Biological and computational evaluation of resveratrol inhibitors against Alzheimer’s disease

Abstract

It has been reported that beta amyloid induces production of radical oxygen species and oxidative stress in neuronal cells, which in turn upregulates β-secretase (BACE-1) expression and beta amyloid levels, thereby propagating oxidative stress and increasing neuronal injury. A series of resveratrol derivatives, known to be inhibitors of oxidative stress-induced neuronal cell death (oxytosis) were biologically evaluated against BACE-1 using homogeneous time-resolved fluorescence (TRF) assay. Correlation between oxytosis inhibitory and BACE-1 inhibitory activity of resveratrol derivatives was statistically significant, supporting the notion that BACE-1 may act as pivotal mediator of neuronal cell oxytosis. Four of the biologically evaluated resveratrol analogs demonstrated considerably higher activity than resveratrol in either assay. The discovery of some “hits” led us to initiate detailed docking studies associated with Molecular Dynamics in order to provide a plausible explanation for the experimental results and understand their molecular basis of action.

• Alzheimer’s disease
• BACE-1
• docking
• free energy calculations
• MM–PBSA
• molecular dynamics
• resveratrol analogs

Introduction

Alzheimer’s disease (AD) is a neurodegenerative disease of the central nervous system associated with progressive memory loss resulting in dementia. Two pathological characteristics are observed in AD patients at autopsy: extracellular plaques and intracellular tangles in the hippocampus, cerebral cortex and other areas in the brain essential for cognitive function. Plaques are formed mostly from the deposition of amyloid β (Aβ), a peptide derived from amyloid precursor protein (APP)1. β-secretase (BACE-1) is an enzyme that belongs to the family of aspartic proteases and selectively cleaves APP to create Aβ. BACE-1 is regarded as a key target for therapeutic interventions in AD because it is the principal BACE-1 in neurons responsible for Aβ generation in the brain, and targeted deletion of BACE-1 in APP transgenic mice completely abolishes the production and deposition of Aβ2.

It has been reported that beta amyloid induces production of radical oxygen species and oxidative stress in neuronal cells, which in turn upregulates BACE-1 expression and beta amyloid levels, thereby propagating oxidative stress and increasing neuronal injury3. In line with these findings, it has been reported that BACE-1 expression in neuronal cells is induced by 4-hydroxynonenal, a mediator of oxidative stress that is detected in the brain of patients with AD4, and that upregulation of neuronal BACE-1 expression by beta amyloid is inhibited by antioxidants5. It has also been reported that acrolein, a by-product of oxidative stress, increased APP as well as BACE-1 expression and induced cell death in HT22 cells6, an immortalized neuronal cell line believed to model oxidative stress-induced cell death, and that caffeic acid and its phenethyl ester, known to possess antioxidant and neuroprotective properties, prevented acrolein-induced upregulation of BACE-1 expression in these cells7. Notably, glutamate-challenged HT22 cells have been extensively used to study oxidative stress-induced neuronal cell death (oxytosis) and in the discovery and development of potentially neuroprotective antioxidants8–11. The two major drivers of AD progression, accumulation of beta amyloid-containing extracellular aggregates and oxidative stress, could be attributed to interactions of beta amyloid with metal ions12. Agents that are designed to prevent these interactions, such as metal ion chelators, are therefore thought to be able to delay the course of AD13. In accordance with this idea, exposure of HT22 to ferrous ions (Fe⁺⁺) was recently reported to increase lipid peroxidation and cellular oxidant levels14. Hence, HT22 cells are useful for evaluating the neuroprotective activity of agents potentially acting as metal chelators and/or direct radical scavengers as well as BACE-1 inhibitors.

For this study, we have synthesized new analogs and have evaluated the BACE-1 inhibitory potency of a series of 11 resveratrol analogs using a homogeneous TRF assay and their anti-oxidant neuroprotective efficacy using glutamate-challenged HT22 cells. We found that several of the compounds examined acted as inhibitors of BACE-1 and HT22 cell oxytosis. In addition, we found that the BACE-1 and oxytosis inhibitory activity of resveratrol analogs are significantly correlated. The above findings suggest that beneficial effects of resveratrol analogs against AD may be due to their dual action as BACE-1 and oxytosis inhibitors. Our results also support the notion that BACE-1 and beta amyloid peptide are pivotal mediators of oxytosis.

