 column (3.5 mm, 4.6 mm Â 50 mm) and a UVevisible photodiode array detector (l ¼ 190e700 nm).MS were acquired in an electrospray ionization (ESI) interface working in the positive or negative-ion mode.Microwave driven reactions were performed in a Biotage Initiator 2.5 reactor.Unless otherwise stated, products were purified by automatized flash column chromatography using an IsoleraOne (Biotage) equipment, with different cartridges of silica gel Biotage ZIP KP-Sil 50 mm.Alternatively, we used preparative TLC on Merck silica gel 60 F254 plates or semipreparative HPLC on a Waters Autopurification system with UVevisible photodiode array detector (l ¼ 190e700 nm) coupled to a quadrupole mass spectrometer (3100 Mass Detector).Purity of all compounds (!95%) was determined by HPLC on a Waters 2690 equipment, at a flow rate of 1.0 mL/min, with a UVevisible photodiode array detector (l ¼ 190e700 nm), using a SunFire C 18 column (3.5 mm, 4.6 mm Â 50 mm).The gradient of the mobile phase consisted of H 2 O:ACN with formic acid (0.1%), and the gradients time (g.t.) are indicated for each compound.Melting points (mp) (uncorrected) were determined in a MP70 apparatus (Mettler Toledo).Nuclear magnetic resonance ( 1 H NMR and 13 C NMR) spectra were obtained in the following spectrometers: Varian INOVA-300, Varian INOVA-400, Varian Mercury-400 or Varian Unity-500.Chemical shifts (d) are reported in parts per million (ppm) relative to internal tetramethylsilane scale, and coupling constants (J) are expressed in hertz (Hz).2D NMR experiments namely, homonuclear correlation spectroscopy (H,HeCOSY), heteronuclear multiple quantum correlation (HMQC) and heteronuclear multiple bond correlation (HMBC) were acquired to assign protons and carbons of new structures.High resolution mass spectra (HRMS) analyses were carried out by using an Agilent 1200 Series LC system (equipped with a binary pump, an autosampler, and a column oven) coupled to a 6520 quadrupole-time of flight (QTOF) mass spectrometer.ACN:H 2 O (75:25, v:v) was used as mobile phase at 0.2 mL/min.The ionization source was an ESI interface working in the positive-ion mode.The electrospray voltage was set at 4.5 kV, the fragmentary voltage at 150 V and the drying gas temperature at 300 C. Nitrogen (99.5% purity) was used as nebulizer (207 kPa) and drying gas (6 L/min).

General procedure for the synthesis of 2e15
A mixture of 1 (1 equiv.),K 2 CO 3 (1.2equiv.)and the corresponding halide (1.2 equiv) in acetone (7 mL/mmol) was heated under MW-irradiation at 120 C for 10 min.Solvent was evaporated under reduced pressure, EtOAc (10 mL/mmol) was added and the organic layer was washed with H 2 O (x3) and brine and then, dried over MgSO 4 , filtered, and evaporated to dryness under reduced pressure.The crude was purified by column chromatography using the appropriate eluent to afford the corresponding alkyl derivative 2e15.

