From Kinase Inhibitors to Multitarget Ligands as Powerful Drug Leads for Alzheimer's Disease using Protein‐Templated Synthesis

Abstract Multitarget directed ligands (MTDLs) are arising as promising tools to tackle complex diseases. The main goal of this work is to create powerful modulating agents for neurodegenerative disorders. To achieve this aim, we have combined fragments that inhibit key protein kinases involved in the main pathomolecular pathways of Alzheimer's disease (AD) such as tau aggregation, neuroinflammation and decreased neurogenesis, whilst looking for a third action in beta‐secretase (BACE1), responsible of β‐amyloid production. We obtained well‐balanced MTDLs with in vitro activity in three different relevant targets and efficacy in two cellular models of AD. Furthermore, computational studies confirmed how these compounds accommodate adequately into the long and rather narrow BACE1 catalytic site. Finally, we employed in situ click chemistry using BACE1 as protein template as a versatile synthetic tool that allowed us to obtain further MTDLs.


1/[S]
Comp. 5 [ Table S1. Chemical structure and in vitro activity of the precursors of the fragments used in multitarget compounds 8, 14-15 in their respective protein kinases and BACE1.   Figure S8. Number of interactions contacts (Hydrogen bonds, Pi-Pi stacking and Pi-cation) along the simulations. The number of interactions of compounds 14 and 32 are much higher than those of the MBC-2137 fragment, because compounds 14 and 32 are larger and can establish more interactions with the active site residues of BACE1. This is reflected in the ΔGBind binding energy ( Figure  11). Figure S9. Interaction area between ligands and BACE1. Due the larger structure of 14 and 32, the interaction area between the BACE1 binding site and these compounds is larger than the interaction area between the fragment MBC-2137 and BACE1. Having a larger interaction area allows 14 and 32 to establish a higher number of interactions (and more stable interactions) with the residues of BACE1 binding site. Due to the nature of the binding site of BACE1 (a very large site), it is important to have a large interaction area in order to anchor to the active site and thus have a longer residence time, which can be reflected in increased activity (potency).

Code
Figure S10. BACE1-ligand interaction contacts during the 500ns simulations. Green lines: pi-pi interactions; pink lines: H-bonds (and water bridges); red lines: pi-cation interactions. Hydrophobic, polar, and charged residues are displayed as green, cyan, and purple spheres. Interactions that occur more than 10% of the simulation time (500ns) are shown. More stable interactions are observed for compound 14 than for compound 32 and the MBC-2137 fragment. For example, Pi-Pi interactions with Tyr71, or hydrogen bonds with Thr232 remain for more than 50% of the simulation of compound 14 interacting with BACE1. This indicates that 14 stabilized very well in the binding site and that through these interactions it remains anchored in the pocket. For compound 32 there were no interactions above 30%, indicating that the compound moves in the large binding site and does not establish stable interactions throughout the simulation Figure S11. Time dependence of the RMSD for ligand atoms (red) and BACE1 backbone atoms (black) during the 500 ns unrestrained MDs. There are not significant changes in the BACE1 protein during simulations, which allows us to see that in all 3 simulations without energy restrictions the protein behaved stably. Regarding the changes in the atoms of the ligands. It is observed that the active compound 14 slowly changes it conformation in the binding site and after ~100ns it stabilizes and remains stable. Compound 32 undergoes several significant changes at the beginning of the simulation and after ~110 ns it stabilizes until ~350ns, where it undergoes another conformational change. Compound MCB-2137 rapidly changes conformation, moves around the large binding site at the beginning of the simulation, where it stabilizes after ~100ns. Figure S12. BACE1ligands interactions throughout the simulation. Interactions are categorized into five types: Hydrogen Bonds, Water Bridges, Hydrophobic, Pi-Pi, and Pi-cation. The stacked bar charts are normalized over the course of the trajectory: for example, a value of 0.5 suggests that 50% of the 500ns simulation (~250 ns) the specific interaction is maintained. Values over 1.0 indicated that the protein reside make multiple contacts of same subtype with the ligand. The active MTD compound 14 establish strong integrations with BACE1 residues at the binding site, manly with Tyr71, Lys107, and Thr232. These interactions possibly allow it to bind with higher affinity to BACE1 (-70.02 ± 9.90 kcal/mol), which is reflected in its gain of activity against the protein. It is observed that the active compound 14 at 0 ns is located at the top of the binding site ( Figure S13-A). After 250ns stable and strong interactions have already been established, mainly Pi-Pi interactions with the Y71 residue in the flap, and hydrogen bonds with NT232 and N233 at the bottom of the cavity ( Figure S13-B). This allows compound 14 to spread throughout the cavity with high affinity, stabilized also by other hydrophobic interactions. After 500ns the binding mode of compound 14 remains stable ( Figure S13-C). The inactive compound 32, at the beginning of the simulation is interacting in the peripheral part ( Figure S13-D), then it changes its conformation and moves towards the interior of the cavity ( Figure S13-E), there it is observed that it folds in itself establishing intramolecular Pi-Pi interactions. Finally, it moves folded towards the bottom of the cavity and establishes mainly hydrophobic interactions. It does not interact with Y71 at the flap. The MBC2137 fragment at the beginning of the simulation ( Figure  S13-G) is found interacting at the top-peripheral part of the BACE1 binding site. Later it enters the binding pocket a little but always remains at the top (Figures S13-H and S13-I). Because it is a fragment of relatively small size (with respect to 14 and 32) it is not possible for it to establish interactions along the cavity, this is reflected in its lack of activity and affinity for BACE1.