Results and discussion

BACE-1 has been regarded as a key target for therapeutic intervention in AD. The first BACE-1 inhibitors were peptide and peptidomimetic compounds showing a nanomolar affinity for BACE-1. The crystal structures of these inhibitors in complex with the enzyme have been utilized for structure-based projects that have led to the discovery of several classes of compounds with improved pharmacokinetics properties. As BACE-1 hydrolyzes APP and generates Aβ primarily within the endosomes of brain neurons, a clinically effective BACE-1 inhibitor must have the ability to penetrate the blood–brain barrier (BBB) and the neuronal membranes. Enzyme inhibitors with therapeutic potential should preferably be smaller than 700 Da15. Thus, the secondary metabolites of plants, which have relatively low-molecular weights and high lipophilicity, may act as archetypes of drugs against AD.

In recent years, several reports on the neuroprotective effects of various naturally occurring compounds in AD16, such as flavonoids17–21, catechins22, coumarins15, ellagic acid23, punicalagin23, resveratrol and resveratrol analogs, have been published24–26. Regarding the therapeutic potential of resveratrol in AD, Marambaud et al. suggested that it does not inhibit β-amyloid production since it has no effect on the β-amyloid-producing enzymes, β- and γ-secretases, but instead promotes the clearance of beta-amyloid from the brain27. In contrast, Choi et al. demonstrated that resveratrol and resveratrol oligomers significantly inhibit baculovirus-expressed BACE-1 in a dose-dependent manner, as assessed by FRET assay in vitro24^,26.

In this study, we report on the BACE-1-inhibitory activity of 11 previously synthesized trans-resveratrol analogs, bearing one (tert-butyl, 1-ethylpopyl) or two bulky electron donating groups ortho to 4′-OH9^,28.

In silico BBB permeation and Caco-2 permeability

BACE-1 and APP are both endocytosed to endosomes for cleavage. Endosomes are likely to be the major site for BACE-1 processing because of the acidic pH optimum of the enzyme activity. Consequently, orally active BACE inhibitors should penetrate across at least (i) the intestinal barrier when absorbed, (ii) the BBB when entering the brain, (iii) the cell membrane when reaching the site of action, and (iv) the endosomal membrane when binding to BACE29. The need for highly active and permeable compounds prompted us to predict the BBB and Caco-2 permeability of the compounds under study using the VolSurf30 program.

The PLS score space of the BBB model (Figure 1) is divided into the following: (i) a region (left) in which BBB ranges from negative values up to −0.3, this is the region in which compounds show no ability to cross the BBB; (ii) a small region (central) for BBB values from −0.3 to +0.3 (in between red and blue lines) where compounds show moderate permeability; and (iii) a region (right) in which BBB values ≥0.3 are found, this is the region in which compounds show ability to cross the BBB. From the BBB plot (Figure 1a), it can be deduced that all the studied compounds can cross the BBB with the exception of the resveratrol (1).

Figure 1. Projection of resveratrol analogs at (a) BBB and (b) Caco-2 model.

Regarding Caco-2 cell permeability, the 2D PLS score model can discriminate between permeable and less permeable compounds. When the spectrum color is active, red points refer to high permeability and blue points refer to low permeability. From Figure 1(b), it is predicted that most of the studied compounds are located in the region of high permeability, implying that they can be transported across the intestinal epithelium.

Biological assays

The studied compounds were initially evaluated against BACE-1 using a homogeneous TRF assay (Table 1). It was observed that four resveratrol derivatives displayed potencies higher than resveratrol (1) with the rank order: 11 > 10 > 3 > 5, which is in agreement with the corresponding calculated lipophilic values (LogP) with the VolSurf30 program 3.58, 6.47, 5.93, 4.84, 4.77 and 3.58. All the monomethoxy derivatives were inactive. The oxytosis inhibitory potencies (EC₅₀; Table 1) of resveratrol (1) and analogs thereof (2–9) have been reported previously9. The rank order of EC₅₀ of Table 1, 11 > 10 >> 5 ≥ 3 >> 1 (resveratrol), suggests that all the non-methoxylated analogs are moderately (3 and 5) or highly active (10 and 11) at the nanomolar range, while all the methoxy derivatives are inactive and that resveratrol is weakly active (EC₅₀ = 4.67 μM). Notably, the bi-substituted derivatives 10 and 11 displayed more than 10-fold higher inhibitory potency against oxytosis compared to the mono-substituted 3 and 5 and the latter more than 10-fold higher inhibitory potency compared to resveratrol for reasons discussed previously9.