5-[1-(5-Methoxy-1H-indol-3-yl)propan-2-yl]-1,3,4oxadiazol-2(3H)-one (21)
Entitled compound was prepared in a multi-step synthesis starting from a mixture of 5-methoxy-1H-indole-3-carbaldehyde (18, 1.0 g, 5.7 mmol) and (carbethoxyethylidene) triphenylphosphorane (1.3 equiv) in DCM (50 mL) that was refluxed overnight.Solvent was evaporated under reduced pressure to give intermediate 19 [68] as a white solid of mp 116e117 C (1.4 g, 95% yield) after a column chromatography (pure hexane to hexane: EtOAc 1:1).Then, the a,b-unsaturated ester 19 (500 mg, 1.93 mmol) in EtOH (20 mL) was treated with a catalytic amount of PdeC (5%) under N 2 atmosphere.N 2 was displaced by H 2 and the flask was sealed up with a septum.A balloon with H 2 was connected with a needle to stir at rt overnight.Then, PdeC was removed by filtration and the solvent was evaporated to dryness under reduced pressure, obtaining 20 [69] as a white solid (500 mg, > 99% yield), which was used without further purification.A mixture of ester 20 (360 mg, 1.38 mmol) and hydrazine hydrate in EtOH (15 mL) was heated to 150 C for 45 min under MW irradiation.After the solvent was evaporated, DCM (10 mL) was added to the crude and the organic layer was washed with saturated aq.solution of NaHCO 3 (3x10 mL), brine (10 mL) and dried over MgSO 4 .Solvent was filtered and evaporated under reduced pressure to obtain the corresponding hydrazide, which was not isolated but identified by its HPLC-MS data (15:95-g.t.5), t R 2.97 min, m/z ¼ 248.17 Derivative 23 was prepared in a multi-step synthesis as explained below.To a solution of the commercial aldehyde 18 (500 mg, 2.8 mmol) in pyridine (20 mL), malonic acid (2 equiv) and piperidine (13 mL/mmol) were added and the mixture was stirred at 70 C overnight under N 2 atmosphere.Solvent was evaporated under reduced pressure and EtOAc was added.The organic layer was washed with H 2 O (3x10 mL) and 10% citric acid solution (10 mL), dried over MgSO 4 , filtered, and evaporated to dryness, to give 22 (97% yield) as a white solid of mp 208e211 C (bibl [70].197e200 C).Then, to a suspension of the a,b-unsaturated acid 22 (500 mg, 2.30 mmol) and activated 4 Å molecular sieves in anhydrous ACN (40 mL) at rt under N 2 atmosphere, HOBt (1.2 equiv), EDC$HCl (1.2 equiv) and DMAP (0.12 equiv) were orderly added.The mixture was stirred for 3 h until the complete acid activation and then, N 2 H 4 $H 2 O (excess) was added.Once finished the reaction, H 2 O was added, extracted with DCM and washed with saturated NaHCO 3 (aq).The organic layer was separated and evaporated under reduced pressure to obtain the corresponding N- Alkylation of 23 (60 mg, 0.23 mmol) with propargyl bromide (1.2 equiv) in the presence of K 2 CO 3 (1.2equiv) under MWirradiation at 120 C for 10 min, using acetone (7
4.4.2.Binding assays in quinone reductase-2 Evaluation of the affinity of compounds for QR2 in the hamster brain was performed by a radioligand binding assay (Eurofins-CEREP, France, catalog ref. 0088) [39].Membrane homogenates of brain (750 mg protein) are incubated for 60 min at 4 C with 0.1 nM [ 125 I]iodomelatonin in the absence or presence of the test compound in a buffer containing 50 mM Tris-HCl (pH 7.4) and 4 mM CaCl 2 .Nonspecific binding is determined in the presence of 30 mM melatonin.Following incubation, the samples are filtered rapidly under vacuum through glass fiber filters (Filtermat B, Wallac) presoaked with 0.5% PEI (pH 7.4) and rinsed several times with an ice-cold buffer containing 50 mM Tris-HCl and 2 M NaCl using a 48sample cell harvester (Mach II, Tomtec).The filters are dried then counted for radioactivity in a scintillation counter (Betaplate 1204, Wallac) using a solid scintillator (Meltilex B/HS, Wallac).

Methodology and expression of the results of binding assays in hMT 1 R, hMT 2 R and QR2
Radioligand displacements were measured at a fixed concentration of compound in each receptor subtype in 3 independent experiments.Then, compounds displaying a radioligand displacement above 80% were re-evaluated using a range of 5 different concentrations in 3 independent experiments, to determine the binding constants (K i ).The standard reference compound was melatonin, which was tested in each experiment at several concentrations to obtain a competition curve from which its IC 50 was calculated.
The IC 50 values (concentration causing a half-maximal inhibition of control specific binding) and Hill coefficients (nH) were determined by non-linear regression analysis of the competition curves generated with mean replicate values using Hill equation curve fitting where Y ¼ specific binding, A ¼ left asymptote of the curve, D ¼ right asymptote of the curve, C ¼ compound concentration, and nH ¼ slope factor.The inhibition constants (K i ) were calculated using the Cheng Prusoff equation where L ¼ concentration of radioligand in the assay, and K D ¼ affinity of the radioligand for the receptor.A scatchart plot is used to determine the K D .In all cases, K i s are expressed as the mean ± SEM of three independent experiments.

