Donepezil and Memantine Derivatives for Dual-Function and Prodrug Applications in Alzheimer’s Disease

Abstract

The treatment of Alzheimer’s disease by acetylcholinesterase (AChE) and N-methyl-d-aspartate receptor (NMDAR) inhibitors is limited by the narrow therapeutic window and adverse side effects of the drugs. This study aims to increase the efficacy and limit the side effects of donepezil, an AChE inhibitor, and memantine, an NMDAR inhibitor, through the addition of amyloid-β (Aβ)-targeting fragments to create dual-function compounds. The incorporation of the amyloid-targeting fragments successfully produced compounds with affinity for Aβ fibrils, and that can stain amyloid plaques in the brains of 5xFAD mice. The donepezil-based compounds showed significant changes in AChE inhibition compared to donepezil due to the incorporation of the Aβ-targeting fragment and as confirmed by molecular docking studies. The memantine-derived compound showed good brain uptake in 5xFAD mice but lacked compatibility with NMDAR inhibition based on in vitro assays and molecular docking. Importantly, the memantine-derived compound acts as a prodrug in vivo, releasing memantine within a pharmacologically relevant time frame. Overall, these findings suggest that dual-function compounds may be useful as drug delivery agents that can be metabolized to release an active drug in areas of the brain rich in amyloid plaques and thus could lead to improved treatments for Alzheimer’s disease.

Graphical Abstract

INTRODUCTION

Alzheimer’s disease is now, and has been for many years, the most common cause of dementia worldwide.^1,2 Despite the accessibility of donepezil, galantamine, and rivastigmine as acetylcholinesterase (AChE) inhibitors and memantine as an N-methyl-d-aspartate (NMDAR) inhibitor since the early 2000s, the prevalence of Alzheimer’s disease (AD) as a cause of death has continued to rise.^1,3–6 The FDA approval of aducanumab in 2021 marked the beginning of approvals for several disease-modifying antibody-based therapies for AD, although the inhibitors mentioned previously are still widely used drugs.^7–9 Hence, the development of small molecule inhibitors of NMDAR and AChE has continued in efforts to improve treatment efficacy and reduce iatrogenic effects.¹⁰

AChE is responsible for converting acetylcholine into choline and acetate, terminating a signal induced by binding of acetylcholine to receptors. In AD patients, AChE is active at levels higher than in healthy individuals, leading to decreased concentrations of acetylcholine in the brain and decreased signal transduction.¹¹ It has additionally been shown that AChE can drive amyloid-β (Aβ) fibril formation, and ultimately, the inclusion of AChE within Aβ fibrils further exacerbates their toxicity.^12,13 Donepezil, an AChE inhibitor, interacts with the peripheral anionic site (PAS) of the AChE, which is believed to participate in Aβ fibrillization, in addition to the catalytic active site (CAS) where acetylcholine is converted to choline and acetate.^11,14,15 However, treatment with donepezil and other AChE inhibitors has varied long-term efficacy, leading to continued development of AChE inhibitors.^10,16 Since donepezil first received FDA approval, its relatively simple chemical structure and low IC₅₀ have made it attractive for derivatization. Additional functionalities that have been incorporated into the resulting multifunctional compounds include antioxidant, metal chelation, anti-inflammation, and inhibition of Aβ aggregation. Focusing on donepezil derivatives that also interact with Aβ, derivatization of both the indanone and N-benzyl fragments of the molecule has been previously reported. The indanone moiety has been replaced with coumarin-based, heterocyclic, and imidazole-derived structures, all of which showed tolerable AChE inhibition and decent inhibition of Aβ aggregation.^17–20 Derivatives resulting from the replacement of the benzyl piperidine moiety with oxidibenzene, trifluoromethylbenzyl-pyridine, and extended alkanes linked to methoxybenzyl-amines, all of which caused the derivatives to exhibit inhibition of Aβ aggregation in addition to AChE inhibition.^21–23 While the described derivatives showed promise for alleviation of Aβ toxicity or aggregation in addition to reasonable AChE inhibition, the clinical applications of the compounds remain limited due to toxicity and/or efficacy issues.^17–23

NMDARs regulate Ca²⁺ concentrations in postsynaptic neurons. Excessive activation can cause high Ca²⁺ concentrations that lead to neuronal death, while reduced activation can reduce Ca²⁺ concentrations and inhibit central nervous system function.^24,25 In AD brains, Aβ aggregates prevent the uptake of excess glutamate by astrocytes, leading to increased NMDAR activation.²⁶ Memantine, an NMDAR inhibitor, interacts with activated receptors by preventing the movement of ions through the channel, preventing Ca²⁺ influx into the postsynaptic neuron, and preventing neuronal death.^26–28 However, memantine treatment has varied long-term efficacy in AD patients, leading to continued development of NMDAR channel blockers for AD treatment.^10,29 Many memantine derivatives have been synthesized in an attempt to develop additional NMDAR inhibitors. Such derivatives incorporate moieties with antioxidant, anti-inflammatory, neuroprotective, AChE inhibitive, and Aβ aggregation disruptive properties.^29,30 The two most common functionalities that have been incorporated in the literature are antioxidant and AChE inhibition, with Aβ aggregation inhibition being observed for several compounds that were originally designed as antioxidant derivatives of memantine. There are fewer examples in the literature of memantine derivatives with purposeful incorporation of Aβ-targeting moieties.^29,30

Attempts to increase the efficacy of donepezil and memantine treatments have resulted in the development of several multifunctional compounds, several of which have reported Aβ interactions; however, many of these compounds significantly alter the structure of donepezil or memantine. In this work, donepezil and memantine derivatives were synthesized to contain the parent drug fragment along with an added Aβ-targeting functionality.^31–33 The compounds were designed to target the amyloid plaques while also inhibiting AChE or NMDAR, in order to reduce off-target effects and increase treatment efficacy at low doses. Two novel Aβ-targeting donepezil derivatives were synthesized and evaluated by determining their affinity for Aβ fibrils in histological staining experiments. The AChE inhibition of the compounds was also investigated by IC₅₀ determination and molecular docking. Moreover, a novel Aβ-targeting memantine derivative was synthesized and similarly evaluated for Aβ affinity, and NMDAR inhibition activity was assessed by molecular docking and in vivo metabolism investigation.