Table 1. Chemical structure of resveratrol analogs and inhibition data. .

Compounds	R	X	Y	N	ΒΑCE-1 inhibition IC50 (μΜ)*	Oxytosis inhibition EC50 (μM)†
1	H	H	H	1	28 ± 9.0	4.667 ± 0.200
2	CH3	H	H	1	>100	>10
3	H	t-Bu	H	1	13 ± 2.3	0.484 ± 0.188
4	CH3	t-Bu	H	1	>100	>10
5	H	CH(Et)2	H	1	20 ± 4.4	0.432 ± 0.126
6	CH3	CH(Et)2	H	1	>100	>10
7	CH3	H	H	2	>100	>10
8	CH3	t-Bu	H	2	>100	>10
9	CH3	CH(Et)2	H	2	>100	>10
10	H	CH(Et)2	CH(Et)2	1	10 ± 0.7	0.030 ± 0.005
11	H	CH(Et)2	CH(Et)2	2	3 ± 0.6	0.012 ± 0.003

*Mean ± SEM of three independent experiments.

†EC₅₀ values (mean ± SEM of at least three independent experiments) are test compound concentrations required to maintain viability of glutamate-challenged HT22 cells to a level equal to 50% of that of not challenged cells and have been reported previously10.

The higher BACE-1 inhibitory potency of resveratrol analogs 11, 10, 3 and 5 compared to resveratrol (1) suggests that introduction of one or two bulky substituents ortho to 4′-OH increases activity. In particular, analogue 10 displayed higher activity (IC₅₀ = 10 μΜ) than 5 (IC₅₀ = 18 μΜ), which indicates that the introduction of a second ethyl propyl group, ortho to 4′-OH, increases activity. Introduction of a second double bond increased activity further: compound 11 displayed ca. 3-fold higher activity compared to 10 (IC₅₀ = 3 and 10 μΜ, respectively). It is worth noting that the same trend was also observed for oxytosis inhibition (rank order: 11 > 10 > 5 with EC₅₀ = 0.012, 0.030 and 0.432 μM, respectively). However, substituting a tert-butyl group for an ethyl propyl group ortho to 4′-OH affected activity non-significantly, since 3 displayed less than 2-fold higher potency against BACE-1 (IC₅₀ = 12 and 18 μΜ, respectively) and similar oxytosis inhibitory potency compared to compound 5 (EC₅₀ = 0.484 and 0.432 μM, respectively). Using resveratrol’s inhibitory potency as cutoff to classify compounds into high (BACE-1 inhibition IC₅₀ < 28 μM – oxytosis inhibition EC₅₀ < 4.667 μM) and low activity groups (which transforms BACE-1 IC₅₀ and oxytosis EC₅₀ to dichotomous variables) revealed a correlation statistically significant at the 0.01 level (2-tailed, N = 11 pairs, Spearman’s correlation coefficient = 1.000). Pearson’s correlation of oxytosis inhibitory potency with BACE-1-inhibitory potency (EC₅₀ values versus IC₅₀ values of Table 1) was also significant (N = 5 pairs, Pearson’s correlation coefficient = 0.822, p = 0.04, 1-tailed). These correlations are in line with the notion that BACE-1 and beta amyloid peptide are pivotal mediators of oxytosis.

Docking studies

Molecular docking studies were performed in an attempt to understand the molecular interactions between the compounds under study and BACE-1. The crystallographic structure of BACE-1 in complex with the hydroxyethylamine inhibitor LO1 (pdb code 1W51)31 was used to dock the derivatives. The docking procedure was validated by removing the crystallographic compound LO1 from the binding site and redocking it to the binding site of BACE-1.

The active site of BACE-1 contains a catalytic Asp dyad, Asp32 and Asp228, which can adopt multiple protonation states (unprotonated, monoprotonated and diprotonated). Due to the low resolution of the crystallographic structures, the protonation states of the Asp dyad cannot be determined in the presence of inhibitors. Thus, the determination of the exact protonation state of these Asp residues in the presence of specific inhibitors has become an area of intensive research both in understanding the reaction mechanism and in guiding the design of drugs against AD32–39.