Functional assay at human MT 2 R
Evaluation of the agonist activity of 44 at hMT 2 R expressed in transfected CHO cells was performed by measuring its effects on cAMP modulation using the HTRF detection method (Eurofins-CEREP, France, catalog ref.G028-2092) [38].The cells were suspended in HBSS buffer (Invitrogen) complemented with 20 mM HEPES (pH 7.4) and 500 mM IBMX, then distributed in microplates at a density of 7.10 3 cells/well in the presence of either of the following: HBSS (basal control), the reference agonist at 10 nM (stimulated control) or various concentrations (EC 50 determination), or the test compound.Thereafter, the adenylyl cyclase activator NKH 477 is added at a final concentration of 5 mM.Following 10 min incubation at 37 C, the cells are lysed and the fluorescence acceptor (D2-labelled cAMP) and fluorescence donor (anti-cAMP antibody labelled with europium cryptate) are added.After 60 min at room temperature, the fluorescence transfer is measured at lex ¼ 337 nm and lem ¼ 620 and 665 nm using a microplate reader (Envison, PerkinElmer).The cAMP concentration is determined by dividing the signal measured at 665 nm by that measured at 620 nm (ratio).
The EC 50 value (concentration producing a half-maximal response) was determined by non-linear regression analysis of the concentration-response curves generated with mean replicate values of three independent experiments using Hill equation curve fitting where Y ¼ response, A ¼ left asymptote of the curve, D ¼ right asymptote of the curve, C ¼ compound concentration, and nH ¼ slope factor.
The standard reference agonist is melatonin, which was tested in each experiment at several concentrations to generate a concentration-response curve from which its EC 50 value was calculated.

Oxygen radical absorbance capacity assay (ORAC)
Antioxidant activities were measured using the ORAC method [43], in a Polarstar Galaxy plate reader (BMG Labtechnologies GmbH, Offenburg, Germany) with 485-P excitation and 520-P emission filters.The equipment was controlled by the Fluorostar Galaxy software (version 4.11-0) for fluorescence measurement.2,2 0 -Azobis-(amidinopropane) dihydrochloride (AAPH), (±)-6hydroxy-2,5,7,8-tetramethylchromane-2-carboxylic acid (trolox) and fluorescein (FL) were purchased from Sigma-Aldrich.The reaction was carried out in 75 mM phosphate buffer (pH 7.4) and the final reaction mixture was 200 mL.Antioxidant (20 mL) and FL (120 mL; 70 mM, final concentration) solutions were placed in a black 96-well microplate (96F untreat, Nunc).The mixture was preincubated for 15 min at 37 C and then, AAPH solution (60 mL, 12 mM, final concentration) was added rapidly using a multichannel pipette.The microplate was immediately placed in the reader and the fluorescence recorded every minute for 80 min.The microplate was automatically shaken prior each reading.Samples were measured at eight different concentrations (0.1e1 mM).A blank (FL þ AAPH in phosphate buffer) instead of the sample solution and eight calibration solutions using trolox (1e8 mM) were also carried out in each assay.All the reaction mixtures were prepared in duplicate, and at least three independent assays were performed for each sample.Raw data were exported from the Fluostar Galaxy Software to an Excel sheet for further calculations.Antioxidant curves (fluorescence vs. time) were first normalized to the curve of the blank corresponding to the same assay, and the area under the fluorescence decay curve (AUC) was calculated.The net AUC corresponding to a sample was calculated by subtracting the AUC corresponding to the blank.Regression equations between net AUC and antioxidant concentration were calculated for all the samples, measured in three independent experiments.ORAC values were expressed as trolox equivalents by using the standard curve calculated for each assay, where the ORAC value of trolox was taken as 1.0.

NRF2 induction by luciferase activity in AREc32 cells
AREc32 cells were kindly provided by Prof. Roland Wolf (University of Dundee, UK) [48].AREc32 cells were cultured in DMEM with glutamax and high glucose, supplemented with 1% penicillinstreptomycin (10,000 units), geneticin (0.8 mg mL À1 ) and 10% FBS, at 37 C in a 5% CO 2 air atmosphere.AREc32 cells were sub-cultured in 96-well white plates at 2 Â 10 4 cells/well density.Cells were incubated with increasing concentrations of tested compounds in duplicate for 24 h.Luciferase expression was assayed with the Luciferase Assay System by luminescence in an Orion II microplate luminometer (Berthold, Germany).Luciferase activity increase was normalized to basal conditions considered as 1.CD (concentration that duplicates the luciferase activity) values are expressed as the mean ± SEM of four independent experiments in duplicate.