RESULTS AND DISCUSSION

Design, Synthesis, and Characterization of Donepezil-Based Dual-Function Compounds.

Two Aβ-targeting fragments were utilized to synthesize the donepezil-based dual-function compounds investigated herein: Thioflavin-T-derived 2-methoxy-4-benzothiazolylphenol (Benz) and distilbene Pre-LS-4 (Stil) (Figure 1).^34–36 The AChE inhibition fragment of donepezil was chosen due to interaction with the PAS and CAS sites of the enzyme, a lower IC₅₀ value in comparison to rivastigmine and galantamine, and chemical availability of donepezil precursors for synthetic purposes.^37,38 Donepezil derivatives were designed with an incorporation approach to minimize the molecular weight of the compound. The Aβ-targeting fragments contained phenol moieties that replaced the benzyl moiety of donepezil to create two dual-function compounds. 2-methoxy-4-benzothiazolylphenol and Pre-LS-4 were synthesized as previously reported.^36,39 Reaction of desbenzyl donepezil, paraformaldehyde, and 2-methoxy-4-benzothiazolylphenol under Mannich conditions produced Done-Benz in 83% yield. A Mannich reaction of desbenzyl donepezil, paraformaldehyde, and Pre-LS-4 produced Done-Stil in 41% yield.

Figure 1.

Structures of previously reported Aβ-interacting small molecules 2-methoxy-4-benzothiazolylphenol and Pre-LS-4, protein-inhibiting drugs memantine and donepezil that have been approved by the FDA for the treatment of Alzheimer’s disease, and novel dual-function compounds designed in this work with linkage or incorporation moieties shown in red.

As a primary measure to determine the propensity of the compounds to cross the blood-brain barrier, the lipophilicity of the compounds was determined. Lipophilicity is an important consideration in the blood-brain barrier (BBB) permeability of compounds, with log D values of approximately 1.5 or greater being more likely to be brain permeant.⁴⁰ Additionally, the lipophilicity has previously been observed to play an important role in the affinity of the compounds for Aβ species, with amphiphilic compounds exhibiting increased affinity for oligomeric over fibrillar Aβ.³⁶ Done-Benz and Done-Stil have log D_oct values of 2.5 ± 0.2 and 3.0 ± 0.1, respectively (Table S1). The compounds show promising lipophilicity for the blood-brain barrier permeability.

The increased fluorescence of 2-methoxy-4-benzothiazolylphenol and Pre-LS-4 upon interaction with Aβ species is useful for investigating the interaction of the dual-function compounds and Aβ₄₂.^35,36,41 While each of the compounds shows an increase in fluorescence following incubation with Aβ₄₂ fibrils, the fold-increase of the observed fluorescence is minimal (Figure 2a). Done-Benz has significant background fluorescence, and the fluorescence fold-turn-on of Done-Benz is less than 2 at 445 nm following incubation with Aβ₄₂ fibrils. Done-Stil has significantly lower background fluorescence and a nearly 4-fold fluorescence increase at 470 nm following incubation with Aβ₄₂ fibrils. The fluorescence spectra of Done-Stil also blue shifts from 510 to 470 following incubation with Aβ₄₂ fibrils, indicating an interaction between the fluorescent stilbene fragment of the molecule and the fibrillar species, which alters the fluorescent properties of the compound.

Figure 2.

(a) Fluorescence turn-on of dual-function compounds upon interaction with Aβ₄₂ fibrils. [Done-Benz] = [Done-Stil] = 10 μM, [Aβ₄₂] = 25 μM. For the Done-Benz, λ_ex = 320 nm; for Done-Stil, λ_ex = 360 nm. (b) Histological staining of 11-month-old male 5xFAD mice with Done-Benz and Done-Stil (blue and green, respectively) and costaining with AF594 fluorescently labeled antibody HJ3.4 (red). [Compound] = 25 μM; [HJ3.4-AF594] = 1 μg/mL. Scale bar: 125 μm. R value indicates Pearson’s correlation coefficient. (c) Brain sections of 6-month female 5xFAD mice injected for 10 days with 1 mg/kg/day of Done-Benz and Done-Stil (blue and green, respectively). Each section was histologically stained with 1 μg/mL HJ3.4-AF594 antibody prior to imaging. Scale bar = 125 μm. R value indicates Pearson’s correlation coefficient. Additional correlation coefficients can be found in panel b (Table S3) and panel c (Table S4).

The affinity with which the dual-function compounds interact with Aβ₄₂ fibrils was quantified with K_i studies using Thioflavin T (ThT). In this assay, constant concentrations of Aβ₄₂ fibrils and ThT are incubated with increasing concentrations of a dual-function compound. As the compound displaces ThT from fibrillar interactions, the fluorescence of ThT decreases in a sigmoidal dose-responsive manner (Figure S4). Analysis of the resulting fluorescence curves allows for quantification of the dual-function compound affinity with Aβ₄₂ fibrils with respect to ThT affinity. The strongest interaction was observed for Done-Benz, for which the K_i is 0.7 ± 0.2 μM, and Done-Stil has slightly weaker interactions as evidenced by a slightly higher K_i value of 1.3 ± 0.3 μM (Table S2). These values indicate modest interactions with Aβ₄₂ fibrils.

