Alzheimer’s disease (AD) is a progressive neurodegenerative disorder characterized by deteriorating memory, declining language skills, and a range of cognitive impairments in older adults [
1]. As of now, over 55 million individuals worldwide are diagnosed with AD, a figure projected to rise to 78 million by 2030 [
2]. The disease exerts substantial social and economic impacts. Currently, there are no effective therapeutic options available for its treatment.
Despite considerable research efforts, the precise etiology of AD remains unclear. Pathological alterations observed in the brains of AD patients include decreased levels of acetylcholine (ACh), the accumulation of amyloid-
β (A
β) deposits, heightened oxidative stress, hyperphosphorylated tau neurofibrillary tangles, dysregulation of biometal ions, and inflammation [
2,
3]. The complexity inherent in the etiology of AD has prompted researchers to investigate multi-target-directed ligands (MTDLs) as a promising therapeutic strategy. MTDLs are designed to interact simultaneously with multiple disease-related targets, aiming to produce a synergistic effect that enhances clinical outcomes [
4–
6].
The classical "cholinergic hypothesis" posits that the neurotransmitter ACh is critical for nerve cell function related to learning and memory. Abnormally low levels of ACh in the hippocampus and neocortex are directly linked to cognitive decline in AD [
7]. Acetylcholinesterase (AChE) and butyrylcholinesterase (BuChE) are two key enzymes in this context, with BuChE typically acting as a backup to AChE. In AD patients, AChE levels can decline by up to 90%, while BuChE levels may increase to around 165% of normal [
8,
9]. This shift suggests that BuChE compensates for the diminished AChE activity in ACh hydrolysis. Inhibition of BuChE has shown promise in reducing the aggregation of fibrillogenic A
β1–42 peptides, and studies utilizing BuChE knockout models have revealed no significant physiological deficits, indicating its potential as a therapeutic target in advanced AD [
10–
12]. Many selective BuChE inhibitors have been developed and successfully applied for treating AD [
8,
13]. Therefore, developing MTDLs with BuChE inhibition represents a promising approach for treating AD.
Natural products are extensively researched for their potential in developing anti-AD drugs, owing to their significant advantages, such as structural diversity and inherent multifunctional biological activities [
14–
17]. For instance, natural flavonoid compounds are investigated for their antioxidant, anti-inflammatory, and neuroprotective properties; however, their clinical application is hindered by low bioavailability and poor permeability across the blood-brain barrier. Rivastigmine, a pseudo-irreversible, carbamate-type, brain-selective dual inhibitor of AChE and BuChE, has been approved by the Food and Drug Administration (FDA) for the treatment of AD [
18]. The carbamate moiety of rivastigmine is recognized as the pharmacophore responsible for its cholinesterase inhibitory activity [
19–
21]. By integrating the pharmacophore of rivastigmine with the core scaffolds derived from natural products, it is anticipated that multi-target BuChE inhibitors can be developed as promising candidates for AD therapy.
In this study, we propose a strategy that integrates combinatorial compound design, machine learning (ML)-based virtual screening, and structure-activity relationship (SAR) studies for the
de novo design of BuChE inhibitors and the assessment of their anti-AD activity [
22–
24]. As shown in
Fig. 1A, the overall research framework consists of the following steps: (ⅰ) designing a combinatorial compound library based on natural products and rivastigmine; (ⅱ) constructing ML-based activity prediction models; (ⅲ) identifying hit compounds through multistep virtual screening; and (ⅳ) evaluating anti-AD drug efficacy
in vitro and
in vivo studies. Employing this approach, we identified a novel BuChE inhibitor, designated as compound
3c (half maximal inhibitory concentration (IC
50) = 0.43 µmol/L). Docking simulations and molecular dynamics studies provided a rational basis for its enhanced activity. The significant efficacy of compound
3c warranted comprehensive investigations into its reversible inhibition, neuroprotective effects, drug-likeness, pharmacokinetics, and safety profile.
In vivo studies were conducted using AD zebrafish models and scopolamine-induced mice. These findings represent a substantial advancement in the design and development of highly specific and potent BuChE inhibitors, positioning compound
3c as a promising candidate for the treatment of AD.