Surface plasmon resonance (SPR) assays
SPR experiments were performed at 20 C with a Biacore X-100 apparatus (Biacore, GE) in PBS-T with 1 mM DTT and 2% de DMSO (running buffer) at 25 C.The KEAP1 Kelch domain protein was immobilized on a CM5 sensor chip (Biacore, GE) following standard amine coupling method [75].The carboxymethyl dextran surface of the flow cell 2 was activated with a 7-min injection of a 1:1 ratio of 0.4 M EDC and 0.1 M NHS.The protein was coupled to the surface with a 7-min injection at 100 mg/mL in 10 mM sodium acetate, pH 5.0.The unreacted N-hydroxysuccinimide esters were quenched by a 7-min injection of 0.1 M ethanolamine-HCl (pH 8.0).The level of immobilization was 7000 RUs.Flow cell 1 treated as a flow cell 2 (amine coupling procedure) without protein was used as a reference.Prior to use 10 mM stock solutions of compounds were diluted several times until 100 mM final concentration in the running buffer.Typically, a series of different compounds was injected onto the sensor chip a flow rate of 45 mL/min for a period of 1 min followed by a dissociation period of 1 min.After the dissociation process an extra-wash treatment was made over the flow cells with a 50% DMSO solution.No regeneration was needed.Sensograms data were double-referenced and solvent corrected using the Biaevaluation X-100 software (Biacore, GE).

Inhibition of human monoamine oxidases (hMAO-A and hMAO-B)
MAO inhibition measurements were evaluated following the general procedure previously described [76].Briefly, test drugs and adequate amounts of recombinant hMAO-A or hMAO-B (Sigma-Aldrich, Spain) required and adjusted to oxidize 165 pmol of ptyramine/min in the control group, were incubated for 15 min at 37 C in a flat-black-bottom 96-well microtest plate (BD Biosciences, Franklin Lakes, NJ) placed in the dark fluorimeter chamber.The reaction was started by adding 200 mM Amplex Red reagent (Molecular Probes, Inc., Eugene, OR), 1 U/mL horseradish peroxidase, and 1 mM p-tyramine.Then, the production of resorufin was quantified at 37 C in a multidetection microplate fluorescence reader (FLX800, Bio-Tek Instruments, Inc., Winooski, VT) based on the fluorescence generated (excitation, 545 nm; emission, 590 nm).The specific fluorescence emission was calculated after subtraction of the background activity, which was determined from wells containing all components except the hMAO isoforms, which were replaced by a sodium phosphate buffer solution.IC 50 values were the mean ± SD of three independent experiments performed in triplicate.

Inhibition of human lipoxygenase-5 (hLOX-5)
The fluorescence-based enzyme method developed by Pufahl et al. was followed [77] emission: 520 nm) were performed on a FLUOstar OPTIMA (BMG LABTECH, Offenburg, Germany).IC 50 is defined as the concentration of compound that inhibits enzymatic activity by 50% over the control of untreated enzyme.Data were the mean ± SD of three independent experiments performed in triplicate.