In the presence of LO1 inhibitor, the pK_a values computed using the PROPKA server40^,41 are 7.27 for Asp32 and 3.02 for Asp228. The higher pKa value of Asp32 indicates that at the crystallographic pH (pH = 6.6), it exists in a monoprotonated state, while Asp228 remains unprotonated. This is in accordance with experimental and theoretical studies on other aspartic proteases, such as HIV-1 protease42–44. Thus, the docking calculation was performed considering two protonation states for Asp32: Asp32i (i = inner oxygen, OD1) and Asp32o (o = outer oxygen, OD2). At both protonation states, the root mean square deviation (RMSD) between the predicted conformation and the X-ray crystallographic conformation of LO1 was lower than 2 Å (RMSD = 1.345 Å at Asp32i and RMSD = 1.335 Å at Asp228o). Namely, extra precision (XP) GScores values were similar (ΧP GScore: −11.732 when applied with Asp32i and ΧP GScore: −11.850 when applied with Asp32o). In Figure 2, the predicted and the crystallographic conformations of LO1 are depicted in the active site of BACE-1.

Figure 2. Poses of the predicted conformation of LO1 (yellow) and the crystallographic structure at Asp32o (green).

Glide XP docking was carried out for the derivatives of resveratrol in the active site of the enzyme considering two protonation states of the catalytic dyad, either Asp32i or Asp32o. The optimal binding modes of the most active resveratrol analog 11 and the inactive analog 9, are depicted in Figure 3(a and b), respectively. It is observed that both derivatives are placed in the same location as compound LO1 in the crystal structure. However, they do not interact with the amino acids Asp32 and Asp228. The distances between O2 of compound 11 and the two Ca of aspartic acids derived from docking results and molecular dynamics (MD) calculations, which will be described below, are: Asp32/Ca-comp11/O2 (average 11.81 ± 1.63Å) and Asp228/Ca-comp11/O2 (average 14.25 ± 1.84Å). The distances between O27 of compound 9 and the two Ca of aspartic acids derived from docking results and MD calculations are: Asp32/Ca-comp11/O27 (average 11.40 ± 0.48Å) and Asp228/Ca-comp11/O27 (average 9.68 ± 0.66Å). These distances clearly show that the two molecules are not in a spatial proximity with the two key catalytic aminoacids of the protein.

Figure 3. Best binding pose of (a) active resveratrol analog 11 and (b) inactive resveratrol analog 9, hydrogen bond interactions of resveratrol analogs 11 and 9 (left) and hydrophobic interactions (right) with the depicted aminoacids of the active site. Hydrophobic aminoacids are shown as green spheres).

Crystallographic studies of the enzyme-inhibitors complexes conducted by Steele et al.45 and Bowers et al.46 indicated that the catalytic aspartate residues were not engaged in the inhibitor binding. For this reason, the biological effect can be explained by studying the interactions of compounds 11 and 9 with amino acids adjacent to them. In particular, one hydroxyl group of the aromatic ring of derivative 11 forms hydrogen bonds with the –NH group of Ser328 (2.080 Å) and the hydroxyl group of Thr329 (2.247 Å). Hydrophobic interactions also contribute to XP GScore. One 1-ethylpropyl group of derivative 11 establishes hydrophobic interactions with Ile110 and the second 1-ethylpropyl group with Phe108 and Ile118 and the amino acid Tyr71 of the flap (Figure 3a, right).

In Figure 3(b), the interactions of resveratrol analog 9 are illustrated. The phenolic hydroxyl group of Tyr198 forms a hydrogen bond with the oxygen of one methoxy group of the aromatic ring A (1.953 Å), while the hydroxyl group of Thr329 establishes a hydrogen bond with the oxygen of the second methoxy group (2.316 Å). The phenyl ring bearing the two methoxy groups forms hydrophobic interactions with the amino acids Ile226 and Val332, while Tyr198 is located in spatial proximity to a methoxy group of the same ring. The second phenyl ring forms hydrophobic interactions with Phe108 and Ile118, whereas the methoxy group forms hydrophobic interactions with Leu30 and Trp115. It is worth noting that both rings are involved in hydrophobic interactions with the amino acid Tyr71 of the flap.