Herein, 156 natural products were selected as core scaffolds. Considering the carbamate moiety as a critical pharmacophore in Rivastigmine, carbamate modifications were introduced at the
ortho, meta, and
para positions of the selected scaffolds. This strategy led to the construction of a library containing 389 fragment-based combinatorial compounds. To develop an accurate ML predictor for BuChE inhibitory activities, we created a database of 2689 compounds with varying inhibition levels, sourced from the literature. After removing duplicates and ambiguous data, 2361 compounds remained. We then applied five ML models-support vector machine (SVM), logistic regression (LR), k-nearest neighbor (KNN), random forest (RF), partial least squares (PLS), and a deep feedforward neural network (DFNN). Based on prediction accuracy and balanced test set performance (
Figs. 1B and
C), we selected KNN as the optimal model. This choice was supported by robust statistical metrics:
R2 of 0.800 and mean squared error (MSE) of 0.308 for the training set, and
R2 of 0.718 with MSE of 0.432 for the test set, with a relative error under 25%, demonstrating high reliability.
We first used the optimal KNN model to predict the activity of 389 combinatorial compounds, resulting in the identification of 58 compounds with predicted pIC
50 values greater than 5, indicating micromolar-level activity. These 58 compounds were then prioritized for further analysis through molecular docking and virtual screening, which led to the selection of 6 promising candidates for chemical synthesis and bioactivity testing (
Figs. 1D and
E). The optimized compounds with their detailed properties were summarized in
Fig. 1E. The results demonstrated that the ML model provided accurate predictions of compound activity, reflected by the consistent correlation between predicted and observed pIC
50 values. Among these, compound
HK-1 stood out with its submicromolar inhibitory activity, making it a top candidate for further optimization as a lead compound.
To develop more effective BuChE inhibitors, we conducted structural optimization and SAR studies based on the lead compound HK-1 (Fig. S1 in Supporting information). The synthetic routes for the novel target derivatives were illustrated in Schemes S1–S3 (Supporting information). In brief, as shown in Schemes S1 and S2, commercially available starting materials, honokiol (1a) or magnolol (1b) (2.0 mmol), were treated with excess N,N-disubstituted carbamoyl chlorides (2a–2d) (5.0 mmol), including N-ethyl-N-methylcarbamoyl chloride (2a), dimethylcarbamoyl chloride (2b), diethylcarbamoyl chloride (2c), and diisopropylcarbamoyl chloride (2d), in the presence of K2CO3 in CH3CN at 60–65 ℃, yielding target compounds 3a–3c and 4a–4d, respectively. Subsequently, as depicted in Scheme S3, magnolol (1a) (2.0 mmol) was reacted with excess N,N-disubstituted carbamoyl chlorides (2a–2d) (3.0 mmol), respectively, under the same reaction conditions to obtain target compounds 5a–5d.
In order to evaluate the AChE/BuChE inhibitory activity of target derivatives, the modified Ellman’s method was applied using
Electrophorus electricus AChE (
eeAChE) and equine serum BuChE (
eqBuChE) [
25]. As listed in
Table 1, when changing
N-ethyl-
N-methylcarbamate of
3a (
HK-1) with dimethylcarbamate to get compound
3b, the BuChE inhibition slightly decreased to 1.2 µmol/L, while replacing
N-ethyl-
N-methylcarbamate of
3a with diethylcarbamate to afford compound
3c, the BuChE inhibitory potency significantly improved to 0.43 µmol/L. Moreover, when replacing the honokiol skeleton of
3a (
HK-1) with magnolol skeleton to obtain compound
4a, demonstrating weak BuChE inhibition (10.3% at 12.5 µmol/L) and moderate AChE inhibition (IC
50 = 12.5 µmol/L). And the SAR exhibited that replacing
N-ethyl-
N-methylcarbamate of
4a with dimethylcarbamate, diethylcarbamate and diisopropylcarbamoyl to get derivatives
4c–
4d, respectively, indicating weak BuChE inhibition and AChE inhibition (except
4d, IC
50 = 6.8 µmol/L). Further, replacing bissubstituted carbamates of
4a–
4d with monosubstituted carbamates to afford derivatives
5a–
5d, respectively. The data showed that compounds
5a–
5d indicating moderate to good
eeAChE inhibitory activity with IC
50 values from 8.9 µmol/L to 17.9 µmol/L, and weak
eqBuChE inhibitory potency. Among of these derivatives, compound
5a with monosubstituted
N-ethyl-
N-methylcarbamate derivative showed moderate
eeAChE inhibition (IC
50 = 8.9 µmol/L) and weak
eqBuChE inhibition (IC
50 = 13.8 µmol/L), which did not reach up to our expectations. The results suggested that magnolol was not ideal skeleton to develop BuChE inhibitor. As a whole, through multiple rounds of compound efforts, we finally obtained the most promising selective
eqBuChE inhibitor
3c (IC
50 = 0.43 µmol/L), deserving to further investigate.