To assess the ability of the dual-function compounds to interact with native Aβ, histological brain section staining experiments were conducted.^42–45 Brain sections of 11-month-old 5xFAD mice were stained with the compound of interest, followed by a costain of HJ3.4 antibody, which interacts with all forms of Aβ aggregates, to determine the extent and specificity of Aβ interaction.⁴⁶ Done-Stil had a lower overall correlation with HJ3.4 antibody than Done-Benz (Figure 2b, Table S3). This is due in part to the bright fluorescence at the cores of Aβ plaques observed for Done-Stil staining, while Done-Benz-stained plaques are less fluorescent but more diffuse. The previously observed fluorescence turn-on intensity of Done-Stil may explain in part the distinct fluorescent spots observed for Done-Stil histological staining, while the lesser fluorescence turn-on of Done-Benz explains the dimmer fluorescent spots observed for Done-Benz histological staining. The fluorescence emission of compounds Done-Benz and Done-Stil, as determined via histological staining experiments, indicates that the intrinsic fluorescence of the compounds could be utilized to assess the BBB permeability of the compounds. Female 5xFAD mice six months of age were injected intraperitoneally with 1 mg/kg of each compound for 10 days before the brains of the mice were retrieved and treated with HJ3.4 antibody. While the fluorescence of the compounds is dimmer than that observed in the in situ staining experiments, sufficient fluorescent signal can be observed to demonstrate BBB permeability and assess Aβ interaction (Figure 2c). Notably, the morphology of the Done-Benz fluorescence spots is similar to the morphology observed in topical staining experiments, while the morphology of the Done-Stil fluorescence spots is significantly more diffuse than the morphology observed in topical staining experiments. The change in the Done-Stil staining morphology significantly increases the correlation between the compound-based fluorescence and HJ3.4 antibody, with Done-Stil having a higher correlation than Done-Benz for these experiments (Figure 2c, Table S4). Overall, the lipophilicity, Aβ affinity, Aβ staining, and BBB permeability indicate the Aβ-targeting profile of the donepezil-based compounds is sufficient for use in dual-function compounds.

Acetylcholinesterase Inhibition of Donepezil-Based Dual-Function Compounds.

The inhibition capacity of the donepezil derivatives for AChE was assessed through a modified Ellman’s assay.^34,47 Of the two compounds, Done-Benz is a slightly better inhibitor with an IC₅₀ value of 50 ± 4 μM than Done-Stil with an IC₅₀ value of 83 ± 9 μM (Figure 3). Both compounds were observed to have a nearly thousand-fold decrease in IC₅₀ compared to donepezil hydrochloride, for which values as low as 10 nM have been reported by similar protocols but was measured to be 35 ± 1 nM by our method (Figure S7).^48,49 The incorporation of an Aβ-binding fragment at the benzyl piperidine into the donepezil framework likely contributes significantly to the decrease in inhibition capacity and increase in IC₅₀ values. This result is consistent with the previously reported interaction of the benzyl piperidine moiety of donepezil with the catalytic active site of human AChE. Derivatization of this moiety directly impacts the inhibitory capacity of the compound if it significantly alters the interaction with the CAS. It is therefore reasonable that the compound with the larger Aβ-binding fragment, Done-Stil, has the lower inhibition capacity of the compound in comparison to the compound with the smaller Aβ-binding fragment, Done-Benz.^14,15

Figure 3.

AChE activity as assessed by a modified Ellman’s assay in the presence of (a) Done-Benz (IC₅₀ = 50 ± 4 μM) and (b) Done-Stil (IC₅₀ = 83 ± 9 μM). [AChE] = 0.03 U/mL, [DTNB] = 1.5 mM, and [ATChI] = 0.3 mM.

To further investigate and describe the interaction of the donepezil-based compounds with AChE, docking studies were performed with Schrödinger Glide (Maestro 14.2). The interactions of the compounds with AChE are described by both docking score, an empirical approximation of ligand binding energy, and Epik penalties, and the extent to which binding pocket residues interact with the compounds.^50–52 The structure of eel AChE (eeAChE, PDB: 1C2B) was used as it corresponds to the in vitro experiments.⁵³ Done-Benz, Done-Stil, donepezil, 2-methoxy-4-benzothiazolylphenol, and Pre-LS-4 were docked individually onto the crystal protein structure (Figure 4). Looking at the interactions of residues in the binding pocket with donepezil, it is of note that the known interactions of the indanone moiety with the AChE peripheral anionic site (PAS) and the piperidine benzyl site near the catalytic active site (CAS) are reversed.^14,15 This is likely due to computational limitations of molecular docking compared to more complex calculations such as molecular dynamics simulations. Additionally, the removal of water molecules will greatly affect the docking residues, as water molecules have been shown to participate in the interaction of donepezil with AChE. However, residues in both PAS and CAS are seen interacting with donepezil by molecular docking. Each of the five compounds investigated interacts with at least five residues at either the PAS or CAS of eeAChE. The only CAS residue all five compounds interact with is serine 203, the nucleophile of the catalytic triad and therefore an essential residue for AChE to maintain esterase function (Figure 4b). Two PAS residues, tyrosine 341 and tyrosine 124, interact with all five compounds, suggesting that each of the observed compounds interacts with the binding pocket of donepezil at somewhat similar residues. The docking score of donepezil (−13.622) indicates that it has a higher affinity to the enzyme than the dual-function compounds, corresponding to the significantly lower IC₅₀ value obtained for donepezil than the dual-function compounds (Table S5). The docking score of Done-Benz (−11.991) indicating higher affinity to the enzyme than Done-Stil (−5.479) supports the observed IC₅₀ values of the dual-function compounds, for which Done-Benz showed more inhibition capacity than Done-Stil. Overall, the significant alteration in binding residues and docking scores from donepezil versus Done-Benz or Done-Stil supports the in vitro data, for which significant decreases in inhibition capacity were observed. Therefore, it was determined that the donepezil-based, dual-function compounds were not suitable for further investigation as therapeutic agents for Alzheimer’s disease.