In vitro bloodebrain barrier permeation assay (PAMPA-BBB)
Prediction of the brain penetration was evaluated using a parallel artificial membrane permeation assay (PAMPA-BBB), in a similar manner as previously described [27,29,78e80].Pipetting was performed with a semi-automatic pipettor (CyBi®-SELMA) and UV reading with a microplate spectrophotometer (Multiskan Spectrum, Thermo Electron Co.).Commercial drugs, phosphate buffered saline solution at pH 7.4 (PBS), and dodecane were purchased from Sigma, Aldrich, Acros, and Fluka.Millex filter units (PVDF membrane, diameter 25 mm, pore size 0.45 mm) were acquired from Millipore.The porcine brain lipid (PBL) was obtained from Avanti Polar Lipids.The donor microplate was a 96-well filter plate (PVDF membrane, pore size 0.45 mm) and the acceptor microplate was an indented 96-well plate, both from Millipore.The acceptor 96-well microplate was filled with 200 mL of PBS: ethanol (70:30) and the filter surface of the donor microplate was impregnated with 5 mL of porcine brain lipid (PBL) in dodecane (20 mg mL À1 ).Compounds were dissolved in PBS: ethanol (70:30) at 100 mg mL À1 , filtered through a Millex filter, and then added to the donor wells (200 mL).The donor filter plate was carefully put on the acceptor plate to form a sandwich, which was left undisturbed for 120 min at 25 C.After incubation, the donor plate is carefully removed and the concentration of compounds in the acceptor wells was determined by UVeVis spectroscopy.Every sample is analysed at five wavelengths, in four wells and at least in three independent runs, and the results are given as the mean ± standard deviation.In each experiment, 11 quality control standards of known BBB permeability were included to validate and normalize the analysis set.

Neurogenic assays
Ten three-month-old male C57BL/6 mice were used throughout this study.All experimental procedures were previously approved by the Ethics Committee for Animal Experimentation of the CSIC in accordance with the European Communities Council, directive 2010/63/EEC and National regulations, normative 53/2013.Special care was taken to minimize animal suffering.Neural stem cells (NSCs) were isolated from the SGZ of the hippocampal dentate gyrus and prepared following previously published methods [58,81].The isolated NSCs were seeded into 12-well dishes at a density of ~40,000 cells per cm2 in DMEM/F12 (1:1) containing N2 medium.NSCs were grown as free-floating cells under proliferative conditions (10 ng/mL epidermal growth factor þ 10 ng/mL fibroblast growth factor) allowing the formation of neural stem cellsenriched growing spheres also called neurospheres (NS).After 7 days, with all NS having the same stage and size, cultures were treated daily for 7 days with vehicle or 10 mM of compound 6 (10).Now, NS were adhered onto 100 mg mL À1 poly-L-lysine-coated coverslips and treated for 3 additional days under differentiation conditions (medium containing 1% fetal bovine serum and without exogenous growth factors) [82].Once neurospheres were differentiated (72 h), coverslips were processed as previously described 41  and used for immunofluorescence analysis.Then the expression of neuronal markers was analysed.The following antibodies linked to neurogenesis were used: b-III-tubulin polyclonal antibody (TuJ clone; Abcam), a protein expressed at early stages of neurogenesis and a monoclonal microtubule-associated protein type 2 (MAP-2) antibody, a classical marker of late neuronal maturation.To visualize primary antibodies Alexa-fluor-labelled secondary antibodies (Molecular probes) were used.Nuclei were stained with DAPI.Fluorescent representative images from four independent experiments were acquired in a LSM710 laser scanning spectral confocal microscope (Zeiss).Confocal microscope settings were adjusted to produce the optimum signal to noise ratio.

In silico model building of ligands and proteins
The molecular graphics program PyMOL [83] was employed in the construction of the initial molecular models, trajectory visualization and 3D figure generation.Ab initio geometry optimization of the derivatives studied was achieved using a 6-31G* basis set and the Hartree-Fock method, as implemented in the Gaussian 09 program [84].The SCRF ¼ (Solvent ¼ Water) option was used to place the solute in a cavity within the solvent reaction field and mimic hydration effects.For melatonin, 2-iodomelatonin and resveratrol, the optimized geometries were also written in Crystallographic Information File (CIF) format [85] using eLBOW [86] for re-refinement of several human quinone reductase II (QR2) X-ray crystal structures retrieved from the Protein Data Bank (entries 2QWX, 2QX8 [87], and 1SG0 [60]) against structure factor amplitudes using Phenix [88,89] and Coot [90].Six macrocycles were used in each refinement round and concerted motions of atomic groups were taken into account by means of the translationlibration-screw (TLS) model [91].PDB entries 6ME3 [62] and 6ME6 [63] containing the XFEL structures of chimeric MT 1 and MT 2 receptor constructs, respectively, in complex with 2phenylmelatonin were used as templates to model the complexes of the studied compounds with the wild-type human receptors with the aid of the Swiss-Model server [92,93].Consequently, our MT 1 and MT 2 receptor models (UniProt [94] identifiers P48039 and P49286, respectively) contained the original truncations at both termini but had all the introduced point mutations replaced by the natural amino acids and a short added sequence bridging the ICL3 gap (residues 219e227 and 232e240 in MT 1 and MT 2 receptors, respectively).The conserved disulfide bridge between C100 3.25 and C177 ECL2 in MT 1 and between C113 3.25 and C190 ECL2 in MT 2 were explicitly defined.
Atom-centered charges for all ligands were calculated on the 6-31G*-optimized geometries with the sqm program [95] in Amber-Tools17 and the default AM1-BCC option [96], so as to reproduce the molecular electrostatic potential in a way consistent with the ff14SB force field parameter set [97] in AMBER 16 (http://ambermd.org).