The active site of BuChE is a narrow hydrophobic cavity, with Ser198 and His438 serving as the catalytic active sites, while Trp231 and Leu286 form the acyl binding pocket (
Fig. 2A). Herein, to accurately evaluate the binding free energy of hit compound
3c with BuChE, we performed binding free energy studies [
26,
27]. The binding stability of compound
3c with BuChE was evaluated using GROMACS 2022.5. During the 50 ns molecular dynamics (MD) simulation, the system remained stable (Fig. S2 in Supporting information). The root mean square deviation (RMSD) of the complex and protein fluctuations remained below 2 Å during the last 20 ns, indicating that the binding of
3c with BuChE had stabilized (
Fig. 2B and Fig. S3 in Supporting information). The lowest energy conformation revealed that compound
3c snugly fitted into a hydrophobic pocket formed by residues Trp82, Thr120, Glu197, Trp231, Phe329, Tyr332, and His438 (
Fig. 2C). It was worth noting that diethylcarbamate was embedded in a narrow hydrophobic cavity formed by Trp231 and Val228, where it formed van der Waals interactions. The total binding free energy was calculated as −32.93 kcal/mol, with van der Waals interactions contributing significantly to the ligand-receptor binding (−49.18 kcal/mol) (
Fig. 2D). Key residues such as Phe329 (−2.20 kcal/mol), Tyr332 (−1.84 kcal/mol), and Glu325 (−1.50 kcal/mol) played crucial roles in the interaction between
3c and human BuChE (hBuChE), emphasizing their importance in stabilizing the ligand binding (
Fig. 2E).
To assess the reversible inhibition of BuChE by compound
3c, we conducted a time-course evaluation of BuChE activity recovery following a dilution protocol [
28]. As depicted in
Fig. 3A, treatment with 0.1 × IC
50 of donepezil resulted in a significant restoration of BuChE activity, increasing from 47.8% to 77.8% over 240 min, indicating its characteristic reversible inhibition. In contrast, BuChE activity in the presence of 0.1 × IC
50 of rivastigmine remained relatively stable throughout the 240-min observation period, with no significant recovery. Notably,
3c exhibited a profile similar to that of rivastigmine, suggesting that it might act as a
pseudo-irreversible inhibitor of BuChE.
Compound
3c was selected to evaluate its anti-inflammatory activity by assessing the production of nitric oxide (NO) and interleukin-6 (IL-6) in lipopolysaccharide (LPS)-induced BV-2 microglial cells [
29]. As shown in
Fig. 3B, compound
3c exhibited a favorable safety profile with minimal cytotoxicity at concentrations below 10 µmol/L. Moreover, as illustrated in
Figs. 3C and
D, compound
3c significantly reduced the production of NO and the secretion of IL-6 in a dose-dependent manner. Collectively, these findings suggested that compound
3c had potential as an anti-inflammatory agent. The effects of
3c on self-induced A
β1–42 aggregation were evaluated using the thioflavin T (ThT) fluorescence assay, with curcumin serving as a positive control [
30]. As summarized in
Table 1, at a concentration of 25 µmol/L, compound
3c exhibited a significant inhibitory effect on self-induced A
β1–42 aggregation, achieving an inhibition rate of 63.4%, surpassing that of curcumin, which was 47.3%.
The neuroprotective effects of compound
3c were evaluated in Glu-/A
β25–35-induced HT22 cell injury using the cell counting kit-8 (CCK-8) assay, with rivastigmine included as a positive control [
30]. As shown in
Fig. 3E,
3c displayed minimal cytotoxicity in HT22 cells at concentrations below 40 µmol/L.
Fig. 3F illustrated that exposure to 500 µmol/L l-glutamic acid significantly reduced cell viability to 50.0% (
P < 0.01). In contrast, treatment with various concentrations of compound
3c and rivastigmine (1, 2, 4, 8, and 16 µmol/L) resulted in a significant dose-dependent increase in cell viability, indicating that the neuroprotective potential of
3c was markedly superior to that of rivastigmine. Furthermore, as demonstrated in
Fig. 3G, compound
3c provided significant neuroprotection against A
β25–35-induced injury in HT22 cells in a similar dose-dependent manner compared to rivastigmine. Collectively, these findings suggested that compound
3c offered substantial neuroprotective effects against Glu- and A
β25–35-induced cellular injury, underscoring its potential therapeutic relevance in neurodegenerative disorders.