Figure 4.

Molecular docking images of AChE (PDB: 1C2B) with 2-methoxy-4-benzothiazolylphenol, Pre-LS-4, donepezil, Done-Benz, and Done-Stil. (a) Expanded view of each compound’s interaction with a single AChE subunit. (b) Narrow view of each compound’s interaction with a single AChE subunit, with labeled residues representing the peripheral anionic site and catalytic active site residues within 4 Å of the compounds. Schrödinger Glide (Maestro 14.2) was used for structural optimization and docking studies, and PyMOL 3.1 was utilized to create images presented in this figure.

Design, Synthesis, and Characterization of a Memantine-Based Dual-Function Compound.

Thioflavin-T-derived 2-methoxy-4-benzothiazolylphenol (Figure 1) was utilized as an Aβ-targeting moiety to create a memantine-based dual-function compound.^34,35 The NMDAR inhibiting fragment was chosen to be memantine due to FDA approval of memantine for the treatment of AD.⁵ Derivatization with a linkage approach was utilized due to the presence of a primary amine in memantine, allowing for simple linkage between memantine and the Aβ-targeting moiety to create a dual-function compound. The 2-methoxy-4-benzothiazolylphenol precursor was synthesized as previously reported.³⁹ A Mannich reaction involving 2,4-dimethyladamantanamine, paraformaldehyde, and 2-methoxy-4-benzothiazolylphenol produced Mem-Benz in 14% yield.⁵⁴

The linkage method utilized in the design of Mem-Benz allows the dual-function compound to break down into its memantine and phenylbenzothiazolylphenol building blocks. In vivo, this means that the 2-methoxy-4-benzothiazolylphenol fragment could serve as a targeting vector, directing the compound to the area of the Aβ plaques. The compound could act as a prodrug by breaking down into its components, releasing active memantine in the area of Aβ plaques where NMDAR inhibition can be most effective with the fewest off-target effects.⁵⁵ The field of prodrugs has expanded greatly in the last several decades, with a large number of examples utilizing prodrug strategies for central nervous system applications having been reported.^56,57 Indeed, several examples of memantine prodrugs have been previously reported, including reactive Schiff bases designed for in vivo delivery of memantine and H2S-memantine hybrids for the dual delivery of memantine and reactive H2S.^58,59 Additionally, the N-alkylation of primary amines has previously been reported as a CNS prodrug approach to increase brain uptake of the compounds, for which MAO-B-mediated dealkylation was observed in vivo.⁶⁰ Together, this precedent provides a mechanism by which Mem-Benz, a novel structure linking an Aβ-targeting fragment to memantine, may be dual-functional as an Aβ-targeting memantine prodrug.

The lipophilicity of Mem-Benz was determined to be compatible with blood-brain barrier permeability due to a log D_oct value of 1.9 ± 0.1 (Table S1). In fluorescence turn-on studies with Aβ₄₂ fibrils, Mem-Benz showed a significant amount of background fluorescence and minimal fluorescence turn-on of approximately 2-fold at 415 nm (Figure 5a). The compound had higher background fluorescence and a lower fold-increase of fluorescence than both donepezil-based dual-function compounds. Upon quantification of Mem-Benz interaction strength with Aβ₄₂ fibrils, a modest strength interaction was shown by a K_i value of 1.5 ± 0.3 μM, a slightly weaker interaction than what was observed for the donepezil-based dual-function compounds (Table S2). This is likely the result of the adamantyl fragment of the memantine structure limiting the interaction of the benzothiazole-based fragment with Aβ plaques. This impact may also be observed in brain section staining experiments. Despite a high correlation between Mem-Benz and HJ3.4 antibody, the low fluorescence turn-on of Mem-Benz following interaction with Aβ₄₂ fibrils is evident by the very dim fluorescence of Mem-Benz in the blue channel. Since the fluorescence turn-on of ThT-based compounds, including 2-methoxy-4-benzothiazolylphenol, is typically attributed to the inhibition of rotation by a planar binding pocket of Aβ₄₂ fibrils, the inclusion of the adamantyl-based memantine fragment may limit fluorescence turn-on by this mechanism, reducing the fluorescence of the compounds on brain sections despite Aβ interaction having been shown by K_i experiments.^41,61–64

Figure 5.