Molecular docking
The protein in the X-ray crystal structure of KEAP1 Kelch Domain was prepared with the Protein Preparation procedure as implemented in Maestro.All hydrogens were added and the charge states of titratable amino acids were defined using the PROTKA procedure, followed by H-bond optimization and constrained relaxation to 0.2 rmsd tolerance.The resulting structures were used to generate docking grids and perform docking studies with compounds 23 and 24 using GLIDE [98].The ligands were prepared with the LigPrep module using the parameters of OPLS3 Force Field.Generation of possible tautomers was done with Epik routine in Maestro at pH range of 7 ± 2. All the modelling was done with Maestro 11.5 Release 2018-1 (Schrodinger LLC).

Molecular dynamics simulations
The complexes of 44 with MT 1 R and MT 2 R were embedded in a lipid bilayer made up of 240 1,2-dioleoyl-sn-glycero-3phosphocholine (DOPC) and 21 cholesterol molecules with the aid of the web-based CHARMM-GUI Membrane Builder graphical user interface [99].The system size along the Z axis was determined by specifying a 10 Å thickness of 150 mM NaCl aqueous solution on each side of the membrane.The AMBER ff14SB [97] and lipid14 [100] force fields were used for protein and lipids, respectively.
For each complex studied, all hydrogens and water molecules were first reoriented in the electric field of the solute and then all protein residues, ligands and waters were relaxed by performing 25 000 steps of steepest descent followed by 25 000 steps of conjugate gradient energy minimization.The resulting geometry-optimized coordinate sets were then heated up to 300 K using the Langevin thermostat; thereafter the temperature was maintained at 300 K whereas the system's dimensions and density were equilibrated at a constant pressure of 1 atm by means of an anisotropic Berendsen weak-coupling barostat, essentially as described [100].The uMD simulations were run for 200 ns using the pmemd.cudacode [101], as implemented in AMBER16, running on single Nvidia® GPUs.The coupling constants for the temperature and pressure baths were 1.0 and 0.2 ps, respectively.The application of SHAKE to all bonds allowed an integration time step of 2 fs to be used.A cutoff distance of 9 Å was selected for the nonbonded interactions and the list of nonbonded pairs was updated every 25 steps.Periodic boundary conditions were applied and electrostatic interactions were represented using the smooth particle-mesh Ewald method [102] with a grid spacing of 1 Å.Trajectory snapshots were saved every 0.1 ns for further analysis.Besides, a simulated annealing procedure [103] was followed every 5 ns to cool the system down from 300 to 273 K over a 1-ns period.Thereafter the geometries of these "frozen" complexes were optimized by carrying out an energy minimization until the root-mean-square of the Cartesian elements of the gradient was less than 0.01 kcal mol À1 Å À1 .A similar procedure was used to simulate the fully solvated complexes of human QR2 with compounds using our refined structure from PDB entry 1SG0; in these cases, however, we found it necessary to add distance restraints (100 kcal mol À1 Å À2 ) to the Zn 2þ coordination sphere and weak positional restraints (1 kcal mol À1 Å À2 ) to the Ca trace of the protein.The highly converged ensemble of 40 low-energy configurations thus obtained for each complex is expected to represent more accurately the global energy minimum than isolated trajectory frames [104,105].

Potentiometric pK a determination
Titrations were carried out at 25 C in 0.15 M KCl solution (aq) in a SiriusT3 equipment (Sirius Analytical Instruments Ltd, East Sussex, Britain) equipped with an Ag/AgCl double junction reference pH electrode, a Peltier temperature system and a turbidity sensor.Standardized 0.5 M KOH and 0.5 M HCl were used as titration reagents.KOH solution was standardized by potassium phthalate.The pK a value is the mean of 3 titrations ± SD [106].