A comparison of XP GScores of the two derivatives, in the protonation state Asp32o, suggests a higher inhibitory BACE-1 activity for compound 11, since it displayed higher XP Gscore (–7.011) than compound 9 (–5.721). This tendency was confirmed by the biological results. Similar results were observed by comparing XP GScores for the protonation state Asp32i.

In Figure 4, the poses of 11 and the crystallographic structure of LO1 in the active site of the enzyme are illustrated. It is observed that the two 1-ethylpropyl groups are located in the pockets S1 and S3 where the phenyl ring and one of the two N-propyl groups of inhibitor LO1 are placed. It is also observed that the phenyl group of compound 11 bearing the two hydroxyl groups is extended along the flap (Figure 4a). Furthermore, the phenyl group is located near the amino acids of the 10 s loop (amino acids 9–11), which is located at the base of the pocket S3 (Figure 4b). This loop is open to allow the interaction of 1-ethylpropyl group with the amino acids of the pocket S3.

Figure 4. (a and b) Best binding pose of resveratrol analog 11 (green) and pose of crystallographic structure of LO1 (blue) in the active site of the enzyme.

MD and molecular mechanics–Poisson Boltzmann surface area calculations

Conformational analysis

In order to investigate the role of receptor flexibility on the binding mode of the active resveratrol derivative 11 and the inactive resveratrol derivative 9, the docked poses of the two compounds were subjected to MD simulations using the AMBER 11 software package47.

For the binding mode of compound 11, root-mean-square deviation (RMSD) for the Cα atoms of the protein and all atoms of the ligand is shown as a function of the 15 ns run time in Figure 5(a). A structural change takes place for the enzyme in the beginning of the simulation and remains lower than 2 Å throughout the 15 ns trajectory, suggesting that there are no large conformational changes in the complex. A less-pronounced change was observed for compound 11, which also appeared stable throughout the simulation.

Figure 5. Root mean-square deviation (RMSD, in Å) versus MD simulation time for BACE-1 enzyme (top) and: (a) active resveratrol analog 11 (first from the top) and (b) inactive resveratrol analog 9 (first from the top). In both Figures, RMSD versus MD simulation time for the two dyad aminoacids surrounding the two catalytic aspartic acids (Asp32, Thr33 and Gly34 (first from the bottom) and Asp228, Ser229 and Gly230 (bottom) is displayed.

Figure 5(b) shows the RMSD values for the simulation of BACE-1-compound 9. By contrast, the initial conformational change for the protein and for compound 9 was greater with respect to compound 11, suggesting a greater impact of compound 9 on the structure of the protease compared to compound 11. Furthermore, from the variation in RMSD values, it is indicated that the compound 9-bound protein and compound 9 are more flexible than in the case of the active compound 11.

In Figure 5(a and b), the RMSD versus MD simulation time for the two catalytic triplets (Asp32, Thr33 and Gly34 (green line) and Asp228, Ser229 and Gly230 (magenta line)) is observed to be constant and ranging from 0 to 0.4Å. This illustrates the high stability of the active site during the 15 ns trajectory time.

Hydrogen bonding analysis

Hydrogen bonds involving resveratrol analog 11 bound to BACE-1 are illustrated in Figure 6. In Table 2, the percent hydrogen bond occupancy (the percent of time that a hydrogen bond is observed during the trajectory) for each hydrogen bond formed between compound 11 and the enzyme is reported. In accordance with the docking scheme, the hydrogen bond between the oxygen atom in Ser328 and the hydrogen of hydroxyl group of 11 is maintained during the MD simulation, but accounts only for 23% of the hydrogen bond occupancy, which is the lowest value observed. The molecule is stabilized by the formation of two other hydrogen bonds with Gln326 and Lys107, which account for 57% and 35%, respectively. The interaction between compound 11 and Thr329 (Figure 3a) was potentially unstable and did not appear throughout the MD run. Resveratrol analog 9 did not form any hydrogen bonds during the simulation, contrary to the docking results, which suggested two such interactions with Tyr198 and Thr329.

Figure 6. Complex of resveratrol analog 11 with BACE-1 after MD simulation. Hydrogen bonds are depicted with dashed red line.