(a) Fluorescence turn-on of Mem-Benz upon interaction with Aβ₄₂ fibrils. [Mem-Benz] = 10 μM, [Aβ₄₂] = 25 μM, λ_ex = 320 nm; (b) Histological staining of 11-month-old male 5xFAD mice with Mem-Benz (blue) and costained with AF594 fluorescently labeled antibody HJ3.4 (red). [Mem-Benz] = 25 μM; [HJ3.4-AF594] = 1 μg/mL. Scale bar = 125 μm. R value indicates Pearson’s correlation coefficient. (c) Brain sections of 6-month female 5xFAD mice injected for 10 days with 1 mg/kg/day of Mem-Benz (blue). Each section was histologically stained with 1 μg/mL of HJ3.4-AF594 antibody prior to imaging. Scale bar = 125 μm. R value indicates Pearson’s correlation coefficient. Additional correlation coefficients can be found in panel b (Table S3) and panel c (Table S4).

Interaction of Memantine-Based Dual-Function Compound with NMDAR.

Molecular docking studies with NMDAR were performed as described above for eeAChE with the human GluN1-GluN2A NMDA receptor (PDB: 7EU7). In this instance, the binding pocket differences among memantine, 2-methoxy-benzothiazolylphenol, and Mem-Benz are of such significance that they play a greater role in the interpretation of the molecular docking data analysis than docking scores (Figure 6, Table S6). Memantine is shown to interact with the known binding location in a slightly different orientation, where the amine is not facing the key asparagine 616 residue. This orientation difference is likely due to the cocrystallization of the protein with ketamine, which slightly alters the protein structure to suit the different electrostatic properties of that ligand rather than memantine, causing the binding pocket, including Asp616, to be electrostatically suboptimal for memantine. Nonetheless, the memantine control fits well into the known binding pocket. Rather than interacting with the memantine binding pocket, Mem-Benz and 2-methoxy-4-benzothiazolylphenol prefer to interact with the more hydrophobic exterior α helices of the protein. The hydrophobic nature of the α helices that anchor the protein in the cell membrane explains why the likewise hydrophobic 2-methoxy-4-benzothiazolylphenol and Mem-Benz interact with that region of the protein.⁶⁵ This molecular docking screen indicates that Mem-Benz does not bind to NMDA receptors in the same region as memantine, where inhibition would be most effective. Rather, Mem-Benz binds to an alternate location on the protein, suggesting it may have promise as an Aβ-targeting memantine prodrug. Importantly, Mem-Benz is predicted to be inactive as an NMDAR inhibitor, which is ideal for the prodrug design.

Figure 6.

Molecular docking images of N-Methyl-d-Aspartate Receptor (NMDA Receptor, human GluN1-GluN2, PDB: 7EU7) with 2-methoxy-4-benzothiazolylphenol, memantine, and Mem-Benz. (a) Expanded view of each compound’s interaction with the NMDA receptor; cyan arrow indicates location of compound on receptor. (b) Narrow view of each compound’s interaction with the NMDA receptor, with labeled residues representing all residues within 4 Å of the compounds. Schrödinger Glide (Maestro 14.2) was used for structural optimization and docking studies. PyMOL 3.1 was utilized to analyze the binding pocket and create the images presented in this figure.

In Vivo Behavior of Mem-Benz.

To probe the in vivo behavior of Mem-Benz, 2 mg/kg of the compound was intravenously administered to 5xFAD mice of 8 to 9 months of age. Brains of injected mice, sacrificed one- or eight h postinjection, were homogenized and then analyzed by UltraHPLC-MS/MS for concentrations of Mem-Benz, memantine, and 2-methoxy-4-benzothiazolylphenol. Mem-Benz exhibited significant brain uptake (0.71 ± 0.08 %ID/g) 1 h postinjection, followed by a gradual wash out resulting in lower brain concentrations (0.031 ± 0.003 %ID/g) 8 h postinjection (Figure 7). The active drug memantine was also observed in mouse brains 1 h postinjection (0.30 ± 0.11 %ID/g), with slower wash out than Mem-Benz evident by higher brain concentrations (0.069 ± 0.005 %ID/g) remaining 8 h postinjection. The 2-methoxy-4-benzothiazolylphenol Aβ-targeting fragment was not observed at the time points measured (Table S9). This indicates that Mem-Benz breaks down in vivo, releasing memantine and a benzothiazole-derived metabolite. However, the large fraction of Mem-Benz remaining intact in the brain 1 h postinjection indicates that the development of a more reactive linker may increase the delivery efficacy of memantine.

Figure 7.

Average time-dependent concentrations (% ID/g) of Mem-Benz (gray) and memantine (red) in brains 8–9-month-old 5xFAD mice injected intravenously with 2 mg/kg Mem-Benz. Three brains per time point were analyzed. Mem-Benz concentrations were 0.71 ± 0.08 %ID/g one h postinjection and 0.031 ± 0.003 % ID/g 8 h postinjection. Memantine concentrations were 0.30 ± 0.11 % ID/g 1 h postinjection and 0.069 ± 0.005 % ID/g 8 h postinjection. Errors were calculated as the standard deviation of the average. 2-methoxy-4-benzothiazolylphenol concentrations were observed to be below the level of quantification for each time point.

CONCLUSIONS

In summary, three novel dual-function compounds were synthesized and evaluated as Alzheimer’s disease therapeutics. All three compounds include previously reported amyloid-β binding fragments to enhance Aβ affinity and FDA-approved drug fragments to incorporate AChE or NMDAR inhibition. The two donepezil-based hybrids, Done-Benz and Done-Stil, were synthesized and showed low micromolar affinity for Aβ₄₂ fibrils. However, the inhibition of the dual-function compounds toward AChE was notably reduced from the donepezil parent, with the dual-function compounds having mid-micromolar IC₅₀ values in comparison to donepezil’s mid-nanomolar IC₅₀ toward AChE. Molecular docking studies indicate that the incorporation of the amyloid-targeting fragments at the benzyl piperidine moiety of donepezil alters binding to the protein, likely contributing significantly to the observed decrease in AChE inhibition. This indicates that the size of the amyloid-targeting fragments incorporated in this work is not well tolerated on this location of the donepezil framework.