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

The 5 -
methoxyindole-oxadiazolamine 16 inhibited both hMAO-B and hLOX-5 (IC 50 ¼ 37.0 and 60.6 mM, respectively).Introduction of a double bond in the linker changed this profile, as the 5methoxyindole-oxadiazole 23 displayed moderate inhibition of MAO-A and MAO-B (IC 50 ¼ 53.0 and 68.3 mM, respectively) although it was inactive in LOX-5 (IC 50 > 100 mM).This positive effect of the double bond in the linker was not valid for the indoleamide and the naphthalene-oxadiazole series, because 25, 26, 44, and 45 were inactive in the three enzymatic assays.Surprisingly, the introduction of a propargyl substituent in the oxadiazolone ring of the MAO-A/B inhibitor 23 to give 24 led to a total lack of activity on these enzymes (data not shown).Contrariwise, the dihydronaphthalene-oxadiazolone 33 inhibited hMAO-A, hMAO-B and hLOX-5 with IC 50 values of 40.6, 35.7 and 12.5 mM, respectively (Table

Fig. 3 .
Fig. 3. Melatonin-based compounds promote neuronal differentiation on SGZ-derived neurospheres.Neurospheres were cultured in the presence of compounds (10 mM) for 7 days and later adhered for 3 days to allow differentiation in the presence of compounds.Confocal images show the expression of the neuronal markers TuJ clone (early neurogenesis) in green and MAP-2 (mature neurons) in red.DAPI was used for nuclear staining.Fluorescent representative images from four independent experiments were displayed.Scale bar ¼ 100 mm (images of inactive compounds are not shown).(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 4 .Fig. 5 .
Fig. 4. Global energy conformations of indoleeNH-oxadiazolone 23 (A) and naphthaleneeNH-oxadiazolone 44 representative molecules (B), together with the rotational barriers (in vacuum, blue; using an implicit solvent model for water, orange) upon rotation about the bonds connecting atoms labelled 2 and 3 (top) and 4 and 5 (bottom).(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 6 .
Fig.6.Model of a membrane-embedded hMT 2 R (rainbow coloured) in complex with naphthaleneeNH-oxadiazolone derivative 44 (sticks with C atoms coloured in grey).The enlarged region in the box shows the superposition of eight cooled and energy-minimized snapshots from the last 170e200 ns of unrestrained molecular dynamics (uMD) simulation.The starting reference structure was PDB entry 6ME8.

1 .
Assays for human MT 1 R and MT 2 R subtypes These experiments were performed at Eurofins-CEREP (France), using hMT 1 R and hMT 2 R stably transfected in Chinese hamster ovary cells (catalog refs.1538 and 1687, respectively)[37,38].In the case of hMT 1 R cell membrane homogenates (about 80 mg protein) are incubated for 240 min at 22 C with 0.01 nM [ 125 I]iodomelatonin in the absence or presence of the test compound in a buffer containing 50 mM Tris-HCl (pH 7.4), 5 mM MgCl 2 and 0.1% BSA.For hMT 2 R cell membrane homogenates (6 mg protein) are incubated for 120 min at 37 C with 0.05 nM [ 125 I]iodomelatonin in the absence or presence of the test compound in a buffer containing 50 mM Tris-HCl (pH 7.4) and 5 mM MgCl 2 .

Table 1
Binding constants at hMT 1 R, hMT 2 R and QR2 (K i , nM), or percentage of radioligand displacement at the specified concentration (in brackets), antioxidant properties (ORAC, trolox equiv.)and NRF2 induction capability (CD, mM) of new indole and naphthalene derivatives.
(100 mU/well), bovine glutathione peroxidase (GPx, 25 mU/well) and reduced glutathione (GSH, 1 mM).Compounds to be tested were added to the test solution prior to AA and ATP, and preincubated for a period of 10 min at room temperature.Then, the AA and ATP substrates were added; the enzymatic reaction allowed to progress for 20 min and ended by the addition of 40 mL of aceto- nitrile.The fluorescence measurements (excitation: 485 nm;