The permeability of the blood-brain barrier (BBB) is a critical factor in the development of effective therapeutic agents for AD [
31]. In this context, compound
3c was evaluated for its BBB permeability using a parallel artificial membrane permeation assay for the BBB (PAMPA-BBB). Initial validation of the PAMPA-BBB approach involved 11 well-characterized drugs with known BBB permeability profiles, as reported in our previous studies [
32]. The following permeability classifications were established:
Pe < 1.61 × 10
−6 cm/s indicated weak BBB permeation; 1.61
< Pe < 3.44 × 10
−6 cm/s indicated uncertain BBB permeation; and
Pe > 3.44 × 10
−6 cm/s signified high BBB permeability. As outlined in
Fig. 4A, the calculated
Pe value for compound
3c was 13.7 × 10
−6 cm/s, indicating that compound
3c had a high capacity to cross the BBB
in vitro, thereby supporting its potential as a therapeutic candidate for AD.
The exceptional biological characteristics of zebrafish provide a valuable complement to traditional drug development models [
33]. Initially, we assessed the safety profile of compound
3c using six different concentrations (10.0, 5.0, 2.5, and 1.25 µg/mL) in a zebrafish model (Fig. S4 in Supporting information). Notably, no mortality was observed at 1.25 µg/mL over a 48-min exposure period. Following this, we conducted a more detailed evaluation of developmental toxicity by examining pericardial edema relative to body length and the swim bladder area at two concentrations (1.25 and 0.625 µg/mL). The results indicated a significant reduction in both pericardial edema/body length and swim bladder area at the higher concentration of 1.25 µg/mL. Conversely, at 0.625 µg/mL, there were no statistically significant changes observed.
The AlCl
3-induced zebrafish AD model is a well-established platform for the rapid screening of potential anti-AD agents [
30,
34]. Based on the previously determined maximum tolerable concentration (MTC) of 0.625 µg/mL for compound
3c, we subsequently selected three lower concentrations (0.01, 0.02, and 0.04 µg/mL) for further evaluation following preliminary screening. To assess the therapeutic effects of these treatments, a series of behavioral tests were conducted under alternating light and dark conditions over a 60-min period (
Fig. 4B). The experimental design included seven groups: control group,
3c treatment group, AlCl
3-induced model group, AlCl
3 + donepezil group, and AlCl
3 +
3c treatment groups at concentrations of 0.01, 0.02, and 0.04 µg/mL. As shown in Figs. S5A–C (Supporting information), zebrafish exposed to AlCl
3 demonstrated a statistically significant reduction in swimming distance compared to the control group under both light and dark conditions. In contrast, treatment with donepezil and compound
3c at concentrations of 0.01, 0.02, and 0.04 µg/mL resulted in a marked increase in swimming distance, indicating an ameliorative effect on locomotor activity. Furthermore, the cumulative analysis of total swimming distance over the 60-min period (Fig. S5C in Supporting information) revealed that both donepezil and compound
3c at all tested concentrations significantly enhanced the total swimming distance when compared to the AlCl
3-induced zebrafish model group. These findings suggest that compound
3c exhibits potential anti-AD activity that warrants further investigation.
Furthermore, swimming speed significantly decreased in the AlCl3-induced zebrafish AD group compared with the control group during light conditions, respectively (Fig. S5D in Supporting information). Following treatment with donepezil and 3c (0.01, 0.02, and 0.04 µg/mL), the swimming speed remarkably improved compared with the model group. And this trend was also observed under dark conditions (Fig. S5E in Supporting information). Additionally, the total swimming speed over the 60-min period demonstrated that both donepezil and 3c (0.01, 0.02 and 0.04 µg/mL) significantly increased swimming speed compared with the AlCl3-induced zebrafish AD group (Fig. S5F in Supporting information). During the transition from light to dark (Fig. S5G in Supporting information), donepezil and 3c (0.04 µg/mL) significantly enhanced reaction capacity compared to the AlCl3-induced zebrafish AD group. Further, the switch from dark to light revealed that both donepezil and 3c (0.01, 0.02 and 0.04 µg/mL) significantly enhanced reaction capacity compared to the AlCl3-induced zebrafish AD group (Fig. S5H in Supporting information).