A memantine-based hybrid, Mem-Benz, was synthesized and shown to have a low micromolar affinity for Aβ₄₂ fibrils. Excitingly, Mem-Benz was able to enter the brains of 5xFAD mice and accumulate in amyloid-rich areas of the brain, as confirmed by post-mortem staining of the brains of mice injected with the compound; however, molecular docking studies indicated Mem-Benz may not be suitable for NMDAR inhibition. Instead, the prodrug activity of Mem-Benz was investigated. Mem-Benz and memantine were observed in the brains of 5xFAD mice injected with Mem-Benz 1 and 8 h postinjection, suggesting Mem-Benz breaks down in vivo to release memantine and a benzothiazole-based metabolite. However, the large fraction of intact Mem-Benz in the brains of mice 1 h postinjection indicates that future design of memantine prodrugs may involve optimization of the linker to increase the fraction of the prodrug that is metabolized to release memantine. By including an Aβ-targeting fragment in a memantine prodrug such as Mem-Benz, the bifunctional compound exhibits appreciable Aβ affinity, a characteristic not innate to memantine. Importantly, the post-mortem brain section staining experiments in 5xFAD mice indicate that the prodrug accumulates in the brain areas rich with Aβ plaques, and a similar accumulation of the prodrug in the brains of AD patients may decrease off-target inhibition observed for memantine and therefore reduce the side effects of clinical treatment.^66,67 Overall, the development of Aβ-targeting prodrugs with linkers of varied sensitivity may be beneficial in increasing the therapeutic efficacy and decreasing the off-target effects of drugs currently utilized for the treatment of Alzheimer’s disease.

METHODS

Synthesis and Characterization.

Mem-Benz.

2,4-dimethyladamantanamine (80 mg, 0.45 mmol) and paraformaldehyde (30 mg, 0.75 mmol) were combined in 4 mL of ethanol and brought to reflux for 1 h. 2-methoxy-4-benzothiazolylphenol (100 mg, 0.39 mmol) was added to the mixture in 6 mL of EtOH.³⁹ The mixture was stirred under refluxing conditions for 3 days before cooling. Solvent was removed by rotary evaporation before column chromatography (Hex/EA). The product fraction was additionally recrystallized in hexanes to remove trace starting material to yield a light yellow solid product (27.8 mg, 16% yield). ¹H NMR (CDCl₃, 500 MHz): δ (ppm) 8.00 (1H, d, 7.9 Hz), 7.86 (1H, d, 7.4 Hz), 7.54 (1H, d, 2.2 Hz), 7.45 (1H, t, 7.1 Hz), 7.33 (2H, m), 4.10 (2H, s), 3.99 (3H, s), 1.58 (2H, s), 1.43 (1H, s), 1.40 (1H, s), 1.36 (1H, s), 1.33 (4H, m), 1.25 (1H, s), 1.15 (3H, m), 0.86 (6H, m). ¹³C NMR (CDCl₃, 126 MHz): δ (ppm) 168.73, 154.44, 151.73, 149.04, 135.00, 124.39, 123.89, 121.73, 120.06, 53.68, 50.92, 48.38, 44.33, 42.93, 40.71, 32.72, 30.37. ESI-MS [M + 1]⁺: calc 449.2263, found 449.2261.

Done-Benz.

Desbenzyl donepezil (25.4 mg, 0.078 mmol) and paraformaldehyde (4.7 mg, 0.156 mmol) were combined in 15 mL of acetonitrile and left to reflux for 1 h. 2-methoxy-4-benzothiazolylphenol (10 mg, 0.039 mmol) was added in 10 mL of acetonitrile, and the mixture was left to reflux for 24 h.³⁹ Solvent was removed by rotary evaporation before reversed-phase column chromatography (H₂O/MeCN/0.1% TFA). The product was obtained as a light yellow solid product (21.8 mg, 83%). ¹H NMR (CDCl₃, 500 MHz): δ (ppm) 8.04 (1H, d, 8.1 Hz), 7.90 (1H, d, 7.9 Hz), 7.59 (1H, s), 7.49 (1H, t, 7.1 Hz), 7.38 (1H, s), 7.29 (1H, s), 7.20 (1H, s), 6.88 (1H, s), 4.04 (3H, s), 3.99 (3H, s), 3.94 (3H, s), 3.86 (2H, s), 3.30 (1H, dd, 17.4, 8.1 Hz), 2.73 (2H, m), 2.24 (2H, t, 11.8 Hz), 1.96 (1H, m), 1.84 (2H, d, 12.7 Hz), 1.68 (1H, m), 1.45 (3H, m), 1.28 (2H, s). ¹³C NMR (CDCl₃, 126 MHz): δ (ppm) 207.64, 168.61, 155.80, 154.44, 151.12, 149.77, 148.88, 148.61, 135.05, 129.48, 126.46, 124.96, 124.68, 122.92, 121.74, 120.76, 110.03, 107.60, 104.67, 61.49, 56.49, 56.42, 56.37, 53.69, 45.49, 38.71, 34.23, 33.82, 32.72, 32.08, 29.96. ESI-MS [M + 1]⁺: calc 559.2267, found 559.2269.

Done-Stil.