General Information
Reagents were obtained from the commercial sources and used without further purification. Purifications of crudes were performed with the indicated solvent as eluent by flash column chromatography carried out at medium pressure using silica gel (E. Merck, Grade 60, particle size 0.040-0.063 mm, 230-240 mesh ASTM) or IsoleraOne flash purification system from Biotage. 1     The presence of the compound was confirmed using HPLC-MS and it was used in the next reaction.

N-(Benzo[d]thiazol-2-yl)-2-(4-(bromomethyl)phenyl)acetamide (10).
To a solution of aminobenzothiazole (600 mg, 4.0 mmol) in THF (7.5 mL) the acid chloride 9 (990 mg, 4.0 mmol) was added at r. t.. This mixture was heated under microwave radiation at 100 °C for 15 min. Then, the mixture was cooled to r.t., water (35 mL) and CH2Cl2 (65 mL) were added. The solvent of the organic phase was then removed under reduced pressure and the remaining solid was purified by flash chromatography using ethyl acetate/hexane 9:1 as eluents to afford a white solid used in the next reaction.
The solid was washed with cold MeOH to afford a white solid (957 mg, 56% yield).
This compound was used in the following synthetic step without further characterization.

Computational studies.
To appropriately select the correct crystal structure for BACE1 computational studies we created a virtual library of all the ligands (excluding peptide-like compounds and ions) co-crystallized with the BACE1 in the Protein data base (PDB) and prepared them with LigPrep at pH 7.0 ± 2.0 selecting only one conformer per compound. Then the ligands were compared with compound 8 using the ligand-based virtual screening LiSiCA software, 3 which uses the Tanimoto coefficient to search 3D similarities between a given set of compounds.
This step gave us 5I3Y 3D-crystallografic structure, 4 which is the BACE1 target co-crystallized with a ligand that is very similar to our reference compound, so the binding site is adapted to this compound. Performing docking simulations with this crystalline structure increases the probability of finding acceptable solutions for the compounds studied in this work.
The 5I3Y was prepared with the Maestro module Protein Preparation Wizard, the hydrogen bonds were optimized using PROPKA at pH 7 and the co-crystalized ligand was removed for further docking simulations. The grid box was centered at Asp32 and Asp228 catalytic residues, with a size large enough to include the   Figure S14. Linear correlation among experimental and reported permeability of commercial drug using the PAMPA-BBB assay.