Table 2. Occurrence of hydrogen bonds between BACE-1/compound 11.*

BACE-1 COMPLEX
Residues in H-bonds	Compound 11
Gln326	56.6/22.67†
Lys107	34.87
Ser328	25.73

*Occurrence is defined as the percentage of simulation time that a specific interaction exists, interactions occurring less than 10% of the simulation are not shown.

†Multiple hydrogen bonds are formed between different atoms of compound 11 and the specific residue.

Free energy calculations with the Molecular Mechanics–Poisson Boltzmann Surface Area method

In order to estimate the energetic contributions on the binding of resveratrol analogs 11 and 9, the molecular mechanics–Poisson Boltzmann surface area (MM–PBSA) method as implemented in AMBER was applied. Table 3 lists calculated binding free energies averaged over 1500 snapshots.

Table 3. Contributions to ΔG_bind for resveratrol derivatives 11 and 9 complexed with BACE-1, computed with the MM–PBSA method.

Energetic analysis	Resveratrol derivative 11	Resveratrol derivative 9
ΔΕelec	−14.50 ± 0.14*	−9.09 ± 0.09*
ΔΕvdW	−33.96 ± 0.09	−40.12 ± 0.10
ΔΕMM = ΔΕelec+ΔΕvdW	−48.45 ± 0.17	−49.21 ± 0.13
ΔGPB	31.44 ± 0.12	35.24 ± 0.18
ΔGsolv	27.39 ± 0.12	31.15 ± 0.17
ΔGNP = ΔGsolv−ΔGPB	−4.05 ± 0.17	−4.09 ± 0.25
ΔGelec (tot) = ΔΕelec+ΔGPB	16.95 ± 0.18	26.15 ± 0.20
ΔH	−21.07 ± 0.12	−18.06 ± 0.09
−TΔSsolv	17.09 ± 1.29†	19.66 ± 2.32‡
ΔGbind	−3.98 ± 1.29	1.6 ± 2.32

*Errors are shown as standard errors of mean (SEM). SEM = standard deviation/N^1/2, where N is the number of trajectory frames used in MM–PBSA calculations (N = 50† or 25‡ for entropy calculations, and N = 1500 for all enthalpy terms).

According to Table 3, the calculated ΔG_bind value is −3.98 kcal/mol for resveratrol analog 11, whereas a positive value of ΔG_bind was computed for 9 (ΔG_bind = 1.6 kcal/mol), which implies that compound 11 exhibits a stronger potency to BACE-1 than analog 9. This result is in accordance with docking and biological results. For both analogs, the major contribution to ΔG_bind is mainly van der Waals interactions (ΔΕ_vdW = −33.96 kcal/mol for 11 and ΔΕ_vdW = −40.12 kcal/mol for 9). For analog 11, it is observed that the enthalpic term (ΔH = −21.07 kcal/mol) contributes mostly to the binding energy compared to the entropic term (−TΔS = 17.09 kcal), whereas for analog 9, the entropic term (−TΔS = 19.66 kcal) is slightly higher than the enthalpic term (ΔH = −18.06 kcal/mol). Furthermore, nonpolar and Coulomb electrostatic interactions contribute favorably to complex formation, in both cases (ΔΕ_elec = −14.50 kcal/mol for analog 11, ΔΕ_elec = −9.09 kcal/mol for analog 9).

Conclusions

In this study, the BACE-1 inhibitory activity of 11 resveratrol analogs was examined. Four resveratrol analogs (11, 10, 3 and 5) demonstrated considerably higher activity than resveratrol (1). Docking and MD calculations for the active resveratrol analog 11 and the inactive analog 9 further confirmed the above results. In addition, we examined the anti-oxidant neuroprotective efficacy of these compounds using glutamate-challenged HT22 cells. Our findings suggested that beneficial effects of resveratrol analogs against AD may be due to their dual action as BACE-1 and oxytosis inhibitors.

DSC data previously reported from our group have shown that the three compounds, 11, 10 and 5, exerted the most significant effects as expressed by the lowering of T_m, broadening of the main phase transition and abolishment of pretransition of DPPC bilayers48. As DSC thermal profiles reflect the lipophilicity orientation and location of a drug in the membrane bilayers, this appears important for both receptor binding and oxytosis inhibition. LogP results obtained for resveratrol analogs are also in accordance with DSC and experimental data showing that lipophilicity is an important factor for their bioactivity.