Desbenzyl donepezil (25.4 mg, 0.078 mmol) and paraformaldehyde (4.7 mg, 0.156 mmol) were combined in 15 mL of acetonitrile and left to reflux for 1 h. Pre-LS-4 (10 mg, 0.039 mmol) was added in 10 mL of acetonitrile, and the mixture was left to reflux for 24 h.³⁶ Solvent was removed by rotary evaporation before reverse-phase column chromatography (H₂O/MeCN/0.1% TFA). The product was obtained as a light yellow solid product (21.8 mg, 83%). ¹H NMR (CDCl₃, 500 MHz): δ (ppm) 8.04 (1H, d, 8. One Hz), 7.90 (1H, d, 7.9 Hz), 7.59 (1H, s), 7.49 (1H, t, 7.1 Hz), 7.38 (1H, s), 7.29 (1H, s), 7.20 (1H, s), 6.88 (1H, s), 4.04 (3H, s), 3.99 (3H, s), 3.94 (3H, s), 3.86 (2H, s), 3.30 (1H, dd, 17.4, 8.1 Hz), 2.73 (2H, m), 2.24 (2H, t, 11.8 Hz), 1.96 (1H, m), 1.84 (2H, d, 12.7 Hz), 1.68 (1H, m), 1.45 (3H, m), 1.28 (2H, s). ¹³C NMR (CDCl₃, 126 MHz): δ (ppm) 207.80, 155.82, 150.39, 149.75, 148.97, 148.37, 137.39, 136.38, 130.20, 129.45, 129.36, 128.77, 128.28, 127.82, 126.69, 126.53, 126.09, 124.29, 112.74, 112.16, 108.86, 104.61, 104.68, 56.49, 56.37, 56.22, 51.08, 51.05, 45.53, 40.73, 33.86. ESI-MS [M + 1]⁺: calc 673.3636, found 673.3620.

Log D_oct Determination.

DMSO stocks (5 or 10 mM) of the compound of interest were diluted in 1 mL of 1:1 mixtures of PBS-saturated octanol:octanol-saturated PBS to a final concentration of 250 μM in a microcentrifuge tube. The solution was vortexed for 1 min, then centrifuged at 10,000 rpm for 5 min. The octanol layer was removed to a well plate, and the PBS layer was transferred to a clean microcentrifuge tube. An equal volume of PBS-saturated octanol was added to the PBS layer before vortexing for 1 min and centrifugation at 10,000 rpm for 5 min. The octanol layer was removed to a well plate, and the fluorescence of the two octanol layers was measured. For benzothiazole-based compounds, λ_ex = 320 nm. For stilbene-based compounds, λ_ex = 360 nm. The log D_oct was calculated as the log of the ratio of fluorescence of the first octanol layer extraction to the second octanol layer extraction.

Amyloid-β₄₂ Fibril Formation.

Amyloid-β₄₂ (Aβ₄₂) monomer lyophilized powders were purchased from GL Biochem. Monomeric film aliquots were created by dissolving 1 mg of Aβ₄₂ in 1 mL of HFIP, aliquoting desired masses of Aβ₄₂ into microcentrifuge tubes, evaporating under vacuum overnight, and drying by vacufuge. Aβ₄₂ fibrils were created by dissolving 0.25 mg of monomeric films in 10 μL of DMSO and then diluting in 540 μL of PBS to a final concentration of 100 μM. The solution was incubated in an Eppendorf ThermoMixer C for 72 h, 37 °C, at 1000 rpm.

Fluorescence Turn-On.

DMSO stocks (5 or 10 mM) of the compounds of interest were diluted in PBS to a final concentration of 25 μM in the presence of 10 μM Aβ₄₂ fibrils. The mixture was incubated on an orbital shaker at room temperature for 15 min before fluorescence measurements were obtained. For benzothiazole-based compounds, λ_ex = 320 nm. For stilbene-based compounds, λ_ex = 360 nm.

K_i Determination.

K_i values were determined by incubating constant concentrations of Aβ₄₂ (5 μM) and Thioflavin T (ThT, 2 μM) with increasing concentrations of the compound of interest (0.01–100 μM, with maximum concentrations determined by the extent of aqueous solubility, even if binding saturation was not achieved). To a 96-well plate were added 5 μL of 40 μM ThT, the desired amount of the compound of interest, and PBS to bring each well to a total volume of 95 μL. To each experimental well was added 5 μL of 100 μM Aβ₄₂ fibrils; to each control well was added 5 μL of PBS instead of Aβ₄₂ fibrils. Each triplicate of experimental wells had one corresponding control well. The well plate was incubated on an orbital shaker at room temperature for 15 min before fluorescence measurements were obtained (λ_ex = 435 nm, λ_em = 485 nm). K_i values were calculated by a one-site competition nonlinear fit model with a ThT K_d value of 500 nM using GraphPad Prism version 8.0.2 for Windows, GraphPad Software, Boston, Massachusetts, www.graphpad.com.

Histological Brain Section Staining.

HJ3.4 antibody was purchased for the Holtzmann laboratory at Washington University. The antibody was labeled with a ThermoFisher Scientific Alexa Fluor 594 Antibody Labeling Kit (catalog number A30008) and stored at 4 °C until use. Brain sections were obtained from 9-month-old male 5xFAD mice perfused with PBS (0.1% heparin) prior to sacrifice and brain extraction. Brains were fixed with paraformaldehyde and stabilized in sucrose before cryosectioning at 40 μm thickness and then stored at −20 °C until use. Brain sections utilized for staining were washed with PBS (3 × 5 min) and cleared with 70% ethanol (1×, 5 min). Blocking in bovine serum albumin (BSA) in PBS (2 wt %, 10 min) was conducted before sections were transferred to a PBS solution of the compound of interest (25 μM, 30 min; for postinjection histological brain section staining, this step was omitted). The sections were transferred to a solution of AF594-labeled HJ3.4 antibody (HJ3.4-AF594) to costain (1 μg/mL, 30 min). Poststain blocking in BSA in PBS (4 min) and washing in PBS (3×, 2 min) were conducted before sections were mounted on microscope slides with nonfluorescent antifade mounting media. Images were obtained using an EVOS FL Auto 2 fluorescence microscope. Image analysis and colocalization by Pearson’s (R) and Mander’s (tM) coefficients were conducted using Fiji ImageJ.⁶⁸

Acetylcholinesterase Inhibition Assays.