Following behavioral assessments, we conducted histopathological analyses using hematoxylin and eosin (HE) staining, as depicted in
Figs. 4C and
D. The findings indicated that treatment with AlCl
3 resulted in a significant increase in the number of apoptotic cells within the zebrafish brains. In contrast, administration of compound
3c at concentrations of 0.01, 0.02, and 0.04 µg/mL led to a significant reduction in apoptotic cell counts in a dose-dependent manner, yielding results comparable to those observed with donepezil. Furthermore,
Figs. 4E and
F illustrated that compound
3c elevated ACh levels by inhibiting BuChE, thereby exerting anti-AD effects. Additionally, compound
3c was associated with a decrease in the levels of amyloid precursor protein (APP), as shown in
Fig. 4G, suggesting a potential mechanism for mitigating amyloidogenesis. Moreover,
3c demonstrated substantial anti-inflammatory activity, evidenced by the reduction of pro-inflammatory cytokines, including human tumor necrosis factor alpha (TNF-
α,
Fig. 4H), interleukin-6 (IL-6,
Fig. 4I), IL-1
β (
Fig. 4J), and IL-10 (
Fig. 4K). In summary, the cumulative findings from this study indicated that
3c significantly ameliorated the effects of the AlCl
3-induced zebrafish AD model by inhibiting BuChE activity and elevating ACh levels. Additionally,
3c exhibited remarkable anti-inflammatory properties and reduced APP levels, underscoring its potential as a therapeutic agent in the management of AD.
Compound
3c was selected to evaluate its effects on memory deficits induced by scopolamine in mice, using the step-down passive avoidance task as the primary assessment method. Animal studies were conducted in accordance with the approved protocol by the Animal Ethics Committee of the Sichuan Academy of Traditional Chinese Medicine (IACUC) (approval number: R-20240721–01). To establish the safety profile of
3c, Kunming mice (18–22 g) were administered 500 mg kg
−1 day
−1 via gavage for 14 days. The treatment did not result in any significant changes in movement behavior, mental state, or body weight, indicating a favorable safety profile. Subsequently, the efficacy of
3c in alleviating scopolamine-induced memory deficits was assessed [
25,
28]. As shown in
Fig. 5, administration of 5.0 mg/kg scopolamine resulted in a significant decrease in step-down latency to 51.7 s in the model group (
P < 0.01), compared to 239.8 s in the control group. Treatment with 5.0 mg/kg donepezil significantly improved step-down latency, increasing it to 186.0 s (
P < 0.01) relative to the model group. Furthermore, treatment with compound
3c at doses of 5.6 and 16.8 mg/kg resulted in step-down latencies of 90.7 and 183.5 s (
P < 0.01), respectively, demonstrating that
3c significantly mitigated scopolamine-induced memory impairment.
AD is a chronic, progressive neurodegenerative disorder primarily observed in the elderly population. To date, there are no effective treatments available. The MTDLs strategy is considered one of the most promising approaches for addressing the complex pathogenesis of AD. Natural products have been extensively investigated for their potential anti-AD effects due to their diverse chemical structures and multiple pharmacological activities. However, their clinical applications are often limited by low bioavailability and poor permeability across the BBB. Rivastigmine, an FDA-approved dual cholinesterase inhibitor for the treatment of AD, contains a carbamate moiety that is crucial for its pharmacological activity. To develop natural product-derived multi-target compounds with BuChE inhibitory activity, we constructed a combinatorial compound library based on natural products, ultimately identifying a novel BuChE inhibitor, HK-1, as the lead compound through multi-step virtual screening.
Subsequently, HK-1 underwent several rounds of structural optimization, resulting in candidate compound 3c. Results demonstrated that compound 3c exhibited favorable BuChE inhibitory activity (IC50 = 0.43 µmol/L) and potential anti-inflammatory properties. Furthermore, compound 3c displayed significant inhibitory capacity against Aβ1–42 aggregation and remarkable neuroprotective effects. In vivo studies revealed that 3c significantly ameliorated the effects of AlCl3-induced zebrafish AD model by inhibiting BuChE activity and elevating ACh levels. Additionally, 3c demonstrated notable anti-inflammatory effects and reduced levels of APP. Importantly, compound 3c also improved scopolamine-induced memory impairment. Collectively, these findings suggested that compound 3c was a promising artificial intelligence (AI)-driven multifunctional agent with BuChE inhibition potential for the treatment of AD.
PDF
Full
77
44
关闭全屏
No. 86 Xueyuan South Road, Haidian District, Beijing
010-62662699(Operation and Technical Support)
qkjq@cast.org.cn