Experimental

Chemistry

Synthesis of resveratrol analogs

The synthesis of resveratrol analogs has been reported previously9^,28.

Biological assays

BACE-1 assay

The inhibition of BACE-1 was determined in a homogeneous TRF assay (True Point kit, Perkin-Elmer). The assay buffer contained sodium acetate, CHAPS, Triton-X 100 and EDTA, pH 4.5. The substrate used was the Swedish mutant sequence Eu-EVNLDAEFK-Quencher. Inhibitor dilutions were made in DMSO at 30 × final assay concentration; 1 μl per well of inhibitor dilutions were added to 15 μl per well of 20 nM BACE-149 in assay buffer on a black half area 96-well plate. The plate was covered and incubated at room temperature for 30 min. The assay was initiated by the addition of 15 μl per well of 400 nM substrate in assay buffer (final concentration in assay 200 nM), and the plate was read for 60 min at room temperature in a Wallac Victor-2 (Waltham, MA, U.S.). The excitation wavelength was 340 nm, and the emission was monitored at 615 nm. The percent inhibition was calculated from initial velocities of the inhibited reactions relative to the uninhibited control. IC₅₀ values were determined using the following relationship: % inhibition = 100[I]/([I]+IC₅₀).

Oxytosis inhibition assay

The potency of the new aroylhydrazones to inhibit oxytosis of HT22 cells was assessed as already described18^,22, with minor modifications. Briefly, HT22 cells were plated in 96-well flat bottom transparent plates at a density of 4000 cells per well in 100 µL of DMEM (low glucose) containing 10% fetal bovine serum. Twenty-four hours after plating, the cells were exposed to 5-mM glutamate in the absence or presence of increasing concentrations (0.03–10 µM) of test compounds in fresh medium for 24 h prior to using conversion of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide to colored formazan as a means to indirectly assess relative cells numbers. Complementarily, relative cell numbers were assessed using crystal violet (CV) or sulforhodamine B (SRB) as already described50 with similar results. The difference in optical density at 550 nm (or 510 and 570 nm for SRB and CV assays, respectively), determined using a Safire II plate reader (Tecan), was taken to measure cell number. Cells exposed only to vehicle (DMSO) with or without glutamate served as controls. The number of cells in the presence of glutamate, expressed as % of that in its absence, in the different test compound concentrations, was used to calculate EC₅₀, i.e. test compound potency to protect HT22 cells from oxytosis.

Statistics

Pearson and Spearman’s correlation analyses were carried out using the SPSS 10.0 statistical package. Correlations were considered significant for values of p ≤ 0.05.

Computational studies

Structures generation. Prediction of BBB permeation and Caco-2 permeability

The structures were generated using SYBYL51 molecular modeling package running on a Linux operating system, and their energies were minimized using the Powell method with a convergent criterion provided by the Tripos force field52. BBB53 permeation and Caco-2 cell permeability54 of the studied compounds were predicted using VolSurf (version 4). We used the probes water (OH2), hydrophobic (DRY) and H-bonding carbonyl (O) to generate the 3D interaction energies and a Grid space of 0.5 Å.

Docking studies

Molecular docking studies were performed using Glide XP protocol55^,56 from Schrödinger. The X-ray crystal structure of BACE-1 in complex with compound LO1 (PDB ID: 1W51)31 was obtained from Protein Data Bank57. The PROPKA server41^,42 was used to predict the pKa of the Asp dyad in the presence of LO1. In order to determine the protonation state of the ligand molecules, their pKa values were computed at experimental pH = 4.5 using the MoKa program58^,59. For resveratrol and its analogs, a neutral state was predicted.

The default values of the van der Waals scaling factor (1.00) and charge cutoff (0.25) were used for the generation of the grid. The active site was defined within a 10 Å radius around the ligand present in the crystal structure. The ligands were docked to the active site using the “extra precision” glide docking (Glide XP), which considers the ligand flexible. No constraints were applied for all the docking studies. The best poses of each ligand with the lowest Glide scores were chosen for further analysis.