Stock solutions of AChE from Electrophorus electricum (Millipore Sigma, catalog number C3389), Ellman’s reagent (5,5′-dithiobis(2-nitrobenzoic acid), DTNB), and acetylthiocholine iodide (ATChI) were prepared fresh daily in DMSO at 2 U/mL, 3 mM, and 3 mM, respectively.³⁴ In a clear bottom 96-well plate, a mixture of compound of interest (1 nM to 350 μM, maximum concentrations determined by extent of aqueous solubility, even if binding saturation was not achieved), 0.03 U of AChE, and 1.5 mM DTNB were combined in PBS and allowed to interact at room temperature with no agitation for 5 min. Each experimental concentration was tested in triplicate, and a corresponding control well lacking AChE was created for each experimental concentration. ATChI was added to each well to a final concentration of 0.3 mM, and the plate was allowed to incubate for 30 min at room temperature in the dark before the absorbance was measured at 405 nm. IC₅₀ values were calculated by a variable slope-four parameter nonlinear fit model using GraphPad Prism version 8.0.2 for Windows, GraphPad Software, Boston, Massachusetts, www.graphpad.com.

Molecular Docking.

All molecular docking studies were performed with the Schrödinger Suite software using AChE model eel 1C2B or NMDAR model human 7EU7 imported from the RCSB database and optimized by minimal minimization with the OPLS4 force field using the Protein Preparation Workflow program.^53,69 Molecular docking studies were conducted for the structures of donepezil, Done-Benz, and Done-Stil with PDB structure 1C2B, while memantine and Mem-Benz studies were conducted with PDB structure 7EU7. These ligands were prepared using Ligprep, and the pH was set as 7.0 ± 2.0 using Epik Classic. The different protonation states of the compounds were obtained and used for docking studies. The grid size was set using Receptor Grid Generation to include the following: the entire optimized protein structure in each direction for 1C2B and the minimal transmembrane portion, which includes the known memantine binding site, for protein structure 7EU7 (grid coordinates (x,y,z) | (125.130.90)).⁶⁵ The final molecular ligand docking was performed by Glide. The calculated poses were ranked by both the docking score and the Glide e-model energy. For each structure, the best pose was determined by the e-model, and the compounds were ranked against each other by docking scores. The compounds with the best docking scores and control compounds were rendered in PyMOL 3.1.

Metabolomic Investigation of Mem-Benz.

Three mice 8–9 months of age were intravenously injected with 2 mg/kg Mem-Benz (formulated in PBS supplemented with 10% Cremophor-EL-30 and 10% ethanol to aid solubility for a 2.5 mL/kg volume). Prior to injection, mice were anesthetized to a lack of pain response using isoflurane, then proparacaine hydrochloride was topically applied. Following injection, mice were allowed to awaken before CO₂ and cervical dislocation sacrifice with subsequent brain extraction 1, 8, or 24 h postinjection. The tissues were flash-frozen on dry ice and stored at −80 °C. Brain samples were analyzed for the target compound using liquid chromatography–mass spectrometry (LC-MS) by the Carver Metabolomics Core Facility of the Roy J. Carver Biotechnology Center, University of Illinois Urbana–Champaign. The samples were thawed, and 1 mL of methanol, along with 5 ceramic beads, was added to each. The samples were then homogenized for 5 min using a bead mill. After centrifugation, 100 μL of supernatant was transferred to a vial prespiked with 5 μL of 1 μg/mL chlorophenylalanine. A calibration curve ranging from 0.1–100 ng/mL of compounds were prepared in 70% methanol. Using an Agilent 1290 Infinity II UHPLC system (Agilent Technologies, Santa Clara, CA), target compounds were isolated using an Agilent Poroshell 12- EC-C18 (1.9 μm, 2.1 × 100 mm) column (Agilent Technologies, Santa Clara, CA) via a gradient method consisting of mobile phase A: water with 0.1% formic acid and mobile phase B: acetonitrile with 0.1% formic acid. The gradient was 0–0.2 min = 10% B; 0.2–3 min = 98% B; 3–4 min = 98% B; 4–4.1 min = 10%; 4.1–7 min = 10%. The flow rate was 0.45 mL/min, with the column temperature set at 40 °C. A Sciex 5500 triple quadrupole MS instrument (Sciex, Framingham, MA) operating in positive ionization mode used multiple reaction monitoring (MRM) to screen for the target compounds. Quantification was performed with Sciex Multi-Quant 3.1 (Sciex, Framingham, MA) using the area ratio of target compounds and chlorophenylalanine, the spiked standard. The detection limit of memantine was 0.1–100 ng/mL, and Mem-Benz and 2-methoxy-4-benzothiazolylphenol were 0.1–50 ng/mL.

Supplementary Material

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acschemneuro.5c00493.

Additional experimental details, materials, methods, ¹H NMR, ¹³C NMR, and high-resolution ESI-MS characterization data of all compounds (PDF)