MD simulations in water

MD simulations in explicit solvent were performed on BACE-1/resveratrol analogs 11 and 9 complexes obtained from docking studies using the particle mesh Ewald (PME) Molecular Dynamics module under the AMBER 11 software package47. The force field ff99SB was used to assign force field parameters for the protein60. Force field parameters for the ligands were assigned using the ANTECHAMBER module (GAFF force field with AM1-BCC charges)61. An explicit solvent treatment was applied to model the effects of solvation. Simulations used the TIP3P water model62 and each structure was solvated in a truncated octahedral water box to allow for at least 10 Å between each atom of the protein and the edge of the periodic box. Crystal water molecules were kept in the structure, and 10 792 water molecules were added with LEaP. Furthermore, 11 Na⁺ counterions were added to neutralize the system. Long range electrostatic interactions were calculated using the PME method63. As explained above, BACE-1 was considered monoprotonated. The starting step was the minimization of the systems over 5000 steps. For the first 2500 steps, the steepest descent method was used, whereas for the next 2500 steps, a conjugate gradient algorithm was employed. The next procedure involved the gentle heating of each complex under constant volume with the gradual increase of the temperature from 0 to 300 K (time step: 2 fs). A 200 ps constant pressure equilibration followed to observe the gradual increase of the density, which converged after ≈50 ps. Subsequently, a 15-ns long MD production simulation in constant pressure was performed for resveratrol analog 9-BACE-1 and resveratrol analog 11-BACE-1 using a Langevin dynamics temperature scaling with a collision frequency of 2 ps⁻¹64. During the MD simulations, all bonds involving hydrogen atoms were constrained to their equilibrium distance65, thus allowing for a 2 fs time step to be used. RMSD and HB analyses were performed on the resulting trajectories with the ptraj module under AMBER. A 3.5 Å donor acceptor distance cutoff and a cutoff of 120° for the donor hydrogen acceptor angle have been used to define HB interactions.

MM–PBSA calculations

The binding free energies (ΔG_bind) of the receptor/ligand complexes using the MM–PBSA method were calculated66^,67. The method involves calculations on a series of snapshots produced by the MD. A total of 1500 snapshots of the BACE-1/ligand complexes were considered. The dielectric constants for the solute and solvent were set to 1 and 80, respectively. All counterions and water molecules were stripped. For each snapshot, a free energy is calculated for the complexes, BACE-1 and ligand (11 and 9), and the ligand-binding free energy is calculated by the following Equation (1(1) ): (1) G_complex, G_BACE-1, G_ligand are the free energies for the complex, BACE-1, and the ligand, respectively.

The binding free energy contains an enthalpic and entropic contribution: (2)

The enthalpy of binding ΔH is composed of ΔG_MM, the change in the molecular mechanics free energy upon complex formation and ΔG_solv, the solvated free energy contribution.

The molecular mechanics free energy is calculated as: (3) ΔE_elec and ΔE_vdW represent coulomb and van der Waals interactions, respectively.

The solvation of free energy is composed of two terms (4) ΔG_PB and ΔG_NP define the polar and the nonpolar contribution to the sοlvation, respectively.

The polar term of energy was calculated by solving the Poisson–Boltzmann equation using for MM–PBSA method68, whereas the non polar contribution to the salvation-free energy was determined as a function of the solvent-accessible surface area (Ų) (SASA)69. (5) where γ represents the surface tension and β is a constant. The values of γ = 0.00542 kcal/mol Ų and β = 0.92 kcal/mol were used70. ΔG_NP was computed via Equation (5(5) ), with the linear combinations of pairwise overlaps method71.

The entropic term –TΔS was calculated by normal mode analysis using the NMODE module72 over only 150 equally spaced snapshots in order to save computational time. The entropy is divided in translational, rotational and vibrational contributions: (6)

Acknowledgements

The authors are grateful to Prof. Gabriele Cruciani (Laboratory of Chemoinformatics and Molecular Modeling, School of Chemistry, University of Perugia, Italy) for kindly providing us the VolSurf and MoKa programs.

Declaration of interest

This research has been co-financed by the European Union (European Social Fund – ESF) and Greek national funds through the Operational Program “Education and Lifelong Learning” of the National Strategic Reference Framework (NSRF) – Research Funding Program: Heracleitus II. Investing in knowledge society through the European Social Fund. This work was supported, in part, by the European Commission through the FP7-REGPOT-2009-1 Project “ARCADE” (Grant agreement no. 245866) and the Marie Curie Early Stage Training project “EURODESY” (contract MEST-CT-2005-020575).